ii PUBLISHED BY THE PRESS SYNDICATE OF THE UNIVERSITY OF CAMBRIDGE The Pitt Building,Trumpington Street,Cambridge,United Kingdom CAMBRIDGE UNIVERSITY PRESS The Edinburgh Building,

Agriculture: The PotentialConsequences of Climate Variability

and Change for the United States

ii

PUBLISHED BY THE PRESS SYNDICATE OF THE UNIVERSITY OF CAMBRIDGEThe Pitt Building,Trumpington Street, Cambridge, United Kingdom

CAMBRIDGE UNIVERSITY PRESSThe Edinburgh Building, Cambridge, CB2 2RU, UK

40 West 20th Street, New York, NY 10011-4211, USA10 Stamford Road, Oakleigh,VIC 3166,Australia

Ruiz de Alarcón 13, 28014 Madrid, SpainDock House,The Waterfront, Cape Town 8001, South Africa

http://www.cambridge.org

First published 2001

Printed in the United States of America

ISBN paperback

The recommended citation for this report is:

Reilly, J., et al. 2001. Agriculture: The Potential Consequences of Climate Variability and Change forthe United States, US National Assessment of the Potential Consequences of Climate Variability and

Change, US Global Change Research Program.Cambridge University Press, New York, NY, 136 pp.

iii

Agriculture Sector Assessment Team

Co-Chairs

John M. Reilly, Massachusetts Institute of Technology

Jeff Graham, US Department of Agriculture (through September 1999)

James Hrubovcak, US Department of Agriculture (from October 1999)

Assessment Team

David G.Abler, Pennsylvania State University

Robert A. Brown, Pacific Northwest National Laboratory

Roy F. Darwin, US Department of Agriculture

Steve E. Hollinger, University of Illinois

R. Cesar Izaurralde, Pacific Northwest National Laboratory

Shrikant S. Jagtap, University of Florida – Gainesville

James W. Jones, University of Florida – Gainesville

John Kimble, US Department of Agriculture

Bruce A. McCarl,Texas A&M University

Linda O. Mearns, National Center for Atmospheric Research

Dennis S. Ojima, Colorado State University

Eldor A. Paul, Michigan State University

Keith Paustian, Colorado State University

Susan J. Riha, Cornell University

Norman J. Rosenberg, Pacific Northwest National Laboratory

Cynthia Rosenzweig, Goddard Institute for Space Studies

Francesco N.Tubiello, Goddard Institute for Space Studies

Additional Contributors: Richard Adams,Department of Agricultural and Resource Economics,Oregon State University, Chi-Chung Chen, Department of Agricultural Economics, National ChungHsing University, Taiwan, Keith Fuglie, CIP-ESEAP-Indonesia, Tasana Gillig, Department ofAgricultural Economics,Texas A&M University

iv

This report was prepared for the US Federal Government as part of the US National Assessment: The PotentialConsequences of Climate Variability and Change. The US Department of Agriculture provided the principal sourceof funding through the Global Change Program Office, Office of the Chief Economist. The US Department ofEnergy provided substantial funding for participation of the Pacific Northwest National Laboratory. The FarmFoundation and the Economic Research Service co-sponsored the initial stakeholder meeting.

Photo Credits: Crops in front of snow covered mountains,Alaska Division of Agriculture, Department of NaturalResources, State of Alaska; Irrigation, Melody Warford, Stone Soup, Inc.; U.S. long grain rice, USDA,ARS Photo byKeith Weller; Erosion in Whitman County,Washington, USDA Photo by Ron Nichols; Maize, USDA,ARS Photo byKeith Weller;York County, MD, USDA Photo by Tim McCabe; Field ready for harvest on the Stockebrand Farm inKansas, USDA Photo by Anson Eaglin;Threshers, USDA Photo by Ron Nichols; Hawaiian (Carica) papayas, USDA,ARS Photo by Scott Bauer.

Cover design: Melody Warford, Stone Soup, Inc., www.stonesoupinc.comInterior design: Michael Shibao, Imaging and Design Center, University Corporation for Atmospheric ResearchEditor: David M. Stearman,WordMaven Editorial Services,Annandale,VA

v

Table of Contents

Foreword............................................................................................................................................................................vii

Preface .................................................................................................................................................................................ix

Summary............................................................................................................................................................................xi

Chapter 1: Changing Climate and Changing Agriculture..................................................................................1

Chapter 2: Assessment Approach: Building On Existing Knowledge ............................................................15

Chapter 3: Impacts of Climate Change on Production Agriculture and the US Economy...............................................................................................................................35

Chapter 4: Impacts of Variability on Agriculture................................................................................................67

Chapter 5: Agriculture and the Environment: Interactions with Climate ................................................83

Chapter 6: Conclusions and Implications...........................................................................................................109

References .......................................................................................................................................................................125

Appendix 1: Agriculture Sector Assessment Working Papers ........................................................................135

Appendix 2: Abbreviations.......................................................................................................................................136

vii

Assessment efforts of this type offer an opportunityfor researchers to apply their research tools andexpertise to issues of national importance. We cameinto this effort hoping that the years spent analyz-ing, modeling, and studying will provide some mea-sure of useful guidance to those who have commis-sioned the assessment. The efforts provide anopportunity to compare results among colleaguesand to deepen one’s understanding of the findingsof other disciplines. I learned much from my col-leagues, who graciously and enthusiastically accept-ed the invitation to serve on the team.The fundingavailable for the assessment was adequate to sup-port specific modeling tasks and essential travel.Team members generously contributed time wellbeyond the tasks that were specifically funded. Forthis I am grateful. It is my hope that members foundthe experience rewarding and thus found participa-tion worthwhile.

This report represents the combined efforts of theAgriculture Sector Assessment Team but I would beremiss if I failed to point out the substantial contri-butions of the individual team members. FrancescoTubiello coordinated the crop model scenarios pro-duced by the suite of crop models run by GoddardInstitute for Space Studies (GISS), the University ofFlorida, and the Natural Resource EcologyLaboratory at Colorado State.The protocols and sitedata developed at GISS by Cynthia Rosenzweig forprevious assessments were graciously made availableto the teams of crop modelers. In addition toFrancesco Tubiello at GISS, Shrikant Jagtap, JimJones, Keith Paustian, and Dennis Ojima composedthe crop modeling teams that developed compre-hensive and consistent scenarios for the two climatescenarios evaluated. The Pacific Northwest NationalLaboratory (PNNL) team of Cesar Izaurralde andNorman Rosenberg and assisted by Robert Brownapplied a model with more geographically compre-hensive coverage for several crops for one climatescenario. This provided an opportunity to assess thedifferences that arose from methodological differ-ences of this approach compared to the detailed site

Foreword

approach used by the other teams. Keith Paustianand Dennis Ojima organised a crop modeling work-shop to compare, in more depth, the performance ofthese models at selected sites to further understandthe types of uncertainties that differing model struc-tures could introduce. Linda Mearns contributed hercrop modeling expertise as well as her expertise onvariability and extreme events. A separate study shewas leading that was funded by the National ScienceFoundation provided critical coverage for cotton.

Bruce McCarl developed national yield changesbased on the site results from the crop studies andsimulated economic effects. He, with several co-authors, also investigated several other aspects ofthe problem including the dependence of pesticideexpenditures on climate, economic effects ofchanges in El Niño, and he interacted with the WaterSector Assessment to assure that our water supplyassumptions were consistent with their estimates.Roy Darwin provided results on impacts on tradebased on recent analyses that he has conductedwith his global model. This large effort was possiblewithin the short time-frame and restricted budgetbecause of the tremendous expertise and experi-ence of these team members.

In other aspects of the assessment, the analyticaltools and approaches for conducting an integratedassessment have not yet been fully developed. Herewe relied on modeling case studies, creative evalua-tion of historic data, and judgment of experts. SteveHollinger studied data on crop variability over thepast 100 years to provide an historical perspectiveon adaptation. Keith Fuglie, CIP-ESEAP-Indonesia,while not a part of the team provided additionalanalyses of historical variability of crop yields.David Abler applied a newly developed model of theeconomics of water quality in the Chesapeake BayRegion and summarized potential environmental/agro/climatic interactions. Eldor Paul and JohnKimble evaluated potential effects of climate changeon soils. Susan Riha provided a summary of our cur-rent understanding of carbon dioxide effects on

viii

plant growth and the potential to develop new cropvarieties as a response to climate change andincreased ambient CO2 levels.These efforts pushedinto some new, but critical territories, lending per-spectives we otherwise would not have.

I am also grateful for the time our SteeringCommittee took from their busy schedules to guidethe effort. I know we have not answered all of thequestions they raised but hope that we haveanswered at least some of them. My thanks also toJeff Graham of USDA. He left USDA before thereport was completed, but left his mark on theeffort. Margot Anderson, Director of the GlobalChange Program Office at USDA, was our initial con-tact, secured funding from the USDA agencies, anddid her best to keep us on track and responsive tothe goals of the assessment before leaving USDA inearly 2000. Jim Hrubovcak and William Hohensteinpicked up where Jeff and Margot left off and contin-ued to provide support for the effort. The studybenefited from the steady support of USDA frominception through the review process even as thepeople changed. USDA remained true to the con-cept of the National Assessment as a scientific andpublic participation activity—providing funding,personnel, technical support, and scientific review.A multi-institution activity of this type also requiredarrangement of travel, meetings, and contracting.Much of this support came from the UniversityCorporation for Atmospheric Research (UCAR). Mythanks to Tara Jay, Gene Martin, Jim Menghi,AmySmith, and Kyle Terran.

Finally, our effort was possible because of the sup-port of the National Assessment CoordinationOffice, led by Michael MacCracken. We were fortu-nate that NACO provided at our disposal two excep-tional staff members who decided to start out theirpromising careers by becoming involved in theNational Assessment. Justin Wettstein provided sup-port through the planning and initiation stages ofthe agricultural assessment in 1998 and 1999.LaShaunda Malone took Justin’s place in the latesummer of 1999 and provided a seamless transitionand continued support through the completion ofour report. Both worked many extra hours and on aday-to-day basis kept the agricultural assessment pro-cess moving while providing support for manyother aspects to the National Assessment. Manythings that we discovered needed to be done late inthe afternoon on a Friday appeared in e-mail earlythe following Monday. Their help was especiallycritical and greatly appreciated.

John ReillyFebruary, 2001

ix

This report is part of the US National Assessmentprocess, the Potential Consequences of ClimateVariability and Climate Change, published byCambridge University Press (National AssessmentSynthesis Team, 2000, 2001). In addition to summa-rizing scientific understanding about the potentialconsequences for the Agriculture sector, the reportprovided input for the two-part national level reportentitled, Climate Change Impacts on the UnitedStates which has been aimed at evaluating theimpacts of climate change and climate variability onthe United States, across its various regions andincluding sectors beyond agriculture.

The US National Assessment was undertaken as ajoint activity of the federal government with aca-demic institutions, local governments, and publicand private groups to understand the implications ofclimate change and climate variability for the nation.Periodic assessments of global change research andthe implications of global change for the US weremandated by Congress when the US Global ChangeResearch Program (USGCRP) was authorized by theGlobal Change Research Act (GCRA) of 1990.TheUSGCRP initiated the National Assessment activity tofulfill, in part, the requirement for a periodic assess-ment in the GCRA. Details about the NationalAssessment beyond those provided here and links toother related sites can be found at http://www.nacc.usgcrp.gov.

The National Assessment includes regional assess-ment activities that are intended to make researchresults relevant and useful to conditions, issues, andconcerns that vary across the country. Sector assess-ment activities also are incorporated, being designedto integrate the analysis across issues of national sig-nificance that bridge across regions and that mayinvolve topics such as interregional and internation-al trade and competitiveness. In addition to the agri-culture sector assessment, the National Assessmenthas sponsored sectoral assessments for forests, waterresources, human health, and coastal areas andmarine resources.Although this list of sectors andactivities affected by climate variability and climatechange is not comprehensive, the sector assessmentactivities cover some of the sectors and systems that

are presently perceived as most sensitive to climate.An important goal of the National Assessment hasbeen that it be participatory and seek to engagestakeholders and the public.This philosophy flowsfrom the belief that applied science must be applica-ble to the needs of those who are expected to useit. Research is far more likely to be applicable if theusers and potential users are involved throughoutthe assessment process. In this spirit, the agriculturesector assessment sought a steering committeecomposed of stakeholders and potential users of theresearch to guide its activities. The individuals whoaccepted this responsibility and actively and gra-ciously participated are listed below. The full reportof the initial meeting of the steering committee andsector assessment team is available at http://www.nacc.usgcrp.gov/sectors/agriculture/workshop-report.pdf.

In carrying through these responsibilities, the agri-culture sector assessment team has made an effortto coordinate closely with those regional assessmentactivities that have included a significant agricultureassessment (to the extent that the schedules of theefforts were compatible). Indeed, several membersof the agriculture sector assessment team areresponsible for agriculture assessment activities invarious regional assessments. Similar efforts havebeen made to coordinate with the other sectorassessments.Where possible, results from other sec-tors (for example, changes in water availability andtheir impact on irrigation water supplies) have beenused as input into our evaluation of agriculture.Furthermore, as with regions, the agriculture sectorassessment team overlaps with other sector assess-ment teams (i.e., water and forests), and so coopera-tive interactions have occurred quite naturally.

As part of the National Assessment, some aspects ofthe approach were under our direction and somewere guided by the need for consistency across thevarious assessment activities. For example, withregard to future climate scenarios, our guidance wasto focus on using the Canadian Climate Centre andHadley Centre climate scenarios as one basis for pro-jecting possible future climate change; in addition,we have used records of historic climate variability

Preface

x

to evaluate resilience to recurrence of such condi-tions and considered whether possible thresholdsmight be important.The National Assessment alsoprovided some guidance on future socioeconomicscenarios.We did not, however, develop numericalagroeconomy scenarios consistent with the econom-ic scenarios; rather we imposed climate change onthe agricultural economy as it exists today becauseprojecting agricultural developments might wellintroduce additional uncertainties.

In keeping with the purpose and goals of theNational Assessment, the agriculture sector assess-ment report has two broad objectives:

• To respond to the goals of the GCRA (Section106: Scientific Assessment), which directs theNational Science and Technology Council toconduct, on a periodic basis, an assessment that

– “integrates, evaluates, and interprets the find-ings of the Program and discusses the scientif-ic uncertainties associated with such findings;

– analyzes the effects of global change on thenatural environment, agriculture, energy pro-duction and use, land and water resources,transportation, human health and welfare,human social systems, and biological diversity;and

– analyzes current trends in global change, bothhuman-induced and natural, and projectsmajor trends for the subsequent 25 to 100years.”

• To bring useful scientific results to decisionmakers in agriculture, with the aim of providinginformation for better decisions.

While we were not able to accomplish all that wewanted nor to resolve every question, we providethese results with confidence that they can helpagricultural stakeholders more fully appreciate thepotential consequences of climate variability andchange.

Richard AdamsOregon State UniversityCorvallis, OR

Bruno AlesiiMonsanto CorporationSt. Louis, MO

Walter ArmbrusterFarm FoundationOak Brook, IL

Gary BaiseBaise and Miller, P.C.Washington, DC

Chuck BeretzAmerican Farmland TrustWashington, DC

Otto DoeringPurdue UniversityWest LaFayette, IN

Jeff EisenbergThe Nature ConservancyArlington,VA

David ErvinWallace InstituteGreenbelt, MD

Richard GadyConAgra, Inc.Omaha, NE

John HickmanDeere and CompanyMoline, IL

Carl MattsonFarmerChester, MT

John McClellandNational Corngrowers

AssociationWashington, DC

Stephanie MercierUS Senate,Agricultural

CommitteeWashington, DC

William OemichenWisconsin Department

of AgricultureMadison,WI

Susan Offutt US Department of

AgricultureWashington, DC

Albert PeterlinUS Department of

AgricultureWashington, DC

Debbie ReedUS Department of

Agriculture Washington, DC

William RichardsRichards Farms, Inc.Circleville, OH

John SchnittkerPublic VoiceWashington, DC

Richard StuckeyCouncil for Agricultural

Science and TechnologyAmes, IA

Ann VenemanFormer Commissioner of

AgricultureState of CaliforniaSacramento, CA

Robert WhiteOffice of Senator

Richard LugarUS Senate,Washington, DC

Robert M WolcottUS Environmental Protection

AgencyWashington, DC

Stuart WoolfWoolf Farming, Co.Huron, CA

David ZilbermanUniversity of California-

BerkeleyBerkely, CA

Agriculture Sector Assessment Steering Committee*

*Note: Individual affiliations are consistent with the time when the assessment was under review. Some individuals are nolonger associated with the listed affiliation.

xi

It is likely that climate changes and atmosphericCO2 levels, as defined by the scenarios examined inthis Assessment, will not imperil crop production inthe US during the 21st century. The Assessmentfound that, at the national level, productivity ofmany major crops increased. Crops showing general-ly positive results include cotton, corn for grain andsilage, soybeans, sorghum, barley, sugar beets, andcitrus fruits. Pastures also showed increased produc-tivity. For other crops including wheat, rice, oats,hay, sugar cane, potatoes, and tomatoes, yields areprojected to increase under some conditionsand decline under others.

Not all agricultural regions of the United States wereaffected to the same degree or in the same directionby the climates simulated in the scenarios. In gener-al the findings were that climate change favorednorthern areas. The Midwest (especially the north-ern half),West, and Pacific Northwest exhibited largegains in yields for most crops in the 2030 and 2090timeframes for both of the two major climate sce-narios used in this Assessment, Hadley and Canadian.Crop production changes in other regions varied,some positive and some negative, depending on theclimate scenario and time period. Yields reductionswere quite large for some sites, particularly in theSouth and Plains States, for climate scenarios withdeclines in precipitation and substantial warming inthese regions.

Crop models such as those used in this Assessmenthave been used at local, regional, and global scalesto systematically assess impacts on yields and adap-tation strategies in agricultural systems, as climateand/or other factors change. The simulation resultsdepend on the general assumptions that soil nutri-ents are not limiting, and that pests, insects, diseases,and weeds, pose no threat to crop growth and yield.One important consequence of these assumptions isthat positive crop responses to elevated CO2, whichaccount for one-third to one-half of the yield increas-es simulated in the Assessment studies, should be

Summary

regarded as upper limits to actual responses in thefield. One additional limitation that applies to thisstudy is the models’ inability to predict the negativeeffects of excess water conditions on crop yields.Given the “wet” nature of the scenarios employed,the positive responses projected in this study forrainfed crops, under both the Hadley and Canadianscenarios, may be overestimated.

Under climate change simulated in the two climatescenarios, consumers benefited from lower priceswhile producers’ profits declined. For the Canadianscenario, these opposite effects were nearly bal-anced, resulting in a small net effect on the nationaleconomy. The estimated $4-5 billion (in year 2000dollars unless indicated) reduction in producers’profits represents a 13-17 percent loss of income,while the savings of $3-6 billion to consumers repre-sent less than a 1 percent reduction in the con-sumers food and fiber expenditures. Under theHadley scenario, producers’ profits are reduced byup to $3 billion (10 percent) while consumers save$9-14 billion (in the range of 1 percent).The majordifference between the model outputs is that underthe Hadley scenario, productivity increases weresubstantially greater than under the Canadian, result-ing in lower food prices to the consumers’ benefitand the producers’ detriment.

At the national level, the models used in thisAssessment found that irrigated agriculture’s needfor water declined approximately 5-10 percent for2030 and 30-40 percent for 2090 in the context ofthe two primary climate scenarios, without adapta-tion due to increased precipitation and shortenedcrop-growing periods.

A case study of agriculture in the drainage basin ofthe Chesapeake Bay was undertaken to analyze theeffects of climate change on surface-water quality. Insimulations for this Assessment, under the two cli-mate scenarios for 2030, loading of excess nitrogeninto the Bay due to corn production increased by17-31 percent compared with the current situation.

xii

Pests are currently a major problem in US agriculture.The Assessment investigated the relationship betweenpesticide use and climate for crops that require rela-tively large amounts of pesticides. Pesticide use isprojected to increase for most crops studied and inmost states under the climate scenarios considered.Increased need for pesticide application varied bycrop – increases for corn were generally in the rangeof 10-20 percent; for potatoes, 5-15 percent; and forsoybeans and cotton, 2-5 percent. The results forwheat varied widely by state and climate scenarioshowing changes ranging from approximately –15 to+15 percent. The increase in pesticide use results inslightly poorer overall economic performance, butthis effect is quite small because pesticide expendi-tures are in many cases a relatively small share of pro-duction costs.

The Assessment did not consider increased croplosses due to pests, implicitly assuming that all addi-tional losses were eliminated through increased pestcontrol measures. This could possibly result inunderestimates of losses due to pests associatedwith climate change. In addition, this Assessmentdid not consider the environmental consequences ofincreased pesticide use.

Ultimately, the consequences of climate change forUS agriculture hinge on changes in climate variabili-ty and extreme events. Changes in the frequencyand intensity of droughts, flooding, and storm dam-age are likely to have significant consequences.Such events cause erosion, waterlogging, and leach-ing of animal wastes, pesticides, fertilizers, and otherchemicals into surface and groundwater.

One major source of weather variability is the ElNiño/Southern Oscillation (ENSO). ENSO effectsvary widely across the country. Better prediction ofthese events would allow farmers to plan ahead,altering their choices of which crops to plant andwhen to plant them. The value of improved fore-casts of ENSO events has been estimated at approxi-mately $500 million per year. As climate warms,ENSO is likely to be affected. Some models projectthat El Niño events and their impacts on US weatherare likely to be more intense. There is also a chancethat La Niña events and their impacts will bestronger. The potential impacts of a change in fre-quency and strength of ENSO conditions on agricul-ture were modeled. An increase in these ENSO con-ditions was found to cost US farmers on averageabout $320 million per year if forecasts of theseevents were available and farmers used them to planfor the growing season. The increase in cost wasestimated to be greater if accurate forecasts werenot available or not used.

1

Introduction

Agriculture on the North American continent haschanged rapidly and continuously at least sinceEuropean colonization. All evidence suggests thatagriculture will continue to change rapidly in thefuture. One of the forces to which future Americanagriculture will likely have to adapt is changing cli-mate induced by the accumulation of greenhousegases in the atmosphere. The impacts and adapta-tions that may occur in response to changing cli-mate are the primary topics of this assessment. Wealso consider weather variability and its impact onagriculture, focusing on some of the implications foradapting to climate change.

In this chapter we begin by identifying key questionswe address in this assessment. We then provide abroad overview of American agriculture: its past, cur-rent conditions, and trends that will take it into thefuture. We conclude this chapter with a report ofthe interests of agricultural stakeholders with whomwe met as part of our assessment. These stakehold-ers include those who are in the business of produc-ing food and fiber and related input and processingindustries, those who are particularly concernedwith the environmental attributes of agriculture, andthose who are involved in public policy and programmanagement in agriculture.

Within the agricultural community there is a greatdeal of interest in the effects of climate change miti-gation policies on agriculture. There are potentialcosts (higher energy prices and costs of controllingnon-CO2 greenhouse gases such as methane andnitrous oxide) and potential opportunities (receivingpayments for sequestering carbon in soils) for agri-culture. Evaluating these costs and opportunities isnot within the scope of this report. Interested read-ers are referred to the reports Economic Analysis ofU.S. Agriculture and the Kyoto Protocol and

Changing Climateand Changing Agriculture

Economic Potential of Greenhous Gas EmissionsReductions: Comparative Pole for Soil Sequestrationin Agriculture and Forestry.1

We focused on answering four questions identifiedas important to stakeholders in our assessment:

• What are the key stresses and issues facingagriculture?

• How will climate change and climate variabilityexacerbate or ameliorate current stresses?

• What research priorities are most important tofill knowledge gaps?

• What coping options can build resiliency intothe system?

These objectives and questions guided the agricul-ture sector assessment. We address the first ques-tion in succeeding sections of this chapter. Wereview results from previous assessments in chapter2 to the extent they contributed to answering eachof these questions. We address the second andfourth questions in chapters 3–5. We organize ourconclusions in chapter 6 to review our answers tothese four questions. To address these questions, wemet with agricultural stakeholders, reviewed rele-vant research and recent assessments, and conduct-ed a program of modeling and research.

Any research effort must operate within budgetaryconstraints. In general, we tried to build on pastwork rather than repeating previous exercises.Several assessments of climate change and agricul-ture within the past four years have involved litera-ture review. We summarize the findings of thesereviews and provide a more detailed discussion ofour methods in chapter 2. Stakeholders identifiedquestions that were much broader and more far-reaching than those covered in recent assessments,however (see “Stakeholder Interests,” below). We

Chapter 1

1Economic Analysis of US Agriculture and the Kyoto Protocol was prepared by the Office of the Chief Economist, Global Change ProgramOffice of the US Department of Agriculture (USDA) with technical input from the Economic Research Service (http://www.usda.gov/oce/gcpo/gcponews.htm). Economic Potential of Greenhouse Gas Emissions Reductions: Comparative Pole for Soil Sequestration inAgriculture and Forestry was prepared by McCarl, Schneider, Murray,Williams, and Sands (http://www.agecon.tamu.edu/faculty/mccarl/mitigate.html).

2

Agriculture: The Potential Consequences of Climate Variability and Change

focused our new research on some of these topicsfor which the research tools to conduct quantitativeassessment were adequate. In many cases, however,answers to these questions would require moreaccurate forecasts and projections than we canachieve—or the development of new assessmenttools. The best we could do with regard to thesetopics was to identify them as open questions andoffer some brief observations about the potentialimplications of climate change. Undoubtedly, thisshortcoming will leave readers with these interestsless than fully satisfied.

Our analysis is a fairly comprehensive treatment of thecountry, with details on individual crops and regions.Because of our limitations of funding and resources,however, the level of detail certainly is inadequate forstate and local decision makers. The job of interpret-ing and deepening the analysis falls to the regionalassessment efforts, which are composed ofresearchers with a firm understanding of the localcontext. This local assessment is particularly crucialfor understanding coping strategies that are relevantto farmers whose conditions vary. In this regard, wefollow a tradition in agricultural research and exten-sion that relies on state and county experts to provideguidance that is directly relevant to local farmers.

Agriculture: Past,Present, Future

Our only guide to the future is what we know aboutthe current state of agriculture and the trends andresponses we have evaluated from the recent past.Part of this knowledge consists of trends in develop-ment and adoption of state-of-the-art technologies.By understanding the technological forefront, wehope to see a decade or two ahead; such assessment,however, is still based on current knowledge and his-torical experience with adopting new technology. Inthis section, we do not attempt to describe agricul-ture comprehensively, and we certainly do not offerprecise predictions for US agriculture for the next100 years. Throughout the report we use the term

“projection” rather than “prediction” indicating thatthe climate scenarios we evaluate and the results wederive from them are based on the mathematical for-mulation of models that capture key interactions aswe understand them. Precise prediction of futureeconomic and social conditions is, of course, not pos-sible. Given the limited exploration of the possiblerange of climate, technological, and socio-economicconditions we were able to explore, even probabilis-tic predictive statements (e.g. there is at least a 70percent chance of an outcome exceeding a givenlevel) were not possible. For those interested ingreater detail on the US agriculture sector, theEconomic Research Service (ERS) of the USDA regu-larly surveys and reports on the current status ofAmerican agriculture, its relationship to the rest ofthe world, its use of natural resources and theenvironment, and the health and nutritional status ofthe US population. Myriad data, reports, and assess-ments conducted by ERS are available at www.ers.usda.gov. Our goal here is to provide a broad-brushoutline of the American agricultural system: what ithas learned from the past, where it is now, andwhere it may be in the next century.

Our focus in this review is on identifying some ofthe important connections between agriculture andweather and climate. Any effort to cope with cli-mate change and climate variability in the futurewill grow out of and react to the perception ofthe success or failure of past efforts to manage agri-culture. As impossible as summarizing Americanagriculture may be, something of a description isneeded to provide a context for studies of theimpact of climate change and variability.

100 Years of Change

US agriculture has undergone vast changes over thepast century. In 1900, 60 percent of the US popula-tion lived in rural areas; there were 6.4 millionfarms, and the average farm size was 132 acres. By1990, only 25 percent of the population lived inrural areas; there were 2 million farms, and the aver-age farm was 435 acres.2 Even in 1900, however, US

2Data on the rural population are from the Census Bureau (http://www.census.gov/population/censusdata/urpop0090.txt). Data on farm-land are from the USDA National Agricultural Statistical Service, based on census data (1999 Agricultural Statistics available at http://www.usda.gov/nass/) and computed from Tables 9.7 and 9.8 at this WEB address. Definitions of farms—and thus land in farms and number offarms—have varied. Other tables give slightly different estimates.

3

Chapter 1: Changing Climate and Changing Agriculture

agriculture was an export industry. Cotton, tobacco,and wheat crops were exported to Europe. As vastas the country seemed at the time, some observerspredicted that the bounty would be exhausted soonby an ever-growing population. Sir William Crookes,of Great Britain, writing in 1900, concluded that “it isalmost certain that within a generation the ever-increasing population of the United States will con-sume all the wheat grown within its borders, andwill be driven to import, and like ourselves, willscramble for the lion’s share of the wheat crop ofthe world” (quoted in Dalrymple 1980). Relativescarcity of supply compared with demand, andstrong overall economic conditions over the nextcouple of decades did, indeed, produce some of themost prosperous times for farming. The years1912–1913 were later regarded as the last pointwhen farmers received a “fair” price for farm prod-ucts. This period of relative prosperity for the farmsector became a benchmark in the 1940s for post-war farm programs whose goal was to revive thefarm economy after the Depression of the 1930s.Through a series of economic downturns and eveneconomic booms, agricultural prices seemed primar-ily to go down. Whereas the US economy boomedafter World War II, agriculture seemed to be mired inlow prices. In the early part of the century, econom-ic development in rural areas lagged behind that in

urban areas. The trend of declining prices repre-sented a success for productivity and productionand reflected overall declining costs of production;in and of itself, this price trend cannot be consid-ered a cause of the economic hardship in rural areasor the farm sector. In fact, the income of farmhouseholds relative to nonfarm households in theUS improved over the latter half of the century evenas prices continued to decline. Declining prices do,however, put continual pressure on individual farm-ers to constantly reduce costs to keep up with mar-ket trends. As in any sector in which there is rapidlychanging technology (e.g., computer production),many producers fall by the way as others lead in theadoption of successful technology.

Oddly (for those who had looked ahead to see foodso scarce that hunger and famine would spread),even as more farmers left the farm, more food wasproduced, and commodity prices continued to fall.Evidence of this worldwide trend since the 1950sincludes falling real prices for food commodities(Figure 1.1) and steadily increasing agricultural out-put. Indices of real prices for all food products andfor cereals fell more than 60 percent from the early1950s to the early 1990s.3 Worldwide food produc-tion growth over the past three decades also hasbeen relatively consistent—increasing by 2.7 percent

per year during the 1960s, 2.8 percentduring the 1970s, and 2.1 percent duringthe 1980s.

How did this happen? In the UnitedStates, the great dams and water projectsof the West made the deserts bloom. Newand more powerful machinery enabledthose remaining on farms to till hundredsof acres instead of tens of acres. Startingmid-century, crop breeding began to pro-duce a constant supply of new varietiesof plants that have increased yields formore than 50 years. Although the rates ofgrowth have varied since 1939, annualrates of growth in yield for corn, potatoes,and sorghum have been on the order of

3The real price declines were 63 percent for total food and 62 percent for cereals, using as a base the 5-year average for 1948 through 1952as compared with the 1992–1996 period. Five-year averages were used to minimize the impact of choice of base year—which can be sub-stantial, given the volatility of commodity prices.

Index of World Food Prices

Figure 1.1: From the 1950s to the early 1990s, indices for real food prices for allfood products and for cereals fell more than 60 percent. Dashed line: cereals; solidline: all food commodities (1970 = 100). Source: Reilly and Schimmelpfennig 1999.

4


2.5–3.0 percent; for rice, wheat, barley, and cotton,the increase has been on the order 1.8–2.2 percent;for soybean, oats, sunflower, and flaxseed, theincrease has been on the order of 1.0–1.25 percent(Reilly and Fuglie 1998). Improved varieties thatwere the basis for these increases required (or wereable to take advantage of) high levels of nutrients,which were supplied by cheap, inorganic fertilizers.Inorganic fertilizers were part of the chemical revo-lution that also brought new ways to control weeds,insects, and diseases in crops. Livestock also under-went improved productivity through breeding, bet-ter veterinary products, improved farm managementpractices, and increasing mechanization. Agriculturaleconomists, observing these forces in the 1950s,termed technical change in agriculture the “technol-ogy treadmill.” The technology treadmill is notunique to agriculture; in this interpretation, howev-er, farmers must adopt the new cost-saving andyield-enhancing technologies just to stay even.Whoever failed to keep up with technology wouldbe run off the treadmill into an abyss of economiclosses as neighbors, the farmers in the next state, orcompetitors around the world kept running.Improved shipping and transportation reduced everfurther any edge a farmer might have in supplyinglocal markets.

Concern about farm income and prices coming outof the Great Depression of the 1930s led to enduringfarm programs. Weather-induced variability was onejustification for these massive programs. The ideawas that the government would buy up commoditieswhen harvests were large—keeping prices up—andsell these stocks when there were crop failures,thereby preventing prices from skyrocketing. In thishopeful view, farmers and consumers would benefitfrom stable prices. After nearly a half-century ofthese programs, however, analysts still argue aboutwhether government intervention may have, instead,increased variability. In addition to the desire toeven out prices, there was a desire to assure a rea-sonable income for farmers. Thus, the prosperousdays of 1912–1913 became the benchmark for parityprices—prices (or some proportion of prices) thatfarm programs would seek to assure farmers. Highprices brought more output, however, and increasing

surpluses that depressed prices. In trying to fightthese market forces, farm programs incorporated acomplex combination of foreign (and domestic) foodaid, acreage reduction programs, commodity stock-piles, and an array of payment mechanisms that weremeant to provide income support without bringingon gluts of production. Farm legislation in 1996—the Federal Agriculture Improvement and Reform(FAIR) Act—was intended to transition US agricultureover a period of seven years toward full reliance onmarkets, ending once and for all this system of incen-tives and counter-incentives.

Dating to the Morrill Act of 1862, which grantedstates and US territories scrip to land that theycould sell to develop colleges that would offer prac-tical instruction in agriculture and the mechanicalarts, a nationwide system of agricultural experimentstations has turned out increasingly high-yieldingvarieties, farm management assistance, and improvedlivestock. Publicly funded research, freely providedto farmers, was a major force behind yield improve-ments. The role of public funding has changed asthe private sector has increased research and devel-opment—taking over much of the applied and prod-uct development research and using intellectualproperty rights protection to recoup the invest-ment. Fuglie et al. (1996) provide a comprehensiveanalysis of public and private research in agricul-ture. Productivity growth—measured as the growthin output less the growth in inputs—has been highsince at least the 1950s, averaging more than 1.9percent per year (Ahearn et al. 1999). Output hasdoubled over that period, while input use remainedessentially unchanged.

Changing regional competitiveness also has been afeature of changing agriculture. The changing com-petitiveness and fortunes of different regions cannotbe traced to a single factor. The opening of canals;building of railroads; construction of large waterprojects; shifting population; changing technology;introduction of new pests; resource degradation oropening of new, more productive areas; and environ-mental considerations have all come into play.Woven into this dynamic was the nature of the peo-ple who took up farming in different regions or

Area Area Variation in crop yield from trend (%, with standard error in parentheses)Commodity harvested irrigated 1870–1994 1900–1994 1950–1994

in 1997 in 1997 Mean Trend in Mean Trend in Mean Trend in(000 ha) (%) variation variation variation variation variation variation

Corn 28,258 15.2 7.77 -1.271E-2 7.24 1.553E-2 6.97 2.357E-1

(0.58) (1.62E-2) (0.68) (2.48E-2) (0.89) **(5.938E-3)

Wheat 23,820 6.8 6.28 -2.834E-2 5.86 -3.122E-2 4.92 -5.662E-4(0.45) ** (0.51) * (0.63) (4.719E-2)

(1.230E-2) (1.81E-2)

Potato 549 79.0 5.75 -8.159E-2 4.40 -7.608E-2 2.42 -4.076E-3(0.52) ** (0.46) ** (0.30) (2.211E-2)

(1.237E-2) (1.457E-2)

5


chose to move on or out rather than adapt. Milkproduction shifted from New York and Pennsylvaniato Wisconsin and then to California and Florida; ithas left behind fading red barns and reforested hills.Cotton shifted from the South to the Southwest andWest as pests, depleted soils, and irrigation waterchanged the fortunes of different regions. Most fruitand vegetable crops, which once were locally pro-duced and were available only seasonally, are nowavailable in supermarkets year-round, with world-wide suppliers. Processed vegetable production alsohas shifted. Cheap and widely available transportgradually has increased the regional specialization ofcropping and livestock production to areas that areespecially favorable for a particular crop or unfavor-able for everything else. As competitivenessdemands greater management, farmers fare best ifthey focus on one or a few complementary cropsthat do well under the climatic and resource condi-tions they face. Farmers and the input suppliers andproduct processors can reap economies of scalefrom large and regionally concentrated production.

We asked two questions about the past 100 yearsthat have a bearing on climate and agricultureinteractions:

• Has yield variability changed over the past century?

• Has the production of major crops relocatedgeographically and climatically?

We found that long-run yield variability did notincrease for corn and fell significantly for wheat andpotatoes (Table 1.1). There were, however, substan-tial geographic shifts in production of these threemajor crops over the past 100 years (see Figures 1.2through 1.4).

Is climate change responsible for these changes?There is some evidence of climate change over thepast century that might have affected crop variabili-ty and the location of production. More rain has fall-en in heavy precipitation events; on average, rainfallhas increased across the United States, with morecloudy days (Karl and Knight 1998). Temperaturevariability increased during the period 1973–1993compared with 1954–1973 (Parker et al. 1994) butdecreased on times scales of one day to one year(Karl et al. 1995). Based on a fitted linear trend, theaverage frost-free season has increased by 1.1 daysper decade (Easterling, in press); the average temper-ature increased by 0.6°C through the early 1990s(Karl et al. 1996).

Table 1.1: Long-term Trends in Variability in US Crop Yields

Yield variation is measured by V = absolute value of (Xt – Xtrend)/Xtrend, where Xt is crop yield in year t and Xtrend is the 9-year moving aver-age of yield centered on year t, using annual crop yield data from 1866 to 1998. The trend in yield variation is the estimate of coefficient βfrom the linear regression model V = α + β t. * Significant at 10% level ** Significant at 5% levelSource: Data are from the US Department of Agriculture. Statistical estimation was conducted by Keith Fuglie of CIP-ESEAP, Indonesia.

6


Figure 1.2: The US corn production geographical center (productionweighted) moved westward from Indiana in 1871 into Missouri by theearly 1900s. Since then production has shifted northward throughIllinois.

Our findings tell us as much about the complexityof climate-agriculture-social interactions as they doabout how the agricultural system might respond toclimate change in the future. The result of nochange or an actual decrease in yield variabilitydespite climate change leaves three competinghypotheses:

• The various climatic forces have had coinciden-tally offsetting effects on yield variability.

• The time period coincidentally shows no changein variability. Indeed, one can pick sets ofdecades that show differences in variability, asin the 1950–1994 period (see Table 1.1).

• Yield variability is a function of economic andsocial acceptance of risk, which may be relative-ly constant over time. This hypothesis recog-nizes that farmers have variability-reducing tech-nologies such as irrigation, shorter-maturingvarieties, changes in type of crop grown, andabandonment of the area to cropping, and thatthey can make an economic and personal calcu-lation about what to do. Lewandrowski andBrazee (1993) have shown, for example, that thestructure of federal farm programs has affectedfarmers’ decisions about risk.

Sorting out these competing theories requires farmore sophisticated empirical evaluation than wewere able to undertake.

Figure 1.3: The US geographical center of wheat production (production weighted) shifted mainly westwardfrom Illinois in the late 1800s as a result of expansion of production in the far west.

Movement of Center of US Corn Production, 1871–1992

Movement of Center of Wheat Production, 1871–1990

7


The northward movement of corn production couldbe a signal of warming and increased frost-free days.Notably, however, the mean temperature at whichcorn was grown fell by 4°C between 1935 and1965—the period when the geographic centroidshifted north.4 If corn production were shifting inresponse to warming, one would expect the meantemperature to remain unchanged. If there had beenno response, the mean temperature would have risenrather than fallen. As a whole, these results indicatethat some other process—such as the developmentof shorter maturing varieties or relocation of produc-tion as a result of economic factors—was causing anorthward movement of production that is greaterthan can be explained by changes in temperature.

The westward migration of crops is more likelycaused by expansion of irrigation in the semi-aridwestern United States over this period. Again, theseresults indicate the difficulty of separating climatesignals from easily observable aggregate indices thatare partially the result of climatic trends but alsoare heavily determined by socioeconomic factors.The movement of the centroid of production, how-ever, does indicate that the geography of crop pro-duction has not been static and that relocationbecause of climate change over the next century, ifit occurred, would not in itself be an unprecedent-ed social change.

Throughout the century, agriculture also has beensubject to inclement weather. Droughts, cold, lateand early frosts, extreme heat, and storms affectsome areas of US agriculture in almost any givenyear. At times they are widespread or catastrophicenough or affect a large enough constituency thatthe weather and its effect on agriculture brieflygains media attention and are broadcast to the 99percent of Americans who are not farmers. For themost part, however, whatever disaster befalls thefarmer, the American consumer is little affected. Formany Americans, the decision about whether to leavea 10, 15, or 20 percent tip at the restaurant probablyhas more economic consequences than any impact

Figure 1.4: The US geographical center of soybean production(production weighted) shifted from southern Indiana northwestinto Illinois.

they will see from adverse weather effects on farmproduction. A widespread drought or weather catas-trophe might increase retail food prices by three tofour percent; yet Americans spend more than one-half their food dollars eating out. Moreover, the farmgate cost of food is a small fraction of the final costto consumers. The nature of agricultural demand(highly inelastic in economic terms) means that awidespread drought can improve the bottom line ofthe farm economy—raising prices more than supplyis cut back and thus increasing farm revenue. On theother hand, a localized drought (such as the one inthe mid-Atlantic in 1999) combined with near-idealgrowing conditions throughout large growingregions in the Midwest can lead to financial lossesfor most farmers. Lower prices resulting from good

Movement of Center of Soybean Production, 1930–1990

4Authors' calculations based on spring and summer state average temperature and precipitation data from 1900 to 1997 (MWRCC 1999) andstate yield data from 1865 to 1997 for corn and wheat, and from 1924 to 1997 for soybean from USDA (1999d).The mean temperature andprecipitation for the growing region of each crop were computed by weighting the mean temperature and total precipitation for each stateby the ratio of the state harvested acreage to the total acreage harvested in the U.S. Crop production centroids were computed from weight-ing the geographic centroid of each state by the ratio of the production in each state to the total national production.

8


production overall can mean that even farmers whoexperience excellent growing conditions and highyields do not cover costs. Those who suffer yieldloss because of drought face low prices combinedwith reduced production.

Weather is so central to farming that most of thetechniques used in farming, agribusiness, and thefood industry somehow reflect a desire to overcomeweather. It is not stretching the facts to observe thatthere is no such thing as a “normal” year. A year char-acterized by 30-year means for all months of thegrowing season and showing an “average” pattern ofextremes would be truly abnormal. Conquering vari-ability is manifest in nearly every dimension of farmmanagement. Included are technologies such ascrop drying, irrigation, drainage and tiling, and stor-age; shading and cooling for livestock; selection andbreeding of livestock and crops that are hardy orhardier under a wider range of climatic conditions;financial and farm management such as financial sav-ings, borrowing, crop insurance, diversified produc-tion strategies, and off-farm income; market instru-ments such as forward markets and contractproduction that shifts and pools risk; prediction andoutlook on weather and economic conditions; andgovernment policy such as disaster assistance, farmprograms, and government involvement in the insur-ance markets.

As We Enter a New Century

Agricultural production is very diverse. This diversi-ty bespeaks an industry undergoing rapid change.We excerpt below a verbatim summary of highlightsfrom the 1999 Family Farm Report, a document pro-duced each year by the ERS under Congressionalmandate. The summary and details on ordering thereport are available at http://www.ers.usda.gov/epubs/htmlsum/aib735.htm. The report finds that

• More than 2 million US farms produced agricul-tural commodities that generated an average of$74,000 in gross value of sales per farm in 1994.Still, 73 percent of farms had gross value of salesunder $50,000 (noncommercial farms), althoughthey accounted for just 11 percent of total USfarm sales.

• Gross cash farm income (adjusted to exclude theshare of production accruing to landlords andcontractors) averaged nearly $69,000. However,gross cash farm income for the nation’s largestfarms (sales of $1 million or more) averagedalmost $2 million, so fewer than 1 percent offarms accounted for 23 percent of gross cashfarm income. Commodity sales accounted for 84percent of total gross cash farm income; govern-ment payments added 5 percent and other farmincome added 11 percent.

• Acreage per farm, which has tripled over thepast six decades, averaged 448 acres operated in1994, but half of all farms were smaller than 180acres. Livestock farms producing some combi-nation of beef cattle, hogs, and sheep accountedfor the largest share of farms grouped by farmtype. Although these farms had larger acreagethan the US average, they had lower averagegross cash farm income and gross value of sales.

• Half of all farms cash-rented or share-rentedsome or all of the land they operated in 1994.Farm operators who owned all the land theyoperated but had a rental arrangement formachinery, buildings, or livestock (5 percent offull owners) had income and sales five times ashigh as full owners who rented nothing.

• More than 90 percent of farm businesses werelegally organized as individual operations; 6 per-cent of farms were partnerships, and 4 percentwere corporations (most of which were family-owned).

• Farms organized as individual operations aver-aged more than $50,000 in gross value of salesand had farm assets that averaged more than$350,000.

• Although 13 percent of all farm operatorsreported having some contractual arrangementfor production and/or marketing of farm com-modities, farms with marketing contracts out-numbered farms with production contracts bymore than 4 to 1.

9


• Use of contracting arrangements varied by farmcharacteristics such as sales class and type ofproduction. For example, more than 60 percentof poultry farms had production contracts.

• Net cash farm income averaged $11,696 forfarms nationwide but ranged from negative forfarms with sales under $50,000 to more than$380,000 for farms with sales of $1 millionor more.

• Farm assets generally increased with sales class,but even farms with sales under $50,000 hadfarm assets averaging more than $250,000.Farms with gross value of sales of $1 million ormore used assets valued at more than $3 millionto generate $2 million in gross cash income.These large farms also had the highest debt-to-asset ratio (0.25).

• In 1994, 61 percent of farms were in a favorablefinancial position, with a low debt-to-asset ratio(0.40 or less) and positive net farm income.Another 34 percent of farms had a low debt-to-asset ratio but were unable to generate enoughincome to offset expenses, so net farm incomewas negative—putting them in the marginalincome category. Most of these operations werenoncommercial farms.

• Only 4 percent of farms were in a vulnerablefinancial position, with a high debt-to-asset ratio(0.40 or more) and negative net farm income thatthreatened the long-term survival of the business.

• More than a third of farms received income fromgovernment payments, averaging $9,306 perreceiving farm. Almost two-thirds of commercialfarms (gross value of sales $50,000 or more),compared with one-fourth of noncommercialfarms, received government payments.Government payments accounted for less than 3percent of gross cash farm income for commer-cial farms, however, compared with 41 percentfor noncommercial farms.

• More than 40 percent of the nation’s farmoperators reported farming or ranching as theirprincipal occupation. Their farms accounted formore than 80 percent of gross cash farmincome and gross value of sales. Households ofoperators with a principal occupation of farm-ing had average total household income thatwas about 85 percent of the US average. Abouta third of total income for these householdscame from earnings from farming activities, andtwo-thirds from off-farm sources.

• Operators younger than 35 years old accountedfor 9 percent of all operators, whereas operators65 years old and older accounted for 24 per-cent. The youngest operators, however, generat-ed their proportionate share of total US grosscash farm income and gross value of sales(based on number of farms), whereas theoldest group generated about half their pro-portionate share.

• About 13 percent of all farm operators usedelectronic information services to get farm busi-ness information. Use of this new technologyincreased with farm size and operator educa-tional attainment level (20 percent of operatorswho completed college, compared with 10 per-cent of those who completed only high school).

• More than 60 percent of farm operators rankedgetting out of debt and improving crop yield orlivestock production as very important businessgoals. Commercial farm operators ranked thesegoals higher than noncommercial farm operators.

• Mean household income from all sources forfarm operator households was near the US aver-age. On average, 90 percent of total operatorhousehold income came from off-farm sources.For almost half of all farm households, earningsfrom farming activities (farm self-employmentincome plus other farm-related earnings) werenegative, but total household income was posi-tive because off-farm income exceeded the loss.

10


• As farm sales increased, household dependenceon earnings from farming activities increasedand household income relative to the US aver-age increased.

• Operator households associated with farms thathad a gross value of sales of $500,000 or morehad average household income 3.5 times the USaverage, and earnings from farming activitiesaccounted for 75 percent of total operatorhousehold income.

• Noncommercial farm operators worked half oftheir annual working hours on the farm; theirspouses worked about one-fourth of their work-ing hours on the farm.

• Commercial farm operators worked 88 percentof their total work hours on the farm; spousesaveraged almost half of their total work hourson the farm.

• Rankings of eight selected measures of farmbusiness success showed that having farmincome sufficient to support the household wasmost important to operators reporting theirprincipal occupation as farming.

• More than half of farm operators reported thatpassing the operation on to the next generationas very important.

The remarkable aspect of this summary is the diver-sity among farm operations. Indeed, the enterpriseof farming appears to have divided into at least sev-eral broad categories. The bulk of commodities areproduced on large commercial farms with large rev-enues, whose operators rely principally on the farmas a source of income and who earn a familyincome above the average of the US household. Asecond group of farm operators run small farms; netincome from the farm is very small or negative, andthe income of the household is determined by theoff-farm earnings of the household members.Another group of farmers is near retirement or insemi-retirement. The farmers in this group typicallyown outright all the land they operate; in fact, theymay rent most of their land or have it enrolled in along-term easement program such as theConservation Reserve Program that pays farmers of

highly erodible land to maintain permanent coveron the land.

A fourth group comprises farmers who own mid-sized farms. This group of farmers is most vulnera-ble to the century-long trend toward larger farms.Farmers in this group are most likely struggling toearn an income from the farm operation and supple-menting household income with off-farm employ-ment, working for the day when the family cansurvive on farm income alone. They face difficultyin affording expensive new technologies, such asprecision farming, that involve onboard computermonitoring guided by global positioning systems(GPS). As profit margins tighten, the scale of opera-tion must grow if the main occupation is to remainthe primary source of family income. One alterna-tive for this group, already widespread in the poultryand hog industries, is to produce under contractwith processors. The processor bears more of therisk and provides specific guidelines for production.This approach represents a major cultural changefor many farmers who value the independent lifethat farming traditionally has represented.

The trend that has emerged in the past decade isthat the smallest farm categories have exhibitedincreasing numbers. The reasons for this trend arehighly diverse as well. People returning to farm—supplementing farm income with off-farm employ-ment and perhaps no intention of fully supportingthemselves through farming—are one group.Others have found niche and local markets wherethe profit margins are higher than for bulk com-modities. Operators have found success with every-thing from organic foods and herbs sold to moreexpensive, high-end restaurants to Christmas trees,selling to local farmers markets, or inviting con-sumers to “pick their own”—perhaps combinedwith some form of entertainment for the urbandweller seeking a farm experience.

Climate change is unlikely to alter this dynamic inany fundamental way. The regional effects of climatechange may vary considerably (as we detail in subse-quent chapters). An increasing percentage of bulkcommodities probably will be produced on the

11


largest farms. Middle-sized farms will continue to besqueezed if they must compete in bulk commodityproduction. Niche producers for local markets canbe successful if the products are cleverly chosen, theenterprises are well-run, and the products deftlymarketed. Success, however, inevitably will invitecompetition If climate change is somehow benefi-cial to a region, it may slightly ease the pressureexerted on mid-sized farms. If climate change hasadverse consequences for the region, it may tightenthe grip on the most vulnerable producers.Agriculture is a highly variable enterprise, however,with relatively low barriers to entry relative toindustries such as automobile and pharmaceuticalmanufacturing. Hence, any evidence of increasedprofitability will tend to draw more producers, bidup land prices, and keep the profit margin tight.Thus, even if climate change is somehow beneficialwith respect to production, it is unlikely to manifestitself as widely perceived windfall profits. Likewise,if climate change is adverse, the gradual change isunlikely to be perceived as windfall losses. In bothcases, downturns in commodity prices will continueto take out the vulnerable farmers, and upturns willencourage production expansion.

Forces Shaping the Future

A few broad forces will shape the future forAmerican agriculture over the next few decades:

• Changing technology. Biotechnology andprecision agriculture are likely to revolutionizeagriculture over the next few decades, just asmechanization, chemicals, and plant breedingrevolutionized agriculture over the past centu-ry—although public concerns and environmen-tal risks of genetically modified organisms couldslow development and adoption of crops andlivestock containing them. Biotechnology hasthe potential to improve adaptability, developresistance to heat and drought, and change thematuration schedule of crops. Biotechnologyalso will give rise to entirely new streams ofproducts and allow the interchange of character-istics among crops. Precision farming—theincorporation of information technology (e. g.,

computers and satellite technology) in agricul-ture—will improve farmers’ ability to manageresources and to adapt more rapidly to changingconditions.

• Global food production and the globalmarketplace. Increasing linkages are the ruleamong suppliers around the world. These linksare developing in response to the need toassure a regular and diverse product supply toconsumers. Meat consumption is likely toincrease in poorer nations as their wealthincreases, which will place greater pressure onresources. Climate change could exacerbatethese resource problems. Trade policy, trade dis-putes (e.g., over genetically modified organ-isms), and the development of intellectual prop-erty rights (or not) across the world could havestrong effects on how international agricultureand the pattern of trade develops.

• Industrialization of agriculture. The ever-faster flow of information and the developmentof cropping systems that can be applied acrossthe world will transcend national boundaries.Market forces are encouraging various forms ofvertical integration among producers, proces-sors, and suppliers, in part driven by the need toproduce uniform product and assure supplydespite local variations induced by weather orother events.

• Environmental performance. Agriculture’senvironmental performance is likely to be agrowing public concern in the future, whichwill require changes in production practices.Significant environmental and resource con-cerns related to agriculture include water quali-ty degradation caused by soil erosion, nutrientloading, pesticide contamination, and irrigation-related environmental problems; land subsi-dence resulting from aquifer drawdown; degrad-ed freshwater ecosystem habitats resulting fromirrigation demand for water; coastal water degra-dation caused by run-off and erosion; waterquality and odor problems related to livestockwaste and confined livestock operations; pesti-cides and food safety; biodiversity impacts from

12


landscape change (in terms of habitat andgermplasm); air quality, particularly related toparticulate emissions; and landscape protection.Tropospheric ozone is increasingly recognizedas an industrial/urban pollutant that negativelyaffects crops. Agricultural use of land also canprovide open space, habitat for many species,and—with proper management—a sink for car-bon. These positive environmental aspects arelikely to be increasingly valued.

In the past, it has been possible to summarize theforces shaping agriculture as a competition betweenincreased demand for food driven by a larger andhigher-income population and increased supply driv-en by new technology. Over a period of a fewdecades, a tenth of a percent difference in exponen-tial growth rates of population and technologicalprogress can make the difference between ever-falling or ever-rising prices. The ability to predictthese rates of change with the degree of accuracynecessary to resolve the difference does not exist.Most likely, the rate of technical progress responds todemand pressure as well as opportunities created byimprovements in basic research. As historical peri-ods of rapid commodity price rises indicate, agricul-tural supply has tremendous ability to respond overthe course of a few years. The best bet is that com-modity prices will continue their long-run decline, asseveral major global forecasting efforts have suggest-ed (see Reilly and Schimmelpfennig 1999).

The specific nature of technical change and what itmeans for different regions and farming systemsremains elusive. Over the next few decades, thereare no obvious biological limits on yields that wouldprevent continued increase (Reilly and Fuglie 1998).In the longer term, far greater changes are possible.Industrialization of agriculture could mean that rawbiomass is processed into livestock feed and pro-cessed food products, using biotechnology-generat-ed microbial organisms—greatly reducing the needfor conventional crop production as we now recog-nize it. As we try to look forward 50 and 100 years,it is not clear whether the crops that will be grownthen will resemble the crops grown today.Although such changes are possible with new

technology, we must also look back to the fact thatcivilizations have relied on the major grain crops forcenturies, and breaking from this trend would repre-sent an epochal change in human history.Nevertheless, stretching our thinking is worthwhileif we are to imagine what the American agriculturecould look like in the year 2100.

Farm policy will work around the edge of thesebroad forces, but in most respects it is unlikely toalter much the inevitable push of technology andfundamental market forces. This conclusion, if any-thing, is the lesson of farm policy over the secondhalf of the 20th century. Well-meaning policiesdesigned to improve the income of farmers createdincentives to produce that overwhelmed the marketand drove prices down, filling public granaries.Attempts to hold prices up in the face of technologythat reduced costs and increased production alsogenerated surpluses. Thus, ultimately, the federal pro-grams were forced to liquidate stocks and lower thetarget price they hoped to maintain.

We cannot easily predict how policy will try toblunt the adjustments and dislocations that theseforces will bring. On the near-term agenda, agricul-tural policy is evaluating the effects of the FAIR Actof 1996. The basic background is that the FAIRAct—passed with much fanfare and intended tobring an end to an era of farm programs—is beingreconsidered. Reconsideration of FAIR is likelybecause low prices caused new financial stresses foragriculture, and most observers see little prospectthat prices will improve in the next few years.Given this background, the major issues likely tocome up in Congress over the next few years arethe following:

• Congress may revisit the 1996 Farm Bill tostrengthen the safety net for farmers. Generalobservations by many observers in the agricul-tural policy community include the following:

1. Assistance for economic disasters must bethought through.

2. There is a sense that planting flexibility inFAIR worked and will be retained.

13


3. There is an interest in improving crop insur-ance, but there are widely different ideasabout what “improved” means.

4. There is a sense that the shift away fromcounter-cyclical program payments was notwell thought out by Congress in the 1996bill.

5. Federal support for agriculture will contin-ue. Decoupling payments—the underlyingapproach in FAIR—was better in theorythan in practice.

• Some unresolved issues include linking of envi-ronmental performance to farm payments; caus-es of the hog price collapse that caused somuch stress in 1998/99; and fundamental con-cerns with the structure of contracting and pric-ing in agriculture. In terms of internationaltrade, the United States probably will continueto seek further reductions in barriers to tradewithin the World Trade Organization. Specificissues will be state trading and trade in geneti-cally modified organisms. A problem facing fur-ther trade barrier reduction is that convincingfarmers that freer trade is good for them isincreasingly difficult.

• Environmental pressures, as they relate to agri-culture, are likely to become more important inthe future.

These policy considerations take us only a few yearsinto the future (at most) and are subject to rapidchange. They do, however, provide insight into theunderlying concerns of the policy community thatare likely to endure. These concerns are not muchdifferent from those that have driven farm policy forseveral decades.

Stakeholder Interests

The agriculture sector assessment developed a steer-ing committee to provide input on the interests ofthe many and varied stakeholders in agriculture.Included among this group were small farmers,representatives from agribusiness, members of the

agricultural research community, representativesfrom environmental groups, staff members ofCongress, and those involved in implementing poli-cy in the federal and state governments.

The comments received from stakeholders could besummarized into nine broad issues:

• Agriculture is diverse. We must speak to thediverse elements that exist. Different concernsrequire that the assessment activity take differ-ent cuts on agriculture.

• Agriculture is changing rapidly; biotechnology,computers, GPS, information technology, and thechanging structure of production have collec-tively altered the sector. Farming is becomingan increasingly specialized, technology-drivenenterprise, which means that farmers need ahigh level of training to operate successfully.

• The assessment should be more integrated thanprevious efforts. Interrelated issues such aswater, pests, land use, and ozone levels must bedealt with effectively.

• Variability is a major concern; it wreaks havocon farmers.

• Environmental links are unexplored but could bevery important. Opportunities for win-win solu-tions exist and should be further investigated.

• The policy environment will be affected by cli-mate change and will affect the ability of agri-culture to adapt.

• Deep thought should be given to the structureof the assessment: Learn from past efforts.

• Further research is needed to assess the accura-cy of scenarios and analyses, as well as wherethe errors are.

• The assessment will be useful if we identify therange and breadth of issues (potential surpris-es), even if we cannot quantify all of them.

Stakeholders also had specific questions they hopedthe assessment could tackle.The following 16 ques-tions reflect the observations of Robert White (from

14


the legislative staff of Senator Richard Lugar) towardthe end of the stakeholder meeting held in January1999:

1. How will crops and livestock be affected? Theassessment should consider not only the abilityto genetically alter crops and livestock inresponse but also the effects of diseases andpests.

2. How will growing degree days change? Willthe distribution of the current patternschange?

3. Will climate change result in changes in com-petition for land? How? What will the futurebaseline competition look like if climatechanges?

4. Consider changes in the structure of agricul-ture: How will climate change affect opera-tions? Will it make it harder or easier to enterinto agriculture?

5. How will international competitiveness beaffected?

6. Consider changes in variability and the pre-dictability of weather and climate: Can wepredict better? Cash is on the line for farmersif they act on predictions.

7. Consider direct and indirect effects—the inter-play of water and nutrients—especially as theyaffect water availability and water quality.

8. How will climate change affect the environ-ment via agriculture, and will it affect thestructure of natural resource management?

9. Where will the regions that gain competitiveadvantage be?

10. What will be the effects on transportation,ports, lock and dam structures? They are cur-rently in bad shape. Where should we buildor abandon?

11. Where will processing plants exist? Do theyneed to co-locate with production? What ifproduction shifts?

12. What about risk management strategies interms of agricultural credit services? Whatwill agricultural creditors demand as proof ofability to repay loans of farmers if productionis much more variable?

13. How will federal, state, and local policymakingbe affected? For example, the local tax base isdependent on property values—how will thistax base change? How will this change affectschool systems through tax base erosionand/or a declining population? Will there be areturn to price supports at a federal level?How important or necessary are current feder-al policies with respect to risk? Will there bemore regulations at the state, federal, or inter-national levels, and what might their impactbe?

14. What will be the effect on the labor supply foragriculture? Labor is already tight in thissector.

15. Will there be adequate funding for research?What research should be funded?

16. Where will the new customers be so that bet-ter marketing strategies can be designed?

With this guidance, we undertook the researchdescribed in the following chapters. We could notaddress all of these questions with quantitative anal-ysis, but we have tried to provide information anddiscussion on the main topics.

15

Assessment Approach:Building on Existing Knowledge

Chapter 2

Introduction

In this chapter we provide a review of previousassessments of the impacts of climate change onUS agriculture. We also describe the methods andapproaches used in the agricultural sector assess-ment.

We begin with a brief review of climate changeimpact studies, focusing on efforts that have soughta comprehensive assessment or relatively compre-hensive review of the literature. Our goal is to sum-marize the main findings, identify as extensively aspossible where some of the climate-agriculture linksexist, and as a result indicate which links have notbeen explored. We then describe the method andapproaches we have used to fill some of these gaps.Our purpose is to help the reader who may be unac-quainted with past assessments to understand thecontext for our findings, what is new, and what rein-forces previous work.

Past Assessments:General Findings

Several assessments of agriculture that include theUnited States or cover major parts of the UnitedStated have been conducted over the past 20 years.As the bibliographies of these reviews and assess-ments attest, there are many detailed studies on vari-ous aspects of climate change; numerous papersreport experimental results of the impact of elevatedambient levels of CO2 on crops, for example. Thisfundamental research is absolutely critical for devel-oping and improving assessment models, assessmentresearch, and ultimately assessments of this type.There are two aspects of this type of research thatare critical to understand:

• Assessment inevitably involves scaling up resultsof bench-, site-, or field-level experiments to a

farm, a region, the entire country, or world mar-kets. There are two very broad concerns in scal-ing up. First, will a mix of independently con-ducted site studies be representative of thescaled-up area, and are they based on consistentassumptions and approaches? Second, are there“fallacies of composition” that occur in simplyadding together effects? The most obviousexample is that a farm-level model of the impactof climate change on farm profits is irrelevantby itself; production changes across the countryand the world will result in changes in marketprices. These changes can be far more impor-tant for farm profitability than the direct effectof climate on farm yields.

• Assessment usually involves translating resultsobtained under controlled, experimental condi-tions to conditions observed on the farm. Theconcerns here involve at least three issues. First,are the environmental controls in these experi-ments a reasonable approximation of open-fieldconditions? If not, are the estimated responsesrelevant to real-world conditions? Second, dothese experiments consider complex interac-tions with the environment (e.g., changes inpests, soils, and other environmental factors)? Ifnot, is there some validity in considering justone element at a time? Can one, for example,consider response to CO2 independent of tem-perature, moisture, nutrients, salinity, tropo-spheric ozone and other factors? Third, howdoes farm management affect these results?For example, how do farmers change applica-tions of water, nutrients, and other managementpractices in response to physiological changesin plants?

Broader assessments—those that attempt to simulateimpacts of climate change on the agricultural econo-my—address the foregoing issues in a variety ofways. Sometimes they do so by making simplifiedassumptions (e.g., that an average CO2 response,

16


independent of other factors, can be used). In othercases, the effects are simply ignored (e.g., changes inthe distribution of pests, in soils, or in variability)because there are no quantitative methods forassessing the problem or on the assumption thateffects are small. In other cases, the method usedmay implicitly capture the effect under some condi-tions. For example, statistical evidence drawn fromcross-section data can embody all of the effects asso-ciated with climatic conditions that vary acrossregions. Also implicit, however, is the assumptionthat climate change will involve the wholesale shiftof climatic regimes with these associations intact.For example, this assumption would imply thatpests, soil conditions, and farming practices wouldall change at the same rate as climate. Anotherapproach is to use expert judgment. Experts alsolikely weigh a variety of evidence—perhaps includ-ing the potential effects of pests and diseases, forexample, to come up with a judgment about cropyields under a changing climate.

Conclusions from Previous Assessments

We do not attempt to review here much of thedetailed scientific literature that is the backgroundfor agricultural assessments. Excellent reviews oncrops and livestock effects, pests, and soils—as wellas discussion of global and regional impacts—areincluded in a special edition of the journal ClimaticChange,“Climate Change: Impacts On Agriculture”(Reilly (ed.), 1999). The five articles included in theedition contain more than 500 citations, providing adetailed guide to the literature for readers soinclined. Instead, we provide a short summary of themajor assessments, by approximate date over whichthe assessment occurred.

1976–1983: National Defense University

A National Defense University project (Johnson1983) produced a series of reports focusing on agri-culture.The final report integrated yield and econom-ic effects. It focused on the world grain economy inthe year 2000, considering warming and cooling ofup to approximately 1ºC for large warming or cool-

ing and 0.5ºC for moderate changes, with associatedprecipitation changes on the order of ± 2 percent.These estimates varied somewhat by region. Thebase year for comparison purposes was 1975. Thestudy relied on an expert opinion survey for yieldeffects; it used these effects to create a model ofcrop-yield response to temperature and precipitationfor major world grain regions. There was no explicitaccount of potential interactions of pests, changes insoils, or livestock or crops such as fruits and vegeta-bles. No direct effects of CO2 on plant growth wereconsidered because the study remained agnosticabout the source of the climate change (natural vari-ability or human-induced). Economic effects wereassessed with a model of world grain markets. Cropyields in the United States were estimated to fall by1.6–2.3 percent as a result of moderate and largewarming and to increase by very small amounts (lessthan 0.3 percent) with large cooling and even small-er amounts with moderate cooling. Warming wasestimated to increase crop yields in the (then) SovietUnion, China, Canada, and Eastern Europe, with cool-ing decreasing crop production in these areas. Mostother regions were estimated to gain from coolingand suffer yield losses from warming. The net effectwas a very small change in world production and onworld prices. The study assigned subjective probabil-ities to the scenarios, attempted to project ranges ofcrop yield improvement in the absence of climatechange, and compared climate-induced changes tonormal variability in crop yields and uncertainty infuture projections of yield. A summary point high-lighted the likely difficulty in ultimately detectingany changes due to climate given the year-to-yearvariability and the difficulty in disentangling climateeffects from the effects of new varieties and otherchanging technology that would inevitably be intro-duced over the 25-year period.

1988–1989: US EnvironmentalProtection Agency (EPA)

The EPA (Smith and Tirpak 1989) evaluated theimpacts of climate change on US agriculture as partof an overall assessment of climate impacts on theUnited States. The agricultural results were pub-lished in Adams et al. (1990). The study evaluated

17

Chapter 2: Assessment Approach: Building on Existing Knowledge

warming and changes in precipitation based on dou-bled CO2 equilibrium climate scenarios from threewidely known general circulation models (GCMs),with increased average global surface warming of4.0–5.2º C. In many ways the most comprehensiveassessment to date, this effort included studies ofpossible changes in pests, and, in a case study ofCalifornia, interactions with irrigation water. Themain study on crop yields used site studies and a setof crop models to estimate crop yield effects. Theseeffects were simulated through an economic model.Economic results were based on imposition of cli-mate change on the agricultural economy in 1985.Grain crops were studied in greatest detail; a simplerapproach was used to simulate impacts on othercrops. Impacts on other parts of the world werenot considered. The basic conclusions summarizedin Smith and Tirpak (1989) were as follows:

• Yields could be reduced, although the combinedeffects of climate and CO2 would depend on theseverity of climate change.

• Productivity may shift northward.

• The national supply of agricultural commoditiesmay be sufficient to meet domestic needs, butexports may be reduced.

• Farmers would probably change many of theirpractices.

• Ranges of agricultural pests may extendnorthward.

• Shifts in agriculture may harm the environmentin some areas.

1988–1990: IPCC First Assessment Report

The first assessment report of the IntergovernmentalPanel on Climate Change (IPCC) (Parry 1990a, 1990b)briefly addressed North American agriculture. Theassessment was based mainly on literature reviewand, for regional effects,expert judgement. NorthAmerican/US results mainly summarized the earlierEPA study. Among the main contributionsof the report were that it identified the multiplepathways of effects on agriculture, including effects

of elevated CO2, shifts of climatic extremes, reducedsoil water availability, changes in precipitation pat-terns such as the monsoons, and sea-level rise. It alsoidentified various consequences for farming, includ-ing changes in trade, farmed area, irrigation, fertilizeruse, control of pests and diseases, soil drainage andcontrol of erosion, farming infrastructure, and interac-tion with farm policies. The overall conclusion of thereport was that “on balance, the evidence suggeststhat in the face of estimated changes of climate, foodproduction at the global level could be maintained atessentially the same level as would have occurredwithout climate change; however, the cost of achiev-ing this was unclear.” As an offshoot of this effort, theERS (Kane, Reilly, and Tobey 1991; Kane 1992; Tobey,Reilly, and Kane, 1992) published an assessment ofimpacts on world production and trade, includingspecifically the United States. The study was based onsensitivity to broad generalizations about the globalpattern of climate change as portrayed in doubled-CO2 equilibrium climate scenarios, illustrating theimportance of trade effects. A “moderate impacts sce-nario”brought together a variety of crop model studyresults, based on doubled-CO2 equilibrium climatescenarios and expert judgments for other regions thatwere the basis for the IPCC study. In this scenario,the world impacts were very small (a gain of $1.5 bil-lion in 1986 dollars). The United States was a verysmall net gainer ($0.2 billion); China, Russia,Australia,and Argentina also benefited, whereas other regionslost. On average, commodity prices were predicted tofall by 4 percent, although corn and soybean pricesrose by 9–10 percent.

1990–1992: DOE Missouri, Iowa, Nebraska,Kansas Study

In the Missouri, Iowa, Nebraska, Kansas (MINK)study (Rosenberg 1993; Easterling et al. 1993), thedust bowl of the 1930s was used as a surrogate forclimate change for the four-state region. Climatechange in the rest of the world was not considered.Unique aspects of the study included considerationof water, agriculture, forestry, and energy impactsand projection of regional economy and crop vari-ety development to the year 2030. Crop responsewas modeled by using crop models, river flow with

18


historical records, and economic impacts by usingan input-output model of the region. Despite thefact that the region was “highly dependent” on agri-culture compared with many areas of the country,the simulated impacts had relatively small effects onthe regional economy. Climate change losses interms of yields were on the order of 10–15 percent.With CO2 fertilization effects, most of the losseswere eliminated. Climate impacts were simulatedfor current crops as well as “enhanced” varietieswith improved harvest index, photosynthetic effi-ciency, pest management, leaf area, and harvest effi-ciency. These enhanced varieties were intended torepresent possible productivity changes from 1990to 2030; they increased yield on the order of 70 per-cent. The percentage losses resulting from climatechange did not differ substantially between the“enhanced” and current varieties. Despite relativelymild effects on the agriculture sector of the regionas a whole, locally severe displacements couldoccur. For example, irrigation in western Kansasand Nebraska would be untenable and would moveto the eastern ends of these states.

1992: Council on Agricultural Scienceand Technology

The Council on Agricultural Science and Technology(CAST) report (CAST 1992) commissioned by theUSDA did not attempt any specific quantitativeassessments of climate change impacts. It focusedinstead on approaches for preparing US agriculturefor climate change. It used a portfolio approach toresponding to climate change, recognizing that pre-diction with certainty was not possible. Attentionwas directed to reform of agricultural policy, improv-ing energy and irrigation efficiency, maintaininginput supply and export delivery infrastructure, pre-serving genetic diversity, maintaining research capa-bility, developing alternative cropping systems,enhancing information systems, attending to devel-opment of human resources, harmonizing agricultur-al institutions, and promoting freer trade. Althoughthe study did not provide quantitative assessments, itdid conclude with a relatively optimistic view of USagriculture’s ability to cope. The study also

addressed opportunities to mitigate agriculturalgreenhouse gas emissions.

1992: National Research Council

The National Research Council (NRC) of the NationalAcademy of Sciences undertook a broad assessmentof the policy implications of greenhouse warmingwith regard to mitigation and adaptation. The reportincluded a discussion of climate change impacts onagriculture and the effect of elevated CO2 on crops(NRC 1992).

1992–1993: Office ofTechnology Assessment

The Office of Technology Assessment (OTA) study(OTA 1993), like the CAST study for agriculture,focused on steps that could prepare the UnitedStates for climate change rather than estimates ofthe impact. The study’s overall conclusions for agri-culture were that the long-term productivity andcompetitiveness of the US agriculture were at riskand that market-driven responses might alter theregional distribution and intensity of farming. Thestudy found institutional impediments to adaptation,recognized that uncertainty made it hard for farmersto respond, and saw potential environmental restric-tions and water shortages, technical limits to adapta-tion, and declining federal interest in agriculturalresearch and education. The study recommendedremoval of institutional impediments to adaptation(in commodity programs, disaster assistance, andwater-marketing restrictions), improvement ofknowledge and responsiveness of farmers to speedadaptation, and support for general agriculturalresearch and research targeted toward specific con-straints and risks that might be related to climatechange (e.g., drought or heat stress).

1992–1994: EPA Global Assessment

A global assessment (Rosenzweig and Parry 1994;Rosenzweig et al. 1995) of climate impacts on worldfood prospects expanded the method used in the EPAstudy for the United States to the entire world. The

19


global assessment was based on the same suite ofcrop and climate models and applied these models tomany sites around the world. It used a global modelof world agriculture and the world economy to simu-late the evolving economy to 2060, assumed to be theperiod when the doubled-CO2 equilibrium climateswould apply. The global temperature changes were4.0–5.2º C. Scenarios with the CO2 fertilization effectand modest adaptation showed global cereal produc-tion losses of 0–5.2 percent. In these scenarios, devel-oped countries showed cereal production increasesof 3.8–14.2 percent; developing countries showedlosses of 9.2–12.5 percent. The study concluded thatthere was a significant increase in the number of peo-ple at risk of hunger in developing countries becauseof climate change. The study also considered differ-ent assumptions about yield increases resulting fromtechnology improvement, trade policy, and economicgrowth. These assumptions and scenarios had equallyimportant or more important consequences for thenumber of people at risk of hunger.

Other researchers simulated the yield effects estimat-ed in this study through economic models, focusingon implications for the United States (Adams et al.1995) and world trade (Reilly et al. 1993, 1994).Adams et al. (1995) estimated economic welfaregains for the United States of approximately $4 and$11 billion for two of the three climate scenarios anda loss of $16 billion for the other scenario (1990 dol-lars). The study found that increased exports fromthe United States, in response to high commodityprices resulting from decreased global agriculturalproduction, led to benefits to US producers ofapproximately the same magnitude as the welfarelosses to US consumers from high prices. Reilly et al.(1993, 1994) found welfare gains to the United Statesof $0.3 billion under one GCM scenario and$0.6–$0.8 billion in losses in the other scenarios, sim-ulating production changes for all regions of theworld through a trade model. They also found wide-ly varying effects on producers and consumers, withproducers effects ranging from a $5 billion loss to a$16 billion gain. Reilly et al. (1994) showed that inmany cases, more severe yield effects produced eco-nomic gain to producers when world prices rose.

1994–1995: IPCC SecondAssessment Report

The IPCC’s Second Assessment Report included anassessment of the impacts of climate change onagriculture (Reilly et al. 1996). As an assessment basedon existing literature, it summarized most of the fore-going studies. The overall conclusions included a sum-mary of the direct and indirect effects of climate andincreased ambient CO2, regional and global produc-tion effects, and vulnerability and adaptation. Withregard to direct and indirect effects, the conclusionswere as follows:

• The results of a large number of experiments toresolve the effect of elevated CO2 concentra-tions on crops have confirmed a beneficialeffect. The mean value yield response of C3crops (most crops except maize, sugar cane, mil-let, and sorghum) to doubled CO2 is +30 per-cent, although measured response ranges from–10 percent to +80 percent.

• Changes in soils (e.g., loss of soil organic matter,leaching of soil nutrients, salinization, and ero-sion) are a likely consequence of climate changefor some soils in some climatic zones. Croppingpractices such as crop rotation, conservationtillage, and improved nutrient management arevery effective in combating or reversing delete-rious effects.

• Changes in grain prices, changes in the preva-lence and distribution of livestock pests, andchanges in grazing and pasture productivity, aswell as the direct effects of weather, will affectlivestock production.

• The risk of losses from weeds, insects, and dis-eases is likely to increase.

With regard to regional and global productioneffects, the conclusions were as follows:

• Crop yields and productivity changes will varyconsiderably across regions. Thus, the pattern ofagricultural production is likely to change inmany regions.

20


• Global agricultural production can be main-tained relative to base production under climatechange, as expressed by GCMs under doubled-CO2 equilibrium climate scenarios.

• Based on global agricultural studies usingdoubled-CO2 equilibrium GCM scenarios, lower-latitude and lower-income countries are morenegatively affected.

With regard to vulnerability and adaptation, the con-clusions were as follows:

• Vulnerability to climate change depends notonly on physical and biological response butalso on socioeconomic characteristics. Low-income populations that rely on isolated agricul-tural systems, particularly dryland systems insemi-arid and arid regions, are particularly vul-nerable to hunger and severe hardship. Many ofthese at-risk populations are in sub-SaharanAfrica, South Asia, and Southeast Asia; they alsoinclude some groups in Pacific Island countriesand tropical Latin America.

• Historically, farming systems have responded toa growing population and adapted to changingeconomic conditions, technology, and resourceavailabilities. There is uncertainty aboutwhether the rate of change of climate andrequired adaptation would add significantly tothe likely disruption from future changes in eco-nomic conditions, population, technology, andresource availabilities.

• Adaptation to climate change is likely; the extentdepends on the affordability of adaptive mea-sures, access to technology, and biophysical con-straints such as water resource availability, soilcharacteristics, genetic diversity for crop breed-ing, and topography. Many current agriculturaland resource policies are likely to discourageeffective adaptation and are a source of currentland degradation and resource misuse.

• National studies have shown incremental addi-tional costs of agricultural production under cli-mate change that could create a serious burdenfor some developing countries.

Material in the 1995 IPCC Working Group II reportwas reorganized by region with some updated mate-rial in a subsequent special report. Included amongthe chapters was a report on North America(Shriner and Street 1998).

1995–1996: Economic Research Service (ERS)

The ERS (Schimmelpfennig et al. 1996;Lewandrowski and Schimmelpfennig 1999) provid-ed a review and comparison of studies that it hadconducted or funded, contrasting them with previ-ous estimates. The assessment used the same threedoubled-CO2 equilibrium scenarios that Rosenzweigand Parry (1994) used but also added a fourth, cool-er model that produced a global average surfacetemperature increase of 2.5°C.

Two of the main new analyses reviewed in thestudy used cross-section evidence to evaluate cli-mate impacts on production. One approach was adirect statistical estimate of the impacts on landvalues for the United States (Mendelsohn et al.1994); the other (Darwin et al. 1995) used evi-dence on crop production and growing seasonlength in a model of world agriculture and theworld economy. Both imposed climate change onthe agricultural sector as it existed in the base yearof the studies (1982 and 1990, respectively). Amajor result of the approaches that were based oncross-section evidence was that impacts of climatewere far less negative for the United States and theworld than had previously been estimated withcrop modeling studies. Although the studiesshowed economic effects that were similar tothose of previous studies, they included no directeffect of CO2 on crops, which in previous studieshad been a major factor behind relatively smalleconomic effects. Hence, if the direct effect ofCO2 on crop yields had been included, the expect-ed result would have been significant benefits. Themore positive results were attributed to the adapta-tion implicit in cross-section evidence that had notbeen completely factored into previous analyses.

21


The assessment also reported a crop modeling study(Kaiser et al. 1993a,b) with a complete farm-leveleconomic model that more completely simulatedadaptation response. It, too, showed more adapta-tion than previous studies. A summary of thisreview was subsequently published asLewandrowski and Schimmelpfennig, 1999. Theassessment also reported a study on the agriculturaleffects of climate change in developing countries(Winters et al. 1999) that found gross domesticproduct (GDP) losses for low income, cereal-import-ing countries in Africa, Latin America, and Asia. Thisfinding was supported by a subsequent ERS-supported study (Darwin 1999), which found thatclimate change would have negative impacts onagricultural land and thereby reduce overall eco-nomic welfare in Southeast Asia, western and south-ern Asia, Latin America, and Africa. The latter studyalso showed that Southeast Asia, which is primarilyin the tropics, is much more adversely sensitive towarming than the United States. Neither studyincluded the direct effects of CO2 on crops.

1996–1999: Electric Power Research Institute

The Electric Power Research Institute (EPRI) fundeda study on the impacts of climate change on all mar-ket sectors in the continental United States. Threedifferent approaches were used to analyze agricul-ture. All three explored a range of hypothetical cli-mate scenarios combining 1.5°, 2.5°, and 5.0° Cwarming with 0 percent, 7 percent, and 15 percentprecipitation increases. The studies explored a 1990economy and a 2060 economy. Carbon dioxide lev-els were assumed to be 530 ppmv. Overall, the stud-ies found substantial benefits for the United Statesresulting from climate impacts on US agriculture.Adams et al. (1999b) used a crop productionapproach in conjunction with a linear programmingmodel to predict effects across major crops in theUnited States. The study adapted the agriculturalmodel constructed for the EPA (Adams et al. 1990)to include a more complete accounting of farmeradaptation, livestock, and warm-loving crops. TheAdams et al. (1999b) study found substantial

benefits, with 1.5° and 2.5° C warming leading tobetween $32 billion and $54 billion in 2060. Thesebenefits were reduced with a 5° C warming tobetween $9 billion and $32 billion. The study wasunique in finding significant net economic benefitsacross the range of scenarios examined. When cli-mate change was imposed on a 1990 economy, themagnitude of benefits was similar to the magnitudeof benefits found in earlier studies for at least somescenarios. The relatively large benefits for 2060reflects the fact that the underlying agriculturaleconomy was considerably larger as a result ofassumptions about growth in productivity.

Segerson and Dixon (1999) used cross-sectional datafrom the Midwest Plains to analyze grain crops.They relied on a production function model to esti-mate crop climate sensitivity. They found that cropsensitivity was slightly less than what Adams et al.(1999b) had assumed. These lower sensitivitieswere then introduced into the Adams et al. (1999b)model, which then generated slightly higher benefitsfrom warming.

Mendelsohn, et al. (1999) explored cross-sectionalanalysis across all counties in the continental UnitedStates that had agriculture. The model accounted forfarm value per acre and the fraction of land used forfarming. The model also accounted for climatenorms and climate variation. The study found thatincluding variation changed the measured sensitivityof crops to warming. With variation in the model,warming is more beneficial. Climate variation itself,however, was highly damaging. The cross-section(Ricardian) analysis suggested net benefits fromwarming that were similar to the Adams et al.(1999b) study for the United States.

EPRI also has funded Ricardian studies in Brazil(Sanghi and Mendelsohn 1999) and India (Dinar etal. 1998; Sanghi et al. 1998); the World Bank also sup-ported the latter. The Brazilian and Indian studiesreveal that the Ricardian model works well in devel-oping countries. Warmer winters and summers areharmful in both of these countries—as they are inthe United States. Brazil and India, however, appear

22


to be more sensitive to warming than the UnitedStates. Even adjusting for their different initial tem-peratures, the developing countries appear to bemore temperature sensitive (Sanghi and Mendelsohn1999).

1998–1999: Pew Center

As part of a series on various aspects of climatechange aimed at increasing public understanding, thePew Center on Global Climate Change completed areport on agriculture (Adams, Hurd, and Reilly 1999a).The report series is based on reviews and synthesisof the existing literature. The major conclusions wereas follows:

• Crops and livestock are sensitive to climatechanges in positive and negative ways.

• The emerging consensus from modeling studiesis that the net effects on US agriculture associat-ed with doubling of CO2 may be small; regionalchanges may be significant, however (i.e., therewill be some regions that gain and some thatlose.) Beyond a doubling, the negative effectsare more pronounced in the United States andglobally.

• Consideration of adaptation and human responseis critical to an accurate and credible assessment.

• Better climate change forecasts are a key toimproved assessments.

• Agriculture is a sector that can adapt, butchanges in the incidence and severity of pests,diseases, soil erosion, tropospheric ozone, vari-ability, and extreme events have not been fac-tored into most existing assessments.

General Results and Conclusions from Past Assessments

Several general results and conclusions arise amongpast assessments; for those who have been involvedin the research, they have become common wisdom

or consensus conclusions. There are, however,important caveats and limitations to existing assess-ments. These limitations exist not becauseresearchers have not recognized them but because,for one reason or another, overcoming these limita-tions in ways that have been convincing to mostother researchers has proved difficult or impossible.Until more convincing evidence is marshaled onone side or the other, these limitations introduceuncertainty in the conclusions. We list first themajor conclusions and then the major limitations ofassessments to date.

Major Agreement and Consensus

• Over the next 100 years and probably beyond,human-induced climate change as currentlymodeled will not seriously imperil aggregatefood and fiber production in the United States,nor will it greatly increase the aggregate cost ofagricultural production. Most assessments havelooked at multiple climate scenarios. About halfof the scenarios in any given assessment haveshown small losses for the United States(increased cost of production); about half haveshown gains (decreased cost of production).1

• There are likely to be strong regional produc-tion effects within the United States, with someareas suffering significant loss of comparative(if not absolute) advantage to other regions ofthe country. With very competitive economicmarkets, whether a particular region gains orloses absolutely in terms of yield matters little;what matters is how it fares relative to otherregions. The south and southeastern UnitedStates are persistently found to lose relative toother regions and absolutely. The effects onother regions within the United States are lesscertain. Although warming can lengthen grow-ing seasons in the northern half of the country,the full effect depends on precipitation, whichclimate models predict notoriously poorly.

1Assessments have used several different "yardsticks" for measuring effects.These yardsticks include measures such as total grain productionin tons or value of production, commodity prices, and economic welfare.The latter concept is generally favored among economists as show-ing the true economic cost.Although there are many differences among these measures, the basic conclusion here is not particularly sensi-tive to which measure is used.

23


• Global market effects and trade dominate interms of net economic effect on the US econo-my. Just as climate’s effects on regional compar-ative advantage is important, the relevant con-cern is the overall effect on global productionand prices and how US producers fare relativeto their global competitors or potential com-petitors. The worst outcome for the UnitedStates would be severe climate effects on pro-duction in most areas of the world and particu-larly severe effects on US producers. Consumerswould suffer from high food prices, producerswould have little to sell, and agricultural exportswould dwindle. Although this outcome isunlikely (based on newer climate scenarios),some early scenarios that featured particularlysevere drying in the mid-continental UnitedStates with milder conditions in Russia, Canada,and the northern half of Europe produced amoderate version of this scenario. The UnitedStates and the world could gain most if climatechange was generally beneficial to productionworldwide but particularly beneficial to US pro-ducing areas. Consumers in the United Statesand around the world would benefit from fallingprices, and US producers would gain becausethe improving climate would lower their pro-duction costs even more than prices fell, thusincreasing their export competitiveness. In fact,most scenarios fall close to the middle, with rel-atively modest effects on world prices. The larg-er gainers in terms of production are the morenorthern areas of Canada, Russia, and northernEurope.Tropical areas are more likely to sufferproduction losses. The United States as a wholestraddles a set of climate zones that includegainers (the northern areas) and losers (southand southeast).

• Empirical studies of climate sensitivity willhave to be completed in more developingcountries to get an accurate picture concern-ing climate effects around the world.Specifically, there is very little information aboutAfrica, even though it is likely to be one of themost sensitive areas to warming in the world.

• Effects on producers and consumers often arein opposite directions, which often is responsi-ble for the small net effect on the economy. Thisresult is a near certainty without trade; it reflectsthe fact that demand is not very responsive toprice, so anything that restricts supply (e.g.,acreage reduction programs, environmental con-straints, or climate change) leads to price increas-es that more than make up for the reduced out-put. Once trade is factored in, this resultdepends on what happens to production abroad.

• US agriculture is a competitive, adaptive, andresponsive industry and will adapt to climatechange; all of the foregoing assessments have fac-tored adaptation into the assessment to somedegree. The final effect on producers and theeconomy after adaptation is considered may benegative or positive. The evidence for adaptationis drawn from analogous situations, such as theresponse of production to changes in commodityand input prices, regional shifts in production aseconomic conditions change, and the adoption ofnew technologies and farming practices.

• The relatively small net effect on the US agri-cultural economy across assessments is thecombination of a variety of negative and pos-itive effects. In many of the earlier assess-ments, the direct effect of carbon dioxide onplant growth offset fairly large yield declinesrelated to changes in temperature and precipi-tation. Some later assessments have not includ-ed the carbon dioxide effect at all but have esti-mated a much larger adaptation response andhave found small negative and even positiveeffects despite the omission.

• The agriculture and resource policy environ-ment can affect adaptation. Lack of water mar-kets, agricultural commodity programs, cropinsurance, and disaster assistance can encouragethe continuation of practices that are no longereconomic on a regular basis. The FAIR Act of1996 eliminated farm program payments tied tobase acreage (failure to maintain base acreage ina crop could mean loss of payments, which

24


encouraged continued production of the samecrop). More-effective water markets could trans-fer water to the highest-value uses and encouragegreater irrigation efficiency, but establishment ofmarkets is hampered by water laws dating to the1800s that granted water rights in the far westand open access to subsurface resources in theplains states. The pressure of increasing competi-tion for these resources is leading to someprogress in this regard. Crop insurance and disas-ter assistance can have the perverse effect ofencouraging continued cropping in areas that areprone to crop disasters, essentially subsidizingproduction in areas that are no longer competi-tive. There is growing awareness of the perverseeffect these programs can have and some interestin managing them in ways that minimize or elimi-nate the effect. It appears difficult, however, forCongress and the administration to resist pres-sure to come to the aid of farmers in a time ofneed regardless of whether those in need havethemselves prepared well for the inevitablevagaries of weather and the variability ofcrop prices.

There have been several assessments of the agricul-tural impacts of climate change; the consensus andagreement among the studies is strengthened by thefact that the assessments were conducted by differ-ent teams of researchers, using different methods,and were sponsored by different organizations. Allof these research teams have labored under thesame set of constraints, some quite severe; thus,many of the results are conditioned on these limits.These limitations include the following:

• The climate scenarios on which these results relyhave been very unrealistic representations ofwhat climate might really be like over the nextseveral decades to 100 years. Most climate sce-narios are based on doubled-CO2 equilibrium sce-narios. There is no particular future year to whichthese scenarios apply, and other factors that affectclimate such as sulfate aerosols have not beenincluded. One assessment assumed that climateswere realized in 2060; most others apply the

conditions to today’s agriculture and are silentabout when the effects might be realized. As aresult, there are no estimates of climate impactsfor the next several decades that are based onactual results of climate models and no estimatesof potential consequences in the far distantfuture—beyond a doubled-CO2 environment.

• Detailed predictions of climate models are par-ticularly uncertain; most climate modelersplace little or no confidence in the detailsbecause the processes that control these detailsare not well represented. Clouds and precipita-tion are key concerns. The big climate modelsdo a poor job of representing current variabilityand do not simulate events such as the El Niño-Southern Oscillation (ENSO), hurricanes, andtyphoons—nor do they have any ability to rep-resent changes in small-scale convective storms.

• The climate scenarios used represent atmo-spheric physics as currently understood, almostexclusively constructed for research rather thanassessment purposes. The scenarios had limitedor no interaction with oceans and terrestrial sys-tems and excluded other climate forcings. Forassessment purposes, trying to roughly take intoaccount as many things as we think are impor-tant would be far preferable to being very pre-cise about the things we know well while leav-ing out completely things we suspect but havenot proved. Having a range of scenarios thatbound our uncertainty about these many fea-tures rather than everyone’s version of a centralestimate (central, conditioned on recognizingthat some things were left out completely) alsowould be preferable. Scenarios that could hap-pen with great consequence but with low prob-ability need to be assessed, appropriately dis-counted for the fact that there might only be a 1in 100 or 1 in 1000 chance of occurrence.

• The CO2 fertilization effect will probablyincrease yields, but the magnitude of the effectremains uncertain. Experimental evidence sug-gests an average yield increase of 30 percent formany crops but closer to 7 percent for corn,

2The distinction here is between C3 and C4 crops (referring to the pathways through which carbon is utilized).The C4 crops—corn,sorghum, and sugar cane—experience much less gain.Virtually all other crops of commercial importance are C3 crops.

25


sorghum, and sugar cane2 under doubled-CO2

levels (from~ 300 ppm to ~600 ppm) andimprovements in water-use efficiency. Therange of experimental results of doubled CO2 isfrom –10 percent to +80 percent; some investi-gators would fasten on the low end of thisrange. A wide variety of factors could reducethe anticipated gain considerably. Only abouttwo-thirds of the increase in greenhouse gasforcing may be caused by CO2; other gaseswould cause warming but not have beneficialeffects. Most of this experimental evidence isfrom single plants grown under glass (highlyartificial conditions); the effects could be quitedifferent under open-field conditions, with pes-simists imagining necessarily less effect. TheCO2 effect depends on and interacts with manyother factors—probably explaining, in part, thewide range of experimental results. Grain quali-ty and forage quality may be reduced (less pro-tein) for crops grown under elevated CO2. Notall of these interactions necessarily would leadto a lower fertilization effect. For example, the evi-dence indicates a stronger effect when crops areunder stresses such as water,heat, and salinity—conditions that are more likely to be observedunder commercial conditions than experimentalconditions. Most of the crop models used inassessments apply a very simple multiplier torepresent elevated CO2 rather than model thecomplex interactions explicitly.

• Many broader agroecological (system-wide)effects have not been included in assessments.The dominant “crop model methodology”simu-lates only the short-term and local effects ofessentially different weather on crop growth.Persistent changes in weather (i.e., climate) maylead to changes in soils, pest prevalence, irrigationwater availability, concentrations of other pollu-tants such as tropospheric ozone, and changes inthe ability of farmers to conduct field operations.For the most part, these factors have not beenexplicitly incorporated into assessments.

• The extent, ease, and cost of the adaptationresponse are controversial and unresolved.Although some amount of adaptation is

inevitable, some analysts question whether theanalogous situations that are used as evidence ofadaptability are good analogies for climatechange. Gradual climate change may be difficultto detect. Hence, the producer may not knowthat climate has changed; he or she may inter-pret a string of odd weather as normal variabili-ty and thus experience losses for some timebefore he or she recognizes that climate haschanged. There also is debate about adjustmentcosts—whether climate will change so graduallythat any adaptation can be handled as a part ofnormal replacement of capital or whether adap-tation will require disruptive and costly replace-ment of equipment made obsolete by changingclimate. For adjustment to be costly, local cli-mates likely would have to experience sometype of punctuated change; the global averagechange in temperature is quite slow relative tothe normal rate of capital turnover in agricul-ture. There is little confidence, however, that cli-mate models would capture such types ofchange, if indeed they were a possibility.

• Regional and local predictions remain vaguelyprobabilistic at best. For example, the findingthat the South and Southeast usually have beennegatively affected may not apply to every cor-ner of the region, every crop grown there, orevery climate scenario. The predictability ofdetail at the small geographic levels for manykey dimensions of climate is nearly zero. Theclimate models themselves are only coarselyresolved. Better downscaling methods are beingapplied but have not been broadly used in theforegoing assessments.

Approach of the CurrentAssessment

As the review of past efforts suggests, there are twobroad methods of assessment: Review and synthe-size existing literature, or conduct a broad-scalemodeling/analysis effort centered on a consistent setof scenarios. The IPCC and Pew Center efforts areexamples of the former approach; the EPA and EPRI

26


efforts are examples of the latter. There also are twobroad objectives of assessments: Assess the impact(measured in a variety of ways) of climate change onagriculture, and assess strategies for limiting or avoid-ing negative consequences or take advantage ofopportunities presented by climate change. TheCAST and OTA assessments were examples of the lat-ter; the USDA and EPRI studies are examples of theformer. The second IPCC assessment, which used lit-erature review, included an evaluation of impacts andpotential responses that could limit impacts.Assessments also vary in their attempts to providequantitative information and qualitative conclusions.

This assessment tackles several of the caveats andlimitations—but not all. We use transient climate sce-narios and therefore are able to consider impactsrepresentative for the years 2030 and 2090. Thisapproach is a substantial improvement comparedwith previous analyses; whether and what types ofactions might be taken over the next 5 to 10 yearsdepend on when the climate impacts are expected.We evaluated and include in our assessment thepotential implications of changes in pesticide expen-ditures resulting from climate change. The issue ofpests and climate remain uncertain, but this inclusionadds another dimension to the complex climateagroecosystem interactions we might ultimatelyexpect. We have evaluated a broad group of crops,including the major grains (wheat, corn, sorghum)and soybeans, forage crops (alfalfa and range), andsome of the more important fruits and vegetables(tomatoes, citrus, and potatoes). By including vegeta-bles and fruits, as well as other crops that are heatloving, we help remove a potential bias in some pre-vious work that considered only the major grains; theconcern with some of these studies was thatomitting heat-loving crops that may have benefitedfrom warming could have overestimated damages.We also have considered more completely the effectsof climate change on irrigation water supply. Wewere able to use results of the water sector assess-ment to evaluate more realistic changes in water sup-ply to agriculture.

We provide a brief discussion of the scenarios usedfor the various analyses. Then we provide asummary and overview of models used in the analy-sis. Finally, we provide a brief discussion of sur-prise, uncertainty, and the scope of climate–agroe-cosystem–economic interactions. The ability toassess the complete system in all its complexitydoes not yet exist; conveying a sense of these com-plexities is useful nonetheless.

Scenarios

Socioeconomic Scenarios and Assumptions

Following the pattern of many past assessments ofclimate change impacts, we applied climate changeto the cropping and economic system as it existstoday (i.e., the year 2000). To many observers, thisapproach appears to go against common sense.Crop yields are likely to be higher in the future, agri-cultural prices will be different, land-use patternswill change, the global trade picture will change, andthe entire set of technological options available tofarming will change. Indeed, our steering commit-tee suggested that we must consider climate changeoperating in a future world. Paraphrasing one com-mittee member, the historical response and even theresponse of today’s agricultural system is irrelevantbecause agriculture is changing so fast.

The assessment team chose to take an alternativeapproach, for several reasons. The simple answerwas that developing interesting scenarios of thefuture that differed in ways that are important interms of climate response would have requiredresources beyond those we had. There is no widelydeveloped set of long-term forecasts for agriculture.The ERS produces a 10-year ahead baseline for theUnited States. We require scenarios for 30 and 90years in the future. Several forecasts of world agri-culture try to look out 30 years (for a review, seeReilly and Fuglie 1998); these types of scenarios donot necessarily change the sensitivity of agricultureto climate change, however.

27


The EPA global study and the MINK study devel-oped future scenarios of world agriculture and agri-culture for the MINK region, respectively. Thelessons from these studies and from other futureforecasts are as follows:

• Future prices and other measures of agriculturalshortfall or excess depend almost completelyon the rate of yield growth relative to popula-tion growth.

• Any extrapolation of yield growth at rates likethose experienced over the past few decadeswill result in yields at least 70 percent abovetoday’s yields by 2030; it is hard to imagine orconceive of crops that would maintain suchyield growth through 2090.

• Factors other than climate change are moreimportant for the agricultural economy in thefuture, and these factors are uncertain; chang-ing underlying assumptions within a rangemost experts would accept as bracketing whatmight happen in the future can lead to vastlydifferent and larger effects than those fromclimate change.

• When different assumptions about these otherfactors have been incorporated in climateassessment, they have not changed the climateresponse very much.

For example, after adjusting crop response to gener-ate higher yields, the MINK study still found aboutthe same percentage effect of climate change oncrops. The EPA global study found that for measuresof people at risk of hunger, the absolute numberincreased with population increase; because hungerrisk depended directly on income and food prices,scenarios with higher income or more rapid yieldgrowth produced smaller numbers of at risk people.One analysis used crop yield results from the EPAglobal study imposed on the current (1990) agricul-tural economy. This analysis came to broad conclu-sions that were similar to those of the original studyin terms of areas that win and lose as a result of cli-mate change and in terms of the net effect on theworld food system. As a first approximation,

measuring economic response in terms of producerand consumer surplus is likely to be relatively insen-sitive, in percentage terms, to the scale of activity(more or less production) and even to whetherprices have fallen or risen, unless demand and sup-ply responses are highly nonlinear.

The “noneffect” of other variables on climateresponse to forecast futures is hardly an absolutefinding or certainty, however. It probably reflectsinstead our inability to foresee or create scenariosthat would substantially change the climateresponse. If there were much more irrigation, ormuch less, the response to precipitation wouldchange. If future US agriculture concentrated inparticular areas that were much more beneficiallyor negatively affected by climate change than otherareas, the response would change. By 2090, cropsand production practices may be unrecognizable tous today; perhaps any fast-growing, highly produc-tive crop will be a feedstock for manufactured foodand feed products—eliminating (or nearly so) theneed to produce grain and other specialized crops.Suitable biomass crops might be grown undermany conditions, including freshwater and marineenvironments.

One problem with trying to assess what these differ-ent scenarios might mean for climate change is thatsuch dramatic changes may represent, in part, aresponse to a gradually changing climate. If techno-logical change itself is highly responsive to relativescarcity of land (and the climatic conditions that gowith it), the variety of dramatically different scenar-ios would develop only under some climate scenar-ios but not others. Considerable evidence has beencollected by some researchers (Hyami and Ruttan1985) showing strong endogenous response of tech-nology to relative input prices. In this framework,broadly worsening climate conditions wouldincrease the price of land in the few remaining goodareas, and these price increases would spur techni-cal change to reduce the need for good climate. Forexample, the response might be to generate the pro-duction system outlined above as a possibility for2090, whereby almost any type of biomass cropcould be used as a feedstock for food production.

28


On the other hand, improving climate conditionscould turn many areas into potentially prime pro-ducing areas. This scenario could greatly reduce theneed for yield-enhancing research; improving cli-mate and higher levels of ambient CO2 would pro-duce yield increases without any research effort.Research dollars would be invested more profitablyelsewhere rather than to spur even greater yieldincreases that caused commodity prices to plum-met. The ability to quantify and forecast thisendogenous response over long periods of time isalmost nonexistent at present and presents aformidable challenge for research.

For the foregoing reasons, we chose to impose cli-mate on agricultural markets as they exist today, sup-plementing this modeling work with a discussion ofpossible future changes and how they could alterclimate sensitivity of agriculture.

With regard to the future, our stakeholder meetingidentified several important changes for agriculture.Given their importance, speculating on how thesechanges might interact with climate sensitivity isworthwhile. The first of these changes is the tech-nological change. Precision agriculture and biotech-nology are the two main technological forcesbehind agricultural research at the moment.

Precision agriculture allows farmers to precisely anddifferentially manage (in terms of application ofwater, nutrients, pesticides, etc.) small areas of a fieldby using computer monitoring and global positionsystems. The idea is that much more efficient use ofinputs and higher yields are possible by directingthe right amount of input to each area rather thanusing an average amount of input that is too muchin some areas and too little elsewhere. Althoughprecision agriculture may have such effects, it is notclear that it would reduce climate sensitivity. A cropgrowing with ideal levels of nutrients, water, andpest control would still be subject to losses from cli-mate. Indeed, the current practice tends to entailrelatively high levels of applications of inputs to gethigh yield over most of the field. More-careful moni-toring and faster response to changing conditions

could reduce adjustment costs, however, if farmersare able to detect and respond to changing weatherconditions more rapidly. Clear detection of climatechange, based on pure data analysis of historicweather, is fundamentally limited by the ability toseparate trends from a very noisy record.

Biotechnology offers the possibility to modify cropsand livestock well beyond the limits imposed by thegenetic diversity within varieties that can be inter-bred. Biotechnology appears to be capable of dra-matically changing the technological response.Some broad biological limits remain and concernsabout environmental and health risks may be imped-iments to development of genetically modifiedcrops.Without water, for instance, high levels ofbiomass production per hectare probably are notpossible. Genetic diversity across species, however,could allow improved response to many differentenvironmental conditions. If anything, biotechnolo-gy increases the potential for endogenous techno-logical change to minimize climate effects.

Globalization of markets and industrialization ofagriculture are two additional forces. A major forcebehind globalization is to ensure supply to marketsunder current weather variability. Along these lines,globalization almost certainly will reduce any nega-tive impacts of climate change on commodity andfood markets—minimizing the impact of climate onpeople who obtain their food from these markets.Globalization is likely, however, to amplify regionaleffects on producers and could further marginalizepoor people in developing countries. Already, theglobal market places considerable pressure on pro-ducing areas that have difficulty competing withmore-productive, lower-cost producing areas. With astrong network of interwoven international markets,crop failures in a region need not increase marketprices if the losses are balanced by gains elsewhere.In contrast, in a world with regionally differentiatedmarkets, producers in the failing area would benefitfrom higher regional prices. Food consumers in theregion obviously would pay more.

29


An interesting example of the attempt to shieldregional producers from competitors in otherregions is the milk marketing system that is gradual-ly being dismantled in the United States. Regionalconsumers paid higher prices, but these prices sup-ported a dairy industry in the Northeast againstcompetition from Wisconsin. Also at risk are subsis-tence farmers and consumers around the world.Governments and markets have not been particular-ly kind to traditional and tribal populations whenthey have had the unfortunate luck of being locatedon a resource that became valuable. If climatechange caused world commodity prices to rise,wealthy consumers in developed countries wouldimport food they need, leaving less available forpoorer consumers in developing countries whocould not afford higher prices.

Industrialization of agriculture is a broad idea, incor-porating many different changes in the structure ofthe agriculture sector. In part, it includes theincreasing technological sophistication and preci-sion management of production that allows produc-tion of commodities to meet processing specifica-tions. It also includes the increasing horizontal(across producing entities and regions) and vertical(with input and processing industries) integration ofproduction. One feature of this structural change iscontract production, whereby many smaller farmsproduce under contract with a processor with someform of price guarantee and with greater specifica-tion for inputs and production practices to assureuniformity and timely delivery of the product. Onefeature of this form of production is that the largeprocessor pools risks across many farmers and areas,creating greater assurance of return for farmersunder contract. This broad-scale integration is likelyto reduce further the chance that a local or regionalcrop failure will disrupt supply in the region.Integration also will pool income risks for produc-ers. Contract production could have similar effects,but the relative risk to the producer and contractordepends on the specific terms of the contract.

The other major trend in US agriculture is increasingdemand for improved environmental performance.

We examine many of these issues in more detail inChapter 5. There are three broad issues. One iscompetition between agriculture and the environ-ment for resources—mainly land and water. In thewestern United States, the desire to improve fishhabitat (e.g., salmon spawning areas on rivers) isleading to a rethinking of the allocation of waterand pressure to remove dams that supply water.There is continuing debate and discussion aboutgrazing on federal land and its implications forwildlife habitat. Other concerns about endangeredspecies habitat, wetland preservation, and furtherdemands for parkland and open space will increas-ingly bid for land now in agriculture. We investigatecompetition for groundwater in the Edwards Aquiferin the area including San Antonio,Texas. We alsoexamine overall agricultural resource use implica-tions in Chapter 3.

A second issue involves interactions of agricultureand urban/suburban land in the landscape. There arepositive and negative aspects of this interaction.Farmland can provide green space in the midst ofurban development. Such farmland can provideunique services and products for the local urbanarea, from fresh produce for farmers markets to farmexperiences for urban dwellers. On the negativeside, intensive production, particularly large livestockoperations, has created large concerns about odorand pollution. The positive aspects of this interac-tion have led many states to develop programs topreserve farmland.The negative aspects have led toregulations and prohibitions on farming practices.

A third aspect is production practices that lead topollution—mainly water pollution, but with recentconcerns about air pollution effects. Soil erosionrunoff into lakes and rivers carries with it nutrientsand agricultural chemicals. Irrigation drainage wateralso concentrates chemicals and salts in water bod-ies. Leaching of chemicals applied to crops can leadto groundwater contamination. Climate change hasthe potential to greatly affect these interactions bychanging land use, irrigation water use, and theintensity of rain and wind that is responsible for ero-sion. We consider the impact on land and water use

30


in Chapter 3 and the impact on soils, nutrient runoffinto the Chesapeake Bay, and implications for pesti-cide expenditures in Chapter 4. As our case studiesin Chapter 4 illustrate, the drive to improve agricul-ture’s environmental performance could, by itself,significantly change farming practices—which couldgreatly affect how climate change will affect agricul-ture and the environment.

Climate Scenarios

We used the Hadley and Canadian model simula-tions to develop climate scenarios for the crop mod-eling work. In this regard, we followed previousagricultural assessments and applied monthly meanchanges in climate between the greenhouse gas-forced scenarios and control runs to a 30-year actualrecord of weather for sites at which we ran cropmodels. This approach has been used in the pastbecause, although climate model output broadlyagrees with observed seasonal and spatial patterns ofclimate, the agreement with actual weather at a spe-cific site is very poor. Applying the differences (addi-tive for temperature and as a ratio for precipitation)means, for example, that all days are warmer but thepattern of warm and cool days (i.e., the variance)remains the same. Thus, any change in variance pro-jected by the GCMs is averaged out. We discuss laterin more detail what the climate models indicateabout variance of weather and climate and someresults using changes in variability.

Broadly, the Hadley and Canadian scenarios fall inthe middle and at the high end, respectively, of IPCC(1996) projections of warming by the year 2100; thenewer IPCC projections, however, are somewhat high-er. Both scenarios have increased precipitation atthe global level, consistent with the enhancedhydrological cycle accompanying warming. For thecontinental United States, the Canadian model sce-nario projects a 2.1º C average temperature changewith a 4 percent decline in precipitation by 2030and a 5.8º C warming with a 17 percent increase inprecipitation by 2095. The Hadley scenario projectsa 1.4º C (2030) and 3.3º C (2095) increase in tem-perature with precipitation increases of 6 and 23percent, respectively. Both scenarios indicate more

warming in the winter and relatively less in the sum-mer. The mountain states and Great Plains are alsoprojected to experience more warming than otherregions in both models. The Hadley scenario alsoshows greater warming in the Northwest. Moredetail on the climate scenarios is available athttp://www.cgd.ucar.edu/naco/vemap/vemtab.html.

Agricultural Models

Climate and other factors strongly interact to affectcrop yields. Models have provided an importantmeans for integrating many different factors thataffect crop yield over the season. Scaling-up resultsfrom detailed understanding of leaf and plantresponse to climate and other environmental stress-es to estimate yield changes for whole farms andregions can, however, present many difficulties (e.g.,Woodward 1993).

Higher-level, integrated models typically accom-modate only first-order effects and reflect morecomplicated processes with technical coefficients.Mechanistic crop growth models take into account(mostly) local limitations in resource availability(e.g., water and nutrients) but not other consider-ations that depend on social and economic responsesuch as soil preparation and field operations,management of pests, and irrigation.

Models require interpretation and calibration whenthey are applied to estimate commercial crop pro-duction under current or changed climate condi-tions (see Easterling et al. 1992; Rosenzweig andIglesias 1994); in cases of severe stress, reliabilityand accuracy to predict low yields or crop failuremay be poor. With regard to the CO2 response,recent comparisons of wheat models have shownthat even though basic responses were correctlyrepresented, the quantitative outcome betweenmodels varied greatly. Validation of models has beenan important goal (Rosenberg et al.1992; Olesen andGrevsen 1993; Semenov et al. 1993a,b;Wolf 1993a,b;Delécolle et al. 1994; Iglesias and Minguez 1994;Minguez and Iglesias 1994).

31


To generate results at the national and global level,results from crop models are used in an economicmodel (e.g.,Adams et al. 1995; Reilly et al. 1994).There are two basic types of economic models:

• Those that include costs of many different activi-ties (crops, cropping practices, rotations, etc.)(e.g.,Adams, et al. 1995). With changed condi-tions, such as changed productivity resultingfrom climate changes, such models find the leastcostly way to satisfy demand.

• Those based on statistical estimates of supplyand demand for individual crops (e.g., Reilly etal. 1994). Changes in climate can then be repre-sented as shifts in supply.

The activity type of model tends to have much morespatial and cropping practice detail. We apply theactivity model in this assessment because of the spa-tial and crop detail.

There have been efforts to further integrate cropyield, phenology, and water use with geographic-scaleagroclimatic models of crop distribution (Brown andRosenberg 1999; Kenny et al. 1993; Rötter and vanDiepen 1994), thereby providing greater representa-tion of diverse conditions across a large geographicscale. There also have been efforts to integrate cropmodels and farm-level economic response (e.g.,Kaiser et al. 1993). Simplified representations of cropresponse have been used with climate and soil datathat are available on a global basis (Leemans andSolomon 1993). More aggregated statistical models(Ricardian models) have been used to estimate thecombined physical and socioeconomic response ofthe farm sector (Mendelsohn et al. 1994). There alsohave been efforts to integrate high resolution (e.g.,0.5 degrees latitude by longitude) agroclimatic mod-els with applied general equilibrium economic mod-els to simultaneously capture farm-level adaptationsand the response of the US farm sector to climate-induced changes in other domestic and internationalmarkets (Darwin et al. 1995).

Incorporation of the multiple effects of CO2 in mod-els generally has been incomplete. Some models do

not include any CO2 effects and thus may overesti-mate negative consequences of CO2-inducedchanges in climate. Other models consider only acrude yield effect. More detailed models considerCO2 effects on water use efficiency (e.g.,Wang et al.1992). With few exceptions, most models fail toconsider CO2 interactions with temperature andeffects on reproductive growth. The erosion pro-ductivity impact calculator (EPIC) model incorpo-rates the CO2 effect in a relatively simplified fashion(Stockle et al. 1992a,b).

We use site-level models for our basic analysis, fol-lowing the approach used in many previous assess-ments. To examine the sensitivity of our results tothis modeling approach, we also have applied theBrown and Rosenberg (1999) model. It has fewercrops and is expensive to use, so we simulated itonly with the Hadley scenario. The results arereported in detail in Izaurralde et al. (1999). Themodel projects corn, winter wheat, soybeans, andalfalfa under dryland and irrigated conditions. Thisstrategy allowed us to investigate the extent towhich the projections of this crop modelingapproach differ from the site approach. In both ofthese approaches, the crop models include a CO2

fertilization effect; we also include in our estimateshigher ambient levels of atmospheric CO2 consis-tent with the climate scenarios (specific assump-tions are provided in Chapter 3). The reduced-formstatistical approach (Ricardian analysis) ofMendelsohn et al. (1994) is relatively simple toapply, once the response is estimated. However, itdoes not include a CO2 fertilization effect, and itcaptures all response as change in land value. Thus,there is no detail on specific crops. The case for thisapproach is that it takes better account of farm-levelresponse, at least under long-run equilibrium condi-tions, and includes (implicitly, though not explicitly)all crops that contribute to agricultural land value.

Broadly, our approach has been to try to use severaldifferent approaches and to test results with sensi-tivity analysis. This strategy has allowed us to con-sider the extent to which the results depend on theparticular method used.

Vulnerability, Surprise,Uncertainty

Quantitative analysis of climate change impacts facesmany difficult challenges. The great value of quanti-tative analysis is that it enforces considerable rigoron our thinking about effects. The limitations arethat potential interactions are only partly or poorlyquantified and often are not incorporated in assess-ment models; climate scenarios are uncertain; wehave only a vague idea of what agriculture may looklike in the future, when climate change is expectedto occur; and, with something as far-reaching asglobal climate change, there are likely to be thingsthat happen that we never foresaw or imagined.These set of concerns have caused analysts toapproach assessment in ways other than the linearapproach that typically has been used (e.g., from cli-mate scenario to crop impact to economic impact).

Vulnerability and sensitivity analysis has been onealternative approach. The idea is that climate scenar-ios are so uncertain that one should investigateinstead a wide range of climatic conditions. Suchanalysis identifies climatic conditions that are partic-ularly damaging. Applied to agriculture, the analysismight then identify actions that could be taken toreduce or eliminate these damages. Such anapproach is one way to avoid the narrow range ofclimate conditions simulated by GCMs. The difficul-ty, however, is that imagining disastrous weather isnot difficult, and spending large amounts of moneyto protect oneself against an outcome that isextremely unlikely to occur would not make sense.The usefulness of this approach rests in findingthings that are simple, cheap, and easy to do thatcould insulate one against things that one had notanticipated.

If a probabilistic scenario analysis can be completed,one can include the probability and damage associ-ated with each scenario in an uncertainty/vulnerabil-ity analysis. In principle, one can estimate theexpected cost associated with climate and under-take only actions whose costs were less than theexpected reduction in damages (for a more formaldiscussion, see Reilly and Schimmelpfennig 1999).

For example, avoiding a $10,000 damage that hadonly a 1 in 100 chance of occurring would be worthonly $100. Unfortunately, current climate modelingis unable to generate such probabilistic scenarios.

The other concern is the potential for surprises: cli-mate interactions with agriculture that we neveranticipated. By their very nature, once we havethought of such interactions, they are no longer acomplete surprise. It is easy, however, to make themistake of applying existing assessment approachesand models, implicitly assuming that they contain allof the important interactions. The antidote to thistrap is to rethink fundamental relationships andinteractions, consider broader connections, and con-duct targeted research to investigate some of thelinks where little is known.

What are possible surprises? The most significantsurprise for agriculture would be significantly differ-ent climate scenarios than are now projected by themajor climate modeling centers. Significant increas-es in variability could greatly disrupt agriculture.(We consider this issue in detail in Chapter 4.)Climate scenarios used to date are mainly centraltendency estimates and do not exhibit major nonlin-earities or state changes. Describing the likelihoodor the character of such scenarios is well beyondthe scope of the agriculture assessment, but theimpacts on agriculture of such climatic conse-quences of warming would be far different than anyscenarios evaluated to date, including those in thisassessment. Under such scenarios, rapid change—atleast at a regional level—could occur and bring withit significant adjustment costs.

Within the agricultural system, the development ofnew pests or expanded pest range and greater resis-tance to control methods are possible but difficultto foresee. We know that weather and climatic fac-tors are one critical element of pest range, but weare poorly equipped to evaluate the full set of habi-tat interactions. We will observe climate change as achange in extreme events (more hot days and fewercold days; more heavy rain or longer droughts)rather than changes in the means. One-in-100-year or

32

33


one-in-1,000-year events will always be a surprise.Our ability to identify whether the occurrence ofsuch an event signals a change or is simply a matterof chance will at least partly determine whether wego back to doing the same things or adapt. In thisregard, institutional preparedness and response arenearly impossible to predict. An unwillingness toadapt and change, rigidities in policy, or counterpro-ductive policy responses could increase costs.Faced with loss of comparative advantage andthreats to its local farming community, a regionmight seek federal money to subsidize farming, cre-ate protectionist trade policy, or build huge waterprojects only to maintain regional production. Suchprograms could have huge economic and environ-mental costs and ultimately might fail as climaticconditions continue to change.

Finally, we know very little about how a regional andlocal economy responds to multiple changes. Thelocal tax base, recreation, agriculture, water, andforests would be affected simultaneously. History hasmany cases of regions and communities decliningand depopulating when a critical resource isexhausted, an industry on which a community isbased fails or fails to keep pace with competitors, orother areas are deemed more livable or more fashion-able. On the other hand, many areas have diversified,shifted, and reoriented themselves to take advantageof new conditions.

35

Impacts of Climate Change onProduction Agriculture and theUS Economy

Chapter 3

Introduction

Climate change affects farmers and the US economythrough several different pathways. Analysis of effectssuch as impacts on crop yields, water demand, watersupply, and livestock, using biophysical models, cantell us a great deal about why a particular climate sce-nario causes yields to rise or fall. This analysis alsocan suggest directions for adaptation at the farmlevel. All of these changes occurring together andacross the United States and the entire world meanthat national and global markets can be affected.Thus, assessment of the economic viability of farmingand impacts on consumers and the US economyrequires consideration of the full effect of changes incrop yield, water demand, water supply, pests, andlivestock as they vary across the country and theworld. Many studies have demonstrated that farmerscan suffer economic losses even if crop yieldsimprove because commodity prices fall (see Chapter2). The net effect on the US economy can be positivein this situation because consumers gain from lowerfood prices. Such results are sensitive to how climatechange affects agricultural production in the rest ofthe world.

The techniques and approaches we use in thisassessment build on previous efforts, the mostrecent of which is reported in Adams et al. (1999).The other notable direct ancestor of this work isAdams et al. (1990).

In this analysis, we considered five principal directeffects of climate change:

• Crop yields and irrigated crop water use

• Irrigation water supply

• Livestock performance and grazing/pasture supply

• Pesticide use

• International trade

We combine these effects of climate change in aneconomic model that determines the new set ofprice, consumption, regional production, andresource use levels.

The focus of the analysis was to estimate the conse-quences for the agriculture sector of climate-inducedchanges via each of the foregoing mechanisms interms of the overall level of producer income andthe welfare of agricultural consumption by con-sumers. We also estimated changes in the location ofproduction and utilization of resources as influencedby climate change. We estimated these changes byusing a US national agricultural sector model (ASM)that is linked to a global trade model. In particular,the basic analytical approach entailed introduction ofestimates of climate change-induced alterations inthe five data items and examination of how themodel solution differs from the base solution, with-out climate change. The most important aspect ofthis analysis was the generation of changes in neces-sary inputs into the economic model such as cropyields and water demands for irrigation. Severalteams of crop modelers simulated changes in yieldsand water demand to provide these changes.

The results were simulated for transient scenariosdrawn from the Canadian model and the Hadleymodel. Although the impact analysis was based onthese transient scenarios, it used average climateconditions for the 2020–2039 and 2080–2099 peri-ods. Simulated climates vary from year-to-year. Theintent of using 20-year averages was to produceclimate change scenarios that were more representa-tive of the years 2030 and 2090 than would be anysingle year of simulated climate. We thus refer tothe results generated in this way as results for theyear 2030 and 2090.

The underlying yield and water demand changeswere simulated for crops like those that exist today.Similarly, changes in pesticide use, water supply,

36


livestock changes, and trade scenarios are based onpatterns that exist today. The economic results wereproduced by simulating the impact of climatechange on the agricultural economy as it existed inthe year 2000. We also considered the impacts ofclimate change on a scenario of the agriculturaleconomy projected forward to the years 2030 and2090. These scenarios took advantage of scenariosgenerated under the forest sector assessment (Joyceet al. 2001; Irland et al. 2001). The ASM model weused in this analysis is part of the combined forest-agriculture sector model that was used in the USNational Assessment Forest Sector Assessment. Thus,we were able to simulate the combined effects offorest and agricultural changes on the US economyand consider the implications for land use.

We considered climate change via the five principaldirect effects so that these changes could be intro-duced into the economic model. The economicmodel we used in the analysis does not use climatedata directly. It uses changes in crop yields, waterdemand, water supply, and other factors as they areaffected by climate. The changes are then intro-duced into the ASM model alone or in combinationto evaluate their effect on the agricultural economyand resource use. In the following section wereview the basis for these changes and discuss theadditional assumptions needed to introduce theminto the economic model. In the remainder of thischapter, we describe basic methods and findingsfrom the crop studies; describe the approaches andadditional assumptions needed to use these site-levelresults in a national level economic model; providedetails on the estimation of livestock effects; brieflydescribe the process by which pesticide use wasincluded in the economic estimates (we providegreater detail in Chapter 6); describe the basis forconsidering the effect of changes in production else-where in the world that affect US agriculturethrough international trade; and report the econom-ic and resource use results that we estimated withthe economic model.

Yield and Water-UseChanges

The crop yield and irrigation water-use impactsdeveloped here were based on crop studies con-ducted as part of the agriculture sector assessment.Coordinated site studies were conducted by teamsat the Goddard Institute of Space Studies (GISS), theUniversity of Florida, and the National ResourceEcology Laboratory (NREL) at Colorado StateUniversity; these studies provide the core set ofyield and irrigation water-use estimates in the eco-nomic analysis. The Pacific Northwest NationalLaboratory (PNNL) also produced a set of crop yieldresults; these results, however, were developed onlyfor the Hadley climate scenarios, and they did notinclude as many crops as the coordinated site-levelstudies or consider adaptation. An advantage of thePNNL work, however, is that it estimated impacts foreach of more than 200 representative regions,whereas the detailed site studies were based on 45sites, and not all crops were simulated at all sites.The PNNL analysis also used a different cropmodel—the Erosion Productivity Index Calculator(EPIC)—to estimate yield and irrigation waterdemand effects, different assumptions about ambi-ent carbon dioxide levels, and a somewhat differentmethod to develop weather scenarios. We alsoadapted results from a Southeastern US projectbeing conducted at the National Center forAtmospheric Research (NCAR) in Colorado (led byLinda Mearns) to provide estimates of impacts oncotton—an important crop for which we wereunable to conduct new yield estimates. We describevery briefly here the basic approach and summarizethe principal findings from the core site-level cropstudies. We also review very briefly other relatedcrop studies. Details on each of the studies conduct-ed under the auspices of the agriculture assessmentare included in reports available at the NationalAssessment web site (http://www.nacc.usgcrp.gov).

Using these results in an economic model that cov-ers the entire United States and many crops raisestwo methodological issues: how to treat crops forwhich crop simulations were not conducted and

37

Chapter 3: Impacts of Climate Change on Production Agriculture and the US Economy

how to extrapolate from sites to regional scaleimpacts. We discuss how we did this extrapolationand summarize the national average yield changesbelow. We then provide more detail on the cropmodel simulations and site-level results.

National Average Yield andWater-Use Changes

With regard to omitted crops, the basic issue is thatproduction and resource effects in the economicmodel depend on relative changes in yield andwater use among crops. As a result, the productionof crops omitted from the simulation studies isaffected in the economic model even if no direct cli-mate effects are assumed for them. This problemcould create regional and resource use shifts thatreflect the relative importance of omitted cropsrather than the estimated climate effects. Left unaf-fected by climate change, the omission of impactson some crops could lead to a bias in the estimateof the overall economic impact of climate change.Generally improving conditions would be underesti-mated if no yield increase were included. The con-verse also is true: If conditions were generally wors-ening, the impact would be underestimated. Theseconsiderations lead to the conclusion that, for assess-ment purposes, it is useful to make a best guess forthese omitted crops.

We assumed that, for each omitted crop, one of thecrops for which yields were simulated in the cropstudies could serve as a proxy (a common assump-tion). Crops that were grown in similar areas andsimulated for sites in those areas were used as proxycrops. The use of proxy crops has many limitations;without actually simulating the omitted crops, onecannot easily establish the error involved in usingone crop or another as a proxy. In simulating eco-nomic effects, we follow previous work in assumingthat a crude assumption is better than leaving omit-ted crops unaffected. The latter approach wouldlead to production shifts to (or away) from the crop.For cotton, we adapted the results of NCAR’sSoutheastern US study. We discuss the specificapproach for each omitted crop in the followingpages.

Crop with missing data Crop used as proxy

Hard Red Spring Wheat Spring wheatHard Red Winter Wheat Winter wheatSoft Wheat Wheata

Durham Wheat Wheata

Barley Wheata

Oats Wheata

Silage CornOranges, fresh OrangesOranges, processed OrangesGrapefruit, fresh OrangesGrapefruit, processed OrangesTomatoes, processed TomatoesTomatoes, fresh TomatoesSugar Cane Rice Sugarbeet Hay

Proxy Crops

We used a direct proxy crop approach for the cropslisted in Table 3.1. For example, silage sensitivitywas assumed to be the same as corn sensitivity.

Cotton

We were unable to secure and run a new set ofresults for cotton. Thus, we relied on existing NCARwork undertaken by Linda Mearns that included cot-ton but used a different set of climate scenarios.The NCAR study simulated yield effects by usingmany of the same crop models we used in ourassessment and for several climate scenarios, includ-ing the Hadley scenario. NCAR used a climate repre-sentative of 2060, however, and did not conduct sim-ulations based on the Canadian climate model. Acomparison of yield effects among the NCAR cropsshows that none of the other crops responds simi-larly to cotton, suggesting that no single crop wouldserve as a proxy for cotton. An attempt to use

Table 3.1. Proxy Crops

aFor each ASM region, the dominant variety of wheat (spring orwinter) grown in that region was used for the proxy for thesecrops.

38


multiple regression analysis to statistically relate cot-ton yields to the yields of all other crops verified theconclusion that no single crop—nor any combina-tion of crops—explained the site-level variation inyield impacts of cotton. The approach we adoptedinstead was to adapt the NCAR cotton yield sensitiv-ity data directly, as explained in McCarl (2000).Operationally, this analysis involved extrapolatingthe 2060 spatial distribution of cotton yield andwater-use sensitivity from the NCAR study to 2030and 2090, based directly on the climate in theseyears relative to the Hadley climate for 2060.

The intent of these assumptions was to avoid under-estimating overall economic impacts of climate onthe US agriculture economy by assuming no effectat all on these crops. Crop coverage has been anissue in all assessments of this type. Early agricultur-al assessments often were limited to corn, wheat,rice, and soybeans. More recent assessments, includ-ing this one, have worked to provide broader cropcoverage. Caution obviously is warranted in usingdetailed crop results from the economic modelwhere crop yield effects were not simulated directly.These uncertainties also introduce uncertainties in

CC CC HC HC HC-PNNL* HC-PNNL*

Crop 2030 2090 2030 2090 2030 2090

Cotton 18 96 32 82 32 82

Corn 19 23 17 34 11 16

Soybeans 20 30 34 76 7 9

Hard Red Spring Wheat 15 -4 20 30 17 24

Hard Red Winter Wheat -16 -1 21 55 24 41

Soft Wheat -5 3 8 20 58 68

Durum Wheat 15 -5 21 30 10 18

Sorghum 17 21 15 70 15 70

Rice -2 -8 3 10 3 10

Barley 56 25 83 132 70 124

Oats 23 -2 54 101 158 182

Silage 17 18 15 32 11 24

Hay -10 -1 2 15 43 57

Sugar Cane -5 -5 0 8 0 8

Sugar Beet 7 11 9 24 30 45

Potatoes 7 -25 6 -3 6 -3

Orange, Fresh 32 91 40 69 40 69

Orange, Processed 13 120 28 49 28 49

Grapefruit, Fresh 21 101 33 60 33 60

Grapefruit, Processed 15 112 29 53 29 53

Pasture 3 20 22 38 22 38

Table 3.2a National Average Change in Dryland Yields Without Adaptation (percentage change from base conditions)

* Shaded cells in the final two columns are yields that were not based on PNNL crop yield simulations either directly or indirectly using theproxy crop method; PNNL climate assumptions differed somewhat from the site study climate assumptions as described in the text.

39


1There is no obvious basis for selecting weights within an ASM region; thus, we treat these data as multiple representative draws from thesame population.


the overall economic results. In a very limited way,we explored this uncertainty by simulating the eco-nomic model, using the different approaches wedeveloped for cotton.

With regard to extrapolation from site-level data, theASM model includes 63 regions (see Figure 3.1, withoverlay of the USDA production regions). Not allcrops were simulated at each of the 45 sites. Insome cases, multiple sites were located in a singleASM region. When multiple simulation sitesappeared in a region, we used an unweighted aver-age across those sites.1 We used proxy regions forregions in which no sites were located; in thesecases, we used adjacent regions as proxies. McCarl

(2000) discusses the ASM, the use of adjacentregions as proxies, and other details of the methodswe used in greater detail. Briefly, the ASM is a spatialequilibrium model that includes domestic and for-eign demand for agricultural products and foreignsupplies for agricultural products. Multiple activities(crops, irrigated and nonirrigated, types of livestock,etc.) are represented in the model. Each US produc-tion region has bilateral trade with other domesticand foreign regions. Model solutions involve solvingfor a set of prices for all goods in all markets wherethe quantity demanded for each product is equal tothe quantity supplied. Activity choice is solvedsimultaneously in the determination of equilibriumprices, based on their profitability.


Crop 2030 2090 2030 2090 2030 2090

Cotton 18 96 32 82 32 82Corn 20 24 17 34 11 16Soybeans 39 64 49 97 7 9Hard Red Spring Wheat 20 14 23 36 17 24Hard Red Winter Wheat -9 13 23 59 24 41Soft Wheat -3 4 9 21 58 68Durum Wheat 18 12 22 33 10 18Sorghum 43 87 32 96 32 96Rice 7 4 9 18 9 18Barley 96 133 105 197 70 124Oats 33 24 57 106 158 182Silage 18 20 16 32 11 24Hay -10 -1 2 15 43 57Sugar Cane 6 7 7 16 7 16Sugar Beet 7 11 9 24 30 45Potatoes 8 -20 7 1 7 1Orange, Fresh 32 91 40 69 40 69Orange, Processed 13 120 28 49 28 49Grapefruit, Fresh 21 101 33 60 33 60Grapefruit, Processed 15 112 29 53 29 53Pasture 3 20 22 38 22 38

Table 3.2b National Average Change in Dryland Yields With Adaptation (percentage change from base conditions)

40


Figure 3.1: This figure presents Agriculture Sector Model (ASM) regions as they appear when overlaid with USDA regions. The ASM regionsfollow state boundaries and each state is typically a region except where further disaggregated. For example, the California region is anASM region. The USDA regions are delineated with the heavier black line. For example, one USDA region is the Pacific which includesWashington, Oregon, and California.

As with crop proxies, the lack of direct esti-mates for a site within a region introduces con-siderable uncertainty in estimates for thatregion. Even for regions with site estimates, asample of one or two sites may not be repre-sentative of the region. The PNNL crop yieldmodel results were based on a denser selectionof sites, although the model simulates all cropsat one site in every hydrological basin; thus, theresults for many of the sites (where productiondoes not now occur, nor would it occur in thefuture) are weighted as zero in the national


Crop 2030 2090 2030 2090 2030 2090

Cotton 36 122 56 102 56 102Corn -1 -2 0 7 21 22Soybeans 16 28 17 34 17 34Hard Red Spring Wheat -10 -18 4 6 4 6Hard Red Winter Wheat -4 -6 5 13 5 13Soft Wheat -6 -5 3 9 3 9Durum Wheat -10 -21 5 6 5 6Sorghum -1 -16 1 -2 1 -2Rice -2 -8 3 10 3 10Barley -40 -71 8 15 8 15Oats -17 -31 12 28 12 28Silage 1 0 1 9 26 30Hay 3 2 23 24 37 40Sugar Cane -5 -5 0 8 0 8Sugar Beet 22 23 39 42 41 44Potatoes -6 -28 -3 -13 -3 -13Tomato, Fresh -9 -21 1 -4 1 -4Tomato, Processed -16 -6 -6 -14 -6 -14Orange, Fresh 32 91 40 69 40 69Orange, Processed 13 120 28 49 28 49Grapefruit, Fresh 21 101 33 60 33 60Grapefruit, Processed 15 112 29 53 29 53


Table 3.3a National Average Change in Irrigated Yields Without Adaptation (percentage change from base conditions)

ASM Regions with USDA Regions Overlaid

41


average change. Nevertheless, these results indicate,in part, the uncertainty in estimated impacts thatderive from different approaches for estimating yieldimpacts. As a sensitivity analysis, we therefore usedthe available PNNL results in the economic model asa substitute for the coordinated site-level results.

These assumptions provide the basis for estimatingyield impacts for all crops in each region of theASM. The national average change in yields for dry-land and irrigated crops with and without adapta-tion are listed in Tables 3.2a,b and 3.3a,b. Table3.4a,b lists the national results for changes in wateruse on irrigated crops. We constructed the nationalaverages by weighting ASM regional estimates gener-ated from the crop model results as described aboveby harvested acreage in each ASM region; the

weights were based on data from the 1992 NationalResource Inventory (NRI). McCarl (2000) providesadditional details.

The estimates in Tables 3.2 through 3.4 are a sum-mary of input into the ASM model. Actual nationalproduction depends on changes in the agriculturaleconomy induced by these changes. The estimatesare, however, a useful intermediate result that sum-marizes the crop modeling simulations. The site sim-ulation results by themselves can provide a mislead-ing impression of overall impacts because cropswere simulated at many sites where little of thecrop is grown or sites under dryland conditionswhere the crop is mainly grown only with irriga-tion. Weighting results for the site by area providesa better guide to how climate would affect

CC CC HC HC

Crop 2030 2090 2030 2090

Cotton 36 122 56 102Corn 1 0 1 9Soybeans 23 33 23 40Hard Red Spring Wheat -1 -6 7 10Hard Red Winter Wheat -1 0 8 16Soft Wheat -5 -3 5 11Durum Wheat 2 -4 9 10Sorghum 22 8 22 21Rice 7 4 9 18Barley 3 -16 28 40Oats -6 -15 17 33Silage 3 3 2 10Hay 3 2 23 24Sugar Cane 6 7 7 16Sugar Beet 22 23 39 42Potatoes -4 -21 -1 -8Tomato, Fresh 1 6 10 13Tomato, Processed 10 44 10 17Orange, Fresh 32 91 40 69Orange, Processed 13 120 28 49Grapefruit, Fresh 21 101 33 60Grapefruit, Processed 15 112 29 53

Table 3.3b National Average Change in Irrigated Yields With Adaptation (percentage change from baseconditions)

Table 3.4a National Average Change in Water Use onIrrigated Crops, Without Adaptation (percentage changefrom base conditions)

CC CC HC HC

Crop 2030 2090 2030 2090

Cotton -11 107 36 60Corn -34 -54 -30 -60Soybeans 0 3 -12 -26Hard Red Spring Wheat -28 -22 -17 -21Hard Red Winter Wheat 5 -9 -8 -29Soft Wheat 5 -29 -12 -44Durum Wheat -28 -15 -18 -21Sorghum -7 -23 -9 -35Rice -10 37 -2 -4Barley -98 -90 -61 -85Oats -57 -73 -47 -80Silage -35 -50 -33 -63Hay 2 26 -29 -36Sugar Cane -23 3 -8 -1Sugar Beet -12 40 -28 -28Potatoes -5 7 -1 4Tomato, Fresh -9 14 -5 5Tomato, Processed -3 -6 -4 -4Orange, Fresh -21 94 -6 -6Orange, Processed 0 438 11 24Grapefruit, Fresh -1 324 8 21Grapefruit, Processed 1 401 11 24

42


production. These tables also provide input data forthe ASM, based on PNNL crop results. The PNNLmodeled only corn, alfalfa, wheat, hay, and soybeans.Crops other than these four (and those for whichone of these crops were proxies) have identicalchanges as the core results for the Hadley climatescenario. These entries are shaded in the table. Asshown Table 3.2, the PNNL results differ substantiallyin magnitude for some dryland crops for some peri-ods and thus indicate substantial uncertainty in theestimates of crop yields resulting from these differ-ent methodological approaches. The results of thetwo approaches agree in that both find generally sub-stantial positive yield effects for the dryland crops

considered by both. The PNNL results, in contrast tothe coordinated site studies, generally showincreased yields for irrigated crops (Table 3.3). Wediscuss these differences further below; in general,however, the source of these differences cannot beisolated without highly controlled comparisons ofthese models. As part of the agriculture assessment,we funded a model comparison workshop aimed atestablishing such a comparison (Paustian et al. 2000).

The results vary across crops, time periods, and cli-mate scenarios, but some broad patterns emerge.

• Even without adaptation, the weighted averageyield impact for many crops grown under dry-land conditions across the entire United States ispositive under the Canadian and Hadley climatemodels. In many cases, yields under the 2030climate conditions are improved compared withthe control yields under current climate andimprove further under the 2090 climate condi-tions. These generally positive yield results areobserved for cotton, corn for grain and silage,soybeans, sorghum, barley, sugarbeet, and citrusfruit. The yield results are mixed for other crops(wheat, rice, oats, hay, sugar cane, and potatoes);yield increases under some conditions anddeclines under other conditions.

• Changes in irrigated yields, particularly for thegrain crops, were more often negative or lesspositive than dryland yields. This result reflectsthe fact that precipitation increases were sub-stantial under these climate scenarios.Precipitation increases do not provide a yieldbenefit to irrigated crops because no waterstress occurs; all of the water that is needed isprovided through irrigation. Higher tempera-tures speed development of crops and reducedthe grain filling period, thereby reducing yields.For dryland crops, the negative effect of highertemperatures was counterbalanced by the posi-tive effect of increased moisture.

• Water demand by irrigated crops dropped sub-stantially for most crops. The faster develop-ment of crops resulting from higher

Table 3.4b National Average Change in Water Use onIrrigated Crops, With Adaptation (percentage changefrom base conditions)

CC CC HC HC

Crop 2030 2090 2030 2090

Cotton -11 107 36 60Corn -33 -55 -32 -60Soybeans 18 12 0 -20Hard Red Spring Wheat -12 -15 -12 -15Hard Red Winter Wheat 9 -3 -6 -25Soft Wheat 5 -24 -10 -45Durum Wheat -3 -5 -9 -12Sorghum 3 -19 2 -27Rice 2 48 5 8Barley -40 -57 -41 -61Oats -37 -60 -38 -68Silage -35 -52 -35 -62Hay 2 26 -29 -36Sugar Cane -19 -11 -6 7Sugar Beet -12 40 -28 -28Potatoes -3 10 0 7Tomato, Fresh -8 6 2 13Tomato, Processed 3 -14 -3 -6Orange, Fresh -21 94 -6 -6Orange, Processed 0 438 11 24Grapefruit, Fresh -1 324 8 21Grapefruit, Processed 1 401 11 24

43


temperatures reduced the growing period andthereby reduced water demand more than off-setting increased evapotranspiration because ofhigher temperatures while the crops weregrowing. To a large extent, the reduced wateruse thus reflects the reduced yields on irrigatedcrops. Increased precipitation also reduced theneed for irrigation water.

• Adaptation contributed small additional gains inyields of dryland crops, particularly those withlarge yield increases from climate change.Adaptation options were considered for siteswith losses and those with gains; for the mostpart, however, these adaptations had little addi-tional benefit where yields increased from cli-mate change. This finding suggests that adapta-tion may be able to partly offset changes incomparative advantage across the United Statesthat results under these scenarios. Other strate-gies for adaptation, such as whether to switchcrops or to irrigate or not, are part of the eco-nomic model. The decisions to undertake thesestrategies are driven by economic considera-tions—that is, whether they are profitable undermarket conditions simulated in the scenario. Wedid not consider adaptation for several cropsbecause the measures we considered (such asplanting date) were not applicable to manyperennial and tree fruit crops. We conductedadaptation studies only for a limited numberof sites.

• Adaptation contributed greater yield gains forirrigated crops. Shifts in planting dates canreduce some heat-related yield losses. Withhigher yields than in the no-adaptation case,water demand declines were not as substantial.Again, this finding reflected the fact that theadaptations we considered extended the grow-ing (and grain-filling period), and this extensionmeant a longer period over which irrigationwater was required.

The PNNL results for dryland crop yields showsimilar positive effects to those estimated with themore detailed site-level crop models, although the

magnitude of the impact varies. The differencesbetween the PNNL and site-level models were notconsistently higher or lower. These differences arelikely partly related to the site selection (where thePNNL approach has advantage because of densersampling), differences in the crop models (where thesite-level models have an advantage because theyhave been developed to better represent each crop),and differences in experiment design (i.e., assumedambient levels of ambient CO2 were different). ThePNNL did not consider adaptation. The PNNL alsoconsidered irrigation only for corn and alfalfa. ThePNNL results for these irrigated yields for the cropsthey considered also differ substantially from the site-level models. Whereas the site-level models showyield losses and reductions in irrigation water use,the PNNL results show yield gains. In the site-levelmodels, higher temperatures speed development ofthe crop and reduce yield and water demand. TheEPIC model on which the PNNL results are based donot show this negative effect of temperature; instead,temperature increases yield. These differencesshould be interpreted, therefore, as indicative of thelevel of uncertainty in the estimates contributed bycrop modeling and experimental design; we couldnot conclude that one approach or the other wasclearly superior on all counts.

Crop Model Results and Methods

The national average results presented above werebased on site-level studies for several major crops:wheat, maize, soybean, potato, citrus, tomato,sorghum, rice, and hay. The GISS-Florida-NRELresults were obtained from 45 sites across theUnited States (Table 3.5). Greater details on themethods and results are given in Tubiello et al.(2000). We used a network of major crop growingsites, based on current USDA national and state-levelstatistics. A subset of these sites had been used inprevious work (Adams et al. 1990; Rosenzweig andIglesias 1994; Adams et al. 1999). The study sites weselected do not span the United States homoge-neously; they focus on areas of major productionand importance to national output. We simulatedcrops at current sites of production for winter and

44


Site Crops simulated

1. Abilene, TX Winter Wheat, Sorghum, Pasture2. Alamosa, CO Potato, Pasture3. Bakersfield, CA Citrus, Rice, Pasture4. Boise, ID Winter Wheat, Spring Wheat, Potato, Pasture5. Buffalo, NY Potato, Tomato, Pasture6. Caribou, ME Potato, Pasture7. Columbus, OH Tomato, Winter Wheat, Corn, Pasture8. Columbia, SC Soybean, Sorghum, Tomato, Pasture9. Corpus Christi, TX Citrus, Pasture10. Daytona Beach, FL Citrus, Pasture11. Des Moines, IA Corn, Soybean, Pasture12. Dodge City, KS Winter Wheat, Pasture13. Duluth, MN Corn, Soybean, Pasture14. El Paso, TX Citrus, Rice, Sorghum, Tomato, Pasture15. Fargo, ND Spring Wheat, Potato, Corn, Pasture16. Fresno, CA Rice, Spring Wheat, Tomato, Pasture17. Glasgow, MT Spring Wheat, Pasture18. Goodland, KS Winter Wheat, Sorghum, Pasture19. Indianapolis, IN Potato, Corn, Soybean, Tomato, Pasture20. Las Vegas, NV Citrus, Pasture21. Louisville, KY Soybean, Sorghum, Pasture22. Madison, WI Potato, Corn, Soybean, Pasture23. Medford, OR Potato, Pasture24. Memphis, TN Corn, Soybean, Pasture25. Miami, FL Rice, Citrus, Pasture26. Montgomery, AL Citrus, Rice, Soybean, Sorghum, Tomato, Pasture27. Muskegon, MI Potato, Soybean, Tomato, Pasture28. North Platte, NE Winter Wheat, Corn, Soybean, Sorghum, Pasture29. Oklahoma City, OK Winter Wheat, Sorghum, Pasture30. Pendleton, OR Potato, Pasture31. Peoria, IL Corn, Soybean, Sorghum, Pasture32. Pierre, SD Spring Wheat, Sorghum, Pasture33. Port Arthur, TX Rice, Citrus, Pasture34. Raleigh, NC Soybean, Sorghum, Tomato, Pasture35. Red Bluff, CA Rice, Citrus, Pasture36. Savannah, GA Citrus, Soybean, Sorghum, Pasture37. Scott Bluff, NE Potato, Pasture38. Sioux Falls, SD Corn, Sorghum, Pasture39. Shreveport, LA Rice, Citrus, Pasture40. Spokane, WA Winter Wheat, Spring Wheat, Pasture41. St. Cloud, MN Spring Wheat, Corn, Soybean, Pasture42. Tallahassee, FL Citrus, Tomato, Pasture43. Topeka, KS Winter Wheat, Corn, Soybean, Sorghum, Pasture44. Tucson, AZ Spring Wheat, Citrus, Pasture45. Yakima, WA Potato, Pasture

Table 3.5 Crop Study Sites

45


spring wheat, maize, soybean, potato, hay, and citrus.Some of the sites may not have been ideally locatedfor the remaining crops in this analysis, such as toma-to and rice. In addition, we simulated at morenortherly sites the production of some crops current-ly limited to southern locations, to estimate thepotential for northward shifts under climate warm-ing. Ideally, we would have included a much densernetwork of sites, but our resources for more exten-sive data and time intensive calculations were limit-ed. We contrast these results for the Hadley climatescenario with the PNNL modeling results conductedfor each of the 204 eight-digit Hydrological UnitAreas defined by the US Geological Survey, usingEPIC-based crop models for corn, soybean, winterwheat, and alfalfa.

At each of the 45 sites we examined, we collectedobserved time series of daily temperatures (minimaand maxima), precipitation, and solar radiation forthe period 1951–1990, representing the “baseline” cli-mate for this study. We simulated the crop modelsover this 40-year period to compute an average yieldand water use for the baseline climate. We producedscenarios of climate change according to transientsimulations performed with two general circulationmodels (GCMs): the Canadian Climate Centre (CCC)model and the Hadley Centre (HAD) model. We con-sidered two time periods in this analysis: 2030 and2090—representing changes in climate projected byeach GCM—and calculated by using 20-year averagescentered around the years 2030 and 2090, respec-tively. We used absolute average temperature devia-tions and percentage changes in precipitation toadjust the 40-year historical record to produce analtered climate that was representative of theseyears. We then used the crop models to simulateyields for each of these 40-year altered climates andcompared the average yield and water use with thesimulated baseline yield and water use to computethe impact of climate change. Atmospheric CO2 con-centrations used for the crop model simulations—350 ppm for the baseline; 445 ppm for 2030, and 660ppm for 2090—were based on the IPCC IS92a sce-nario of future emissions. The IS92a emissions sce-nario is approximately consistent with a 1 percentper year increase in total emissions of greenhouse

gases (GHGs). Thus, the CO2 concentrations we usedin the crop models are less than if all of the 1 per-cent increase were CO2—reflecting the fact thatother GHGs will contribute to warming. If all of theincrease were caused by CO2 emissions, CO2 concen-trations would be higher and the crop models wouldshow a larger CO2 fertilization effect.

We downscaled GCM output to each of the studysites by linear interpolation, using the four grid-points closest to each location. We then appliedmean monthly changes in temperature and precipita-tion to the observed baseline meteorological seriesto produce representative weather for the future sce-narios. We used a total of five scenarios in this study:

• the baseline, representing current conditions

• the Hadley climate model for 2030 (HAD 2030)

• the Canadian climate model for 2030 (CCC 2030)

• the Hadley model for 2090 (HAD 2090)

• the Canadian model for 2090 (CCC 2090).

HAD 2030 and CCC 2030 are representative of theclimate and CO2 concentration for the 2020–2039period; HAD 2090 and CCC 2090 are representativeof the climate and CO2 concentrations for the2080–2099 period.

We used a suite of crop models to simulate thegrowth and yield of study crops under the currentand climate change scenarios. The Decision SupportSystems for Agrotechnology Transfer (DSSAT) familyof models was used extensively in this study, to sim-ulate wheat, corn, potato, soybean, sorghum, rice, cit-rus, and tomato (Tsuji et al., 1994). The CENTURYmodel was used to simulate grassland and hay pro-duction (Parton et al., 1994).

All of the models we employed have been usedextensively to assess crop yields across the UnitedStates under current conditions as well as under cli-mate change (Rosenzweig et al. 1995; Parton et al.1994; Tubiello et al. 1999). Apart from CENTURY,which was run in monthly time-steps, all othermodels use daily inputs of solar radiation, minimum

46


and maximum temperature, and precipitation tocalculate plant phenological development fromplanting to harvest; photosynthesis and growth; andcarbon allocation to grain or fruit. All models use asoil component to calculate water and nitrogenmovement, so they were able to assess the effects ofdifferent management practices on crop growth.

The simulations performed for this study consideredrainfed production and optimal irrigation—definedas refilling of the soil water profile whenever waterlevels fall below 50 percent of capacity at 30 cmdepth. Fertilizer applications were assumed to beoptimal at all sites.

The climate change scenarios we used in this studyare more realistic than those previously available.Because they include the effects of sulfate aerosolson future climate change, they result in projectedchanges in temperature and precipitation that aresmaller than those in previous “equilibrium” andtransient climate change simulations, particularly inthe first half of the 21st century. In fact, the temper-ature increases in 2090 become substantial at allsites we considered, as the “masking” effect ofaerosols on climate warming becomes small com-pared to the magnitude of greenhouse forcing.

Additional analyses, independent from the foregoingsite studies, have been developed by other groups inthe United States as part of the assessment effort orin ongoing research with the same or similar climatemodels. There are some important differences inthe assumptions in these analyses, however, thatmake them not directly comparable to the core stud-ies reported above.

Researchers at PNNL developed national-level analy-ses for corn, winter wheat, alfalfa, and soybean, usingclimate projections from the Hadley GCM (Izaurraldeet al. 1999). In the PNNL study, the baseline climatedata were obtained from national records for theperiod 1961–1990. The scenario runs were con-structed for two future periods (2025–2034 and2090–2099). EPIC was used to simulate the behaviorof 204 “representative farms” (i.e., soil-climate-man-agement combinations) under the baseline climate,

the two future periods, and their combinations withtwo levels of atmospheric CO2 concentrations(365 ppm and 560 ppm). This approach differedfrom the core studies that used 2030 and 2090 CO2

levels. The CO2 effect was independent of climateeffects in the PNNL work, however, allowing interpo-lation. The results of the PNNL study with CO2

effects interpolated were used in the economicmodel to compare the approach with that used inthe coordinated site studies.

Another group, coordinated at Indiana University,focused only on corn; this group developed a regionalanalysis for the Corn Belt region, using Hadley modelprojections (Southerland et al. 1999). Baseline climatewas defined by using the period 1961–1990. Severalfuture scenarios were analyzed for the decade of2050, with atmospheric CO2 concentration set at 555ppm. Corn yields were simulated with the DSSATmodel at 10 representative farms. Adaptations studiedincluded changes in planting dates, as well as the useof cultivars with different maturity groups. Althoughthis work was not conducted with funding from theagriculture assessment, it offers some additional site-level information for corn.

Although specific differences in time horizons, CO2

concentrations, and simulation methodologies com-plicate comparison of these additional analyses withthe work discussed herein, model findings overallwere in general agreement with ours. We discussthem briefly, crop-by-crop, in the “results” section ofthis work.

Simulations Under Current Climate

A test of the basic validity of the models involvessimulating yields under current climate and com-pare them, coarsely scaled, to the state level byusing statistical information on percent irrigation.These comparisons generally showed good agree-ment with reported yields variations across theUnited States.

In addition to current practices at each site, we alsosimulated different adaptation techniques for useunder climate change. These simulations consisted

47


largely in testing the effects of early planting—a real-istic scenario at many northern sites under climatechange—and testing the performance of cultivarsthat are better adapted to warmer climates, usingcurrently available genetic stock. In general, earlyplanting was considered for spring crops, to avoidheat and drought stress in the late summer months,while taking advantage of warmer early tempera-tures. New, better-adapted cultivars were tested forwinter crops (e.g., wheat) to increase the time tomaturity (shortened under climate change scenar-ios) and to increase yield potential.

Winter Wheat

We simulated winter wheat at Abilene,Texas; Boise,Idaho; Columbus, Ohio; Dodge City, Kansas; Topeka,Kansas; Goodland, Kansas; North Platte, Nebraska;Oklahoma City, Oklahoma; and Spokane,Washington. The differences in yields between irri-gated and rainfed production for the crop simula-tions were similar to differences between actualcounty-level averages of yield for irrigated and rain-fed production. Record irrigated yields were simu-lated at Boise; all remaining sites produced 4.5–5.5 t/ha. Coefficients of variation (CVs) for irrigatedproduction were 10–15 percent. The largest differ-ences between irrigation and rainfed practice wereat Boise (more 400 percent) and Spokane (150 per-cent). The smallest gains with irrigation were at thewet sites: Columbus and Topeka. Simulated yields atthese sites had been compared previously with cur-rent conditions and were well correlated with pro-duction data at the state level.

Spring Wheat

Spring and durum wheat are grown extensively inNorth and South Dakota and Montana; there also aresome important production centers in theNorthwest, California, and Arizona. We chose a totalof eight sites of importance to US spring wheat pro-duction: Boise, Idaho; Fargo, North Dakota; Fresno,California; Glasgow, Montana; Pierre, South Dakota;St. Cloud, Minnesota; Spokane,Washington; andTucson,Arizona. Simulated irrigated yields were

50–60 percent higher than rainfed yields, with loweryear-to-year variability (CV).The simulated marginal(i.e., additional) returns on irrigation were large atBoise, Spokane, and Tucson, where irrigated yieldswere 100 percent, 300 percent, and 1,000 percenthigher, respectively, than under rainfed conditions.The highest irrigated yields, 7–8 t/ha, were simulatedat Tucson and Fresno; all remaining sites produced3–5 t/ha. Coefficients of variation were 10–15 per-cent for irrigated production and 40–50 percent forrainfed production. Simulated yields at these sitespreviously had been compared with current condi-tions and were well correlated with production dataat the state level.

Maize

Simulated maize yields agreed well with reportedstate-level averages; the highest dryland yields—above 8 t/ha—were simulated at Columbus, Ohio;Madison,Wisconsin; and Indianapolis, Indiana.Production at the remaining sites was in the 5–7t/ha range, with low yields and high CVs simulatedat St. Cloud, Minnesota—currently at the northernmargin of the main US corn production area.

Potato

We chose a total of 12 sites of importance to nation-al potato production:Alamosa, Colorado; Boise,Idaho; Buffalo, New York; Caribou, Maine; Fargo,North Dakota; Indianapolis, Indiana; Madison,Wisconsin; Medford, Oregon; Muskegon, Michigan;Pendleton, Oregon; Scott Bluff, Nebraska; andYakima,Washington. We simulated viable continu-ous rainfed potato production at Buffalo, Caribou,Fargo, Indianapolis, and Madison. Under current cli-mate, crop simulations correlated well with reportedproduction. The highest simulated irrigated yields—slightly above 80 t/ha—were at the Northwesternsites (Medford, Pendleton, and Yakima), where themarginal impact of irrigation was also the greatest(irrigated yields were about 10 times rainfed yields).At all remaining sites, production was between 40and 50 t/ha. Coefficients of variation for irrigatedproduction were 6–9 percent. CVs were between 30and 40 percent under rainfed conditions.

48


Citrus

We conducted simulations for Valencia oranges ateight sites with substantial current production. Ofthese, four sites—Bakersfield, California; CorpusChristi,Texas; Daytona Beach, Florida; and Miami,Florida—corresponded to high-producing areas inthe United States, yielding more than 11 t/ha offruit. One site (Red Bluff, California) representedmid-level production—around 7 t/ha; three sites—Tucson,Arizona; Port Arthur,Texas; and Las Vegas,Nevada—produced 4–6 t/ha, representing marginalproduction levels. We chose an additional five sitesto investigate the potential for citrus expansionnorthward of the current production area: El Paso,Texas; Montgomery,Alabama; Savannah, Georgia;Shreveport, Louisiana; and Tallahassee, Florida.Under current climate, simulations at these sitesyielded 2–2.5 t/ha. Simulated yields at these sitespreviously had been compared with current condi-tions and were well correlated with production dataat the state level.

Soybean

We simulated soybean production across the UnitedStates at 15 sites: Charleston, South Carolina;Louisville, Kentucky; Raleigh, North Carolina; DesMoines, Iowa; Duluth, Minnesota; Indianapolis,Indiana; Madison,Wisconsin; Memphis,Tennessee;Montgomery, Alabama; Muskegon, Michigan; NorthPlatte, Nebraska; Peoria, Illinois; Savannah, Georgia;Saint Cloud, Minnesota; and Topeka, Kansas.Simulated yields at these sites previously had beencompared with current conditions and were wellcorrelated with production data at the state level.

Sorghum

We simulated sorghum production across the UnitedStates at 14 sites: Charleston, South Carolina;Louisville, Kentucky; Raleigh, North Carolina;Abilene,Texas; El Paso,Texas; Goodland, Kansas;Montgomery, Alabama; North Platte, Nebraska;Oklahoma City, Oklahoma; Peoria, Illinois; Pierre,South Dakota; Savannah, Georgia; Sioux Falls, South

Dakota; and Topeka, Kansas. Simulated yields atthese sites compared well to state-level variationsacross the US sorghum production area.

Rice

We selected eight sites, accounting for 48 percent ofUS rice production, to represent the US rice growingregions: Louisville, Kentucky; Bakersfield, California;Des Moines, Iowa; El Paso,Texas; Fresno, California;Miami, Florida; Montgomery, Alabama; Port Arthur,Texas; Peoria, Illinois; Red Bluff, California;Shreveport, Louisiana; and Topeka, Kansas. Wechose these sites to include regions with currentproduction and those where rice production couldbe viable under climate change. The highest simulat-ed yield under current conditions was 9 t/ha (inCalifornia), and the lowest was 5 t/ha (in Louisiana)—in agreement with observed state-to-state yielddifferences.

Tomato

We simulated tomato production across the UnitedStates at 18 sites: Charleston, South Carolina;Louisville, Kentucky; Raleigh, North Carolina; Boise,Idaho; Buffalo, New York; Duluth, Minnesota; ElPaso,Texas; Fresno, California; Indianapolis, Indiana;Montgomery, Alabama; Muskegon, Michigan; NorthPlatte, Nebraska; Oklahoma City, Oklahoma; Peoria,Illinois; Tallahassee, Florida; Topeka, Kansas; Tucson,Arizona; and Yakima,Washington. Simulated yieldsat these sites, compared with current conditions,correlated well with state-level data.

Simulation Results UnderClimate Change

In this subsection, we provide a brief summary ofthe main climate change results, including thoseincorporating adaptation. Many of the yield resultswere positive, particularly for dryland crops atnorthern and western sites and particularly underthe Hadley climate scenario. Results for theCanadian climate scenario differed substantiallyfrom this general result, particularly for 2030 and for

49


crops grown in the Great Plains and southern states.These differences are directly related to differencesin the climate scenario. The Canadian scenarioshowed declines in precipitation for 2030 comparedwith present climate; it also showed greater temper-ature increases than the Hadley scenario. The tem-perature increases and precipitation reductionswere particularly strong in the southern and GreatPlains states. In contrast, the Hadley scenario exhib-ited smaller temperature increases and much greaterprecipitation increases. Furthermore, the Canadianscenario for 2090 (in contrast to the result for 2030)exhibited a large increase in precipitation comparedwith present climate. In the discussions of detailedresults by crop that follow, we note some differ-ences from these generalizations (e.g., for heat-lov-ing crops such as citrus and tomatoes that did wellin the south, for cool-loving crops such as potatoesthat did not do particularly well even at northernsites, or for irrigated crops where additional precipi-tation had no yield benefit because water wasalready available as needed by the crop). The adap-tations we considered in the crop yield simulationswere changes in variety and changes in plantingdates. Other changes that involve economic deci-sions—such as whether to irrigate or not, theamount of water to apply, whether to shift to differ-ent crops or reduce acreage planted, and theamount of labor and other inputs to use—are includ-ed in the economic model.

Winter Wheat

The two climate scenarios we considered in thisstudy gave opposite responses for US wheat produc-tion. The Canadian climate scenario resulted in largenegative to small positive impacts, whereas theHadley scenario generated positive outcomes. Thewarmer temperatures projected under climatechange were favorable to northern site productionbut deleterious to southern sites. Increased precipi-tation in the Northwest and decreased precipitationin the central plains were the major factors control-ling the response of wheat yields to the future sce-narios we considered in this study. We first analyzeresults for production based on current varieties and

planting dates (current management) and then dis-cuss the potential yield effects of changes in vari-eties and planting dates (changes in managementor adaptation).

In agreement with the results presented here, thePNNL study found that “winter wheat exhibitedconsistent trends of yield increase under the[Hadley] scenarios of climate change across the US”(Izaurralde et al. 1999). The PNNL study did notconsider the Canadian climate scenario.

Under rainfed conditions, Columbus, Ohio, was theonly site where all climate scenarios resulted inyield increases: 3–8 percent in 2030 and 16–24 per-cent in 2090. At all other sites, including the majorproduction centers in the Great Plains, the Canadianscenarios resulted in large negative impacts for con-tinuous and fallow production. Grain yieldsdecreased 10–50 percent in 2030 and 4–30 percentin 2090. Most important, at Dodge City, Kansas;Goodland, Kansas; and North Platte, Nebraska, coeffi-cients of variation of yield consistently increased inboth decades, indicating greater variability in yieldfrom year to year and greater risks to producers.Under the Hadley climate scenario, yields increasedat all sites. Rainfed production increased by 6–20percent in 2030 and by 13–48 percent by 2090.Year-to-year variation decreased at most sites.

Irrigated wheat yields increased under both GCMscenarios, although increases were larger under theHadley scenario than under the Canadian scenario.In 2030, yield increases were 2–10 percent. In2090, yields were 6–25 percent greater than undercurrent conditions.At the same time, irrigationwater use decreased by 10–40 percent.

Crop simulations showed no benefit from changingfrom current crop and water management of prac-tices for wheat production under the Hadley sce-nario. Under the Canadian scenario, simulations ofrainfed cultivation were subject to a high frequencyof years with very low yields, suggesting that rainfedproduction may no longer be viable in Kansas ifthese climate conditions are observed in the future.

50


All else being equal, maintenance of current produc-tion would require irrigation.

Adaptation strategies simulated for wheat in the cen-tral plains involved shifting to cultivars that are bet-ter adapted to a warmer climate. Specifically, culti-vars that require less vernalization and have longergrain filling periods could be planted to counterbal-ance the hastening of maturity dates resulting fromwarmer spring and summer temperatures. Forexample, cultivars currently grown in the southcould be planted at northern locations. Projectedyield decreases at North Platte were eliminated byshifting to a southern-grown variety. The same strat-egy did not yield positive results for the Kansas andOklahoma sites we considered in this study becauseof large decreases in precipitation projected by theCanadian model at these sites.

Spring Wheat

Warmer temperatures were the major factor affect-ing spring wheat yields across sites, time horizon,and management practice. Considered alone, theyhastened crop development and affected cropyields negatively.

Despite warmer temperature in 2030, rainfed springwheat production increased by 10–20 percentunder both GCM scenarios because of increasedprecipitation that also reduced CVs and thus year-toyear production risks. This positive trend continuedin 2090 under the Hadley scenario, generating yieldincreases of 6–47 percent. The largest increases (47percent) were simulated at Pierre, South Dakota.The2090 Canadian scenario resulted in significantdecreases in spring wheat yields at current produc-tion sites. Yields decreased at Fargo, North Dakota(16 percent), and Glasgow, Montana (24 percent).The Canadian scenario also generated yield decreas-es at Fresno, California (20 percent). By 2090, theCanadian scenario projected high temperatures at allsites we considered, affecting wheat developmentand grain filling negatively and depressing yieldsdespite the gains from precipitation increases.

Irrigated spring wheat production decreased by1–24 percent at the eight sites we considered underboth scenarios. In 2030, yields decreased at Boise,Idaho (7–17 percent); Spokane,Washington (1–4percent); Tucson,Arizona (3–6 percent); and Fresno(16–24 percent). The same negative trends contin-ued at these sites in 2090, with the largest reductionsimulated at Fresno (30–45 percent).

Under every scenario and at all sites, irrigation wateruse decreased significantly. Daily water consump-tion did not change substantially; instead, the grow-ing period was shortened as higher temperaturesaccelerated growth and there were fewer days whenirrigation was required. Thus, the overall changes inwater use were mainly related to accelerated growthrather than stomatal closure during the growth peri-od. By 2090, simulated yield reductions at all siteswere in the range of 20–40 percent and consistentlyabove 50–60 percent at Fresno.

Simulated rainfed production became increasinglycompetitive with irrigation under all scenarios, as aresult of increased precipitation. For example, atSpokane and Boise—which now are irrigated sites—current production levels could be maintainedunder the scenarios we considered by shifting someirrigated land to rainfed production. By 2090, therewould be no need for irrigated production at Boiseunder the Canadian scenario. At Fargo, NorthDakota, and Glasgow, Montana, additional simula-tions indicated that yields could be maintained atcurrent levels by planting two to three weeks earli-er, compared to current practices.

Corn

Climate change affected dryland corn yields posi-tively. Projected increases in precipitation morethan counterbalanced the otherwise negative effectsof warmer temperatures across the US sites we ana-lyzed. We simulated increases at current major pro-duction sites: Des Moines, Iowa (15–25 percent);Peoria, Illinois (15–38 percent); and Sioux Falls,South Dakota (8–35 percent). We simulated largerincreases at northern sites: Fargo, North Dakota

51


(25–50 percent); Duluth, Minnesota (30–50 per-cent), and St. Cloud, Minnesota, where warmer tem-peratures and increased precipitation contributed toincreased corn yields compared to current levels.We simulated smaller changes—in the range –5 per-cent to +5 percent—at the remaining sites.

The PNNL results were in agreement with the find-ings of the site-level studies for rainfed corn produc-tion, for which “increases were predicted for futureproduction of dryland corn in the Lakes, Corn Beltand Northeast regions of the US” (Izaurralde et al.1999). On the other hand, the PNNL results foundincreases in irrigated corn yields in almost allregions of the country. In contrast, the site-levelresults found that climate change affected irrigatedyields negatively—in the range of –4 percent to –20percent—at the two major irrigated production siteswe considered (in Kansas and Nebraska). As withthe wheat results, higher temperatures resulted in ashorter growing and grain-filling period. At northernsites, simulated irrigated yields—which currently arelimited by cold temperature—increased substantial-ly. For instance, at St. Cloud, Minnesota, simulatedyields under the 2090 Canadian scenario werealmost three times as much as current levels.

Additional simulations suggested that early plantingwould help maintain or slightly increase currentproduction levels at sites experiencing smallnegative yield decreases. In general, dryland cornproduction could become even more competitivethan irrigated corn production, with higher yieldsand decreased year-to-year variability. We simulatedgreat potential for increased production andimproved water management at the northernmostsites, in North Dakota and Minnesota. A study forthe Corn Belt region, conducted at IndianaUniversity (Southerland et al. 1999), was in generalagreement with our findings, projecting increases incorn yields across the northern Corn Belt region.For five southwestern locations in Indiana andIllinois, the Indiana University work projected cornyield decreases in the range of 10–20 percent. Thecoordinated site studies we conducted did not showyield losses in the southern Corn Belt sites, but we

did not have as many sites in this southern portionof the Corn Belt. The PNNL analysis, which providesa much denser sampling, showed yield declines forcorn consistent with the Indiana results for thisarea. There also were differences in the analysis pro-tocol used by the Indiana group that probably led todifferences in results. For reliability at sub-state lev-els, a far denser sampling is needed than the 45 siteswe chose to cover the entire nation.

Potato

Irrigated potato yields generally fell; under rainfedconditions, yield changes generally were positive.

Under rainfed conditions, both climate scenariosconsidered in this study resulted in sizable gains in2030. At four of the five sites we considered, cropproduction increased by an average of 20 percent;the exception was at Indianapolis, where theCanadian scenario projected a 33 percent reductionand the Hadley scenario resulted in a 7 percentincrease. CVs for all sites generally decreased as aresult of increased precipitation. In 2090, theCanadian scenario resulted in large decreases atmost sites; under the Hadley scenario, potato yieldsincreased by 10–20 percent, largely maintaining thegains reached by 2030. Under the Canadian sce-nario, rainfed production decreased by an average ofmore than 20 percent. Smaller effects were simulat-ed at Madison, and the largest effect was at Fargo(63 percent). Under this scenario, large increases intemperature in 2090 counterbalanced the beneficialeffects of increased precipitation.

Irrigated yields decreased in 2030, by 1–10 percent;a few sites registered no change or small percentageincreases. The projected temperature increasesaffected crop production negatively. Under theCanadian scenario, most sites showed simulatedyield reductions of 6–13 percent. Exceptions wereIndianapolis (a 36 percent decrease) and Yakima (a 5percent increase). Under the Hadley scenario, yieldsdecreased by 6–8 percent, although small increases(2 percent) were simulated in Fargo and Yakima.Both GCM scenarios projected 5 percent increasesin yield at Caribou, Maine.

52


The simulated decreases continued in 2090 underboth climate scenarios. Potato yields decreased by10 percent at two of the three major productionsites in the Northwest; water use increased by anaverage of 10 percent. Both GCMs resulted in largerdecreases at Boise, Idaho (30–40 percent), and ScottBluff, Nebraska (27–50 percent), and smaller ones atPendleton, Oregon; Medford, Oregon (10–15 per-cent); and Buffalo, New York (8–18 percent).

As with other crops, simulations suggested that rain-fed production could become more competitivewith irrigated production, compared to today.Cultivar adaptation would do little to counterbal-ance the negative temperature effects in our simula-tions. Current US potato production is limited tocultivars that need a period of cold weather fortuber initiation. The only viable strategy would be achange in planting dates to allow for increased stor-age of carbohydrates and sufficient time for leaf areadevelopment prior to tuber initiation. Additionalsimulations suggested, however, that current produc-tion levels could not be reestablished even with ashift in planting date. For example, moving plantingahead by as much as one month at Boise andIndianapolis helped reduce yield losses under cli-mate change by 50 percent relative to simulationswithout adaptation. This offset is substantial, but itstill leaves sizable losses compared to current yields.

Citrus

Fruit production benefited greatly from climatechange. Simulated yields increased 20–50 percent,while irrigation water use decreased. Crop loss fromfreezing was 65 percent lower, on average, in 2030and 80 percent lower in 2090 (at all sites). Miamiexperienced small increases—in the range of 6–15percent. Of the three remaining major productionsites, we simulated increases in the range of 20–30percent in 2030 and 50–70 percent in 2090. Irrigationwater use decreased significantly at Red Bluff,California; Corpus Christi,Texas; and Daytona Beach,Florida. All sites experienced a decrease in CV, as aresult of the reduction of crop loss from freezing.

Fruit yields increased in Tucson and Las Vegas. Slightto no changes in simulated water use implies, how-ever, that these sites—which currently are at themargin of orange production—will be even lesscompetitive in 2030 and 2090 than they are today.In fact, all of the additional sites we chose to investi-gate the potential for northward expansion of UScitrus production continued to have lower fruityield and higher risk of crop loss from freezing thanthe southern sites of production.

Hay and Pasture

Simulated dryland pasture and hay productionincreased under all scenarios and at most sites—except under the 2030 Canadian scenario, whichresulted in decreases of up to 40 percent in theSoutheast, Delta, and Appalachian regions. Thelargest increases—in the range 40–80 percent—were simulated for the Pacific Northwest andMountain regions. By 2090, both climate scenariosresulted in increases of greater than 20 percent at allsites. Results from the PNNL study were in generalagreement with these findings.

Soybean

Under rainfed conditions and the two climate scenar-ios we considered, soybean yields increased at mostof the sites we analyzed; increased temperaturesfavored growth and yield compared to currentconditions. Notable exceptions were the southeast-ern sites. Under the Canadian scenario, yields in thisarea were reduced in 2030 by 1–36 percent. By2090, losses of more than 70 percent were simulatedat Montgomery,Alabama, and Memphis,Tennessee.Adaptation in this area—by shifting the crop maturi-ty group—reduced losses by more than 50 percent.

At sites in the major producing areas of the CornBelt, rainfed yields increased by 10–30 percent. Atthe three northernmost sites in this study (Duluth,Minnesota; St. Cloud, Minnesota; and Muskegon,Michigan—which currently are at the northern mar-gin of US soybean production), yields increased by

53


more than 30 percent in 2030 and more than 50percent in 2090 as a result the positive effects ofwarmer temperatures.

The PNNL study, using the Hadley climate scenario,also found increases in soybean yields in the Lakestates of Michigan, Minnesota, and Wisconsin, as wellas the Northeast. It also found, however, that “soy-bean yields decreased in the Northern and SouthernPlains, the Corn Belt, Delta, Appalachian, andSoutheast regions” (Izaurralde et al. 1999). Thus,there is considerable disagreement between the twoapproaches for soybeans, particularly for the impor-tant Corn Belt region.

Irrigated soybean yields increased at all sites andunder all scenarios—by 10–20 percent in 2030 andby 10–40 percent in 2090. Again, increasing tempera-ture was the main factor that enhanced soybeanyields in this simulation analysis. In the rainfed case,at the northern sites yields increased by more than 50percent in 2030 and more than 100 percent in 2090.

Sorghum

Under rainfed conditions, the two climate scenarioswe analyzed in this study produced opposite results atmany sites, as a result of differences in projectedchanges in precipitation. Under the Hadley scenario,rainfed production increased at all sites (because ofincreased precipitation with respect to the current cli-mate) by 1–10 percent in 2030 and by 10–60 percentin 2090. Under the Canadian scenario, reductions of10–30 percent were simulated at southern and south-eastern sites. The largest decreases were simulated in2090 at Savannah, Georgia (15 percent); Charleston,South Carolina (20 percent); and Oklahoma City,Oklahoma (30 percent). Under both GCM scenarios,warmer temperatures and, where projected, increasedprecipitation enhanced production at the northern-most sites. Large increases in sorghum yields weresimulated at North Platte, Nebraska (30 percent and80 percent); Pierre, South Dakota (45 percent and 100percent); and Sioux Falls, South Dakota (50 percentand 60 percent) in 2030 and 2090, respectively.

Under irrigated production, the generally negativeeffects of increased temperature on sorghum devel-opment and growth resulted in yield reductions of10–20 percent at most sites, under both scenariosand time horizons. The largest decreases were simu-lated in 2090 at Oklahoma City (38 percent). Yieldsincreased by 10–15 percent at two of the northern-most sites; they decreased at Pierre (3 percent).

Early planting by two to four weeks helped to coun-terbalance the negative effect of warmer tempera-tures at most sites we analyzed.

Rice

Under irrigated production, the two climate scenar-ios we analyzed produced much different projec-tions of future rice yields, largely because of differ-ences in the size of the projected changes intemperature. In 2030, the Hadley scenario resultedin small positive yield increases—in the range of1–10 percent—with larger increases at two north-ern sites, that currently are well outside the US riceproduction region (Peoria, Illinois, and Des Moines,Iowa) but that we considered because of the poten-tial that climate change will make rice productionviable. The Canadian scenario resulted in smallreductions—on the order of –1 percent to –5 per-cent—at major production sites in California and atsites in the Delta region. In 2090, the patterns ofsimulated changes among scenarios, as well astheir geographic distribution, was similar to thatprojected for 2030. Yields increased under Hadleyscenario, except in Bakersfield, California(–12 percent). The Canadian scenario resulted inlarger yield decreases in 2090 than in 2030: up to–20 percent in California and the Delta region andby –50 percent in El Paso,Texas.

We simulated adaptation by planting cultivars thatare better adapted to warmer temperatures, as wellas by early planting. These techniques helped toreduce—but not to counterbalance completely—theyield reductions we simulated under climate changeand no adaptation.

54


Tomato

Under irrigated production, the climate change sce-narios generated yield decreases or small increases,depending on the scenario chosen, at most southernsites. At the northernmost locations analyzed in thisstudy, increased temperatures were highly beneficialin terms of yield.

In 2030 under the Canadian scenario, tomato yieldsdecreased at most southern sites, in the range of10–20 percent. Larger decreases were simulated atOklahoma City, Oklahoma (45 percent), and atTucson,Arizona (37 percent). At northern sites, sim-ulated yields increased: at Boise, Idaho (20 percent);Duluth, Minnesota (80 percent); Muskegon, Michigan(40 percent); and Yakima,Washington (30 percent).This trend continued in 2090 under the CanadianCentre scenario, with larger magnitudes of both pro-jected gains and losses. In 2090, general decreasesat most sites were in the range of 20–40 percent.Decreases of more than 70 percent were simulatedin Oklahoma and Texas. Northern sites continued tobenefit under warmer temperatures; yields increasedby as much as 170 percent at Duluth.

We simulated the same patterns under the Hadleyscenario except that—because of smaller projectedincreases in temperatures—the simulated losses atmost sites and the gains at northern locations weresmaller than those projected under the Canadianscenario. Specifically, under the Hadley scenario,sites in the Delta region and in the Southeastexperienced moderate gains (in the range of 5–15percent) with respect to current production levels.

PNNL Results

The PNNL results were based on slightly differentassumptions and were produced only for the Hadleyscenario, for climates representative of 2030 and2095 (H1 and H2). In addition, the PNNL agricultur-al model includes a simulator that produces randomweather scenarios exhibiting changes in mean con-ditions like those derived from the climate scenar-ios. The detailed site studies used actual historicalweather adjusted by the changes derived from the

climate scenarios. Thus, the specific weather condi-tions simulated in the PNNL study differed from thatused in the site studies. The climate change scenar-ios are applied with two levels of atmospheric CO2

concentration: 365 ppm (current ambient) and 560ppm to represent a CO2 fertilization effect. Theresults are shown in Figures 3.2 and 3.3 (see colorplate section) and are summarized in Tables 3.5 and3.6. The land areas indicated in the figures are the 4-digit (USGS nomenclature) hydrologic basins. Datain the table are aggregated from these regions intoproduction regions as defined by USDA.

Temperatures rise modestly (1–2 ºC) by 2030, andprecipitation increases by 25–125 mm y--1 over mostof the corn-growing region. By 2095, temperatureincreases by 2.0–3.5 ºC, and precipitation increasesby more than 175 mm y-1 over the entire region.Yield in the EPIC model used for this analysis isdirectly proportional to biomass production, whichis favored by a reduction in cold stress and a length-ening of the growing season in the Lake region, theCorn Belt, and the Northeast (Figure 3.1). Table 3.6shows that yields increase at current CO2 concentra-tions and improve still more at higher concentra-tions. Yields are slightly depressed in the Delta,Appalachian, and Southeastern regions, where high-er temperatures shorten the growing season (Figure3.2). With no CO2 fertilization, regional yields arereduced by 2030 and 2095 in the Delta andSoutheast but only in 2030 in the Appalachianregion. Climate-related losses are more than offsetby CO2 fertilization in all cases.

Figure 3.3 shows baseline winter wheat yields anddeviations resulting from the climate and CO2 sce-narios are shown for the Northern and SouthernGreat Plains, Mountain (Great Plains portions ofMontana,Wyoming, and Colorado), and the Westernregions; Table 3.7 summarizes these data by USDAproduction region. Temperatures in these regionsincrease by 1–2 ºC in 2030 but are considerablyhigher by 2095. By 2030, precipitation increases by25–50 mm y-1 over much of the Plains and Mountaingrowing regions, as well as in Washington and Idaho,but is lower in California. By 2095, precipitationincreases still more in the Plains and Mountain

55


regions; it increases in California and is variable inthe northwestern states. CO2 fertilization aloneincreases yields in all regions. The C-3 crops2,including wheat, experience increased photosyn-thetic rates and decreased transpiration rates underelevated CO2. The reduction in transpiration is par-ticularly important for wheat, which generally isgrown in semi-arid regions. Aggregate regional pro-duction increases under all scenarios in the Pacificregion and under most scenarios in the Mountainand Plains regions. Decreases in aggregate produc-tion were projected for the Mountain and Plainsregions when wheat growth was simulated withouta CO2-fertilization effect in 2030 and for theSouthern Plains by 2095 (also without the CO2-fertil-ization effect). Higher temperatures reduce the

frequency of cold stress and increase the length ofthe growing season by shortening the winter dor-mancy period. In more northerly regions, the cropmatures before the extreme heat of summer.

Irrigation Water Supply

Water supply for irrigation is also an important con-sideration. The ASM includes estimates of availableagricultural water supply that is allocated to crops.Estimates of changes in water supply under the cli-mate change scenarios for each ASM region weredeveloped based on total water supply changes byriver basin. The changes in water supply were fromthe Water Sector Assessment (Gleick 2000). The crit-ical assumption made was that the change in water

2C-3 refers to the photosynthetic pathway of carbon in the plant. Another significant group of plants are C-4 plants. The C-3 pathwayproduces a higher photosynthetic response to elevated ambient carbon dioxide than the C-4 pathway. Most crops are C-3. The principalexceptions of importance in the US are corn and sorghum which are C-4.

CO2 / RegionScenario Lakes Corn Belt Delta Northeast Appalachian Southeast

Mg ha-1

B-365 4.57 6.05 6.26 4.16 6.13 5.76B-560 4.95 6.53 6.55 4.54 6.73 6.35H1-365 5.30 6.31 5.84 4.70 5.94 5.34H1-560 5.94 6.98 6.74 5.24 6.70 6.13H2-365 6.04 6.53 5.84 4.81 6.27 5.04H2-560 6.69 7.09 6.32 5.35 6.95 5.76

Table 3.6. Simulated Yields of Dryland Corn Under Baseline Climate (B) and Hadley Centre Projections in 2030 (H1) and2095 (H2), at Two CO2 Concentration Levels (365 and 560 ppm) for Six Major Growing Regions of the United States

CO2 / RegionScenario Pacific Mountain Northern Plains Southern Plains

Mg ha-1

B-365 3.37 1.84 3.09 3.75B-560 4.08 2.44 3.71 4.61H1-365 3.68 1.74 2.90 3.65H1-560 4.45 2.38 3.85 4.66H2-365 3.81 2.42 3.20 3.21H2-560 4.59 3.21 4.21 4.02

Table 3.7. Simulated Winter Wheat Yields Under Baseline Climate (B) and HadleyCentre Projections in 2030 (H1) and 2095 (H2), at Two CO2 Concentration Levels(365 and 560 ppm) for Four Major Growing Regions of the United States

56


supply to agriculture was proportional to thechange in total water supply; i.e. that agricultureand non-agricultural users faced the same propor-tional change in water supply. More detail on thespecific changes and how they were derived fromthe estimates developed by the Water SectorAssessment are provided in McCarl (2000).

Crop Input Usage

Yield changes also can imply changes in someinputs, such as chemical inputs and those related tocrop harvesting, drying, and storage. A larger (orsmaller) yield will require more (or less) of theseother inputs. This association between yield andinput use is evident over time. Technologicalprogress that has increased yield has been accompa-nied by increases in input usage. On the otherhand, yield by definition is per unit of land; otherinputs, such as labor, are more closely related to areathan to yield. As part of the EPRI study that used theASM model, Adams et al. (1999) estimated a yield-input relationship. Land, labor, and water inputswere excluded from the estimation. For most crops,the increase in use of these other inputs was 40 per-cent of the yield change. Thus, if yield decreased by1 percent, crop input use decreased by 0.4 percent.Similarly, a 2 percent yield increase would bematched by a 0.8 percent input usage increase. Thisrelationship was included in the simulations. It hasthe effect of making yield improvements less eco-nomically beneficial than they otherwise would bebecause achieving the increases requires purchasingthese other inputs. Conversely, yield losses are notas economically costly because the purchase ofmaterial inputs is reduced. This type of adjustmentis appropriate for consideration of ongoing climatechange, for which technological change—the basisfor the estimate—provides a good analogy.

Livestock Performance,Grazing and PastureUsage

Much of the work on climate change impacts onagriculture focuses mainly on impacts on crops; itconsiders impacts on the livestock sector only indi-rectly, through changes in crop yields. Temperaturechange also can cause livestock to achieve alteredrates of gain. Heat stress has a variety of detrimentaleffects on livestock, including significant effects onmilk production, meat quality, and reproduction indairy cows. Analyses suggest that the most detri-mental effects would occur during warm periods inalready warm regions (Rötter and van de Geijn1999). The EPRI study (Adams et al. 1999) devel-oped relationships between temperature change,livestock performance, and feedstuff consumption inconsultation with experts on livestock productionand management. These estimates were used as abasis for estimating temperature-related declines inlivestock performance. McCarl (2000) provides theassumed changes in livestock production on a perhead basis.

Altered livestock performance—in terms of alteredending weights of sale animals or sales of livestockproducts—means that animals need differentamounts of feedstuffs to produce that ending weightor volume of products. In this study, we assumedthat feedstuff usage was strictly proportional to thevolume of products, although changes in climatecould change this proportion. Thus, if 10 percentmore milk were produced, 10 percent more feed-stuffs had to be consumed. When the livestock unitproduced multiple products, we used a weightedaverage of the percentage change in output to adjustthe feedstuff usage. The feed usage quantities forwhich we applied these adjustments included notonly traditional grains but also the number of animalunit months required of grazing and the acreage ofpasture required.

57


Other Livestock Input Usage

As with crops (and with the same rationale), weassumed that changes in nonfeed input use was 40percent of the production change.

Pasture Supply

Grazing and pasture use also are important assump-tions in the ASM model and are influenced by cli-mate change. The crop modeling component of theagriculture assessment included estimates ofchanges in grass and pasture growth resulting fromclimate change. Alterations in the growth rate ofgrass changes the available feed supply from a givenarea of pasture. Pasture use and grazing land avail-ability are represented in the ASM model and werechanged to reflect the change in grass and pasturegrowth. Pasture use was adjusted by the change ingrass growth. Thus, if grass growth increased by 10percent, livestock pasture use was multiplied by 0.9(1/1.1). We made this adjustment after changing thepasture required as a result of any change in bodyweight directly related to temperature.

We addressed grazing on western rangelands likethe adjustment for pasture; the availability of suchlands, however, traditionally has been measured interms of animal unit months (AUMs) of grazing. Wedeveloped an estimate of AUM supply sensitivity toclimate change by assuming that the change inAUM supply was the same as the change in grasssupply. Thus, if grass growth increased by 10 per-cent, the AUMs available increased by 10 percent.We did not make any further adjustments relatingto possible changes in forage quality resulting fromCO2 enrichment.

This combination of climate effects on livestockincludes most of the primary effects of climate on thelivestock sector. The principal omissions are directlosses of livestock from extreme storms. Otherpotential changes to the livestock sector that we donot consider here include an increase or decrease infloods or extreme winter weather events.

Pesticide Costs

Change in the incidences and ranges of agriculturalpests represents another likely effect of climatechange. Most insects, weeds, and diseases are sensi-tive to climate; climatic factors are an importantdeterminant of the range of many important agricul-tural pests. No previous assessment of agriculturalimpacts of climate change has explicitly consideredthe effects of climate change on agricultural pestswith the resulting economic effects of thesechanges. Studies based on cross-section evidencesuch as Mendelsohn et al. (1994) and Darwin et al.(1995) and subsequent analyses with theseapproaches implicitly include changes in pests andmany other factors, assuming that entire productionand climate regimes shift, intact, as a result of cli-mate change.

To consider how climate could affect agriculturethrough its affect on pests, we conducted a statisti-cal analysis that related pesticide expenditures to cli-mate. We conducted this analysis on cross-sectiondata; we explain it in greater detail in Chapter 5. Weestimated changes in pesticide expenditures forcorn, cotton, soybeans, wheat, and potatoes withregard to a percentage change in precipitation andtemperature. Based on these statistical relationships,we estimated a change in pesticide costs under eachclimate scenario. The limitations and advantages ofsuch cross-section evidence applied to time-seriesphenomena such as climate change have been dis-cussed in the context of other such efforts—thebroadest such effort being the Ricardian rentmethod developed by Mendelsohn et al. (1994). Amain additional limitation in this context is that, asapplied, this approach implicitly assumes that anyadditional potential damage from pest range andincidence expansion is fully controlled by the use ofadditional pesticides. Thus, the only economic lossto farmers is additional pesticide expenditures. Ifcrop losses were greater, even with additional pesti-cide expenditures, the lost revenue from thereduced sale of crops would be an additional loss.The full interaction of pests, climate change and cli-mate variability, and habitat is complex. Many pests(weeds, insects, diseases) respond to changes in

58


humidity and precipitation and are known to havetolerance limits to extreme temperatures. Pests alsorespond to habitat modification that might beinduced by climate change or by managementchange (such as changing the crop mix or expand-ing irrigation). A comprehensive review is includedin Patterson et al. (1999).

International Trade

Studies that have considered global impacts of cli-mate change have demonstrated that the economicimpact on a country can be heavily affected by howclimate change affects agriculture production inmajor agricultural exporting and consumingcountries (see Chapter 2). The ASM model includesan international sector; thus, in all scenarios, climateimpacts on the United States affect US competitive-ness in export markets. In the base scenarios, how-ever, although US agricultural production is affectedby climate there is no climate impact elsewhere inthe world. Conducting a full assessment of the restof the world was beyond the scope of this assess-ment, however. Our approach is roughly equivalentto assuming that, although there may be positive andnegative impacts of climate change on agricultureelsewhere in the world, the net impact is a “balanc-ing out” (i.e., no change). In fact, in the global stud-ies that have been conducted, the net global effectoften is relatively small because of a combination ofgainers and losers around the world.

To consider the sensitivity of our results to theimplicit assumption of no impact elsewhere in theworld, we constructed three sensitivity scenarios forpotential climate impacts on the rest of the world,based on previous global assessments. Two scenar-ios were developed from work that was based on aneconomic modeling analysis of international yieldchanges based on climate scenarios of the GoddardInstitute for Space Studies (GISS) and UnitedKingdom Meteorological Office (UKMO) climatescenarios (Reilly et al. 1993). Production changes inother regions are given in Tables 3.8a,b. Anotherscenario was based on a previous global modelingexercise that used the Hadley climate scenario. The

US estimates in this scenario were based on a differ-ent approach than the one used here and, therefore,are not directly consistent with our US crop studyestimates. This scenario is based on a model devel-oped at the Economic Research Service (Darwin etal. 1995). The GISS/UKMO climate scenarios are fair-ly old; the impact analysis dates to the early 1990sand involves doubled CO2 equilibrium scenarios.The advantage of these scenarios is that the underly-ing approach used for the crop studies is similar tothe approach we used in this assessment, and thestudy provides details on the major crops and worldregions represented in the ASM model. For theDarwin scenario, we based adjustments on changesin net exports from the United States.

Although none of these scenarios is completely con-sistent with the analysis of the United States that weconducted, they provide a useful way to demon-strate the sensitivity of the economic estimates weobtained to different assumptions about how cli-mate change could affect the rest of the world. Wechose the GISS/UKMO scenarios in part because inthe study from which they were taken they repre-sented the mildest (GISS) and the most severe(UKMO) scenarios among those that consideredboth adaptation and the CO2 fertilization effect.

Economic Results

In the following subsections, we discuss the maineconomic results. We first discuss the results fromthe core scenarios. We then consider sensitivitycases, including the trade sensitivity results, thealternative Hadley scenarios that are based on thePNNL crop modeling, and a set of miscellaneoussensitivities. We report here the major results.Altogether we ran 43 scenarios representing differ-ent impact combinations (e.g., with and withoutadaptation or pest effects) and alternatives (e.g.,alternative trade and crop yield effects), producingresults for aggregate economic effect, regionalproduction, and resource use for each scenario. Inmost cases, the general pattern of change acrossregions and resource use closely reflects differencesin the aggregate economic effect across scenarios.

59


Region Wheat Coarse Grains Rice Other Crops Secondary

Canada 4.5 -6.6 6.0 -7.5 -2.0Western Europe 11.0 9.8 13.5 11.6 -1.4Former Soviet Union -8.1 -6.0 -7.4 -1.4 -0.3Eastern Europe 1.5 3.0 3.1 11.2 -0.3Australia & New Zealand 46.2 19.8 28.0 27.0 -0.3China, Taiwan, & S. Korea 0.9 0.7 2.6 12.9 0.5Other East Asia -15.0 -30.0 -15.7 -10.2 -0.8India -19.8 -36.0 -17.1 -25.6 -1.2Argentina -7.6 -0.6 18.7 17.5 0.0Brazil -28.4 -13.7 -18.6 -7.0 -0.6Mexico -27.2 -33.8 -24.1 -16.1 -0.2Japan 1.6 17.3 8.5 10.4 -1.7Africa (all) & Middle East -12.8 -25.3 8.7 -8.0 -1.6Other Latin America -28.7 -17.6 -15.5 -25.2 0.1

Table 3.8b International Trade Scenarios: Percentage Production Changes, Based on UKMO Climate Scenario

Region Wheat Coarse Grains Rice Other Crops Secondary

Canada 20.0 17.2 2.2 20.3 1.4Western Europe -0.7 3.1 4.5 12.0 0.7Former Soviet Union 23.0 12.0 13.2 17.6 0.1Eastern Europe 6.8 1.3 1.3 13.7 0.1Australia & New Zealand -11.6 10.7 17.1 8.2 0.4China, Taiwan, & South Korea 14.9 0.1 1.1 15.6 -0.1Other East Asia -21.0 -32.9 -5.7 -15.6 0.4India -4.4 -13.9 -2.2 -6.1 0.8Argentina -25.8 8.5 9.8 6.0 0.4Brazil -35.2 -10.3 -11.8 -0.5 0.2Mexico -34.9 -34.8 -18.0 -19.9 0.2Japan -1.9 22.2 11.4 11.2 0.4Africa (all) & Middle East -19.0 -24.0 3.2 -5.3 1.9Other Latin America -29.1 -10.6 -9.7 -18.6 0.1

Table 3.8a International Trade Scenarios: Percentage Production Changes, Based on GISS Climate Scenario

We have tried to highlight here the broad pattern ofresults. Complete tables of results are provided inMcCarl (2000).

Results from Core Scenarios

A value of an economic model such as the ASM isthat it can summarize the net impact of a combina-tion of many different changes. The model also

provides the ability to consider distributional andresource use effects.

Aggregate Economic Impacts

We report aggregate results in terms of a change inwelfare—measured as the sum of producer and con-sumer surplus. Welfare is preferred as a measure ofeconomic impact over measures such as change in

60


agricultural production or consumption because itincludes consideration of the fact that with less pro-duction, fewer inputs are used and the fact that con-sumers, may shift consumption among agriculturalproducts, or might substitute consumption of othergoods. Figure 3.4a displays the results based on theCanadian climate scenario, and Figure 3.4b displaysresults based on the Hadley scenario. Included inthese figures are changes of consumer and producersurplus in the United States, as well as a change intotal surplus. The difference between the two is theeconomic impact on producers and consumersoutside the United States. The scenarios reported inFigures 3.4a,b do not include any direct climate

impact on agriculture outside the United States, butimpacts on foreign producers and consumers occurbecause of changes in prices of internationally tradedcommodities. The figures provide results for 2030and 2090 under three different scenarios.The firstscenario reflects the impact of climate change—including crops, livestock, and water demand—andsupply effects without adaptation. The second addsadaptation, and the third adds, in addition, the effectsof climate on pesticide expenditures.

Given the differences in the climate models and theintermediate crop modeling results, the economicresults are generally as expected. Overall, the effects

on total surplus generally are posi-tive—much more so for the Hadleyscenario. Net economic benefitsrange from about –$0.5 to +$3.5billion (year 2000 dollars) in theCanadian scenario and between 6and 12.5 billion dollars for theHadley scenario. In both climatescenarios, the total and domesticsurplus increases between 2030and 2090. Several analysts have sug-gested that at more extreme levelsof climate change, one shouldexpect losses. We have not con-ducted a full transient cropmodel/economic analysis, so wecannot be sure whether benefits by2090 are declining from some peakexperienced between 2030 and2090 or whether benefits are con-tinuing on a general upward trend.As illustrated by the Canadianscenario, however, the time path ofimpact may not be easily describedby a simple function. In 2030—atleast for the no-adaptation case andfor the US surplus—the net effect isan economic loss that turns to gainsby 2090 for all but the no-adapta-tion case. Given the multitude ofeffects across many differentregions, tracing these results

Figure 3.4b. Economic Impacts of Climate Change, Hadley Centre Climate

14

0

US SurplusTotal Surplus

bill

ions

of

do

llars

2030

2030

with

adap

tatio

n

2030

with

pes

ts

2090

2090

with

adap

tatio

n

2090

with

pes

ts

12108642

Figure 3.4a. Economic Impacts of Climate Change, Canadian Centre Climate

4

3

2

1

0

–1

–2

US SurplusTotal Surplus

bill

ions

of

do

llars

2030

2030

with

adap

tatio

n

2030

with

pes

ts

2090

209

0 w

ithad

apta

tion

2090

with

pes

ts

61


Figure 3.5a. Producer versus Consumer Impacts ofClimate Change, Canadian Centre Climate

conclusively to a specific aspect of the climate sce-narios is impossible. The pattern of results very like-ly reflects the fact that in the Canadian scenario, pre-cipitation decreases in the United States in 2030 andthen increases by 2090. Care must be taken to avoidoverinterpreting this time path or any of the specificresults. Climate models produce variability from yearto year and decade to decade. Even for specific mod-els such as the Canadian or Hadley models, a particu-lar decade of climate drawn from a particular sce-nario must be considered only one possible drawfrom a distribution of possibilities. By 2030, the addi-tional greenhouse gas forcing beyond that of currentclimate is relatively smaller compared with 2090, sothe natural variability on a decadal scale can have alarge effect relative to the signal from greenhousegas forcing.

The distribution of benefits between foreign anddomestic is notably different in the two scenarios.Much of the benefit goes abroad in the Canadianscenario, whereas relatively little flows abroad inthe Hadley scenario. This difference occursbecause of the differential effects on crops whereexports are important versus those that are mainlyconsumed domestically.

As observed for the intermediate crop yield results,adaptation is considerably more important whenthe impacts are adverse than when they are benefi-cial. Although this effect shows up in the compari-son of the two climate scenarios, more research isrequired to assess the robustness of this result. Amore expansive exploration of adaptation optionssuch as double cropping could reveal further gainsin northern regions.

Net pesticide expenditures increase, thereby reduc-ing total economic surplus. This effect is quitesmall. The size of the effect is not surprising, howev-er, given that pesticide expenditures account foronly a few percent of total costs. This estimate mayunderstate losses, however, because it does notinclude any increase in damage that cannot be elimi-nated through increased use of pesticides.

Distributional Effects

The distribution of economic effects between pro-ducers and consumers and among regions varies.Figures 3.5a and 3.5b display the distribution ofeffects between domestic US producers and con-sumers. Across all scenarios, consumers generallygain from lower prices, whereas these lower pricescause producer losses despite the fact that climatechange has improved productivity. The Canadian

Figure 3.5b. Producer versus Consumer Impacts ofClimate Change, Hadley Centre Climate

8

6

2

4

0

–2

–4

Consumer SurplusProducer Surplus

bill

ions

of

do

llars

2030

2030

with

adap

tatio

n

2030

with

pes

ts

–6

2090

2090

with

adap

tatio

n

2090

with

pes

ts

86

24

0–2–4

Consumer SurplusProducer Surplus

bill

ions

of

do

llars

2030

2030

with

adap

tatio

n

2030

with

pes

ts

–6

10121416

2090

2090

with

adap

tatio

n

2090

with

pes

ts

62


scenario produces an approximate balance in termsof domestic consumer gains and producer losses inmost scenarios. In contrast, the Hadley scenario pro-duces large consumer gains. The productivity gainsare so substantial, however, that the output andexport gains to producers nearly offset the pricedeclines. Although the absolute level of change iscomparable between producer and consumers, inpercentage terms the changes to producers aremuch more substantial. For comparison purposes,the total economic benefit derived from food con-sumption in the base is estimated at approximately$1.1 trillion, whereas total producer surplus is onthe order of $30 billion. Thus, the $4–5 billion sur-

plus losses in the Canadian scenario represent13–17 percent loss of surplus to producers, whereasthe gains of $9–14 billion of consumer surplus inthe Hadley scenarios represent only a 1.1–1.3 per-cent gain to consumers. We would expect producerlosses to be realized ultimately as changes in thevalue of land. A 13–17 percent loss in this assetvalue is substantial; to place this in context, howev-er, agricultural land values fell on the order of 50percent between 1980 and 1983.

Figures 3.6a and 3.6b display the regional differ-ences. We report an aggregate index of the produc-tion across crops, with prices used as weights in the

Figure 3.6a. Regional Production Changes, Canadian Centre Climate, Percentage Change in Output (crop and livestockproduction aggregated using price weights)

Figure 3.6b. Regional Production Changes, Hadley Centre Climate, Percentage Change in Output (crop and livestockproduction aggregated using price weights)

Ind

ex, 0

=no

cha

nge

No

rthe

ast

0

50

100

–50

–100

Lake

Sta

tes

Co

rnb

elt

No

rthe

rnP

lain

s

Ap

pal

achi

a

So

uthe

ast

Del

taS

tate

s

So

uthe

rnP

lain

s

Mo

unta

in

Pac

ific

2030

2030 withadaptation

2090

2090 withadaptation

No

rthe

ast

Lake

Sta

tes

Co

rnb

elt

No

rthe

rnP

lain

s

Ap

pal

achi

a

So

uthe

ast

Del

taS

tate

s

So

uthe

rnP

lain

s

Mo

unta

in

Pac

ific

2030

2030 withadaptation

2090

2090 withadaptation

Ind

ex, 0

=no

cha

nge

0

20

406080

100

120140

63


index. The plotted values are percentage changesfrom base production. The figures show substantialregional differences in both scenarios. The basicregional pattern is similar in both scenarios. TheLake, Pacific, Mountain, and the Corn Belt regions (indecreasing order) show large increases in produc-tion—generally between 50 and 150 percentincreases in output. The pattern of absolute (or rela-tive) losers varies more across the scenarios. TheSoutheast, Southern Great Plains, and Delta stateslose absolutely in the Canadian scenario or showthe smallest increases in production in the Hadleyscenario. Appalachia also is more negatively affect-ed. Impacts on the other regions vary substantiallyacross the two climate scenarios and over the twotime periods.

In the Hadley scenario, no region shows a produc-tion decline. With substantial overall producer loss-es in the United States as a result of declining com-modity prices, however, farmers in regions thatshow only modest increases in production clearlyare suffering substantial economic loss. In thesecases, we expect economic losses to show up asdecreases in the value of assets located in theseregions—primarily agricultural land. In theCanadian scenario, several regions show absolutedecreases in production. This adjustment processover the longer term explains why production cancontinue to increase even though the region experi-ences economic loss. Owners ofland may be forced out of business,and the resulting price of land wouldreflect the reduced productionpotential resulting from degradingclimatic conditions. A new buyerpaying the lower price could thenprofit because the asset cost waslower. Thus, production continuesdespite the fact that owners of farm-land take a significant economic loss.In the Canadian scenarios, regionswith production losses also are suf-fering from price decreases, althoughnot as severely as in the Hadleyscenario.

Resource Use

Overall, measures of resource use generally declineacross all categories, both climate scenarios, andboth time periods (Figure 3.7). Irrigated land andwater use decline most, reflecting the overallincrease in production and decline in prices and therelative yield effects between irrigated and drylandagriculture. Overall, the results for these scenariossuggest considerably less pressure on resources as aresult of the overall increase in productivity.

Trade Scenarios

Table 3.9 provides aggregate economic results forthe three different trade scenarios for 2030 and2090. All three foreign trade scenarios were runagainst the Canadian and Hadley domestic US sce-narios. The trade scenarios generally do not lead toa substantial change in total surplus or total US sur-plus. This result reflects the fact that the UnitedStates is a substantial commodity exporter but also asubstantial food consumer. Hence, global pricechanges have roughly offsetting effects—consumersgain from price decreases whereas producers lose.With price increases, these effects are in the oppo-site direction but again roughly offset one another.Thus, the biggest effect of the trade scenarios isreallocation of the total domestic effect betweenproducers and consumers. The Darwin scenario

Figure 3.7. Changes in Resource Use, Canadian and Hadley Centre Climates,without Adaptation

Per

cent

cha

nge

2030 CC

2090 CC

2030 HC

2090 HC

0

10

20

–10–20–30–40–50–60

Irri

gat

edLa

nd

Dry

land

Cro

pla

nd

Pas

ture

AU

Ms

Wat

er

Lab

or

64


creates somewhat greater losses for US producers—the implication being that the impact on productionin the rest of the world for those goods in whichthe United States trades is positive with generallylower world prices than in comparable cases inwhich world production was left unchanged. Forthe GISS and UKMO scenarios, the effect is theopposite:World prices increase—very modestly inthe GISS case and more substantially in the UKMOcase—thereby shifting some of the gains from USconsumers to US producers.

These trade scenarios were not developed consis-tently with the domestic impacts. If the resultsobtained for the United States with these climatescenarios—generally, more positive yield effects thanin past assessments—were observed across theworld, we would expect world prices generally todecline. The result would be further gains by USconsumers and losses by producers—as observed inthe Darwin scenario rather than in the GISS orUKMO scenario. On the other hand, a factor that isundoubtedly important in moderating the climateimpacts on the United States is the cooling effect ofsulfate aerosols in northern temperate regions.Earlier assessments used climate scenarios that didnot include sulfate aerosol effects. Often these

assessments showed warming benefits in morenortherly regions and losses in tropical regions.Sulfate aerosol effects could produce a regional pat-tern of climate change that reduces benefits to somenorthern regions compared with earlier assessment,while leaving unchanged the losses in the tropicalregions. If such a result obtained, the implicationmight be world price increases and a shift of bene-fits from US consumers to US producers. More com-plete global studies with newer climate scenariosare required to resolve this effect.

Alternative PNNL Crop Scenarios

Table 3.9 provides aggregate economic results forthe alternative PNNL crop simulations. These resultswere produced only for the Hadley scenario and didnot include adaptation. We did not include pestchanges in this comparison because the primarypurpose here is to evaluate scenarios for PNNL cropsimulations versus core crop simulations for a com-parable set of scenarios. In terms of aggregateconsumer and producer surplus, the PNNL-basedeconomic results were similar to the site-based eco-nomic results. The PNNL study did not cover allcrops, however, so the PNNL-based economic resultsinclude the site-based yield results for crops not

Year Scenario Consumer Surplus Producer Surplus Foreign Surplus Total Surplus

2030 Base, CC 2.8 -4.2 0.8 -0.62030 Darwin, CC 5.0 -6.4 0.8 -0.62030 GISS, CC 2.2 -3.7 0.8 -0.72030 UKMO, CC 2.8 -4.2 0.8 -0.62030 Base, HC 9.4 -3.6 0.6 6.42030 Darwin, HC 11.0 -5.2 0.5 6.32030 GISS, HC 9.3 -3.4 0.6 6.42030 UKMO, HC 9.4 -3.6 0.6 6.42090 BASE, CC 4.5 -4.5 1.1 1.12090 Darwin, CC 5.1 -5.5 1.1 0.72090 GISS, CC 4.2 -4.3 1.2 1.02090 UKMO, CC 4.5 -4.5 1.1 1.12090 BASE, HC 11.4 -0.8 1.0 11.52090 Darwin, HC 11.8 -1.5 1.0 11.32090 GISS, HC 11.3 -0.8 1.0 11.62090 UKMO, HC 11.4 -0.8 1.0 11.5

Table 3.8 Sensitivity to Trade Scenarios, Without Adaptation

Note: Columns may not sum to total due to independent rounding.

65


considered by PNNL—thereby making the econom-ic results more similar than they might have been ifPNNL had different results for all crops. The totaleconomic welfare gain is somewhat higher in 2030and somewhat lower in 2090. Although the generalresult—increases in productivity for most crops inmany places—obtained for the PNNL and for thesite-based studies, there were some important differ-ences in the results, such as the effect on irrigatedcrops of differences and denser coverage of sites.These differences did lead to differences at theregional level. In the PNNL scenarios, the Southeastdoes not show up as a particularly severely affectedregion; the Southern Plains and Northeast are affect-ed considerably more positively than in the core sce-narios. The Northern Plains appear to be the morenegatively affected region in the PNNL scenario.The Lake States, Corn Belt, and Pacific regions areamong the more positively affected regions in bothscenarios.

Overall, this comparison is reassuring in the sensethat the limited site selection in the core scenariosdoes not appear to have created a substantial bias inaggregate estimates. The aggregate effects offer arelatively weak test, however; several crops were leftunchanged between the core and PNNL resultsbecause the crops were not simulated by PNNL.Clearly, some differences do occur at the regionallevel, emphasizing the uncertainties in producingconsistent projections at the regional level.

Other Scenarios and Sensitivities

We were unable to generate yield changes for cottonwith a cotton crop model. Instead, we adapted resultsfrom another study. We also simulated results byusing soybeans as a proxy for cotton. Soybean results

Consumer Producer Foreign TotalSurplus Surplus Surplus Surplus

HC 2030 9.4 -3.6 0.6 6.4HC-PNNL 2030 11.2 -3.3 0.3 8.2HC 2090 11.4 -0.8 1.0 11.5HC-PNNL 2090 14.0 -4.0 0.7 10.6

Table 3.9 PNNL Crop Yield Simulations, Without Adaptation generally were quite negative in the South in theCanadian scenarios, whereas the alternative cottonscenarios showed more positive effects. As a result,this alternative assumption produced quite differentresults. Notably, under the Canadian scenarios the$0.6 billion total surplus loss in 2030 doubled, andthe approximately $1 billion gain in 2090 changed toa $1 billion loss. Most of this change accrued todomestic and foreign consumers. Producers lossesactually were slightly reduced in 2090 as a result ofhigher cotton prices. Negative production effectswere most substantial in the Delta region.

The results derived by projecting the agriculturaleconomy forward to 2030 and 2090 were not quali-tatively different. Specific quantitative resultsdepend crucially on highly uncertain forward pro-jections. The two basic aspects of these projectionsare yield growth and demand. Projecting historicalyield growth and increases in demand because ofpopulation growth increases the absolute size of theagricultural economy. If we consider yield changesin percentage terms as operating on the new higheryields, the percentage effect is similar. Differencescan arise by virtue of different assumptions aboutyield and demand growth for different crops and dif-ferences in yield impacts among crops.

We also jointly considered the impacts of changes inagriculture and forestry. Because of the long growthcycle of forests, there is a far greater need to lookforward and consider the present value of changesover many years. Forest yield scenarios were basedon the Canadian and Hadley climate scenarios anduse of two ecological process models (Joyce et al.2001; Irland et al. 2001; McCarl 2000;Alig et al.1997). The results suggest that when the agricultureand forestry sectors are considered together, theeffects for the US economy are beneficial. Increasedsupplies from forests lead to reductions in log pricesthat decrease producers’ welfare (profits) in the for-est sector. At the same time, lower forest productprices mean that consumers generally benefit. Thispattern of distributional impacts on forestry produc-ers and consumers is similar to results in the agricul-tural sector. Increases in the net present value oftotal economic welfare (combined forestry and

66


agriculture) ranged between 0.9 and 1.2 percent,with higher positive impacts under the Hadley cli-mate change scenarios. More details on theseresults are provided in McCarl (2000).

Land-use changes between forestry and agriculturaluses are an important avenue of adjustment to cli-mate-induced shifts in production, and there arenotable differences in these adjustments across cli-mate change scenarios. Over the full projectionperiod, the base and Canadian scenarios project anet shift of land from agriculture to forests—the lat-ter at about half the rate of the former—whereasthe Hadley scenarios project a net loss of forest landto agriculture. Yields from the land generallyincrease in both the forestry and agricultural sectorsin all four scenarios. In the Canadian scenarios,these shifts are relatively more favorable for forestryprofits compared to agriculture, whereas the oppo-site is true in the Hadley scenarios.

Summary of Main Economic Results

The main results of the economic analysis are asfollows:

• Climate change as modeled under the climatescenarios that we considered is mostly benefi-cial for society as a whole in terms of agricultur-al impacts, particularly if adaptation is consid-ered. This finding differs from the results ofprevious scenario analyses, in which resultshave been mixed and generally negative in theabsence of adaptation.

• Climate change uniformly shows increases in cropproduction and exports and decreases in cropprices. Livestock production and prices are mixed.

• Climate change is largely detrimental for produc-ers. Climate changes are beneficial for foreignsurplus and for consumers. These results reflectthe overall positive effect on production, whichleads to decreasing prices.

• There are substantial shifts in regional production,with gainers and losers. The Lake states, Mountainstates, and Pacific region show gains in produc-tion; the Southeast, the Delta, the Southern Plains,

and Appalachia generally lose. Results in the CornBelt are generally positive. Results in other regionsare mixed, depending on the climate scenario andtime period. Regional results show broadly thatclimate change favors northern areas and canworsen conditions in southern areas—a resultobtained by many previous studies.

• Our analysis suggests increases in pesticideexpenditures as a result of climate change—apartial offset to the overall benefits.The magni-tude of this effect is relatively small.

• The overall benefits of climate change are greaterin 2090s than in 2030 for the United States as awhole; even for regions with losses, thesechanges generally are less in 2090 than in 2030.Changes in precipitation and atmospheric CO2

probably are the source of this result.

• Climate change largely causes a decrease inresource usage because of expanded productivi-ty. In particular, dryland, total cropland, pastureland, and water usage decline.

• Farm-level adaptation increases the climatechange benefits to society by about $1 billion.Producer losses generally are reduced byadaptation.

• Consideration of climate effects in other coun-tries did not greatly alter the climate changebenefits to society. It can have substantial distri-butional assumptions, depending on how climateaffects the rest of the world.

• Changing the base year does not alter the signof the climate change benefits to societyas a whole.

• The results we obtained by using two differentcrop yield simulation approaches were quitesimilar in overall magnitude.

• Jointly considering forest and agriculturalchanges resulting from climate does not changethe impacts substantially. The net effect on soci-ety of both changes is positive, and the distribu-tion effects are similar; producers suffer surpluslosses because of declining prices, whereas con-sumers benefit.

67

Impacts of Variability onAgriculture

Chapter 4

Introduction

Crop yield variability is the result of many differentfactors. These factors include changing productionpractices such as the introduction of new tools, newhybrids and varieties or cultivars, development ofnew diseases and pests, and government policy.Underlying many of these factors are extremeweather events and the variability of weather fromyear to year.

Extreme weather events such as hurricanes anddroughts have obvious impacts and recently necessi-tated two disaster relief bills for farmers. In the pastdecade, large yield reductions were observed in1988 because of severe drought throughout the mid-section of the United States and again in 1993 whenlarge areas of Illinois, Iowa, Missouri, and otherMidwestern states experienced record rainfall fromearly spring through summer. In the early 1980s,corn surpluses were so large that in 1983 farmerswere paid to remove large acreage from production.

In recent years, climate scientists have improvedtheir ability to identify and predict seasonal to inter-annual climate phenomena such as ENSO 6 to 18months in advance. This improved prediction capa-bility has contributed to increased attention towardidentifying how farmers would or could respond inanticipation of these events. Several studies suggestthat one-fifth of the losses related to such eventsmight be avoided if appropriate changes in croppingpractices were made.

In this chapter, we review and evaluate the impactsof climate variability on crop yields and consequentimpacts on the US agricultural economy, focusingprimarily on how greenhouse gas-induced climatechange could change variability. We first present themethod by which the climate change scenarioswere used in results discussed in Chapter 3 and laterin this chapter. The purpose is to make clear the

extent to which the approach already includes vari-ability and extreme events as they affect agriculture.We also clarify the relationship among changes inthe climate means, the variability of climate, and thefrequency of climatic extremes.

The basic approach in the core site studies inChapter 3 was to apply changes in mean monthlyprecipitation and temperature from the GCM scenar-ios to actual 40-year historical records for the sites.The PNNL approach used changes from the GCMsas seeds for a stochastic weather generator that ispart of their model. Both approaches therebyinclude variable weather. For temperature in thesite studies, we calculated absolute differencesbetween the GCM-modeled mean monthly tempera-ture in the scenario with greenhouse gas forcingand the GCM-modeled climate without forcing(often referred to as the control scenario). Weadded these differences to the daily values in thehistorical record for each site for the applicablemonth for the 40-year historical record. Thus, thevariability of temperature remains the same as in thehistorical record, but the mean is higher.

For precipitation, the standard approach is to useratios of the greenhouse gas-forced climate and thecontrol climate—rather than absolute differences.Ratio adjustments for precipitation are widely usedin crop studies because they avoid the problem ofpotentially negative precipitation. A negative precip-itation estimate can result when there are errors insimulation of precipitation in the control run of theGCM combined with projected decreases in precipi-tation. Some studies have used absolute differencesin precipitation, replacing negative values wherethey occurred with zero precipitation, and foundthat using absolute differences resulted in more pre-cipitation than the ratio approach, particularly overdesert regions (Alcamo et al. 1998; Darwin 1997).The ratio approach changes the variability of thedaily intensity of precipitation. The variance

68


changes as a function of the square of the ratio ofthe climate change to control climate projections(Mearns 1996). This change in variance is the coin-cidental result of using ratios rather than differ-ences; it does not reflect an analysis of how variabili-ty might actually change on the basis of analysis ofGCM results. These methodological differencescould lead to different yield results and water-useresults, particularly in more arid regions where thedifferences are greater. Yield increases in arid areasand reductions in water use would be expected tobe larger if the same difference were observed withthe Canadian and Hadley models. We were unableto consider this alternative method directly.

The PNNL stochastic weather generator also repro-duces weather that varies like that observed in thepast, but the stochastic aspect of the approachmeans that the realized weather has characteristicslike historical weather but is different in each run.The mean and variance calculated over many yearsof simulation are the same across runs. Theseapproaches have been developed because climatemodel results are still too inaccurate on a regionalscale to be used directly.

Thus, the method used here for generating climateinput for the crop models produces a weatherrecord with climate change that includes storms,droughts, and extreme temperatures. In particular,because we used monthly mean changes, the season-ality of climate (e. g., distribution of precipitationand the pattern of warming over the year) canchange. For example, if the GCM scenario projects aprecipitation decrease of 90 percent in the summerand a precipitation increase of 90 percent in thewinter for a location with seasonally balanced pre-cipitation, the yearly total precipitation would notchange but the seasonal distribution would be great-ly altered. This scenario can be regarded as a changein the seasonal cycle (seasonal variability) of precipi-tation. This change is captured by methods appliedin Chapter 3.

Changing the mean temperature and precipitationin this way also changes the frequency ofextremes—for instance, the likelihood that the maxi-mum temperature on any day in the summer willexceed 35°C. In fact, given the usual distribution oftemperature highs for a day, the frequency withwhich the temperature exceeds an absolute thresh-old such as 35°C changes rapidly with a change inthe mean. For example, based on the 30-year weath-er record for Des Moines, Iowa, there is an 11 per-cent chance that the maximum temperature on anyday in summer will exceed 35°C. Moreover, basedon the distribution of high temperatures for DesMoines, if the mean temperature were to increase by1.7°C, the chance that it will exceed 35°C wouldrise to 22 percent. Thus, for a relatively smallchange in the mean maximum temperature, the like-lihood that the temperature will exceed 35°C dou-bles. Again, this increase in the likelihood ofextremes is captured in the methods applied inChapter 3 as well as later in this chapter.

If the variability (i. e., the standard deviation or vari-ance) of the temperature also changed, this variabili-ty would further affect the frequency of extremeevents. For example, if the simulated distribution ofhighs became wider (i.e., the variance increased),the chance that the temperature will exceed 35°C inthe foregoing example would increase by more than22 percent. This aspect of change in variability wasnot incorporated in our scenarios. Similarly, someaspects of potential changes in variability in precipi-tation are excluded as a result of the methodsapplied in Chapter 3. For example, if the historicalrecord shows on average 10 rain events in July andAugust, the climate change scenario developed withthe method in Chapter 3 also will have, on average,10 rain events in July and August. The method alsodoes not account for changes in frequency of pre-cipitation on a daily time scale. Therefore, the resultof GCM projections of an increase in precipitation isthat each rain event has more rain. The methodused in Chapter 3, however, would not include aprojected trend toward fewer rain events or raincoming in heavy downpours rather than slowly overthe course of a day.

69

Chapter 4: Impacts of Variability on Agriculture

Common parlance recognizes that a drought is adrought regardless of whether it is caused by achange in the mean or a change in the variance. Wecan easily imagine, however, that two areas with theexact same climatic means can have very differentagricultural potential. An area with even rainfall andtemperatures through the year could be the bread-basket of a nation. If identical mean conditionsoccurred but precipitation fell in torrential down-pours—followed by months with no rain—and tem-peratures varied from freezing to scorching, theregion would become a wasteland with regard toagricultural potential.

A major point of this discussion is to make clear thatour method produces changes in extremes but doesnot include changes in all aspects of climate variabil-ity that affect the frequency of extremes (e.g., vari-ance). The intent of this chapter is to address morespecifically the impacts of variability, extremeevents, and changes in variability.

We begin by briefly reviewing the evidence from cli-mate modeling on how variability could change.We then review the impact of weather on variabilityin crop yields and discuss possible future responsesto changing variability. We consider the impacts ofclimate change and variability in the context of pro-jecting extreme events, simulating the potentialimpacts of climate variability and extreme events oncrops, relating crop yield variability to climate, andconsidering the economic implications of potentialENSO shifts.

We examine the impacts of climate on the variabilityof US corn, cotton, sorghum, soybean, and wheatyields. We chose these crops because of theirwidespread coverage and important economicvalue. Other regionally important crops also will beaffected by climate change and variability but wedid not analyze them.

Predicting ExtremeEvents

Most of our knowledge of possible changes inextremes comes from climate model experiments offutures with increased greenhouse gases andaerosols. Climate modeling capabilities haveimproved greatly in the past 10 years, and examina-tion of changes in at least certain types of extremessimulated in climate models is more common nowthan it was in the past. The current generation ofcoupled atmosphere-ocean general circulation mod-els (AOGCMs) has improved spatial resolution(about 2.5 degrees latitude), adopts more realisticland surface schemes, and includes dynamic sea iceformulations. These and other improvements, suchas nesting of high-resolution (tens of kilometers)regional models within AOGCMs, have improvedour ability to estimate possible changes in someextremes. In this section, we review what is knownfrom climate models on possible changes in extremeevents in the 21st century.

Temperature

One of the earliest and simplest analyses of possiblechanges in extreme events concerns increased fre-quency of extreme daily high-temperature eventsand decreased frequency of low daily temperatureevents. With an increase in mean (maximum and/orminimum) temperature—assuming no otherchanges in other aspects of temperature (e.g., vari-ability)—there will be an increase in the likelihoodof, for example, days with maximum temperaturesexceeding 35ºC. The change in the probability ofextreme daily temperature events is nonlinear withthe change in mean temperature—that is, a smallchange in mean temperature will produce a relative-ly large change in the probability of a temperatureextreme (Mearns et al. 1984). Changes in tempera-ture variance also contribute to changes in the fre-quency of extremes; on a per degree basis, thesechanges have a greater influence than the change inthe mean (Katz and Brown 1992). In climate modelexperiments investigated to date, however, the meanusually changes more than the variance.

70


Within the literature on climate change, several cli-mate simulations of the future have found that innorthern mid-latitudes, the daily temperature vari-ance increases in summer but tends to decrease inwinter. These changes complement the effects ofthe changes in mean. The increased frequency ofhigh-temperature events in summer is furtherincreased by the increased variability, and decreasesin low extremes in winter are further decreased bythe decreasing variance (Meehl et al. 2000).

Precipitation

Earlier studies of climate models found a tendencyfor increased precipitation intensities; more recentstudies continue to find this result. For example,Zwiers and Kharin (1998) found that mean precipi-tation increased by about 4 percent and precipita-tion extremes increased by 11 percent over NorthAmerica in a doubled-CO2 simulation. Anotherimportant and seemingly robust result from climatemodels is a tendency toward mid-continental dryingin summers—as a result of higher temperature andreduced precipitation—with increases in CO2 (e.g. ,Wetherald and Manabe 1999). Seasonal and regionalchanges in the pattern of precipitation and tempera-ture are accounted for within crop studies describedin Chapter 3; we use them as the basis for economicmodeling. These general regional and seasonal pat-terns are reflected in the regional estimates present-ed in Chapter 3.

Extratropical and Tropical Storms

Although researchers have made steady improve-ments in the ability of climate models to adequatelymodel tropical and extratropical storms, they stillhave relatively low confidence in model simulationsof changes in these features. A growing number ofstudies address possible changes in extratropicalstorm activity, but little agreement is found amongthese studies. Moreover, no consensus has emergedamong global models regarding changes in the fre-quency or intensity of tropical cyclones. Severalstudies have shown increased intensity of tropicalcyclones, but the models are still too coarse to

resolve many important features of such storms(e.g., the eyes of hurricanes).

El Niño/Southern Oscillation andLa Niña

ENSO is a major coupled ocean-atmosphere phe-nomenon that determines the interannual variabilityof climate and thus will be a major determinant ofthe future variability of climate. El Niño is the partof the oscillation when Pacific waters off the coastof South America are warm; La Niña is the coolphase. Current climate models have much improvedsimulations of ENSO, but conclusive evidence ofhow ENSO might change remains elusive. Severalstudies suggest, however, that with a warmer basecondition, precipitation extremes associated with ElNiño events may become more extreme—that is,more-intense droughts and flooding conditions maybe found (e.g., Meehl 1996). There has been consid-erable progress in the realm of seasonal forecastingof ENSO events and its connections with broaderclimate phenomena. The relevance of more-severeENSO events to agriculture is discussed below.

Conclusions

The literature on predicting extreme events indi-cates that our knowledge of changes in extreme cli-mate events in the future remains limited, with theexception of relatively simple single-variableextremes such as those related to daily temperature.Yet many types of extreme events certainly willchange in frequency and possibly intensity in thefuture. Many of these events (temperature and pre-cipitation extremes, droughts, floods) have impor-tant effects on agriculture. Even with little conclu-sive information on how such extremes may change,sensitivity analyses can illustrate how changes inextremes could affect cropping systems and agricul-ture in the United States, suggesting strategies thatreduce losses. Although long-term prediction ofchanges in climate variability because of greenhousegas accumulation may remain elusive, studies ofresponse to variability are useful in identifyingstrategies that could be used as medium-term cli-mate prediction improves.

71


Predicting the Impact ofClimatic Variability andExtreme Events on Crops

Most research regarding potential changes in cropyield resulting from climate change has focused onthe impacts of changes in long-term climatic aver-ages, with the assumption that climate variability astechnically defined will be the same as in the pre-sent climate. Changes in climate variability, howev-er, will affect the frequency of extremes and couldhave important impacts on crop yields. We discussbelow some of the effects of extreme events on agri-culture (independent of whether their probabilitiesare changing), aspects of modeling extreme eventsin crop models, and the effect on interannual eventssuch as ENSO. We then review some recent effortsthat have attempted to separate changes in variabili-ty from changes in the mean. Finally, we discussspatial variability.

Examples of Extreme Events Affecting Crops

Extreme events that affect crops occur on varyingspatial and temporal scales. Events on the interan-nual time scale include seasonal droughts, floods,cold winters, and so forth. Well-known periods ofdrought in the 1930s and again in the 1950s severe-ly decreased crop yields in the United States.

On time scales of hours to weeks, very short-livedextreme events within the cropping season cancause serious damage to crops. For example, manyfield crops suffer after consecutive days of high tem-peratures during sensitive phenological stages. Cornis a very sensitive crop, and several researchers haveidentified damaging events: Shaw (1983) reportedthat damage to corn occurs after 10 days of highmaximum temperatures during silking, and Berbeceland Eftimescu (1973) identified daily maximum tem-peratures above 32°C during tasseling and silking asparticularly damaging. Although soybeans are lessvulnerable than corn, soybeans can suffer from maxi-mum temperatures exceeding 40°C at the onset offlowering (Mederski 1983). Cotton plants abort bolls

when the temperature exceeds 40°C for more thansix hours; in rice, a temperature exceeding 30°C dur-ing anthesis causes spikelet sterility (Acock andAcock 1993). Short-term moisture deficits also cancause loss in yield, depending on the phenologicalstage during which they occur. Most often, repro-ductive stages are the most vulnerable. Excess pre-cipitation also causes problems for crops in the formof lodging, lack of aeration, and increased insect pestinfestation (Rosenzweig and Hillel 1998).

Extreme cold events affect fruit and citrus. Freezingtemperatures (below 0°C) during the winter monthsresult in catastrophic damage to the citrus crops inFlorida,Texas, and California. Extreme winter tem-peratures affect the more cold-sensitive peach cropby killing the flower buds with temperatures below–18°C and killing the peach trees with temperaturesbelow –30°C. A change in the frequency of theseextreme events as a result of climate change couldcause a contraction of the area in which these cropsare grown—if extreme events occur more frequent-ly—or an expansion of the production regionwith a less frequent occurrence of extreme coldtemperatures.

Modeling of Extreme Events inCrop Models

In most crop models, the impact of temperatureoccurs on a daily basis. The simulation of tempera-ture effects in crop models is almost always inde-pendent of the temperature of the preceding day.In other words, the impact of a warm day on growthis the same whether the day before was warm orvery cold. Many models accumulate temperaturestress days, based on high and low prescribedthreshold temperatures. Given the relative successof most crop models, this approach appears to workreasonably well.

Occasionally, crop models simulate more complexsequences of extremes. One example is the model-ing of winter kill in some crop models (e.g., CERES-Wheat), which takes into consideration the harden-ing of the crop (based on temperature accumulation

72


at some prescribed low temperature) and exposureto very low extremes (killing temperatures). If thecrop experiences a rapid oscillation between highand low minimum temperatures, winter kill canresult (e.g. , Mearns et al. 1992).

Crop models generally are less successful, however,at modeling the effects of sequences of days, such asthe effects of five consecutive days of above 35°Ctemperatures during silking in corn. The relativelysmall sample size of such events makes successfulmodeling of the physiology of this effect difficult.Being able to predict the effects of heat waves, forexample, could be more important in a climate-changed world in which the mean and variability ofday-to-day temperatures increased. Current state-of-the-art models probably underestimate the impact ofresultant extremes of climate on crop growth.Thus, although the altered climate scenarios we usecreate a greater likelihood of such heat waves, exist-ing crop models lack specific mechanisms to fullyreflect these types of events.

On the other hand, crop models have long beenconstructed with a view toward modeling theeffects of moisture stress (i.e., a deficit) on crops—and are relatively successful at doing so. Importantdifferences in the details of how moisture stress ismodeled can result in very different responses ofcrop models to the same climate change conditions,however. For example, the sensitivity of crops tomoisture stress tends to be growth stage-specific.Although most crop models use the accumulateddegree-day approach to represent the progressivephenology through a crop season, they can differsubstantially with regard to how detailed this treat-ment is. EPIC, for example, has a relatively crudephenological submodel, whereas the CERES familyof crop models tends to represent more detailedphases of phenological development. In a compari-son of the response of CERES maize and wheat withEPIC maize and wheat for climate change scenariosin the Great Plains, Mearns et al. (1999) found thatthe models projected different magnitudes and

directions of change in yield, primarily as a result ofdifferences in the phenological stage at which simu-lated crops experienced moisture stress.

Although moisture deficit (drought) has been theprincipal concern of crop modeling efforts, excessmoisture also causes significant crop damage.Some crop models (such as EPIC;Williams et al.1989) include the modeling of stress from insuffi-cient aeration, and at least one of the CROPGROmodels (SOYGRO; Boote et al. 1998) includes anexcess moisture factor. There is little information,however, on how realistically these models simulateexcess moisture effects.

Infrequent combinations of weather variables alsocan lead to unusual crop responses. For example,moisture or high humidity after physiological matu-rity has been reached, in combination with warmtemperatures, can cause grain to germinate or sproutbefore harvest. Waterlogging in combination withwarm temperatures in spring can have particularlynegative impacts on crop growth. Crop modelsoften do not simulate the effects of these interac-tions. For example, the EPIC model calculates anaeration stress factor that is based on the water con-tent of the top 1 m of soil, but this factor is notdependent on temperature.

Overall, a major direction of crop modeling istoward understanding of crop response to varyingclimate. Climate can vary in many dimensions, andnot all potential effects are captured. Moreover,most of the testing and validation of crop modelsoccurs in areas where these crops are grown.Although annual variability in climate creates a richset of weather conditions against which to evaluatethese models, climate change could produce combi-nations of climatic conditions that are only infre-quently observed where these crops currently aregrown; thus, our ability to capture these effects maybe limited. Direct comparisons of different modelsof the same crops to the same climate conditionscan produce widely varying results, and running a

73


crop model at a new site can require considerablecalibration before it can estimate realistic yields atthe site. Overall, crop models capture some of thebroad changes fairly well and on average performwell. As we begin to consider more detailed aspectsof climate and attempt to make more precise predic-tions about how to respond to very specific climateconditions, we require more detailed models, experi-mental evidence, and site-level verification so thatthe model can reproduce actual responses to vary-ing conditions.

Interannual Variability: ENSO Events

An example of an increase in climate variability on aninterannual scale would be if precipitation extremesassociated with the El Niño phenomenon becomeeven more severe than they are now. Our under-standing of the influence of ENSO—as well as otherimportant couplings of ocean currents and atmo-spheric dynamics—on climate variability in specificregions has greatly increased in the past decade. Thisdevelopment has enhanced our ability to forecastevents such as El Niño and La Niña years on a region-al basis. The general impacts on crop yield of climateregimes associated with the El Niño phenomenon arereasonably well understood and are captured effec-tively in several crop simulation models. These mod-els have been used to determine specific componentsof the climate that are responsible for yield variations.For example, a study of the impact of El Niño eventson corn yield in the US corn belt, using crop growthsimulation, indicated that water stress in July andAugust is the primary cause of lower corn yields in LaNiña years, along with a shorter period of grain fillingbecause of high temperatures (Phillips et al. 1999).The cooler temperatures and greater rainfall during ElNiño years had less pronounced effects on yield thanthe dryer, warmer La Niña years.

Studies also have been undertaken to determine thevalue of El Niño forecasting to agriculture at thefarm management and industry level. Hammer et al.(1996) compared a fixed management strategy fornitrogen fertilizer application rate and cultivar selec-tion in a wheat cropping system in Australia to a

tactical strategy that depended on the seasonal fore-cast, using the Southern Oscillation Index. An analy-sis of simulated results with the tactical strategyindicated significant increases in profits and reduc-tions in risks compared to the fixed managementstrategy. In another Australian study, phases of theSouthern Oscillation Index were used to make for-ward estimates of regional peanut production(Meinke and Hammer 1997). Because peanut yieldvaries greatly with rainfall, high variability in rainfallis a concern for peanut processors and marketers.One conclusion of this study was that the industrycould profit by using yield forecasts made three tofive months ahead of harvest to strategically adjustfor expected volume of production.

The foregoing studies were conducted to evaluatethe extent to which advanced warning of El Niño orLa Niña events, as well as other important couplingsof ocean currents and atmospheric dynamics, cansignificantly improve farm and agricultural industrymanagement decisions. As these types of analysesimprove, our ability to predict the impacts ofchanges in decadal-scale climate variability on agri-culture will be enhanced. Future studies shouldtake into account, on a regional basis, current agri-cultural systems and feasible alternative systems inthe context of current and possible future economicand policy environments. This type of approach,linked with appropriate climate scenarios, should beuseful in predicting the sensitivity of agricultural sys-tems to changes in decadal-scale climate variability.

Intra-Annual Variability (Weather)

Climate change also may cause changes in the with-in-season variability of temperature and precipita-tion, although most studies of agricultural yieldsunder future climate change scenarios haveassumed that the nature of this variation will be thesame as in the present climate. There could beimportant impacts, however, if within-season vari-ability increases. Such changes would further shiftthe probability of extreme events and also mighthave less-obvious influences on crops, such aschanging the rate of development.

74


Changes in Variability Alone

Several studies, encompassing a variety of crop simu-lation models and regions, have systematically inves-tigated the impact of changing within-season vari-ability of temperature and precipitation (e.g., Mearnset al. 1996; Riha et al. 1996). General conclusionsfrom these studies are that crop yield decreases astemperature variability increases and that the capaci-ty of the soil to store water strongly mediates cropresponse to changes in precipitation variability. Notsurprisingly, sandy soils are far more vulnerable toincreases in rainfall variability.

Riha et al. (1996), who applied EPIC corn and soy-bean models, found that increased variability of tem-perature or precipitation resulted in substantiallylower mean simulated yields; decreased variability oftemperature produced insignificantly small increasesin yield. The implications of this asymmetricresponse to variability in temperature is that relative-ly low variability in temperature is one of the majorfactors that make these Corn Belt areas so produc-tive. The year-to-year variability of yields alsoincreased with increased variability of temperatureand precipitation. The implication for climatechange is that the main risk to these regions is likelyto be the potential for increased variability.

Combined Effects of Mean and Variability Changes

Several studies (Mearns 1997; Semenov and Barrow1997) have examined the effects of climate changescenarios that included changes in the mean and thevariance of climate on simulated crop yields by alter-ing parameters of stochastic weather generators.The negative effects of climate change on cropswere exacerbated by including the effects ofchanges in climate variability.

Spatial Dimensions of Extremes

Extreme events can have spatial characteristics thathave implications for appropriately simulating theirimpact on crop yields over relatively large spatialand temporal scales. Some extreme events are com-mon when large areas are being considered butoccur infrequently in a specific location (e.g., hail).Hail causes damage that can lower yield and, in thecase of horticultural crops, lower the value of thecrop. For a given location (such as an experimentalfarm) where data for crop model development andtesting are being generated, the likelihood of hail inany given growing season may be quite low.Therefore, the impact of such a phenomenon is notconsidered in the simulation of climate impacts oncrop yields. Clearly, if the frequency of occurrenceof such a phenomenon were to increase, it wouldcause damage to a larger proportion of the croppedarea and might reach a point at which regionalyields were significantly affected.

Some extreme events are more likely to occur incertain areas rather than randomly over an areabecause of the interactions of weather with thelandscape. Examples include cold air drainage thatcreates frost pockets, gusting winds that causes lodg-ing, snow pack of variable depth that affects thewinter survival of wheat, and flooding. Some cur-rent crop models can simulate the impact of suchevents on crop growth and field operations, but themore difficult challenge is to predict the spatialextent of these events from terrain and weatherdata. This variability in the spatial dimension usual-ly is not explicitly included as input to crop models.For example, most agronomic crops cannot surviveflooding. Changes in precipitation that result inmore rain during short periods of time could lead tomore flooding; clearly, however, the likelihood andextent will depend on terrain factors, as well asflood management policies.

75


Response of Future Cropsto Extreme Events/Climate Variability

Adaptation to Temperature Extremes

Crop varieties have been developed to avoid tem-perature extremes through selection of plants thatcan complete their life cycle more quickly than tra-ditional varieties. In temperate climates, these vari-eties can be planted late and harvested early toavoid chilling and frost injury. In tropical climates,these varieties can be used to avoid periods of hightemperatures. This type of adaptation generally iswell simulated by crop models. Increases in tem-perature variability alone would be expected to fur-ther reduce the length of the growing season andtherefore require growing a shorter-season varietyor crop. For many crops, however, varieties havebeen developed that can tolerate (not just avoid)heat and cold. This type of adaptation is somewhatmore difficult to simulate because tolerance often islimited to a particular stage of development, such asgermination, emergence, flowering, or grain ripen-ing. These adaptations, though limited, can have sig-nificant impact on growth and yield. For example, aseed’s ability to germinate at temperatures that areeven a few degrees cooler in many cases can signifi-cantly increase the region in which the crop can begrown. Breeding for cold tolerance during germina-tion and heat tolerance during grain filling probablywill mitigate some impacts of increases in tempera-ture variability and some extremes. Crop simula-tion models vary in their ability to simulate thesevarietal adaptations.

Although selected varieties may tolerate tempera-ture extremes better than more traditional varietiesduring specific life stages, if the mean seasonal tem-perature moves outside the optimum range for thecrop, the yield of all varieties generally decreases sig-nificantly. In general, varieties that yield best undernonstressful environments also yield best understressful environments, though the yield is reduced(Evans 1993). This finding suggests that current

breeding strategies will be useful in selecting plantsthat can perform reasonably well even if tempera-ture variability increases.

Adaptation to Drought

Similarly, crop varieties have been developed to avoiddrought through selection of plants that can eithercomplete their life cycle more quickly than tradition-al varieties or are not in phenological stages that aresensitive to stress (such as flowering) when droughtis likely to occur. It is less clear that the ability ofplants to tolerate drought stress has been significant-ly improved in the course of plant breeding, exceptthat breeding for tolerance of high temperatures mayimprove yield under drought. The water use efficien-cy (WUE) of crops, expressed as the ratio of biomassof crop produced per unit mass of water transpired,decreases as temperature increases, assuming radia-tion and vapor density are similar.

Empirical Estimates ofCrop Yield Variability asRelated to Climate

Another approach for evaluating the impact of vari-ability on crops is to use cross-section evidence.The availability of state-level detailed climate andyield data across the United States allowed us toexamine how year-to-year and region-to-region cli-mate variation alters crop yields (Chen et al. 1999b)as part of the agriculture sector assessment.Variability influences of climate were investigatedwith USDA-NASS (1999c) Agricultural Statisticsstate-level yields and acreage harvested for 25 years(1973–1997). State-level climate data matched tothe agricultural output data were drawn fromNOAA (1999), which includes time series observa-tions for thousands of weather stations. The April-to-November average temperature for the publishedweather stations ina state was used.

The approach relies on the ability to separatechanges in variability from changes in means (detailsof which are provided in Chen et al. 1999b). The

76


basic results are in terms of elasticities—that is, howdoes a 1 percent change in temperature or precipita-tion affect yields in percentage terms? We can esti-mate how the 1 percent change in climate affectsthe mean yield and the variability of yield. Resultscan vary, depending on the functional form of theestimated equation.

Table 4.1 reports the mean yield elasticity estimatesfor a linear form and a multiplicative functional form(the specific form is commonly known as a Cobb-

Douglas production function in economics). Interms of changes in the mean, the sign on precipita-tion is positive for corn, cotton, and sorghum cropsand negative for temperature. This result indicatesthat crop yields increase with more rainfall anddecrease with higher temperatures. Elasticities forsoybean and wheat crops are mixed. Sorghum hadthe highest elasticities for rainfall and temperature.

The impact of climate change on variability is report-ed in Table 4.2. In terms of variability, the clearest

Production Corn Cotton SorghumFunction Precipitation Temperature Precipitation Temperature Precipitation TemperatureForm % Change % Change % Change

Linear 0.3273 -0.2433 0.0371 -1.5334 2.8844 -2.0866

Cobb-Douglas 1.5148 -2.9792 0.4075 -0.7476 1.8977 -2.6070

Table 4.1. Response of Mean Crop Yields to Changes in Means of Climate Variables (measured as percentage change inmean yield for a percentage change in climate variable, temperature in ˚C)

Production Soybean WheatFunction Precipitation Temperature Precipitation TemperatureForm % Change % Change

Linear -0.2068 0.0002 -0.1309 -0.5076

Cobb-Douglas 0.34640 N.S. 1.4178 -0.3721

N.S. = not significant.

Yield Corn Cotton SorghumVariability Precipitation Temperature Precipitation Temperature Precipitation TemperatureFunction % Change % Change % Change

Linear -9.7187 7.5058 -0.3028 -10.9386 0.5230 -5.3517

Cobb-Douglas -1.4461 0.8923 -0.0212 -3.5800 0.4802 -2.5633

Yield Soybean WheatVariability Precipitation Temperature Precipitation TemperatureFunction % Change % Change

Linear -0.7932 -0.2739 -2.1572 -0.1035

Cobb-Douglas 0.8194 0.0586 -1.6473 5.0875

Table 4.2. Response of Crop Yield Variability to Changes in Means of Climate Variables (measured as percentage changein yield variability for a percentage change in climate variable, temperature in ˚C)

77


results are obtained for corn, cotton, and sorghum.The results are the same for both functional formstested. Increases in rainfall decrease the variabilityof corn, cotton, and wheat yields. Corn yields arepredictably more variable with higher temperatures.Cotton and sorghum rainfall variability elasticities aresmall; a 1 percent increase in rainfall leads to a 0.5percent or less increase or decrease in yield variabili-ty. Cotton and sorghum have high temperature vari-ance elasticities: a 1 percent increase in temperatureproduces as much as an 11 percent decrease in yieldvariability. Similarly large elasticities are obtained forrainfall effects on corn and wheat yield variability.All of these results are consistent across functionalforms. Soybean elasticities are all less than one, butsign inconsistency across functional forms confoundinterpretation of these results.

We used regional estimates of climate change arisingunder the Canadian and Hadley climate model simu-lations to estimate whether, based on these climateprojections and the statistical models estimated here,

crop yield variability would increase or decreaseusing the estimated Cobb-Douglas functions. Theresults (see Table 4.3) show fairly uniform decreasesin corn and cotton yield variability, with mixedresults for other crops. Wheat yield variability tendsto decrease under the Hadley Center model andincrease under the Canadian model. Soybeanyield variability shows a uniform increase with theHadley model.

The basic conclusion is that these mean climatechanges can produce fairly large changes in variabili-ty, but these changes can be increases or decreases.This analysis considers only the potential forchanges in the mean climate conditions to changeyield variability; it does not consider how changes inclimate variability itself might affect mean yields orthe variability of yields.

Canadian Climate Change Scenario Hadley Climate Change ScenarioCorn Soybeans Cotton Wheat Sorghum Corn Soybeans Cotton Wheat Sorghum

CA -12.84 -11.81CO 34.43 -10.60GA -10.35 -6.92IL -25.71 21.28 -24.73 18.90IN -8.73 8.06 -26.31 20.30IA -36.89 33.14 -26.83 20.90KS -14.39 -0.75 -18.16 3.38LA -13.03 -7.97MN 4.01 10.60MT 32.86 -6.36MS -13.92 -7.73NE 15.30 -4.74 48.22 -16.15 -15.05 11.65 -5.57 -1.72OK 16.34 -9.27 -17.07 2.83SD -21.75 -6.94 -24.37 -19.10TX -13.21 27.86 -10.83 -8.05 2.26 -3.10

Table 4.3. Percentage Change in Crop Yield Variability for 2090, Selected States*

*Percent change in the variance of the yield expected if the change in the simulated mean climate conditions simulated were observed fora series of years.

78


Estimates of EconomicImplications of PotentialENSO Shifts

Some researchers argue that global climate changemay alter the frequency and strength of extremeevents. One marker for extreme events that hasreceived considerable public attention is the ENSOclimatic phenomenon. Timmermann et al. (1999)presented results from a climate modeling studyimplying that global climate change would alterENSO characteristics, causing

• the mean climate in the tropical Pacific regionto change toward a state corresponding to pre-sent day El Niño (warm) conditions;

• stronger inter-annual variability, with moreextreme year-to-year climate variations; and

• more skewed inter-annual variability, with strongcold events becoming more frequent.

There is much debate about these results. We usethem here to illustrate the sensitivity of agricultureto such shifts. Details of the analysis are providedby Chen et al. (1999a), a study conducted as part ofthe agriculture sector assessment. The Chen et al.analysis examines the economic implications of ashift in ENSO frequency and intensity by using thequantitative definition of the shift as developed byTimmermann et al. (1999). Specifically, Chen et al.presents estimates of the economic consequences ofshifts in ENSO frequency and strength on the worldagricultural sector.

According to Timmermann et al. (1999), the currentprobability of ENSO event occurrence (with pre-sent-day concentrations of greenhouse gases) is0.238 for the El Niño phase, 0.250 for the La Niñaphase, and 0. 512 for the neutral (non–El Niño,non–La Niña) phase. They project that the probabili-ties for these three phases, under increasing levels ofgreenhouse gases, will be 0.339, 0.310, and 0.351 forEl Niño, La Niña, and neutral, respectively. In otherwords, they project that the frequency of the ElNiño and La Niña phases would increase, and thefrequency of the neutral phase frequency would

decrease. Although they do not offer specificevidence, they argue that such a frequency changecould be expected to have strong ecological andeconomic effects.

Our analysis investigates more formally and quantita-tively whether such a change would have strongeconomic impacts on the agricultural economy.The ENSO impacts are based on a time series statisti-cal analysis of ENSO impacts on each region of theworld. Thus, we are able to consider how ENSOchanges affect agricultural production across theworld. (For details, see Chen et al. 1999a. ) ENSOevents influence regional weather and, in turn,crop yields.

Several studies have estimated the value to farmersof adapting to ENSO events. The question is, if farm-ers knew ahead of time the ENSO phase, what couldthey do to improve their economic outcome com-pared to the situation in which they operate only onlong-term average climate conditions? Results indi-cate that there is economic value to the agriculturalsector in using information on ENSO events. Interms of aggregate US and world economic welfare,the estimated benefits of using ENSO information inagricultural decision making have been in excess of$300 million annually for current ENSO frequencies.

The model experiments conducted to study theseevents involve different assumptions about the infor-mation with which farmers operate. To consider thevalue of knowing which event would occur, twofundamentally different situations were simulated inthe Agriculture Sector Model (ASM):

• Producers were assumed to be operating with-out any information concerning ENSO phaseand thus choose a crop plan (a set of crops tobe planted on their land base) that representsthe most profitable crop mix across a uniformdistribution of weather events, based on datafor the past 22 years. We refer to this analysisas the “Without use of ENSO PhaseInformation” scenario.

79


• Producers were assumed to incorporate infor-mation regarding the pending ENSO phase andthus choose a set of crops that perform besteconomically across that individual phase.Thus, crop mixes that are optimized for ENSOevents are selected across a distribution of theENSO states, as are crop mixes for other states.Initially, each ENSO event is assumed to beequally likely. We refer to this analysis as the“With use of ENSO Phase Information” scenario.

For the analysis conducted here, we assumed thatforecast information for ENSO is correct. The eco-nomic analysis assumes that all farmers make theoptimal choice, given this correct information.Failure to produce correct forecasts, failure by farm-ers to adjust planting in response to the forecast, orlack of knowledge on the part of farmers aboutresponses to ENSO would reduce the value of fore-casts. Losses could increase if forecasts are subjectto error, if farmers respond to wrong forecasts, or iffarmers do not respond unless they see evidence ofsufficient accuracy. In many ways, this analysistherefore considers the greatest potential value offorecasts, although the management choices includ-ed in the economic model used may not include allpossible management responses.

This analysis was conducted separately from anychange in mean conditions resulting from climatechange (i.e., separate from the analysis conducted inChapter 3) to isolate the effect of change in ENSOintensity and frequency. Like the Chapter 3 analysis,the scenarios are all imposed on an agriculturaleconomy as it exists in the year 2000—but as if dif-ferent ENSO phases had occurred in that year.

In addition to structuring the analysis to vary theresponse of farmers to ENSO information, a secondkey component is varied in the model experimenta-tion. In particular, three ENSO phase event probabil-ity conditions are evaluated.

• The first probability condition represents cur-rent conditions. Specifically, we assume El Niñophases occur with a probability of 0.238, La

Niña with a probability of 0.250, and neutralphases with a probability of 0.512. Within an ElNiño phase, we assume that individual cropyields for five El Niño weather years containedin our data set are each equally likely (i.e., thesame strength), with a comparable assumptionfor the four La Niña events and the 13 neutralyield states.

• The second probability condition incorporatesfrequency shifts suggested by Timmermann etal. (1999). Here, the El Niño phase occurs witha frequency of 0.339, the La Niña phase with aprobability of 0.351, and the neutral phase witha probability of 0.310. Within each of thephases, we again assume that cropping yielddata states are equally likely.

• The third probability condition considers theimpact of stronger or weaker ENSO events. Thethree event types were reclassified into five dif-ferent ENSO events: strong El Niño, weak ElNiño, neutral, weak La Niña, and strong La Niña.The frequency shifts used in this experiment arefrom Timmermann et al. (1999). To evaluateevent strength shifts, we assume that thestronger El Niño and La Niña events occur witha 10 percent higher frequency. Specifically, if the1982–1983 and 1986–1987 El Niños each occurwith a 0.20 probability within the set of five ElNiño events observed in the data set (assuming auniform distribution across the five observed ElNiños in our data set), we shift those probabili-ties to 0.25 and reduce the probabilities of thethree other El Niño years to 0.167. Similarly, thetwo strongest (in terms of yield effects) La Niñaevents have their probabilities raised from 0.25to 0.30, and the weaker two La Niñas have theirprobabilities reduced to 0.20.

The results of this analysis appear in Tables 4.4 and4.5. Table 4.4 provides estimates of aggregate eco-nomic welfare before and after the ENSO probabilityshifts. Table 4.5 contains a more disaggregatedpicture of these economic effects. The economicconsequences are evaluated for both situations

80


regarding producer decision making (i.e., ignore oruse ENSO forecasts). As in Chapter 3, the economiceffect is measured in terms of changes in welfare.The aggregate changes in Table 4.4 are the sum ofdomestic consumer, domestic producer, and foreignsurplus. Table 4.5 provides a breakdown of theseresults between producers, consumers, and foreigninterests. Four major results can be drawn fromthis work:

• First, an increase in ENSO event frequency andintensity causes significant increases in croplosses. Specifically, the welfare loss from the fre-quency shift where farmers operate withoutinformation on ENSO event probability is esti-mated to be $414 million. When both frequen-cy and strength shifts are considered the wel-fare loss increases to $1,008 million. This figureis about 5 percent of typical US agricultural netincome, or about 0.15 percent of total foodexpenditures in the United States. If thestrength shift were more substantial than theone assumed here, it could have substantiallylarger effects.

• Second, there is considerable value to farmers inoperating with better information about ENSOevents, and the value increases if the frequencyand intensity of these events increase. Thevalue of ENSO forecasts under current ENSOfrequency and strength is estimated at $453 mil-lion. This value is very similar to previous work,as estimated by Solow et al. (1998). The value ofENSO forecasts increases to $544 million withthe frequency shift and to $556 million if bothfrequency and intensity shift.

• Third, the additional damage from these moreintense and frequent ENSO shifts is only partial-ly offset by better forecasting. The forecastinggains are greater with a more frequent andstronger ENSO than under the current ENSO fre-quency and strength, but the gains do not offsetthe additional losses from the ENSO shifts. Theuse of ENSO forecasts mitigates some of thenegative economic effects of the shift.

Without use With use Gain fromof ENSO of ENSO of ENSO use

information information information

($ millions)

Current – – 453probabilities

Phase frequency -414 -323 544shift

Phase frequency -1008 -905 556and strength shift

Table 4.4. Change in Aggregate Economic Welfare WithShifts in ENSO Frequencies and Strength

Specifically, the figures in Table 4. 4, column 2show an increase in damage from the currentENSO event frequency and intensity of $323and $905 million, respectively, when ENSO infor-mation is used compared to the figures in col-umn 1 which show an increase in damages of$414 and $1,008 million, respectively whenENSO information is not used.

• Fourth, there are winners and losers from changesin ENSO frequency and intensity (Table 4.5).Specifically, the total welfare loss from the shift inENSO frequencies results in domestic producerand foreign country welfare losses but gains todomestic consumers. Most of these welfare lossesoccur in the foreign markets. These differencesacross groups arise from changes in US and worldprices for the traded commodities. For example,for the commodities evaluated here, there areprice declines as a result of slight increases inworldwide production when phase frequencyshifts. Price declines result in losses to producersand exporting countries but gains to consumers.

The referenced ENSO case of Chen et al. (1999b)that is summarized here confirms the analysis byTimmermann et al. (1999) that climate change-induced shifts in ENSO frequency will haveeconomic consequences. We further find that

81


Phase frequency Phase frequencyshift and strength shift

($ millions)

US Producers -307 -321

US Consumers 591 326

Foreign interests -607 -910

Total -323 -905

Table 4.5. Change in Welfare, by Component, With Use ofENSO Information ($ millions)

those consequences involve changes in agriculturalprices and welfare. Prices and welfare fall, but theseeffects are reduced as producers anticipate andreact to forthcoming El Niño and La Niña events.The projected changes of Timmermann et al.(1999) can be partly offset by producer reactions toENSO information. Again, we caution that there ismuch uncertainty and controversy with regard towhether or how global climate change would affectENSO. Our intent here was simply to consider theENSO shifts as a “what if” scenario.

Implications

The importance of extreme events in the context ofthe impacts of climatic change and variability onagriculture has received increased attention inrecent years. Extreme events and climate variabilityhave documented impacts on agriculture. Farmershave many financial mechanisms with which toaddress variability and extreme events, ranging fromcrop insurance and savings to forward contracting,and an emerging market for weather derivatives.They also can change production practices to makethemselves less vulnerable to variability. Thesemechanisms, however, cannot eliminate the realeffects of variability on costs. Moreover, in the caseof mechanisms such as insurance and forward mar-

kets, the costs of variability are merely pooled orspread, not eliminated or reduced. As demonstratedby analysis of possible changes in ENSO events, bet-ter forecasting can reduce the effects of increasedvariability, but it cannot eliminate all the additionalcosts. The greatest limitation in our understandingof the impacts of variability on agriculture is ourvery limited ability to predict how variability willchange. Our knowledge regarding possible shifts inthe frequencies of extreme events with a new cli-mate regime is limited. Work also remains to bedone to incorporate current information on changesin variability, as represented in climate models, intomethods for assessing impacts on agriculture.

Investigators must distinguish among the relevanttime scales and spatial scales of extreme eventsimportant to agriculture. In general, crop modelsadequately handle extreme events that are longerthan their time scale of operation. For example,crop models operating on a daily time scale can sim-ulate fairly well the effects of a seasonal drought(lasting a month or more), but they will have moredifficulty properly simulating responses to veryshort-term extreme events, such as daily tempera-ture or precipitation extremes. Another difficulty forcrop models is properly representing compositeextreme events such as a series of days with hightemperatures combined with precipitationextremes. Therefore, in considering the possibleeffects of extremes and climate variability on cropsfrom a policy point of view, extreme caution mustbe exercised in interpreting the analyses of climatemodels regarding what types of changes in extremesmight occur in the future and in interpreting theresponses of crop models to extreme climateevents. Research in these areas is likely to continueto develop rapidly, however.

Although predicting the future climate with greataccuracy is impossible, the analysis presented in thischapter provides an indication of the more-favorableand less-favorable future climates. For corn, a wetterand cooler climate is the more favorable; a hotterand drier climate is the less favorable than current

82


climate, resulting in decreased yield and greateryear-to-year yield variability A wetter and warmerclimate would result in the greatest decrease inyear-to-year yield variability; conversely, a drier andcooler climate would result in increased year-to-yearyield variability. Sorghum year-to-year yield varia-bility would be reduced most by a drier andwarmer climate.

The US consumer wins in the case of a future cli-mate with a change in the ENSO phase frequencyand an ENSO phase frequency shift with a change inthe strength of the phases. Agricultural producers,on the other hand, are losers as a result of lowerprices for their crops. Foreign interests also lose.The United States generally is a winner when pro-ducers and consumers are considered.

This analysis does not include all of the potentialeffects of changes in climate; added together, thesechanges may have more profound effects on agricul-tural production than changes to the ENSO phasefrequency and phase frequency shift. Again, theENSO shifts are based on a single study, and muchuncertainty remains about how global climatechange would affect ENSO.

In summary, this chapter documents many of theways in which variability can affect crops and howit may change in the future. The difference, in termsof agricultural productivity, between a moderate andeven climate and one of extremes of hot and cold,wet and dry, can be stark. Challenges also remainfor the agricultural assessment community in evalu-ating the impacts of variability changes.

83

Agriculture and theEnvironment: Interactionswith Climate

Chapter 5

Introduction

Many previous assessments of the potential impactsof climate change and variability on agriculture havefocused solely on agricultural production, foodprices, and farm incomes. Our nation’s interest inagriculture is broader than these issues, however.People in rural and urban areas value agriculturalland as open space and a source of countrysideamenities. Agricultural land is an important habitatfor many remaining wildlife species. Agriculture alsois a source of negative environmental impacts insome areas. Nutrients, pesticides, pathogens, salts,and eroded soils are leading causes of water qualityproblems in many parts of the United States. Inmany parts of the western United States, irrigatedagriculture is a major user of scarce water resources.Our nation also has an interest in agriculturebecause of its potential to serve as a sink for green-house gases.

Agriculture has many environmental impacts—someoccurring on the farm and others off the farm. Forexample, cultivation of crops increases the exposureof the land to the forces of wind and water erosion,which has on-farm and off-farm effects. Soil erosionreduces on-farm soil productivity by depleting soilnutrients and altering the structure of the soil inways that reduce the soil’s capacity to filter andhold water. Farmers bear the costs of lower soil pro-ductivity in the form of diminished production andsales. Thus, farmers can make economic decisionsabout how much of this productivity loss to avoid,given the costs of doing so. That is, farmers have adirect financial stake in the on-farm impacts of soilerosion and other environmental problems.

Off-farm environmental impacts of agricultural pro-duction, such as surface water sedimentation fromeroded soils, are an entirely different matter. Theseimpacts generally do not show up on any farmer’sbottom line. Farmers may be as concerned about

the environment as anyone else (or even more con-cerned), but expecting them to voluntarily reducetheir own incomes for the sake of protecting theenvironment is asking a great deal—particularlywhen they have no reason to believe that their fel-low farmers will follow suit or when the linksbetween their individual actions and water qualityare poorly understood. Off-farm environmentalimpacts have been a motivation for major federalconservation programs such as the ConservationReserve Program (CRP), Environmental QualityIncentives Program (EQIP), and the Wetland ReserveProgram (WRP), as well as voluntary conservationcompliance that is based on farmers’ assessment ofthe payments the programs offer for participation.

For these reasons, the off-farm environmentalimpacts of climate change could be more importantfrom a public policy perspective than the impactson agricultural production, food prices, or farmincomes. Farmers (as well as seed companies, fertil-izer distributors, and other firms that sell productsand services to farmers) will have strong financialincentives to adapt to climate change by minimizingnegative impacts on production and exploiting posi-tive impacts. For off-site effects of agricultural prac-tices where environmental and conservation “goods”are not priced in markets, federal, state, and localgovernments will have to decide if environmentalregulations must be strengthened if climate changeworsens environmental problems.

Consideration of all potential agriculture-environ-ment interactions and how they might be affectedby climate change was beyond the scope of the agri-culture assessment. Much more research and modeldevelopment is needed on these interactions beforethe capacity exists to quantitatively and completelyassess them. Whereas, relatively well-developed dataon current conditions exist for market impacts, forenvironmental concerns we often have very incom-plete information on the extent of current problems

84


and their causes. We considered some specific casestudies that help illuminate the environmental risksthat climate change may present. In most cases, wesought to produce new, quantitative results withmodels that allowed us to simulate results using theHadley and Canadian climate scenarios. The hazardwith case studies is that the cases may not be repre-sentative of what might happen elsewhere or underdifferent climate conditions. Indeed, many environ-mental problems depend on very specific and pre-cise dimensions of climate. Erosion and runoff arehighly nonlinear with rainfall intensity. There maybe little or no erosion with moderate storm events;most erosion may occur during one or two extreme-ly heavy and intense storms. Similarly, waterrecharge and water supply are highly dependent onthe specific character of regional rain events.

We considered the relationship between agricultureand water quality in the Chesapeake Bay region(Abler, Shortle, and Carmichael 2000), potentialchanges in pesticide use that might occur as a resultof changing climate (Chen and McCarl 2000), theinteraction of urban and agriculture demand forgroundwater in the Edwards’ aquifer region near SanAntonio,Texas (Chen, Gillig, and McCarl 2000), andthe potential impacts of climate change on soils(Paul and Kimble 2000). These environmental issuesare important in their own right. In addition, theseenvironmental and conservation concerns are quitedifferent from a physical, biological, economic, andpolicy perspective; they are illustrative of the rangeof environmental and conservation issues thatwould be affected by climate change.

The Chesapeake Bay has been seriously degradedover the years by agricultural production and otherhuman activities. In the following section, we ana-lyze the potential impacts of climate change onnutrient runoff into the Chesapeake Bay, based onnew results from an integrated economic-environ-mental model of corn production in the Bay region.Nutrient runoff during heavy rainfall is the primarymode by which corn production affects the Bay.This dynamic is a case of an environmental external-ity related to agricultural production. There are no

direct market incentives for farmers to controlrunoff of residues into the Bay. The Bay is an open-access, public resource.

In the next section, we examine the interactionbetween climate change and pesticide use. Weaddress how changes in climate might alter pestpopulations and the costs of pest treatment.Decisions to control the effects of pests are internalto the farmer’s decision making, and incentives tocontrol pests are market-driven. Pesticide use raisesmany environmental concerns—such as residues onfood, contamination of water, and consequences forwildlife. We therefore consider here the extent towhich climate change could change the use of pesti-cides. We do not attempt to relate the change inpesticide use to particular changes in exposure ofpeople or wildlife to these chemicals, nor do weconsider all chemicals used on all crops. That wouldbe an immense task. Attempts to estimate the rela-tionship between current levels of chemical use andexposure levels are very uncertain. Even withknown levels of exposure, the health and ecosystemeffects are very uncertain. Nevertheless, we believethe results suggest the possible direction of the envi-ronmental effect.

In the next section we consider intersectoral waterreallocation in the water scarce region around SanAntonio,Texas. Groundwater is a resource that oftenis not well-managed, although recognition thatuncontrolled access to groundwater will lead toexcessive depletion has increasingly led states in aridregions of the country, which rely on groundwater, tostep in and manage withdrawals. Drawdown ofwater levels in aquifers can have effects on wetlandsand water levels in rivers and lakes, thereby threaten-ing wildlife and recreation as well as increasing thecost of pumping water for urban and agriculturalusers. Climate change can affect the demand forwater and the recharge rate of the aquifer.

In the final section of this chapter, we examine inter-actions between climate change and soil properties.We discuss many interactions of soil and climate,including the relationship between soil organic

85

Chapter 5: Agriculture and the Environment: Interactions with Climate

matter and climate. Soil organic matter consistslargely of carbon; hence, the effects of climate onsoil organic matter are a feedback into the climatesystem. Increases in soil organic matter reflectremoval of carbon dioxide by plants and incorpora-tion of the residue into the soil. Decomposition oforganic matter, on the other hand, releases carboninto the atmosphere. The rate of decomposition ver-sus incorporation of organic matter determineswhether the soil of a given area is a net source ornet sink for carbon. Increases in organic matteritself improve soil quality, provide a source of nutri-ents, and thus can improve crop productivity. Theprincipal goal of this section, however, is to discussthe many ways that climate affects soil and hencethe productivity and sustainability of agriculturalproduction. Soil quality and crop productivity arelargely on-site issues; farmers normally would haveincentives to maintain soil quality in an economicmanner. Considerable uncertainty about the crop-ping practices that best maintain soil and the long-term effect of existing practices remain. In view ofthis lack of information, there is a need for data,technical assistance, testing, and monitoring so thatfarmers can better manage their soils toward theirown interest of maintaining the long-term profitabili-ty of their farm.

Agriculture and theChesapeake Bay

In this section we examine agriculture in theChesapeake Bay region as it exists today and as itmight evolve in the first few decades of the 21stcentury. We also examine the potential impacts ofclimate change on agriculture and water quality inthe Chesapeake region, based on new results froman integrated economic-environmental model ofcorn production.

We begin with an overview of the Chesapeakeregion; then we consider agriculture as it currentlyexists in the region, sketch a possible future for agri-culture in the region, and identify how climate maychange in the region. With this background, we

then briefly describe the simulation model wedeveloped and used to investigate the impacts of cli-mate, present the principal results, and offer someimplications for current decisions.

Introduction

The 64,000-square-mile Chesapeake Bay watershedis the largest estuary in the United States(Chesapeake Bay Program 1999). The watershedincludes parts of New York, Pennsylvania,WestVirginia, Delaware, Maryland, and Virginia, as wellas the entire District of Columbia. More than15 million people currently live in the ChesapeakeBay watershed.

The Chesapeake Bay is one of the nation’s mostvaluable natural resources. It is a major source ofseafood, particularly the highly valued blue crab andstriped bass. It also is a major recreational area;boating, camping, crabbing, fishing, hunting, andswimming are all very popular and economicallyimportant activities. The Chesapeake and its sur-rounding watersheds provide a summer or winterhome for many birds—including tundra swans,Canada geese, bald eagles, ospreys, and a wide vari-ety of ducks. In total, the Chesapeake region ishome to more than 3,000 species of plants and ani-mals (Chesapeake Bay Program 1999).

Human activity within the Chesapeake Bay water-shed during the past three centuries has had seriousimpacts on this ecologically rich area. Soil erosionand nutrient runoff from crop and livestock produc-tion have played major roles in the decline of theChesapeake. The Chesapeake Bay Program (1997)estimates that agriculture accounts for about 39 per-cent of nitrogen loadings and about 49 percent ofphosphorus loadings in the Bay. Thus, agriculture isthe single largest contributor to nutrient pollution inthe Chesapeake. Other contributors include sourcessuch as wastewater, forests, urban areas, and atmo-spheric deposition.

Agriculture within the Chesapeake Bay region also isa major source of pollution, compared to agriculturein other parts of the country. Of 2,105 watersheds

86


(defined at the 8-digit hydrologic unit code level) inthe 48 contiguous states, watersheds in southernNew York, northern Pennsylvania, southeasternPennsylvania, western Maryland, and westernVirginia rank in the top 10 percent in terms ofmanure nitrogen runoff, manure nitrogen leaching,manure nitrogen loadings from confined livestockoperations, and soil loss from water erosion (Kellogget al. 1997). Watersheds in southeasternPennsylvania also rank in the top 10 percent interms of nitrogen loadings from commercial fertiliz-er applications (Kellogg et al. 1997).

Agriculture in the ChesapeakeBay Region

Agriculture in the Chesapeake region is character-ized by smaller farms and a wider range of cropsand livestock products than in many other parts ofthe United States. Average farm size in the region isless than 200 acres, compared with more than 500acres for the rest of the country (USDA NationalAgricultural Statistics Service 1999b). Poultry andhog operations within the region tend to be as largeand intensive, however, as those in other parts of thecountry (USDA National Agricultural StatisticsService 1999b).

Major sources of farm cash receipts within theChesapeake region include dairy products, poultry,eggs, hogs, mushrooms, other vegetable and nurseryproducts, apples, and peaches. There also is signifi-cant production of corn, soybeans, and hay; thesecommodities, however, are mainly consumed on thefarm as livestock feed rather than sold.

Crop production in the Chesapeake Bay region isoverwhelming rainfed rather than irrigated. Lessthan 3 percent of crop acreage in the region is irri-gated, compared with about 13 percent in the restof the United States (USDA National AgriculturalStatistics Service 1999b).

Forests are the largest category of land use in theChesapeake region, accounting for about 60 percentof total land use. Agriculture is the second-largestcategory, accounting for nearly 30 percent of total

land use. Urban areas, residential areas, wetlands,and other land uses account for the remainder.Production agriculture accounts for about 2 percentof the total labor force in the Chesapeake region.

Future Agriculture in the Chesapeake Bay Region

Agriculture in the Chesapeake region—like US agri-culture as a whole—has changed radically duringthe past century, and there are few reasons toexpect this rapid pace of change to slow. Tractorsand other farm machinery have virtually eliminatedthe use of draft animals and enable a single farmerto cultivate tracts of land that are orders of magni-tude larger than a century ago. The introduction ofsynthetic organic pesticides in the 1940s revolution-ized the control of weeds and insects. Similarly,there has been tremendous growth in the use ofmanufactured fertilizers and hybrid seeds. Farmershave become highly specialized with regard to thelivestock products and crops they produce; theyalso have become much more dependent on pur-chased inputs. Crops that were virtually unheard-of100 years ago, such as soybeans, are of major impor-tance today. As agricultural productivity has risenand as real (inflation-adjusted) prices of farm com-modities have fallen, substantial acreage in theChesapeake region has been taken out of agricultureand either returned to forest or converted tourban uses.

The basic science of biotechnology is progressingvery rapidly; tens of millions of crop acres in theUnited States already have been planted with geneti-cally modified organisms (GMOs). Plant biotechnol-ogy has the potential to develop crops with signifi-cantly greater resistance to many pests and greaterresilience during periods of temperature and precip-itation extremes—and even cereal varieties that fixatmospheric nitrogen in the same manner aslegumes. Work also is underway to engineer pestvectors into beneficial insects as part of integratedpest management (IPM) strategies. GMOs with tol-erance to specific herbicides also are being devel-oped and released, and concerns have been raisedthat these new crops may promote herbicide usage.

87


Animal biotechnology has the potential to yield live-stock that process feed more efficiently, leading toreduced feeding requirements and fewer nutrients inanimal wastes. Feed also may be genetically modi-fied to reduce nutrients in livestock wastes.Genetically engineered vaccines and drugs couldsignificantly reduce livestock mortality andincrease yields.

Another development already underway is precisionagriculture, which uses remote-sensing, computer,and information technologies to achieve very pre-cise control over agricultural input applications(e.g., chemicals, fertilizers, seeds). Precision agricul-ture has the potential to significantly increase agri-cultural productivity by giving farmers much greatercontrol over microclimates and within-field varia-tions in soil conditions, nutrients, and pest popula-tions (National Research Council 1997). This tech-nology may be accompanied by significantimprovements in computer-based expert systems toaid farmers with production decision making(Plucknett and Winkelmann, 1996). The environ-ment could benefit insofar as precision agriculturepermits fertilizers and pesticides to be applied moreprecisely where they are needed at the times of theyear when they are needed.

Future population increases in the Chesapeake Bayregion may lead to additional conversion of farm-land to residential and commercial uses. Futureincreases in per capita income could manifestthemselves in larger homes and lot sizes and thusmore residential land use—a tendency that hasbecome evident over the past 30–40 years. Studiesof land use confirm that population and per capitaincome are important determinants of conversion offarmland and forestland to urban uses (Hardie andParks 1997; Bradshaw and Muller 1998). Probablefutures for the spatial pattern of development with-in the Chesapeake region are more difficult to assessthan an overall tendency toward urbanization. Onepossible future involves a “fill in” of areas betweenexisting major urban centers, such as the areabetween Baltimore and Washington, D.C. (Bockstaeland Bell 1998). Increases in per capita income alsoincrease the demand for environmental quality.

Economic conditions facing agriculture in theChesapeake Bay region can be expected to continueto change for many other reasons, including changesin global agricultural commodity prices and stricterenvironmental regulations toward agriculture (Ablerand Shortle 2000). There probably will be fewercommercial crop and livestock farms within theregion in the future than there are today, and someof the region’s agricultural production will shift toother regions and countries (Abler and Shortle2000). There may be growth in “weekend,”“hobby,”and other noncommercial farms within the region.Such farms, however, account for only a small frac-tion of total agricultural output. Production perfarm and yields per acre on the remaining commer-cial farms within the Chesapeake Bay region also arelikely to be significantly higher than they are today.

Future Climate in the Chesapeake Bay Region

Climate in the Chesapeake region also is likely tochange. Climate projections for the region differ sig-nificantly, however, depending on the climate modelused. Projections that use the Hadley and GENESISmodels for the Mid-Atlantic region—which includesthe Chesapeake Bay region—suggest increases inaverage daily minimum and maximum temperaturesand increases in average annual precipitation (Polskyet al. 2001). Projections that use the Canadianmodel, however, suggest a much warmer and drierclimate than the Hadley or GENESIS model (Polskyet al. 2001).

Predicting whether extreme weather events (such asdroughts, floods, heat waves, hurricanes, ice storms,blizzards, and extreme cold spells) will occur moreor less often is very difficult. Current trends for themid-Atlantic region suggest a change toward fewerextreme temperatures but more-frequent severethunderstorms and severe winter coastal storms(Yarnal 1999). Whether these trends will continueis unclear.

88


Simulation Model of Climate Change,Agriculture, and Water Quality

To assess the potential impacts of climate change onagriculture’s contribution to water quality problemsin the Chesapeake Bay region, we constructed a sim-ulation model of corn production and nutrient load-ings in six watersheds within the region. The modelcontains economic and environmental modules thatlink climate to productivity, production decisions bycorn farmers, and nonpoint pollution loadings. Cornis an important crop to study because of its impor-tance to the region’s agriculture and because it is amajor source of nutrient pollution; it accounts formore than half of all nonpoint nitrogen loadings.Corn is the most nitrogen-intensive of all majorcrops currently grown within the region. In addi-tion, livestock farms within the region often disposeof manure on corn land.

The economic module projects the choices thatfarmers make with respect to the amount of landdevoted to corn and the usage of fertilizer and otherinputs into corn production. Precipitation, tempera-ture, and atmospheric carbon dioxide levels affectthe uptake of fertilizer and the productivity of landused in corn production. The economic module isbased on previous economic models we construct-ed to examine nonpoint agricultural pollution (Ablerand Shortle 1995; Shortle and Abler 1997). We cali-brated the module to the six watersheds with avail-able state-, county-, and watershed-level data on farmproduction, land use, nutrient applications, andother inputs. Details on the model and the resultsappear in Abler, Shortle, and Carmichael (2000).

Using the farmer decisions projected by the eco-nomic module, the environmental module projectsnitrogen loadings from corn production withineach of the six watersheds. The environmentalmodule is based on the Generalized WatershedLoading Functions (GWLF) model (Haith et al.1992). The GWLF model uses precipitation andtemperature data, combined with data on land use,topography, and soil types, to estimate water runoffand pollutant concentrations flowing into streamsfrom several types of land use, including corn. The

GWLF model was calibrated to field conditions inthe six watersheds by Chang, Evans, and Easterling(2000). The GWLF model projects nitrogen andphosphorous loadings. We found, however, thatphosphorous loadings from corn production werevery highly correlated with nitrogen loadings fromcorn production in each watershed. Thus, we focushere on nitrogen loadings.

The locations of the six watersheds (ClearfieldCreek, Conodoquinet Creek, Juniata/Raystown River,Pequea Creek, Pine Creek, and Spring Creek) withinthe Chesapeake Bay region are shown in Figure 5.1.Statistics on land cover and land use for the water-sheds are provided in Table 5.1, and statistics onnitrogen loadings are provided in Table 5.2. Thewatersheds are diverse in terms of the percentage ofland devoted to agriculture as a whole and to corn.They are similar, however, in that agricultureaccounts for the vast majority of nonpoint nitrogenloadings. Corn alone accounts for more than half ofall nonpoint nitrogen loadings in each watershed.Across the six watersheds, corn accounts for anaverage of 69 percent of all nonpoint loadings.

In the simulation model, the weather is random inthe sense that farmers do not know what tempera-ture and precipitation during the growing seasonwill turn out to be. Therefore they must make plant-ing and production decisions on the basis of average(more precisely, expected) temperature and precipi-tation patterns. Farmers in the model are aware ofclimate change, however, in the sense that theyknow how average temperature and precipitationpatterns are evolving over time in their area.

We consider three climate scenarios in the model.The first is present-day climate (temperature andprecipitation averages for the 1965–1994 period),which serves to establish a reference point. The sec-ond climate scenario is based on projections fromthe Hadley climate model for the 2025–2034 period.The Hadley model suggests increases in averagedaily minimum and maximum temperatures andincreases in average annual precipitation (Polsky etal. 2001). The third climate scenario is based on pro-jections from the Canadian model for the

89


Chesapeake Bay Region and Study Watersheds

Figure 5.1: Sources: Chesapeake Bay Program (1997) andChang, Evans, and Easterling (1999).

Land Area Percentage of Watershed (1,000 acres) Total Land Area

Total All Agriculture Corn All Agriculture Corn

Clearfield Creek 240 33 8 14 3Conodoquinet 321 199 24 62 7Juniata/ Raystown 458 154 40 34 9Pequea Creek 98 70 9 71 9Pine Creek 629 66 27 11 4Spring Creek 44 21 13 49 31All Six Watersheds 1,789 543 120 30 7

Table 5.1. Land Cover/Use in the Six Study Watersheds

Nonpoint Inorganic Nitrogen Loadings Percentage of Total NonpointWatershed (1,000 pounds) Nitrogen Loadings

Total All Agriculture Corn All Agriculture Corn

Clearfield Creek 2,057 1,852 1,453 90 71Conodoquinet 5,102 5,023 2,914 98 57Juniata/ Raystown 4,359 4,261 3,661 98 84Pequea Creek 1,335 1,327 940 99 70Pine Creek 1,623 1,317 981 81 60Spring Creek 709 697 587 98 83All Six Watersheds 15,192 14,481 10,536 95 69

Note: Figures for six watersheds may not add to column totals in the row because of rounding.Sources: Chang, Evans, and Easterling (1999) and authors’ own calculations.

Table 5.2. Nonpoint Nitrogen Loadings in the Six Study Watersheds

2025–2034 period. The Canadian model suggests amuch warmer and drier climate than the Hadleymodel (Polsky et al. 2001). Because the weather israndom in the model, the climate scenarios involvechanges in the means and variances of the model’stemperature and precipitation variables.

We also consider two future baseline scenarios inthe model. These scenarios describe what mighthappen to corn production in the Chesapeake Bayregion in coming decades independent of climatechange. Shortle, Abler, and Fisher (1999) discussprocedures to use in constructing future baselinescenarios. These procedures develop scenarios thatestablish probable upper and lower bounds oneconomic and environmental impacts. Althoughpinpointing the exact magnitude of an impact is

90


impossible, we can say that the impact is likely tofall within a certain range.

With an eye toward establishing probable upper andlower bounds on changes in nitrogen loadings fromcorn production in the Chesapeake region betweennow and the 2025–2034 period, we consider twofuture baseline scenarios. These two scenarios—acontinuation of the status quo (SQ) and an “environ-mentally friendly,” smaller agriculture (EFS)—aredetailed in Table 5.3. The EFS scenario is much moreprobable than any scenario approximating a contin-uation of the status quo, but both scenarios areneeded to establish probable bounds on climatechange impacts. The EFS scenario establishes alower bound on any increase in nitrogen loadingsresulting from climate change because biotechnolo-gy and precision agriculture help to minimize load-ings from any given level of agricultural production.In addition, stricter environmental regulations in theEFS scenario lead farmers to adopt less nitrogen-intensive corn production practices. None of thesefactors occurs in the SQ scenario; therefore, the SQscenario establishes an upper bound on increases innitrogen loadings resulting from climate change.

Scenario Scenario Detailsa

"Environmentally Friendly," • Significant decline in corn production in Chesapeake Bay regionSmaller Agriculture (EFS)

• Significant decrease in number of commercial corn farms in region

• Substantial increase in agricultural productivity via biotechnology and precision agriculture

• Major increase in corn production per farm and corn yields on remaining commercial farms

• Significant decrease in agriculture’s sensitivity to climate variability through biotechnology and precision agriculture

• Continued conversion of agricultural land to urban uses, with someabandonment of unprofitable agricultural land

• Significant decrease in commercial fertilizer and pesticide usage through biotechnology

• Less runoff and leaching of agricultural nutrients and pesticides via precision agriculture

• Stricter environmental regulations facing agriculture

Status Quo (SQ) • Agriculture as it exists today in the Chesapeake Bay region

Table 5.3. Baseline Agricultural Scenarios for the 2025–2034 Period

aNote: For greater detail, see Abler, Shortle, and Carmichael (2000).

With three climate scenarios and two future baselinescenarios, we must analyze a total of six scenariocombinations. Because the weather is random, weanalyzed each combination by using a Monte Carloexperiment in which we took 100,000 randomdraws for the model’s temperature and precipitationvariables. Each of these random draws could be con-sidered an alternative possible growing season with-in a particular climate scenario. The results are themeans of 100,000 random draws.

Results from the Simulation Model

Results from the simulation model for each water-shed and for the six watersheds as a whole are pre-sented in Table 5.4. Results for the six watershedsas a whole also are illustrated in Figure 5.2. Theresults for the SQ baseline scenario suggest that cli-mate change could lead to significant increases innitrogen loadings from corn production. For the sixwatersheds as a whole, nitrogen loadings are morethan 3 million pounds higher in the Hadley climatescenario than with the present-day climate—anincrease of 31 percent. In the Canadian climate sce-nario, nitrogen loadings for the six watersheds as a

91


Nitrogen Loadings from Corn Production

whole are nearly 2 millions pounds higher than withthe present-day climate—an increase of 17 percent.

The results for the EFS baseline scenario, on theother hand, suggest that climate change would leadto more modest increases in nitrogen loadings fromcorn production. For the six watersheds as a whole,nitrogen loadings are about 400,000 pounds higherin the Hadley climate model scenario than with thepresent-day climate, an increase of 19 percent. Inthe Canadian climate model, nitrogen loadings forthe six watersheds as a whole are about 200,000pounds higher than with the present-day climate—an increase of only 8 percent.

The results for the SQ and EFS baseline scenarios dif-fer significantly in part because the EFS scenariostarts from a much lower level than the SQ scenario.Under the present-day climate, total loadings for thesix watersheds as a whole are about 2 millionpounds in the EFS scenario, as opposed to more than10 million pounds in the SQ scenario. Many forcescause fertilizer usage and environmental impacts tobe much lower in the EFS scenario than in the SQscenario. The results for the SQ and EFS scenariosalso differ because agriculture is less climate-sensitivein the EFS scenario than in the SQ scenario.

Watershed Baseline/Climate Scenario Combination

SQ EFSPresent- Hadley Canadian Present- Hadley Canadian

Day Climate Climate Day Climate ClimateClimate Model Model Climate Model Model

Clearfield Creek 1,453 1,913 1,710 313 374 340Conodoquinet 2,914 3,835 3,426 629 750 681Juniata/ Raystown 3,661 4,803 4,294 788 938 855Pequea Creek 940 1,242 1,108 203 243 221Pine Creek 981 1,285 1,150 211 251 229Spring Creek 587 771 689 126 151 137All Six Watersheds 10,536 13,848 12,377 2,270 2,706 2,462

Table 5.4. Nitrogen Loadings from Corn Production under Alternative Scenarios (1,000 pounds)

Note: Figures for each scenario are averages across 100,000 random samples. Figures for six watersheds may not add to column totals inthe row because of rounding.

The SQ and EFS baseline scenarios are in agreement,however, regarding the direction of change in nitro-gen loadings from corn production. In both scenar-ios, climate change leads to increases in loadings. Inpercentage terms, the increase for the six water-sheds as a whole ranges from 8 percent (EFSscenario/Canadian climate model) to 31 percent(SQ scenario/Hadley climate model).

Loadings increase because climate change makescorn production in the six watersheds more eco-nomically attractive. As corn production becomeseconomically more attractive, farmers devote more

Figure 5.2: Nitrogen loadings from corn production for the sixwatersheds are presented here for the continuation of statusquo (SQ) baseline scenario and the environmentally friendly(EFS) baseline scenario.

92


land to corn compared with the baseline (no cli-mate change) and increase their use of inputs peracre to raise yields. As they take these steps, theirusage of nitrogen fertilizer increases—leading toincreases in nitrogen loadings.

Leaving aside for a moment the economic respons-es by farmers, the increase in growth potential ofcorn because of climate change in and of itselfleads to greater uptake of nitrogen by crops, leavingless nitrogen to run off into surface waters or leachinto groundwater. To take full economic advantageof the growth potential of crops, however, farmersapply more nitrogen fertilizer. These increasednitrogen applications result in overall greater nitro-gen loadings.

In the Hadley climate model scenarios, nitrogenloadings also increase because average precipitationduring the growing season increases—washingmore nutrients into streams, rivers, and groundwa-ter. In the Canadian climate model scenarios, on theother hand, average precipitation during the grow-ing season falls. Nevertheless, because of theincreased nitrogen applications by farmers inresponse to the yield effects of climate change,nitrogen loadings from corn production stillincrease in the Canadian climate model scenarios.

Pesticide Use and Climate

An open issue in the climate change arena involvesthe following question: How might changes in cli-mate alter pest populations and, in turn, the costs ofpest treatment? We use an approach that is similarto that employed by Mendelsohn, Nordhaus, andShaw (1994). In that study, the authors used geo-graphic variation to consider the implications of cli-mate for land values and to draw conclusions fromthe statistical model for projected changes of cli-mate in the future. We statistically estimate a rela-tionship between pesticide costs and climatic condi-tions. We use the estimated statistical model to

consider the impact of future climate change onpesticide costs. We estimated a panel data versionof the production function laid out by Just and Pope(1978) that allowed us to estimate average pesticidecosts and the variance of pesticide costs. For moredetail, see Chen and McCarl (2000).

State-level pesticide usage for corn, wheat, cotton,soybeans, and potatoes was drawn from USDA(1991-1997; 1999a). These data give statistical sur-vey-based averages for various insecticide, herbicide,and fungicide compounds by crop and year. Thestates for which data were available vary by crop;they are listed in Table 5.5. We computed a totalcost of pesticides by multiplying the pesticide useby category by annual prices from the USDA (1997).We use aggregate total cost data to reflect pesticidesubstitution as climate and pesticide prices vary.Climate data were drawn from the US NationalOceanographic and Atmospheric Administration(NOAA, 1999). Rainfall data were cropping yeartotals, to reflect not only cropping season supplybut also water stored in soil or irrigation deliverysystems. Temperature data were the March–September average for all crops except for winterwheat areas. In winter wheat areas, we usedOctober–April temperature data. We derived state-level temperature and rainfall data by averaging alldata for weather stations in a region.

Crop State

Corn IL, IN, IA, MI, MN, MO, NE, OH, SD, WI

Cotton AZ, AR, CA, LA, MS, TX

Soybeans AR, IL, IN, IA, LA, MN, MS, MO, NE, OH, TN

Wheat CO, ID, KS, MN, MT, ND, NE, OK, OR, SD, TX, WA

Potatoes CO, ID, ME, MI, MN, NY, ND, OR, PA, WA, WI

Table 5.5. States for Which Pesticide Data Are Availableby Crop

93


Results

The estimated impacts of rainfall and temperature onpesticide cost and its variability by climate are dis-played in Tables 5.6 through 5.9. The estimationresults in Table 5.6 show the relationship betweenpesticide usage costs and climate. Table 5.7 containsthe computed percentage change in cost resultingfrom the percentage change in climate characteris-tics, using the data in Table 5.6. The impacts of pre-cipitation on pesticide usage cost for these five cropsare all positive and significant, except for cotton.This result indicates that increased rainfall increasespesticide costs. For example, when rainfall increasesby 1 percent, we compute that corn pesticide costsincrease by 1.49 percent. We find mixed effectsfrom temperature. A 1 percent temperature increase(measured in degrees Fahrenheit) increases pesticidecosts for potatoes by 2.67 percent. Pesticide costsfor corn, cotton, and soybeans also increase withtemperature, but wheat costs decrease.

The impacts of climate on the variability of pesticideusage cost are more complicated (see Tables 5.8 and5.9). We found that a hotter temperature increased

the variance of pesticide cost for corn, cotton, andpotatoes but decreased the cost variance for soy-beans and wheat. For example, a 1 percent increasein temperature will increase the year-to-year costvariance for corn by 6.96 percent. A rainfallincrease also increases the pesticide cost variabilityfor cotton but decreases the variance for soybeans,wheat, and potatoes.

Under a warmer and wetter climate (and given theestimated relationships), we generally would expectclimate change to increase pesticide use. Someregions may have less rainfall, however, and not allcrops show positive relationships between the cli-mate variables and pesticide usage. For perspective,then, we used the regional estimates of the Canadianand Hadley climate scenarios for 2090 to obtain esti-mates of the effects of projected climate change onpesticide usage cost for selected crops in selectedregions (Table 5.10). The results for states with sig-nificant production of each crop are given in Table5.10. They show increases in pesticide use on corngenerally in the range of 10–20 percent, on potatoesof 5–15 percent, and on soybeans and cotton of 2–5percent. The results for wheat varied widely by

Crop Precipitation Temperature Constant

Corn 0.7351 0.9222 –30.183(25.85) (19.00) (–11.30)

Cotton 0.0059 0.9730 –17.213(0.26) (8.39) (–2.27)

Soybeans 0.0632 0.5523 32.343(3.78) (13.22) (15.04)

Wheat 0.1211 –0.1160 7.7950(29.25) (–21.30) (24.41)

Potatoes 1.3684 2.5914 –89.564(22.76) (11.99) (–7.54)

Table 5.6. Regression Results for Effects of Climate on Per AcrePesticide Cost

Note: Temperature is measured in degrees Fahrenheit and rainfall ismeasured in inches. T-statistics in parentheses indicate significance ofall estimates except for cotton, where precipitation and temperaturecoefficients are insignificant at 5 percent level.

Crop Precipitation Temperature

Corn 1.49 1.87

Cotton – 1.94

Soybeans 0.09 0.78

Wheat 2.86 –2.74

Potatoes 1.41 2.67

Table 5.7. Percentage Change in Pesticide Cost fora 1 Percent Change in Average Climate Measures

Note: Percentage change for pesticide cost is computedby dividing coefficient parameters in Table 5.6 by US aver-age pesticide cost for a crop across all years and places.Results are computed only for estimated parameters witht ratios that exceed 1.9. Temperature percentage changeis based on degrees Fahrenheit; rainfall percentage isbased on inches.

94


state and climate scenario, with changes rangingfrom approximately –15 percent to +15 percent.

Pesticides and Climate: Some Conclusions and Limitations

Regional pesticide cost data show systematic varia-tions that can be related to climate characteristics.Average per-acre pesticide usage costs for corn, soy-beans, wheat, and potatoes increase as precipitationincreases. Similarly, average pesticide usage costs forcorn, cotton, soybeans, and potatoes increase astemperature increases, although the pesticide usagecost for wheat decreases. Climate also affects theyear-to-year variability of pesticide cost. More rain-fall decreases the variability of pesticide costs forsoybeans, wheat, and potatoes but increases the vari-ability of pesticide costs for cotton. Increased tem-perature reduces the variability of pesticide cost forsoybeans and wheat but increases it for corn, cot-ton, and potatoes.

This study is one of the first investigations of therelationship of pests and climate, conducted so thatthe results could be integrated into an economicassessment. There are several limitations in thestudy. For example, we do not consider how altered

Crop Precipitation Temperature Constant

Corn -0.0008 0.1179 -6.2453(-0.22) (19.56) (-19.93)

Cotton 0.0093 0.0497 -2.1377(4.03) (3.65) (-2.42)

Soybeans -0.0190 -0.0500 4.4399(-7.52) (-8.96) (16.33)

Wheat -0.0489 -0.0225 0.4838(-25.45) (-7.15) (2.83)

Potatoes -0.0372 0.1273 -3.4946(-12.00) (8.25) (-4.02)

Crop Precipitation Temperature

Corn – 6.96

Cotton 0.39 3.44

Soybeans -0.83 -3.20

Wheat -1.33 -1.34

Potatoes -1.15 7.14

Note: Percentage change for pesticide variability cost iscomputed by multiplying coefficient parameters in Table5.8 by average precipitation and temperature across allyears and places. Results are computed only for estimat-ed parameters with t ratios that exceed 1.9. Temperaturepercentage change is based on degrees Fahrenheit andrainfall percentage is based on inches.

Note: Temperature is measured in degrees Fahrenheit and rainfall ismeasured in inches.

Table 5.8. Regression Results on Influence of Climate onVariance of Pesticide Usage Cost

Table 5.9. Percentage Change in Variance ofPesticide Usage Cost for a One percent Change inAverage Climate Measures

CO2 could effect pests. Moreover, the approach con-siders how pesticide use changes but not how pestdamage itself changes, implicitly assuming that thecost implications of any change in pests is fully cap-tured by changes in pesticide expenditures. In gen-eral, statistical analyses are limited by data availabili-ty in their ability to capture some of the detailedstructural interactions or to trace increased pesti-cide use to specific pest infestations—in this case,specific pesticides that have different environmentalconsequences. Projections of changes in pesticideuse under future climates are highly speculativebecause few areas of agriculture change as rapidly aspesticides. Pests can quickly develop resistance toparticular control methods, and new control meth-ods are developed. In the future, pest resistancemay be increasingly introduced directly into crops.Nevertheless, these results indicate that for most ofthe crops we considered and for most locations, thefuture climate is likely to increase pest problemsand create the need for more effective control meth-ods. The environmental implications clearly willdepend on the types of methods developed to con-trol pests. The likelihood of increased pest prob-lems creates an added incentive to ensure that pestcontrol methods that do not create environmentalharm are developed and used.

95


Effects of Climatic Changeon a Water-DependentRegional Economy:A Study of the EdwardsAquifer

Global climate change portends shifts in waterdemand and availability. In areas where wateralready is a severely limited resource, potential reduc-tions in supply can pose significant questions withregard to allocation of the remaining resource.Agriculture is the major user of water in mostregions. In this section we summarize the results ofan analysis we carried out as a part of this assess-ment (described in greater detail in Chen, Gillig, andMcCarl 2000). The study examines the implicationsof climate change projections for the San Antonio,Texas, Edwards Aquifer (EA) region, concentrating onthe economy and the water use pattern.

Canadian Climate Change Scenario Hadley Climate Change ScenarioCorn Soybeans Cotton Wheat Potatoes Corn Soybeans Cotton Wheat Potatoes

CA 5.16 4.69CO -10.29 7.33 9.15 13.25GA 4.23 2.66ID 21.03 15.42IL 18.19 3.26 14.23 2.00IN 10.01 2.72 15.07 2.04IA 26.07 3.94 15.66 2.17KS 13.60 12.93LA 5.36 3.12MN 2.25 8.10 1.90 9.67MT -9.85 6.28MS 5.83 3.01ND 5.54 10.67NE 3.35 2.69 -14.54 10.72 2.16 5.83OK -3.48 12.34SD 17.08 8.88 14.73 13.96TX 5.41 -8.78 3.15 0.81WA 13.19 10.68

Table 5.10. Percentage Increase in Crop Pesticide Usage Cost from Reference Levels for Year 2090, by Scenario

We begin with a discussion of the Edwards Aquiferarea; we then provide a summary of the methodswe used to estimate the various impacts of climateon water use in the region, describe the modeland methods we used to consider the implicationsof these effects for the region, discuss the results,and offer some broader conclusions on the basis ofthe study.

The Edwards Aquifer

The Edwards Aquifer supplies the needs of munici-pal, agricultural, industrial, military, and recreationalusers. The Edwards Aquifer is a karstic aquifer. Akarstic aquifer is one that has many characteristicsin common with a river. Annual recharge over theperiod 1934–1996 averaged 658,200 acre feet; dis-charge averaged 668,700 acre feet (USGS 1997).Edwards Aquifer discharge is through pumping andartesian spring discharge. Pumping rose by 1

96


percent per year in the 1970s and 1980s (Collinge etal. 1993) and now accounts for 70 percent of totaldischarge. Pumping in the western Edwards Aquiferis largely agricultural (AG), whereas eastern pump-ing is mainly municipal and industrial (M&I). Springdischarge—mainly from San Marcos and Comalsprings in the east—supports a habitat for endan-gered species (Longley 1992), provides water forrecreational use, and serves as an important supplysource for water users in the Guadalupe-Blanco riversystem. The aquifer is now under pumping limita-tions as a result of actions by the Texas Legislatureand because of a successful suit by the Sierra Clubto protect endangered species (Bunton 1996).

Reduced water availability or increased waterdemand because of climate change could exacerbatethe regional problems that arise in dealing withwater scarcity. This study utilizes an existingEdwards Aquifer hydrological and economic systemsmodel known as Edwards Aquifer Simulation Model(EDSIM) (McCarl et al. 1998), to examine the impli-cations of climate-induced changes in recharge andwater demand.

Effects of Climatic Change in the Edwards Aquifer Region

The Canadian and Hadley model results for theEdwards Aquifer region climate are listed in Table5.11. Changes in regional climatic conditions wouldalter water demand and supply. An increase in tem-perature would cause an increase in water demandfor irrigation and municipal use but would alsoincrease evaporation lowering runoff and in turn

Edwards Aquifer recharge. A decrease in rainfallwould increase crop and municipal water demand,lower the profitability of dryland farming, andreduce available water for recharge.

Recharge Implications

To project climatic change effects on EdwardsAquifer recharge, we used regression analysis to esti-mate the effects of alternative levels of temperatureand precipitation on historically observed recharge.We drew US Geological Survey (USGS) estimates ofhistorical recharge data by county from the EdwardsAquifer Authority annual reports for the years1950–1996. We obtained county climate data for thesame years from the Office of the Texas StateClimatologist and a University of Utah Web page. Weconcluded that the preferred regression model forthis data set was a log-linear model. The significantrecharge regressions coefficients all exhibited theexpected sign. Summary measures of the effect ofthe projected climate changes on annual recharge forthe years 2030 and 2090 under different climate sce-narios are displayed at the top of Table 5.12; thesedata show that the projected climate change causeslarge reductions in recharge for drought years (21–33percent) and wet years (24–49 percent).

Municipal Water Use Implications

Griffin and Chang (1991) present estimates on howmunicipal water demand is shifted by changes intemperature and precipitation. In particular, theyestimate the percentage increase in municipal waterdemand for a 1 percent increase in the numberof days that temperature exceeds 90°F and precipita-tion falls below 0.25 inches. To obtain theanticipated shifts for the 2030 and 2090 climate con-ditions, we took the daily climate record from 1950to 1996 and adjusted it by altering the original tem-perature and precipitation by the projected climateshifts from the climate simulators. We then recom-puted the municipal water demand accordingly.The results are given in Table 5.12; the forecast cli-mate change increases municipal water demand by1.5–3.5 percent.

Climate Change Temperature PrecipitationScenario (°F)

Hadley 2030 3.20 -4.10Hadley 2090 9.01 -0.78

Canadian 2030 5.41 -14.36Canadian 2090 14.61 -4.56

Table 5.11. Projected Percentage Climate Changes forEdwards Aquifer Region, by Scenario

97


Crop Yields and Irrigation Water Use

Changes in climatic conditions influence crop yieldsfor irrigated and dryland crops, as well as irrigationcrop water requirements. For this study, we estimat-ed the shift in water use and yield under projectedclimate changes by using the Blaney-Criddle (BC)procedure (Heims and Luckey 1983; Doorenbos andPruitt 1977, following Dillon 1991). In particular, weused the BC procedure to alter yields and water usefor the nine recharge/weather states of nature pre-sent in the EDSIM model, an economic and hydro-logical simulation model of the Edwards Aquiferregion (McCarl et al. 1998). Summary measures ofthe resultant effects are presented in Table 5.12; thedata show a decrease in crop and vegetable yieldsand an increase in water requirements. For exam-ple, under the Hadley scenario in 2090, the irrigatedcorn yield decreases by 3.47 percent, whereasthe irrigation water requirement increases by31.32 percent.

Hadley Canadian2030 2090 2030 2090

Recharge in drought year -20.6 -32.9 -29.7 -32.0Recharge in normal year -19.7 -33.5 -29.0 -36.2Recharge in wet year -23.6 -41.5 -34.4 -48.9Municipal Water demand 1.5 2.5 1.9 3.5Irrigated Corn Yield -1.9 -3.5 -4.3 -5.6Irrigated Corn Water Use 12.0 31.3 23.5 54.0Dryland Corn Yield -3.9 -6.8 -8.2 -10.8

Irrigated Sorghum Yield -1.8 -3.4 -2.8 -4.2Irrigated Sorghum Water Use 15.1 38.2 42.7 79.4Dryland Sorghum Yield -5.9 -13.1 -10.8 -16.8

Irrigated Cotton Yield -9.1 -15.8 -19.8 -24.6Irrigated Cotton Water Use 16.9 40.8 34.6 71.5Dryland Cotton Yield -7.1 -11.6 -14.0 -17.8

Irrigated Cantaloupe Yield -1.3 -2.3 -2.9 -3.6Irrigated Cantaloupe Water Use 19.0 46.5 41.4 82.7

Irrigated Cabbage Yield -5.6 -12.1 -9.6 -14.7Irrigated Cabbage Water Use 14.8 31.0 36.4 71.3

Table 5.12. Selected Effects in Terms of Percentage Changes from Base Scenario

Methods for Developing Regional Impact

These effects were combined in EDSIM. The modeldepicts pumping use by the agricultural, industrial,and municipal sectors while simultaneously calculat-ing pumping lift, ending elevation, and springflow.EDSIM simulates the choice of regional water use,irrigated versus dryland production, and irrigationdelivery system (sprinkler or furrow) such that over-all regional economic value is maximized. Regionalvalue is derived from a combination of perfectlyelastic demand for agricultural products, agriculturalproduction costs, price-elastic municipal demand,price-elastic industrial demand, and lift-sensitivepumping costs.

In terms of its implementation, EDSIM is a mathe-matical programming model that employs two-stagestochastic programming with recourse formulation.The multiple stages in the model depict theuncertainty inherent in regional water-use decision

98


making. Many water-related decisions are madebefore water availability is known. For example, thedecision about whether to irrigate a particular par-cel of land and the choice of crops to put on thatparcel are decided early in the year, whereas thetrue magnitude of recharge is not known until muchlater in the year.1

Model Experimentation, Regional Results,and Discussion

We considered five scenarios in this study:

• base without climatic change

• change projected by the Hadley model for theyear 2030

• change projected by the Canadian model for theyear 2030

• change projected by the Hadley model for 2090

• change projected by the Canadian model for 2090.

Table 5.13 displays EDSIM results on the economicand hydrological effects of climate change under thebase scenario as actual values; results under theother scenarios are displayed as a percentagechange from the base results. The total water usageis held less than or equal to a 400,000 acre-feetpumping limit mandated by the Texas Senate foryears after 2008. Under the base condition, agricul-ture uses 38 percent of total pumping, and M&Ipumping usage accounts for the rest. Total welfarefor the region is $355.69 million—$11.39 millionfrom agricultural farm income and $337.65 millionfrom M&I surplus. In addition, $6.64 million accruesto the Edwards Aquifer Authority for the water-usepermits. This authority surplus can be regarded asrents to water rights to use some of the 400,000acre-feet available. Comal and San Marcos spring-flows are 379.5 and 92.8 thousand acre-feet, respec-tively—greater than recent average historical levels.

The strongest effect of climate change falls onspringflow and the agricultural sector. Under the

climatic change scenarios, the Comal (the most sen-sitive spring) springflows decrease by 10–16 per-cent in 2030 and by 20–24 percent in 2090. Thischange could require additional springflow protec-tion (see below). In terms of agriculture, the changeresults in a reallocation of water away from agricul-ture. It adds to the cost of pumping because thewater must be pumped from greater depths, and itincreases water demand for irrigation because ofhigher temperatures and less rainfall. Overall yieldsare lower. The result is a reduction in farm incomeof 16–30 percent in 2030 and 30–45 percent in2090. Regionally, income falls by $2.8–5 million peryear in 2030 and $5.8–8.8 dollars in 2090. The pro-jected shift in agricultural water to M&I indicatesthat city users are purchasing water that otherwiseis allocated to agriculture through water markets.

Despite an increase in M&I water use, the M&I sur-plus decreases because of an increase in pumpingcosts that result from an increase in pumping liftderiving from lower recharge. In contrast to thewelfare decrease for agricultural and nonagriculturalpumping users, rents to the authority or water per-mits increases by 5–24 percent. The increaseddemand for water increases water permit prices.Water use in the nonagricultural sector is less vari-able, and a shift to that sector actually makes wateruse slightly greater—with corresponding decreasesin springflow.

The large reduction in springflow would put endan-gered species in the spring emergence areas in addi-tional peril. The projected climate change thereforewould require a lower pumping limit to offer thesame level of protection for the springs, endangeredspecies, and other environmental amenities nowprovided by the 400,000 acre-foot limit. Table 5.14presents the results of an examination of thepumping limit that would be needed to preserve thesame level of Comal and San Marcos springflows asin the current situation. The results indicate that adecrease in the Edwards Aquifer pumping limit of35,000–50,000 acre-feet in 2030 and 55,000–75,000acre-feet in 2090 would be needed. These further

1This uncertainty may be best illustrated by referring to the Irrigation Suspension Program implemented by the EdwardsAquifer authority: Early in the year an irrigation buyout was pursued, but the year turned out to be quite wet in terms ofrecharge.

99


Base 2030 2090

Hadley Canadian Hadley CanadianVariable Units Value (%) (%) (%) (%)

AG Water Usea 1000 af 150 -0.9 -1.4 -2.4 -4.2

M&I Water Useb 1000 af 250 0.6 0.9 1.5 2.6

Total Water Usec 1000 af 400 0.1 0.1 0.1 0.1

Net AG Incomed $1,000 11391 -15.9 -29.4 -30.3 -45.0

Net M&I Surpluse $1,000 337657 -0.2 -0.4 -0.6 -0.9

Authority Surplusf $1,000 6644 3.8 7.1 12.7 21.6

Net Total Welfareg $1,000 355692 -0.6 -1.2 -1.3 -1.9

Comal Flowh 1000 af 379.5 -10.0 -16.6 -20.2 -24.2

San Marcos Flowi 1000 af 92.8 -5.1 -8.3 -10.1 -12.1

Table 5.13 Aquifer Regional Results under Alternative Climate Change Scenarios

a Agricultural water use.b Municipal and industrial water use.c Total water use, including agricultural and nonagricultural water use.d Net farmer income.e Net municipal and industrial surplus.

f Surplus accruing to pumping or springflow limit.g Net total welfare, including agricultural and nonagricultural welfare.h Comal springflow.i San Marcos springflow.

Base 2030 2090

Hadley Canadian Hadley CanadianVariable Units Value (%) (%) (%) (%)

Pumping Limit 1000 af 400 365 350 345 320

AG Water Use 1000 af 150 -16.5 -22.7 -23.7 -46.1

M&I Water Use 1000 af 249 -4.0 -6.3 -7.7 -4.3

Total Water Use 1000 af 400 -8.7 -12.5 -13.7 -20.0

Net AG Income $1,000 11391 -18.4 -33.4 -34.6 -58.3

Net M&I Surplus $1,000 337657 -0.8 -1.3 -1.9 -1.9

Authority Surplus $1,000 6644 32.3 52.5 73.7 68.3

Net Total Welfare $1,000 355692 -0.8 -1.4 -1.6 -2.5

Comal Flow 1000 af 379.5 1.5 0.5 1.2 -1.1

San Marcos Flow 1000 af 92.8 -0.3 -1.1 -1.1 -2.5

Table 5.14. Results of Analysis on Needed Pumping Limit to Preserve Springflows at Base, without Climate Change Levels

Note: Pumping limit under each scenario represents amount of water restriction in Edwards Aquifer regions.

100


decreases in pumping impose substantial additionaleconomic costs beyond those imposed by climatechange alone; welfare would fall by $0.5–0.9 millionin 2030 and $1.1–1.9 million in 2090. The addition-al pumping reduction causes a large impact on agri-culture and a substantial municipal cutback.

Concluding Remarks

Changes in climatic conditions projected by theCanadian and Hadley models cause a reduction inavailable water resources, as well as a demandincrease in the Edwards Aquifer region. The changelargely manifests itself in reduced springflows and asmaller regional agricultural sector. The regionalwelfare loss was estimated to be $2.2–6.8 millionper year. If springflows are to be maintained at thecurrently desired level to protect endangeredspecies, pumping must be reduced by 10–20 per-cent below the limit currently set, at an additionalcost of $0.5–2 million per year.

Global Climate Change:Interactions with SoilProperties

Soil and Society

Soil processes operate on time scales that rangefrom thousands of years (e.g., breakdown of rocksubstrate) to hours (e.g., erosion). For much ofNorth America, the climate is naturally highly vari-able, and this variability has punished us badlywhen we have not been good stewards of the land.Land abandonment after excess growth of cotton inthe Southeast, the loss of soil fertility and acidifica-tion in the Northeast, and the dust bowl of thePrairies can be attributed to soil management prac-tices. These instances stem from failure to recog-nize soil as a resource that is subject to degradationand failure to develop practices that maintain soilunder climate conditions that vary from decadeto decade.

Atmospheric Constituents andSoil Processes

Carbon, nitrogen, oxygen, and hydrogen are thebuilding blocks of life on earth. They also are themost important constituents of soil organic matter.The earth’s carbon and nitrogen cycles have theability to restore and even increase soil organic mat-ter content and tilth if properly established scientificprinciples are applied to implement good land stew-ardship and sustainable agriculture during a time ofglobal change.

Global change scenarios are associated most oftenwith projected increases in temperature and cli-mate instability associated with increased atmo-spheric concentration of gases of carbon and nitro-gen. These radiative gases consist of CO2, CH4, andN2O, which are produced by microbial activities insoils, sediments, surface waters, and animal digestivesystems or through the burning of fossil fuels. Soilmicroorganisms produce CO2 by breaking downplant and animal residues in environments thatcontain oxygen. This process returns to the air the

Category 1990 1996

CO2

Fossil fuel combustion 1,330 1,450Other industrial sources 20 20

CH4

Transportation and industry 60 60Land use and agriculture 50 50Landfills and waste 60 70

N2OTransportation and industry 30 30Land use and agriculture 65 70

HFCs, PFCs, SF6 20 35

Total 1,635 1,785

Table 5.15. Trends in US Greenhouse Gas Emissions(MMT carbon equivalents)

Source: US Environmental Protection Agency.

101


carbon that has been fixed by photosynthesis andin the past has kept the carbon cycle in near bal-ance. Where oxygen is lacking—such as in peatbogs, rice fields, and the stomachs of ruminants—methane (CH4) is produced instead of CO2.

Soil inorganic nitrogen is produced when microor-ganisms “burn off” the carbon of plant and animalresidues or organic matter in their never-ceasingsearch for energy. Other microorganisms oxidize, bythe nitrification process, inorganic nitrogen that isproduced on mineralization or added as fertilizer.This process is leaky and produces N2O. The oxi-dized form of nitrogen, NO3, that is produced duringnitrification can again be reduced under anaerobicprocesses where there is no oxygen. This processalso can result in N2O leakage to the atmosphere.

Methane is 20 times as effective as CO2 in retainingatmospheric heat; N2O is 300 times as effective asCO2. The relative effect of these gases in causinggreenhouse effects is best seen by expressing theemissions as carbon equivalents. In 1996, the UnitedStates released 1,450 million metric tons (MMT) ofcarbon into the atmosphere from fossil fuel con-sumption. This amount is less than one-tenth of theamount released annually from our soils by decom-position; the carbon of decomposition is offset, how-ever, by a nearly equal amount of photosynthesis,whereas the equivalent of about one-half of the car-bon from fossil fuels accumulates in the atmosphere.

A total of 180 MMT of CH4 in carbon equivalentsusing 100-year Global Warming Potentials (GWPs) isreleased from US transportation, industry, wetlands,landfills, and waste. Aerobic terrestrial sites absorbCH4, but cultivated, fertilized soils consume onlyabout one-quarter that of undisturbed sites and wild-lands. Agriculture is the predominant source ofN2O; transportation and industry supply about one-third as much as agriculture. All soils release someN2O, but highly managed soils release more thanwildlands (especially if they have trees). CO2 isincreasing in the atmosphere at 0.5 percent peryear; CH4 is increasing at 0.75 percent, and N2O isincreasing at 0.75 percent.

The clearing of forests, the draining of wetlands, andthe plowing of prairies for agriculture led to signifi-cant increases in atmospheric CO2 as organic car-bon was decomposed. The carbon content of mostagricultural soils is now about one-third less than itwas in its native condition as either forest or grass-land. Using computer simulation models of crop-land in the central United States, Bruce et al. (1998)suggest that soil carbon losses have diminished andsoils are starting to accumulate carbon again, revers-ing the trend of carbon loss that had occurred sincecultivation began. This reversal has come aboutthrough higher yields, the return of greater propor-tions of crop residue to the land, conservation tillagesuch as cover crops, and no till (Lal et al. 1998). Thereturn of considerable acreage to grass in conserva-tion reserve programs and to trees in afforestation offormerly plowed lands also is returning atmosphericCO2 to the land. The eastern United States now has110 million acres of afforested lands that are storingcarbon (Fan et al. 1998). This carbon storage occursas tree growth and in increased soil organic mattercontents (Morris 2001). The other greenhousegases—CH4 and N2O—also can be removed fromthe air by soil microorganisms. Improved pasturesand cover crops on cultivated land lower theamount of inorganic nitrogen in soil and can loweratmospheric radiative gases. Higher-quality cattlefeeds can reduce CH4 emissions from domesticlivestock.

Soil-Biological and Chemical Interactions in Global Change

A large number of agronomic-ecological interactionscould occur in a world with more CO2, higher tem-peratures, and a more variable climate. There is agreat diversity of soil organisms, many of whichhave similar functions and general decompositionreactions. This situation enables us to project thefuture effects of changes in soil temperature andmoisture on the basis of overall controls that applyto most soil types within a major climatic area.Climate change and accompanying extreme eventsundoubtedly will alter soil microbial populationsand diversity. Over time, populations of soil biota

102


can adapt, although cataclysmic occurrences such asfloods and erosion will affect the diversity of micro-bial populations in local areas.

The CO2 content of soil is higher than that of theatmosphere; atmospheric concentrations of CO2 arenot expected to directly alter soil nutrient cycling.Indirect effects have to be considered, however. Theadditional available substrate—the symbiotic part-ners consisting of nitrogen fixers such as rhizobiaand mycorrhizal fungi—may be able to obtain agreater food supply and grow more effectively, witha consequent benefit to the plant, although the ben-efits vary depending on the fungi type. This processcould be especially important in forests and nativegrasslands that are not normally fertilized as theyadapt to global change.

Plants are more sensitive than microorganisms tospecific temperatures. Increased temperature willmove the growth requirements of specific plants200–300 km north for each degree Celsius rise intemperature (equivalent to 60–90 miles for eachdegree Fahrenheit). This factor, along with breedingfor cold tolerance, is moving the Corn Belt into thePrairie provinces of Canada. Insect activity of cold-sensitive insects has been observed to move north-ward with even the slight rise in measured tempera-tures that scientists have observed recently. Withincreased temperatures, we may see cold-temperature soil pathogens and weeds as wellas fire ants in areas of what is now the Corn Belt,although pest interactions with climate and weatherdepend on a variety of factors, including moistureand weather variability.

Many soils contain inorganic carbon as carbonates.The pedogenic phases of these compounds canrelease and sequester CO2. Agriculture is acidifyingin nature; on some soils it requires the addition oflime, which on solubilization releases CO2 to theatmosphere. Soils with carbonate horizons are com-mon in arid and semi-arid regions. Calcium is addedas lime through deposition of wind-blown dust andduring weathering of parent materials. This calciumreacts with CO2, based on the carbonate-bicarbonate

(HCO3-) reactions, to produce carbonates. Soil inor-

ganic carbon constitutes approximately 1,700 Pg Cin the surface layers. This amount is similar to val-ues for organic carbon (Nordt et al. 2000); soil inor-ganic carbon is being leached out of soils at an esti-mated rate of 0.25 Pg per year, whereas rivers arethought to transfer 0.42 Pg C to the oceans annual-ly—providing a net CO2 sink. Although irrigationwater releases some trapped CO2, researchers esti-mate that on a worldwide basis soils sequester0.16–0.27 Pg C yr-1 of atmospheric CO2 (Holland1978; Bouwman and Lemans 1995).

Soil formation will be slowly altered by changes inmoisture and temperature. The United States is nowreceiving 10 percent more rainfall than in previousdecades. If this increase were to continue over hun-dreds of years, higher moisture and temperaturewould result in deeper profiles with more clay eluvi-ation to lower horizons. These effects are slow andwill be overshadowed by changes in management orerosion. A single extreme event, over the course of24 hours, can erode away soil that would take hun-dreds of years to form, with the amount eroded high-ly dependent on management practices. The DustBowl of the 1930s is one such example. Agriculturehas changed drastically since then, in response to thedamage caused; nevertheless, precautions are neededin susceptible areas where multiple-year droughts,associated poor crops, and high winds could againcombine to create conditions for severe wind ero-sion (regardless of whether this erosion is associatedwith specific climate change events).

Flooding affects agricultural and nonagriculturalareas. For example, a wetter climate in Californiawith increased temperatures—as occurs in theHadley and Canadian scenarios—and more oceanicevaporation could result in massive soil movement(as in soil slippage) and local flooding from moresevere local storms. Lal et al. (1998) estimate that0.5 Pg C yr-1 are lost from local soils by erosion.Although much of this soil is deposited within asso-ciated landscapes, 20 percent is thought to be lostto the atmosphere through accelerated decomposi-tion. The fate of the transported carbon is not

103


well-known, however; Trimble (1999) estimates thatrecent water erosion is only one-sixth as severe asthe erosion that occurred during the early years ofagriculture in the Midwest.

Erosion and leaching can move extensive nutrientsto rivers and eventually to estuaries. These nutri-ents—especially nitrogen and phosphorous—cancreate local high-nutrient and thus anoxic events,with serious pollution and local fish kills. This situa-tion now exists in the Mississippi Delta and the Gulfof Mexico, as well as the Chesapeake Bay. The contri-bution of agriculture to such pollution must bedetermined. Potential nutrient losses in a climate-change scenario also must be considered. Nutrientmanagement will have to include lower inputs onnitrogen and phosphorus and more containment oflocal floodwaters so nutrients can soak back into theland. It also must consider the effects of extensiveconcentrations of human and animal waste productson small land areas. This concentration removesnutrients from areas where crops are grown andoften concentrates them in erosion- and flood-proneareas, with the potential for eutrophication and localcontamination if flooding increases with climatechange.

Soil Organic Matter and Global Change

Organic matter constitutes 1–8 percent of most soils(by weight) and nearly all of the dry weight oforganic soils such as peats. Because of the greatweight of soils to the plant rooting depth at whichcarbon accumulates, the world’s soils store1,670,000 MMT (16.7 Pg) of carbon. This amountrepresents a carbon storage capacity that is twicethat of the atmosphere. The annual global rate ofphotosynthesis generally is balanced by decomposi-tion; the annual flux is about one-tenth of the car-bon in the atmosphere or one-twentieth of thecarbon in soils. The United States accounts forabout 5 percent of this storage; because of its higherproportion of peat soils, Canada accounts for up to17 percent (Lal et al. 1998).

Soil carbon is composed of a wide range of com-pounds that decompose at different rates dependingon their chemistry, the soil temperature and mois-ture, which organisms are present, the associationwith soil minerals, and the extent of aggregation(Paul et al. 1996). Plant residues in agricultural soilsdo not represent a large storage pool; their manage-ment influences water penetration, erosion, and theextent of formation of soil organic matter—thusaffecting long-term soil fertility and carbon storage.

Decomposition by soil organisms is relatively insen-sitive to dryness on an annual basis. Most soils havesome periods of time when decomposition canoccur. Decomposition is very sensitive to excesswetness, however, which causes anaerobiosis. In thepast, this decomposition has created high–organic-matter peat soils. Changes in moisture contentresulted in increased decomposition of soil organicmatter when the millions of acres of wetlands in theCorn Belt were tile drained (Lal et al. 1998). Warmertemperatures often are associated with drier cli-mates. Researchers have postulated that this rela-tionship greatly affects peat soils that contain somuch of North America’s soil organic carbon.Drying of peat soils to below water saturationwould greatly increase decomposition rates and CO2

evolution to the atmosphere. Water saturation ofsoils is controlled as much by drainage and topogra-phy as by rainfall and temperature. Projections thatare based on temperature and rainfall alone will notnecessarily be valid relative to decomposition inpeats. One can control the soil moisture of tile-drained soils in the winter by controlling (plugging)tile drainage flows. This plugging creates temporarywetlands and thus retards decomposition. It shouldhave the additional benefit of decreasing nitratesand possibly pesticides in the groundwater, as wellas helping in flood control. Wetland restoration ingeneral has potential for future carbon sequestra-tion—providing greater diversity and havens forwildlife and reducing nitrates in ground water. Itwill lead to some increases in methane, however,and possibly N2O evolution from flooded soils.

104


Grasslands contain approximately one-fifth of theworld’s global carbon reserves; many of the world’sgrasslands have been degraded by overgrazing. Thisovergrazing has resulted in a loss of plant cover,reduced protection against wind and water erosion,and loss of production potential. Soil organic matterdegradation in such conditions has contributed tothe rise in atmospheric CO2. Grazing and othermanagement practices that lower overgrazing havethe potential to increase global carbon sequestrationsubstantially (0.46 Pg C yr-1). This management alsoshould result in increased methane utilization.Fertilizer nitrogen is one suggested means—alongwith better grazing management—of increasinggrassland production and soil carbon sequestration.Production of nitrogen fertilizer uses fossil fuels,however, and application of fertilizer could lead toincreased N2O evolution. Closer coupling of grazingwith intense animal-feeding operations that returnsnutrients for pasture improvement would greatlyreduce problems with pollution when excess rain-fall causes flooding.

The increased CO2 in the atmosphere probably hasenabled farmers to greatly increase yields throughplant breeding, fertilizer additions, and pest control.The continued increase in plant yield of 1.25 per-cent per year (Reilly and Fuglie 1998) will produce asimilar increase in the crop residue applied to soils.At equivalent nitrogen levels, production of carbohy-drates and possibly lignin and polyphenols willincrease. Polyphenols should slow down decompo-sition rates and help build organic matter. Thechanged composition of leaves and roots will affectthe insects and microbiota that feed on plant parts.These insects are a part of a complex food web,often involving numerous layers of predators; thus,the insect response to CO2 should be considered inclimate change scenarios.

The large size of the soil carbon pools and theirslow turnover rate mean that they are fairly wellbuffered against change and that short-termeffects—unless they involve erosion and thusremoval of carbon from the landscape—do nothave immediate effects. Normally, 7–15 years ofmanagement effects would be required to produce

measurable differences in carbon and associated soilfertility and soil tilth. The large size of the carbonpool and the fact that soil carbon is very unevenlydistributed across the landscape make accurate mea-surement of any changes that occur over a few yearsvery difficult.

Total soil carbon is very difficult to measure with theaccuracy required for decision making in globalchange calculations. Soil heterogeneity and changesin bulk density further confound the problem ofmeasuring short-term changes in soil organic matter.Calculation of soil carbon sequestration must bebased on long-term plots that have been under a spe-cific plant management scheme for 10–30 years. Soilfractions that are sensitive indicators of soil carbonchanges are best used in conjunction with modelingthat is based on a knowledge of the controls on soilcarbon dynamics (Figure 5.4). This approach enablesresearchers to project the effect of specific manage-ment on other soil types and landscape areas.

Figure 5.3 Controls on Soil Carbon Turnover andAccumulation

105


Indicators that have been useful for estimating soilcarbon include the light fraction obtained by float-ing soil in water or a more dense liquid. This lightfraction reflects partially decomposed plant residuesthat make up a portion of the active fraction of soilorganic matter. The microbial biomass that feeds onthe residues and on the active and slow fraction ofsoil organic matter is another measurable fractionthat changes rapidly enough to be an indicator oftotal changes.

The partially altered plant materials that are heldwithin aggregates and thus slowly decompose overa period of years constitute part of the slow fractionthat is so essential to soil fertility. This fraction—known as particulate organic matter—can be mea-sured by disrupting the aggregates; it has potentialas an indicator of the overall size of the slow pool inmanagement for sustainable agriculture and in car-bon sequestration calculations. Laboratory incuba-tions of soils from various management treatmentson different soil types and under representative cli-matic conditions enable the natural population ofsoil fauna and soil microorganisms to decomposethe different available fractions over time. Analysisof CO2 evolution curves enables researchers todetermine the size and turnover rate of the activefraction and the slow fraction if the size of the resis-tant pool has previously been determined.

The foregoing biophysical techniques are best uti-lized on well-documented and characterized long-term plots with known management histories,where total carbon and soil bulk density can bemeasured to the rooting depth. If these plots arerepresentative of the different soil types, climate,and management, mathematical models can projectthe carbon content of other soils as well as thelandscape. Projections of future carbon levels arebased on modeling that utilizes information fromlong-term plots. Continuation of research on long-term plots, together with measurements on an arrayof well-distributed validation plots would enableresearchers to plan new approaches and supportpolicy decisions that must be made as we adapt toglobal change.

Soils in a North American Context

The warming of North America is noticeable alreadyin increased growing seasons and the northwardmovement of the limits of corn and soybean growth.The Corn Belt may move into the Canadian Prairies.The soils of northern parts of Minnesota,Wisconsin,Michigan, New York,Vermont, and Maine could beutilized for corn, soybeans, and specialty crops. Thepresent soils in these areas are not especially fertile;from an agroforestry viewpoint and from the aspectof removal of carbon from the atmosphere, theymight better be left in trees. Canada does not have agreat deal of potentially useful agricultural land inthe east, unless the climate becomes so warm thatHudson Bay lowlands would be suitable for agricul-ture. Warming of western Canada will produce moreagricultural land. Alberta and northern BritishColumbia could develop significant underutilizedacreage that would be far from markets.

Sandy soils are much more sensitive to climatic fluc-tuations than loams and clay soils. Fortunately, manydrought-sensitive, sandy soils of the Great Plainsalready have been removed from cultivation. Publicpolicy as well as management by individual opera-tors should continue to protect these fragile soils.The extent and distribution of rainfall is the greatestunknown in future climate scenarios. Researchersproject that because of higher temperatures therewill be more moisture in the atmosphere and thusmore rainfall on land. What is not known is wherethis moisture will fall. Warm periods generally havebeen associated with drought on the prairies. If thatrelationship continues to be the case, increaseddecomposition of soil organic matter because ofhigher temperatures will be offset somewhat bydecreased decomposition from lower moisture.

Field Validation

The overall requirements for soil organic matterresearch and field validation of the role of soil car-bon in global change are as follows:

• Provide analytical background and knowledgeconcerning the effects of agronomic

106


management on different soil types to projectand model their effect on soil organic mattercontents and other greenhouse gases.

• Establish benchmark sites, on a national level,that can provide verification of treatmenteffects. This effort requires field measurements,under different management, of soil types andclimates that are representative of most agricul-tural production; these measurements must beaccurate enough that possible future CO2 emis-sion debits/credits can be validated.

• Provide national inventories of soil carbon stor-age and fluxes of CO2, N2O, and CH4 into andout of soils.

• Participate with available informational systems—such as industry consultants and university andgovernment extension systems—to provide nec-essary information to the public and the agricul-tural industry concerning the present and futurerole of soils in global change.

Adapting to GlobalChange: PolicyImplications

Agriculture has had and will continue to have theability to adapt to new conditions. The ability tochange with a changing climate will depend on astrong research base that can supply required infor-mation. Some of the areas that may benefit most areas follows:

• Crops vary in their response to enriched CO2 inseveral growth characteristics. Research that uti-lizes plant breeding and molecular techniques inconjunction with studies of physiologicalresponses to increased CO2 would increase pro-ductivity. It also would result in increased cropresidue additions to the soil. Improved soilorganic matter levels will sequester CO2, enhancesustainability and reduce soil erosion. Similartechniques could be used to produce plants withincreased roots and biological nitrogen fixation,

as well as plants with higher capacities to takeup nutrients through more efficient mycorrhyza.

• Increased phenolic and lignin contents of plantresidues could decrease decomposition ratesand result in more crop residues at the surface.They also should enhance the formation of slowand resistant carbon pools that are important tocarbon storage. The growth of more perennialcrops could have many benefits, especiallywhen such crops are utilized as a biological,non-fossil fuel energy supply.

• Irrigation efficiency could be improved.Increased oceanic temperatures should result inmore rainfall overall. This rainfall could be uti-lized more efficiently by drip irrigation, waterharvesting, and other techniques.

• Farmers should develop more-efficient nitrogenand phosphorus fertilizer usage, especially inflood-prone areas. Precision farming holdspromise for better nutrient control and pesti-cide application. The nitrogen, phosphorus, sili-con, and carbon cycles need to be considered inan ecosystem context.

• The movement of intensive animal feeding oper-ations to the source of the animal feeds wouldenhance the placement of nutrients and organicresidues back on the soil and stop the develop-ment of these facilities on flood-prone areas.

• Increased soil organic matter would store moreatmospheric carbon and result in greater soil fer-tility, better soil tilth, and greater water-holdingcapacity. It also would make plant/soil systemsmore stress-resistant and thus better able to with-stand the greater projected climatic fluctuations.

• Control of water levels on hydric soils whencrops are not being grown could result in carbonsequestration, improved water quality, flood con-trol, and better wildlife habitat. Potential losses ofCH4 and N2O would have to be avoided.

• Soil pathogen and pest control in a warmer,often more humid climate would have to beconsidered in future management scenarios.

107


• Agricultural research and practice should workto improve pasture management for better car-bon sequestration.

• The agriculture sector should integrate farmwoodlots and riparian strips into overall landmanagement and farm policy programs thatenhance water quality and offer a positiveresponse to global change.

Conclusions

Each of the cases presented in this chapter offersspecific conclusions. In addition, five broader con-clusions also emerge. First, environmental impactscan be highly dependent on the specific characterof climate change. For the Chesapeake Bay, nitrogenloadings from corn production in the Chesapeakeregion differ significantly depending on whether theHadley or Canadian climate scenarios is used.Similarly, Chen, Gillig, and McCarl (2000) find thatavailable water resources in the Edwards Aquiferregion of Texas differ significantly depending onwhether they use projections from the Hadleymodel or the Canadian model. In the ChesapeakeBay region and the Edwards Aquifer region, theHadley model projects more precipitation and lesswarming than the Canadian climate scenario.

Second, environmental impacts also are highlydependent on the ability of crops to productivelyuse higher atmospheric levels of CO2. The opti-mistic conclusion for soils is that climate changecould enhance agricultural sustainability, increasewater-holding capacity, and reduce soil erosiondepends on increases in crop growth as a result ofadditional CO2. Results for the Chesapeake Bayregion that show increased nitrogen loadings fromcorn production also hinge on crop responses toadditional CO2. In and of itself, a higher level of CO2

increases nitrogen uptake by corn plants, leavingless nitrogen to run off into surface waters or leachinto groundwater. Higher levels of CO2 may makecorn production in the Chesapeake region economi-cally more attractive. If corn production becomesmore attractive, farmers may devote more land to

corn and increase their use of inputs per acre toraise yields. If they do these things, their usage ofnitrogen fertilizer may increase, leading to increasesin nitrogen loadings.

Third, additional research is needed on interactionsbetween climate, agriculture, and the environment.The vast majority of research on climate change andagriculture to date has focused on agricultural pro-duction impacts. Very little work has been done onhow climate change might affect the environmentalimpacts of agricultural production and land use.Given the magnitudes of environmental effects inmany areas of the country, this area should be a highpriority for research. In addition, research is neededto understand climate impacts on agriculture’s con-tributions to wildlife habitat, rural landscape ameni-ties, and carbon sequestration.

Fourth, particular effort is needed to investigate thepotential for changes in extreme events and theirconsequent environmental effects. Current climatemodels do not adequately represent extreme weath-er events such as floods or heavy downpours thatcan wash large amounts of fertilizers, pesticides, andanimal manure into surface waters. Changes inextreme events could easily overwhelm the environ-mental effects of changes in average levels of precip-itation or temperature, as well as the effects ofchanging atmospheric CO2 levels.

Fifth, many of these environmental and conservationconcerns involve nonmarket, off-farm effects andrequire actions by local, regional, or federal govern-ments if these resources are to be protected. Thefirst step in many cases is that adequate measuresare needed to protect environmental resourcesunder current climate conditions. Climate changemay mean that managers must be prepared to adaptprotection measures if climate change makes theminadequate. The Chesapeake Bay study indicatesthat current management of these resources may beinadequate. The long-term quality of these resourcesmay be affected by climate change, but improvingagricultural practices under current climate wouldoffer significant improvement under the current

108


climate. Such changes also greatly reduce pollutionunder both climate change scenarios we considered.The other side of this story is illustrated in theEdwards Aquifer study:A pumping limit imposedwith the expectation of maintaining the health ofecosystems and protecting endangered species mayprove inadequate by a significant margin if theclimate changes projected by the scenarios weconsidered come to pass.

109

Conclusions and ImplicationsChapter 6

Introduction

This study was conducted as part of a NationalAssessment effort that was designed to evaluate theimpacts of climate change and climate variability onthe United States across its various regions andincluding sectors beyond agriculture. We set out tounderstand the potential implications of climatechange for agriculture. In chapter 1 we provided anoverview of the goals of the assessment and a broad-brush portrait of forces shaping US agriculture overthe past 100 years, where US agriculture finds itselftoday, and some of the major forces that will shapeagriculture into the next century. In chapter 2 wereviewed previous studies on the impacts of climatechange on agriculture, including some of the keyfindings, how the literature has developed, andwhere some of the major gaps remain.

We reported the substantive new work of the agri-culture sector assessment in chapters 3 through 5.In chapter 3 we considered the impacts of futureclimate change on production agriculture and theUS economy. We reported a series of crop model-ing studies that examine in detail the impacts of cli-mate change on crop yield, with the intent of pro-viding a representative estimate of climate impactson US crop yields under two climate scenarios forclimates: with a CO2 concentration projected torepresent the decade of the 2030s and the 2090s.We then combined these results with estimates ofchanges in water supply, pesticide expenditures,livestock, and international trade resulting from cli-mate change to understand the combined impactson the US agricultural economy, resource use, andthe distribution of impacts in the United States byproducer and consumer and by region.

In chapter 4 we considered the question of climatevariability and extreme events, the chance that cli-mate change may cause the probability of extremeevents to change, and the potential consequences foragriculture. Many books discuss climate variability,yield variability, and how farmers cope with

variability apart from climate change. Crop insur-ance, futures markets, weather derivatives, and tech-nological options such as irrigation, storage facilities,and shelter for livestock are intricate parts of theagricultural system because of weather variability. Inno way have we covered this broad literature; wehave tried to understand the extent to which climatechange could exacerbate or reduce variability.

The subject of chapter 5 was one of the poorlyresearched areas of climate impacts on agriculture:the arena of environmental and resource implica-tions. Soil erosion, the fate of chemical residuals, andthe quality and quantity of soil and water resourcesare highly dependent on climatic conditions. Inchapter 5 we began the process of examining someof these interactions. We focused on some case stud-ies to illustrate the issues and problems that couldarise as we try to manage resource use and agricul-ture’s relationship with the environment under achanging climate. We examined the upper portion ofthe Chesapeake Bay drainage area that extendsthrough Maryland, Delaware, and Pennsylvania; theSan Antonio,Texas, area where the Edward’s Aquiferprovides most of the water supply; pesticide use andits relationship to climate; and the direct impact ofclimate on soil. This list leaves many problems unex-plored, including issues such as soil erosion, thepotential for climate to change the level of pollutantssuch as ozone that are detrimental to crops, the inter-action of agriculture with wildlife habitat, livestockwaste issues, and many others.

One of the goals of the assessment has been torespond directly to questions that stakeholders feltwere important. A considerable gap remainsbetween the questions of the stakeholder communi-ty with whom we interacted and the answers wewere able to provide. Answering many of thesequestions would require a modeling capability andprecision that we do not possess. The most funda-mentally difficult conceptual problem is torepresent completely the dynamics of social,economic, and physical interactions in their full

110


complexity. If we could understand and modelthese dynamics, we could address issues such aswhen climate change will begin to affect the agricul-tural sector, when it will be noticed, by whom, andhow they will react to it. As individuals, organiza-tions, and local governments react—or not—howwill the reactions change the relative economicposition of one farm versus another or one regionversus another? Almost any change provides anopportunity for people who are prepared for it andadjust early—and a threat for those who fail toadjust. Technological change—a force that generallyimproves economic performance—creates losersalong with winners. Although pollution regulationusually is regarded as increasing the cost of produc-tion in the industries targeted, it can create winnersamong companies that have or can create innovativesolutions to meet the environmental regulation,allowing them to win market share against theirslower-to-respond rivals. Regardless of whether cli-mate change generally improves agricultural produc-tivity in the United States—as projected in the sce-narios we investigated—or leads to losses inproductivity, as some previous forecasts have pro-jected, there will be winners and losers.

In this chapter we review our principal findings andtry to draw out the implications of these findings foradaptation and adjustment. We make only a smallstart in this direction. In this regard, the researchand assessment team we assembled for the task ofassessing climate impacts on agriculture was bestsuited to describing the impacts of climate change.Understanding what to do requires a far moredetailed engagement of those who are directlyinvolved—the farmers, legislators, research man-agers, government program managers, and localcommunities who will be affected and whoseincomes, livelihoods, and jobs are on the line. Thus,this report is a start in that process, from a team ofresearchers. We organize this chapter to answer thefour questions we identified in Chapter 1:

• What are the key stresses and issues facingagriculture?

• How will climate change and climate variabilityexacerbate or ameliorate current stresses?

• What are the research priorities that are mostimportant to fill knowledge gaps?

• What coping options exist that can buildresiliency into the system?

Key Stresses and Issues

Agricultural production is very diverse. This diversi-ty bespeaks an industry that is undergoing rapidchange. The enterprise of farming appears to havedivided into several broad categories, and the stress-es facing each group are different. Most commodi-ties are produced on large commercial farms withlarge revenues, whose operators rely principally onthe farm as a source of income, and who earn afamily income above the average of the US house-hold. A second group of farm operators run smallfarms where net income from the farm is very smallor negative; the income of the household is deter-mined by off-farm earnings of household members.Another group of farmers are near retirement or insemi-retirement. Farmers in this group typicallyown outright all the land they operate; in fact, theymay rent most of their land or have it enrolled in along-term easement program such as theConservation Reserve Program, which pays farmersof highly erodible land to maintain permanentcover on the land. A fourth group comprises farm-ers who own mid-sized farms (chapter 1).

The important features of future agriculture aremore like “forces” than “stresses.”At least in commonparlance, stress connotes a negative effect. Changein agriculture has proved to be an opportunity forsome farmers, although it threatens the financial sur-vival of others. In this regard, we identify four broadforces that will shape the future for American agri-culture over the next few decades (see chapter 1):

• Changing technology. Biotechnology andprecision agriculture are likely to revolutionizeagriculture over the next few decades—much asmechanization, chemicals, and plant breedingrevolutionized agriculture over the past centu-ry—although public concerns and environmen-tal risks of genetically modified organisms could

111

Chapter 6: Summary and Implications

slow development and adoption of crops andlivestock containing them. Biotechnology hasthe potential to improve adaptability, increaseresistance to heat and drought, and change cropmaturation schedules. Biotechnology also willgive rise to entirely new streams of productsand allow the interchange of characteristicsamong crops. Precision farming—the incorpo-ration of information technology (e.g., comput-ers and satellite technology) in agriculture—willimprove farmers’ ability to manage resourcesand adapt more rapidly to changing conditions.

• Global food production and the globalmarketplace. Increasing linkages are the ruleamong suppliers around the world. These linksare developing in response to the need toassure a regular and diverse product supply toconsumers. Meat consumption is likely toincrease in poorer nations as their wealthincreases, which will place greater pressure onresources. Climate change could exacerbatethese resource problems. Trade policy, trade dis-putes (e.g., over genetically modified organ-isms), and the development of intellectual prop-erty rights (or not) across the world could havestrong effects on how international agricultureand the pattern of trade develops.

• Industrialization of agriculture. Theaccelerating flow of information and the devel-opment of cropping systems that can be appliedacross the world will transcend national bound-aries. Market forces are encouraging variousforms of vertical integration among producers,processors, and suppliers, all of whom are driv-en, in part, to produce uniform products andassure supply despite local variations inducedby weather or other events.

• Environmental performance. The environ-mental performance of agriculture is likely to bea growing public concern in the future, and itwill require changes in production practices.Significant environmental and resource concernsrelated to agriculture include water qualitydegradation resulting from soil erosion, nutrientloading, pesticide contamination, and irrigation-

related environmental problems; land subsidenceresulting from aquifer drawdown; degraded fresh-water ecosystem habitats resulting from irrigationdemand for water; coastal water degradationfrom run-off and erosion; water quality and odorproblems related to livestock waste and confinedlivestock operations; pesticides and food safety;biodiversity impacts from landscape change (interms of habitat and germplasm); air quality, par-ticularly particulate emissions; and landscape pro-tection. Tropospheric ozone is increasingly rec-ognized as an industrial/urban pollutant thatnegatively affects crops. Agricultural use of landalso can provide open space; habitat for manyspecies; and, with proper management, a sink forcarbon. These positive environmental aspects arelikely to be valued increasingly highly.

Climate Change andCurrent Stresses

Climate change is unlikely to alter the driving forcesof agriculture over the next century in any fundamen-tal way (see chapter 1). Regional relocation andresponse to climate variability has been a part of agri-culture over the past century (chapter 1). We cannotpredict the specific consequences of climate changewith great detail (chapter 2). The regional andresource impacts of climate change may vary consid-erably (chapter 3). The new quantitative work weundertook confirmed in many ways the broad resultsof previous studies (reviewed in chapter 2). We there-fore have reached the following conclusions:

• The next 100 years. Over the next 100years and probably beyond, human-induced cli-mate change as currently modeled is unlikely toseriously imperil aggregate food and fiber pro-duction in the US, nor will it greatly increase theaggregate cost of agricultural production. Ourquantitative results—which are based on newerclimate scenarios and include a broader range ofimpacts, including effects of CO2 fertilization,changes in water resource, pesticide expendi-tures, and livestock—confirm the emerging

112


consensus in the literature and, if anything,suggest significantly more positive results thanprevious studies (chapter 2, chapter 3).

• Regional effects. There are likely to bestrong regional production effects within theUnited States; some areas will suffer significantloss of comparative (if not absolute) advantageto other regions of the country. In the scenarioswe evaluated the Lake states, the Mountainstates, and the Pacific region showed gains inproduction, whereas the Southeast, the Delta,the Southern Plains, and Appalachia generallylost. Results in the Corn Belt were generallypositive. Results in other regions were mixed,depending on the climate scenario and timeperiod. The regional results show broadly thatclimate change favors northern areas and canworsen conditions in southern areas—a resultshown by many previous studies (chapter 2,chapter 3).

• Global markets. Global market effects canhave important implications for the economicimpacts of climate change. The position of theUnited States in the world agricultural economyas both a significant food consumer andexporter means that changes in production out-side the United States lead to consumer benefitsfrom lower prices that roughly balance produc-er losses. The situation is reversed if global pro-duction changes cause world prices to rise. As aresult, the net effect on the US economy did notchange much under different global impactassumptions. The main effect was to change thedistribution of impacts among producers andconsumers. We were unable to conduct a newassessment of impacts on the rest of the world.Trade scenarios drawn from previous workshowed both small increases and decreases inworld prices (chapter 2, chapter 3)

• Effects on producers and consumers.Effects on producers and consumers often arein opposite directions, which often is responsi-ble for the small net effect on the economy. Inthe Canadian climate scenario, the absolute

effects on producers and consumers were near-ly balanced. In relative terms, the $4–5 billionlosses to producers in the Canadian scenariorepresent a 13–17 percent loss of producersincome, whereas the gains of $9–14 billion toconsumers in the Hadley scenarios representonly a 1.1–1.3 percent gain in consumer wel-fare. These losses to producers are substantial;to place these figures in context, however, agood comparison is historical changes in landvalues—the asset that ultimately would beaffected by changes in climate. In the early1980’s farmland values fell by 50 percent.Losses because of climate change are projectedover the course of three to four decades ormore and thus would likely inflict far loweradjustment costs than experienced in the early1980’s (chapter 2, chapter 3).

• Adaptation. US agriculture is a competitive,adaptive, and responsive industry that will adaptto climate change; all of the assessments we havereviewed have factored adaptation into theassessment (chapter 2). Adaptation improvedresults substantially under the Canadian scenariobut much less so under the Hadley scenario, inwhich climate change was quite beneficial toproductivity without adaptation (chapter 3).

• Policy environment. The agriculture andresource policy environment can affect adapta-tion. This conclusion is based primarily on ourreview of the literature. We did not extensivelyconsider the policy environment and its impacton adaptation in the new work we conducted.Among the policies to consider are water mar-kets, agricultural commodity programs, cropinsurance, and disaster assistance (chapter 1,chapter 2)

Several limitations have been present in past assess-ments (chapter 2). We addressed some of the mostserious of these limitations:

• We used more realistic “transient” climate sce-narios that simulated gradual climate change asa result of gradually increased atmospheric

113


CO2 and included the cooling effect ofsulfate aerosols.

• Past studies have often considered only themajor grain crops. We considered a broader setof crops including vegetable and fruit crops. Wealso considered livestock, pasture, and grazingeffects.

• We used site-level crop model results combinedwith a spatial equilibrium economic model togenerate national and regional results thatincluded more than a dozen crops. We com-pared this approach with other approaches thathad less crop detail but had other strengths. Weinvestigated trade links and implications byusing sensitivity analysis, based on previousestimates of impacts around the world.

• Previous studies have not considered changes invariability. We examined scenarios of change invariability and implications for the agriculturaleconomy, as well as the extent to which climateis a factor in existing variability in crop yields.

• We considered a more complete interaction ofeffects such as changes in water resources andpesticide expenditures on production agriculture.

• We conducted case studies of environmental-agricultural interactions to examine the poten-tial effects of climate change on the ChesapeakeBay drainage area and groundwater use in theEdward’s aquifer area.

The most important changes in this study of theeffects of climate change on production agricultureare the direct effects on crop yields (chapter 3) asthese contribute most to the aggregate economicimpacts we estimated. We conducted crop simula-tions studies at 45 sites across the United States thatwe selected to be representative of major produc-tion regions and areas that potentially could beimportant under climate change. We also comparedthese results to a more limited investigation thatused a model to estimate yields at more than 300representative sites, using a simpler crop modelingmethodology. These results reflect changes in

climate and atmospheric concentration of CO2.Effects on crop yields varied by climate scenario andsite but overall were far more positive than for manyprevious studies. Specific results that we foundfrom the two climate scenarios we investigatedinclude the following:

Effects on crop yields varied by climate scenario andsite but overall were far more positive than for manyprevious studies.

• Winter wheat. Yields increased by 10–20 per-cent under the Hadley scenario but decreasedby more than 30 percent at many sites underthe Canadian scenario; yields also were morevariable under the Canadian climate scenario.Adaptation helped to counterbalance yield loss-es in the Northern Plains but not in theSouthern Plains. Irrigated wheat productionincreased under all scenarios by 5–10 percent,on average.

• Spring wheat. Yields increased by 10–20 per-cent in 2030 under both climate scenarios.Under the Hadley scenario, yields generallyincreased up to 45 percent higher by 2090;under the Canadian scenario, however, yields in2090 showed declines of up to 24 percent.Irrigated yields of irrigated crops were negative-ly affected by higher temperatures. Adaptationtechniques, including early planting and newcultivars, helped to improve yields underall scenarios.

• Corn. Dryland corn production increased atmost sites, as a result of increases in precipita-tion under both climate scenarios. Larger yieldgains were simulated in the northern GreatPlains and in the northern Lake region, wherewarmer temperatures also were beneficial toproduction. Irrigated corn production was neg-atively affected at most sites.

• Potato. Irrigated potato yields generally fell—quite substantially at some sites—by 2090; underrainfed conditions, however, yield changes weregenerally positive. Adaptation of planting datesmitigated only some of the predicted losses.

114


There was little room for cultivar adaptationbecause predicted warmer fall and winter tem-peratures negatively affected tuber formation.

• Citrus. Yields largely benefited from thewarmer temperatures predicted under all sce-narios. Simulated fruit yield increased by 20–50percent, while irrigation water use decreased.Crop losses from freezing diminished by 65 per-cent in 2030 and by 80 percent in 2090.

• Soybean. Soybean yields increased at mostsites we analyzed; the increases were 10–20 per-cent for sites of current major production.Larger gains were simulated at northern sites,where cold temperatures now limit cropgrowth. The Southeast sites we considered inthis study experienced significant reductionsunder the Canadian scenario. Losses werereduced by adaptation techniques involving theuse of cultivars with different maturity classes.

• Sorghum. Sorghum yields generally increasedunder rainfed conditions—by 10–20 percent—as a result of increased precipitation predictedunder the two scenarios we considered.Warmer temperatures at northern sites furtherincreased rainfed grain yields. By contrast, irri-gated production was reduced almost every-where because of negative effects of warmertemperatures on crop development and yield.

• Rice. Rice yields under the Hadley scenarioincreased by 1–10 percent. Under the Canadianscenario, rice production was 10–20 percentlower than current levels at sites in Californiaand in the Delta region.

• Tomato. Under irrigated production, theclimate change scenarios generated yielddecreases at southern sites and increases atnorthern sites. These differential regionaleffects were amplified under the Canadian sce-nario as compared with the Hadley scenario.

The factors behind these more positive results var-ied but, because the same crop models were used asin many previous studies, some of the more impor-tant differences are due to aspects of the climatescenarios.

• Increased precipitation in these transient cli-mate scenarios is an important factor that con-tributes to the more positive effects for drylandcrops and explains the difference between dry-land and irrigated crop results. The benefits ofincreased precipitation outweighed the negativeeffects of warmer temperatures for drylandcrops, whereas increased precipitation had littleyield benefits for irrigated crops because waterstress is not a concern for crops that alreadyare irrigated.

• Another important factor that contributes to themore positive effects for crops is the high con-centration (660 ppm) of atmospheric CO2 inthe 2090 transient climate scenarios. Earlierstudies typically used a doubled-CO2 concentra-tion of around 555 ppm. Thus, one feature thatappears important in considering the transientscenarios with continuing increases in atmo-spheric concentration of CO2 is that the temper-ature change lags behind the CO2 increase, dueto heat uptake by the ocean.

• The coincidence of geographic patterns of pre-cipitation and crop production contributed todifferences among crops. Crops grown in theGreat Plains—where drier conditions were pro-jected, at least under the Canadian model—andcrops grown in the Southern portion of thecountry, which already sometimes suffer heatstress, were more negatively affected. Heat-loving crops such as citrus benefited, whereascrops that do well under cool conditions (suchas potatoes) suffered.

• Another factor behind the more positive resultsis that previous studies have been based ondoubled-CO2 equilibrium climate scenarios withlarger temperature increases than those exhibit-ed by these transient scenarios through 2100.

• The crop models and crop modeling approacheswere substantially the same as in previous studies.

115


We combined the crop results with impacts onwater supply, livestock, pesticide use, and shifts ininternational production to estimate impacts on theUS economy (chapter 3). This analysis allowed us toestimate regional production shifts and resource usein response to changing relative comparative advan-tage among crops and producing regions.

• The net economic effect on the US economywas generally positive, reflecting the generallypositive yield effects. The exceptions were sim-ulations under the Canadian scenario in 2030,particularly in the absence of adaptation.Foreign consumers gained in all scenarios as aresult of lower prices for US export commodi-ties. The total effects (net effect on US produc-ers and consumers plus foreign gains) were onthe order of a $1 billion loss to $14 billion gain.

• Producers’ incomes generally fell because oflower prices. Producer losses ranged fromabout $0.1 billion to $5 billion. The largest loss-es were under the Canadian scenario. Underthe Hadley scenario, producers lost because oflower prices but enjoyed considerable increasein exports; the net effect was for only verysmall losses.

• Economic gains accrued to consumers throughlower prices in all scenarios. Gains to con-sumers ranged from $2.5 to $13 billion.

• Different scenarios of the effect of climate changeon agriculture abroad did not change the netimpact on the United States very much but redis-tributed changes between producers and con-sumers. The direction of these changes depend-ed on the direction of the effect on world prices.Lower prices increased producer losses andadded to consumer benefits. Higher pricesreduced producer losses and consumer benefits.

• Livestock production and prices are mixed.Increased temperatures directly reduce produc-tivity, but improvements in pasture and grazingand reductions in feed prices resulting fromlower crop prices counter these losses.

These production changes had important implica-tions for demand on natural resources. We foundthat:

• Agriculture’s demand for water resourcesdeclined nationwide by 5–10 percent in 2030and 30–40 percent in 2090. Land under irriga-tion showed similar magnitudes of decline. Thecrop yield studies generally favored rainfed overirrigated production and showed declines ofwater demand on irrigated land (chapter 3).

• Agriculture’s pressure on land resources general-ly decreased. Area in cropland decreased 5–10percent, and area in pasture decreased 10–15percent. Animal unit months (AUMs) of grazingon western lands decreased on the order of 10percent in the Canadian scenario and increasedby 5–10 percent under the Hadley scenario(chapter 3).

We conducted case studies of interactions of theenvironment and agriculture in the Chesapeake Baydrainage basin and in the Edward’s aquifer region ofTexas. The Chesapeake Bay is one of nation’s mostvaluable natural resources, but it has been severelydegraded in recent decades and is further threatenedby climate change (chapter 5). Soil erosion andnutrient runoff from crop and livestock productionhave played a major role in the decline of the Bay.Potential effects of climate change on water qualityin the Chesapeake Bay must be considered veryuncertain because current climate models do notadequately represent extreme weather events suchas floods or heavy downpours, which can wash largeamounts of fertilizers, pesticides, and animal manureinto surface waters. In our simulations, we foundthat under the two 2030 climate scenarios, nitrogenloading from corn production increased by 17–31percent compared with current climate. Changes infarm practices by then could reduce loadings byabout 75 percent from current levels under today’sclimate or under either of the climate scenarios.

116


The Edward’s aquifer area is another region of thecountry where agriculture and resource interactionsare critical, and these interactions could be intensi-fied with climate change (chapter 5). Agriculturaluses of water compete with urban and industrialuses, and tight economic management is necessaryto avoid unsustainable use of the resource. Again,detailed regional predictions of climate are highlyuncertain but our case study illustrates the potentialvulnerabilities of a region dependent on groundwa-ter for urban, agriculture, and habitat protection. Wefound that climate change causes slightly negativewelfare changes in the San Antonio region as awhole but has a strong impact on the agriculturesector. The regional welfare loss—most of which isincurred by agricultural producers—was estimatedto be $2.2–6.8 million per year if current pumpinglimits are maintained. A major reason for the currentpumping limits is to preserve springflows that arecritical to the habitat of local endangered species. Ifspringflows are to be maintained at the currentlydesired level to protect endangered species, we esti-mate that under the two climate scenarios pumpingwould need to be reduced by 10–20 percent belowthe limit currently set, at an additional cost of $0.5–2million per year. Welfare in the nonagricultural sec-tor is only marginally reduced by the climate changesimulated by the two climate scenarios. Increasingscarcity of water is reflected in steeply rising valuesof water permits and a shift of water from agricultur-al to nonagricultural uses.

Another important resource consideration is theimpact of climate change on soil organic carbon.We were not able to conduct quantitative analysis ofthis interaction and it remains a topic that should beaddressed in future assessments. The concern isthat soil carbon may be reduced because warmingspeeds up decomposition of organic matter;increased yields predicted in many areas maycounter this effect, however, if residue is retained onthe soil surface through reduced or minimumtillage. We judged that changes in soils from climatechange are unlikely to have significant effects oncrop productivity (chapter 5). Moreover, microbialactivity in soils is diverse and therefore probably

resilient to changes in climate. With warming onemight expect a shift northward into Canada formany crops but poor soils in Canada limit the extentof movement of cropping into these areas. Stillanother concern is soil erosion if precipitationbecomes more intense. Soils that are managed withsustainable production practices, such as reducedtillage and retaining residues on the soil, producemore under either drought or excessively wet condi-tions and therefore could be a viable adaptationmeasure if weather becomes more variable. Thiswas illustrated, in part, in our Chesapeake Bay casestudy, where erosion and nutrient run-off wasreduced by 75 percent with a change in productionpractices.

Still another environmental concern is thatincreased pests and increased use of pesticidescould result in more environmental damage. Weconducted new analysis that allowed us to estimatethat pesticide expenditures were projected toincrease under the climate scenarios we consideredfor most crops and in most states we considered.Increases on corn generally were in the range of10–20 percent; increases on potatoes were 5–15percent, and increases on soybeans and cotton were2–5 percent. The results for wheat varied widely bystate and climate scenario, with changes rangingfrom approximately –15 to +15 percent. Theincrease in pesticide expenditures could increaseenvironmental problems associated with pesticideuse, but much depends on how pest control evolvesover the next several decades. Pests develop resis-tance to control methods, requiring continual evolu-tion in the chemicals and control methods used.

The increase in pesticide expenditures results inslightly poorer overall economic performance, butthis effect is quite small because pesticide expendi-tures are a relatively small share of production costs.The approach we used did not consider increasedcrop losses from pests; we implicitly assumed thatall additional losses were eliminated throughincreased pest control measures. This approachmay underestimate pest losses.

117


Another substantial additional contribution of thisassessment was consideration of the potential effectsof climate variability on agriculture (chapter 4). Amajor source of weather variability is the El Niño-Southern Oscillation (ENSO) phenomenon. ENSOphases are triggered by the movement of warm sur-face water eastward across the Pacific Ocean towardthe coast of South America and its retreat back acrossthe Pacific, in an oscillating fashion with a varyingperiodicity. Better prediction of these events wouldallow farmers to plan ahead, planting differentcrops and planting at different times. The value ofimproved forecasts of ENSO events has been estimat-ed at approximately $500 million.

ENSO can vary intensity from one event to the next;thus, prediction—particularly of the details—ofENSO-driven weather are not perfect. And, ENSOhas widely varying effects across the country. Thetemperature and precipitation effects are not thesame in all regions; in some regions the ENSO signalis relatively strong, whereas others it is weak.Moreover, the changes in weather have differentimplications for agriculture in different regionsbecause climate-related productivity constraints dif-fer among regions under neutral climate conditions.At least one (highly controversial) study projectedchanges in ENSO as a result of global warming. Wesimulated the potential impacts of these changes onagriculture and found that

• an increase in the frequency of ENSO couldcause a loss equal to about 0.8–2.0 percent ofnet farm income,

• an increase in frequency and intensity couldcause a loss of 2.5–5.0 percent of net farmincome, and

• there are differential effects on domestic pro-ducers, foreign economies and domestic con-sumers. We find gains to domestic consumersfrom increased ENSO frequency and intensitybut losses to domestic producers and toforeign economies.

More generally, climate variability is responsible forsignificant losses in agriculture and it is changes inthese extremes beyond what is captured in currentmethods of agricultural impact assessment thatmight significantly change our results. Droughts,floods, extreme heat, and frosts can damage crops orcause a complete loss of the crop for the year.Sequential years of crop loss can seriously affect theviability of a farm enterprise. Unfortunately, climatemodels do not predict extreme events and changesin variability well, so producing meaningful esti-mates of impacts is difficult. There also are limits tothe ability of crop models to predict the effects ofclimate variability because yields can depend onvery specific aspects of climate—including, forexample, how many consecutive days of high tem-peratures are experienced or whether the crop hasbeen subject to gradual hardening against cold tem-peratures. Even changes in mean conditions canchange the variability of crop yields becausechanges in means change the chance of extremeevents. We were able to examine this phenomenonexplicitly and found mixed results:

• For corn and cotton, under the climate scenarioswe used, yield variability decreased—largely as aresult of the increase in precipitation.

• Wheat yield variability decreased under theHadley scenario and increased under theCanadian scenario.

• Soybean yield variability shows a uniformincrease under the Hadley scenario but mixedresults in the Canadian scenario.

Will these predicted changes exacerbate or amelio-rate current stresses? Before answering this questiondirectly, we need to add three important caveats.First, we consider only two climate scenarios in thenew work we conducted. Although the results ofthese scenarios confirmed broad patterns that areevident in previous studies, there are large differ-ences even in the two scenarios we used. Second,the ability to predict climate at the detail requiredfor agriculture assessment (i.e., in terms of regional

118


predictions and in terms of specific features such asextreme event probabilities) is extremely limited(chapter 2, chapter 4). Third, we have not been ableto completely study all of the ways in which climatecan affect agriculture. We were particularly limitedin our ability to consider environmental interactionsand the impacts of climate variability (chapter 4,chapter 5).

Given these limits, climate change as currently mod-eled seems more likely to put downward pressureon commodity prices, with negative consequencesfor farm income. This development could putgreater pressure on farmers in marginal crop-pro-ducing regions, particularly if they are adverselyaffected by climate change through increaseddrought. An important consideration is what willhappen to foreign demand for US exports as a resultof climate change or—probably of more impor-tance—agricultural production and populationgrowth abroad. Some scenarios of climate changesuggest deteriorating conditions abroad, thus confer-ring an advantage to US farmers; other scenarios sug-gest the opposite (chapter 3). With regard to otherfactors, a review of 20- to 30-year forecasts of globalproduction by major food agriculture organizationsfound continuing trends toward declining agricultur-al commodity prices. Although these conditionswould create further stress on farm income, theycould reduce stress on resource demands (waterand land), providing more opportunities for devot-ing these resources to wildlife, recreation, or urbanresidential uses—all of which are likely to grow inthe future. Thus, these changes would amelioratewhat might otherwise be increasing competitionover these resources.

Although climate change is highly uncertain, itsgreatest stress on agriculture may be how it affectswater quality in areas that remain under intensiveproduction. This threat could come from increaseduse of pesticides, increased competition for water insome local areas, nitrogen loading of coastal areas,and soil erosion and runoff of manure and agricul-tural chemicals (chapter 1, chapter 5). Such changeswould exacerbate existing environmental problems.

Changes in climate variability also are not well pre-dicted. If variability increased (e.g., if there wereheavier rains and longer or more frequent periods ofdrought), it would further exacerbate these environ-mental problems. Increased variability is also thegreatest threat to production agriculture. Increasesin the intensity and frequency of ENSO or otherchanges in variability would increase losses. On theother hand, we found that changes in mean condi-tions had mixed effects on the variability of cropyield: For several crops, yield variability decreasedunder the climate scenarios we studied. This resultis extremely dependent on the specific scenariosexamined, however.

Research Priorities

Further research is needed in three broad areas: inte-grated modeling of the agricultural system; researchto improve resiliency of the agricultural system tochange; and several areas of climate-agriculture inter-actions that have not been extensively investigated.

Integrated Modeling of Agricultural System

The main methodology for conducting agriculturalimpact models has been to run detailed crop modelsat a selected set of sites and to use the output ofthese site models as input to an economic model.Although this approach has provided great insights,future assessments will have to integrate these mod-els to consider interactions and feedbacks, multipleenvironmental stresses (tropospheric ozone, aciddeposition, and nitrogen deposition), transient cli-mate scenarios, and global analysis and to allowstudy of uncertainty where many climate scenariosare used. The present approach, whereby crop mod-elers run models at specific sites, severely limits thenumber of sites and scenarios that can be consid-ered feasibly.

The boundaries of the agricultural system in anintegrated model must be expanded so that more ofthe complex interactions can be represented.

119


Changes in soils, multiple demands for water, moredetailed analysis and modeling of pests, and theenvironmental consequences of agriculture andchanges in climate are areas that should be incorpo-rated into one integrated modeling framework.Agricultural systems are highly interactive with eco-nomic management choices that are affected by cli-mate change. Separate models and separate analy-ses cannot capture these interactions.

Resiliency and Adaptation

Specific research on adaptation of agriculture to cli-mate change at the time scale of decades to cen-turies should not be the centerpiece of an agricultur-al research strategy. Decision making in agriculturemostly involves time horizons of one to five years,and long-term climate predictions are not very help-ful for this purpose. Instead, effort should be direct-ed toward understanding successful farming strate-gies that address multiple changes and risks—including climate change and climate variability.

There is also great need for research to betterpredict and to make better use of short-term andintermediate-term (i.e., seasonal) weather changes.

New Areas of Research

Much of the existing research has focused more nar-rowly on the effects of changes in moisture, temper-ature and elevated ambient CO2. Many other impor-tant interactions remain under-researched incomparison. Experimentation and modeling ofinteractions of multiple environmental changes oncrops (jointly changing temperature, CO2 levels,ozone, soil conditions, moisture, etc.) are needed.Experimental evidence is needed under realisticfield conditions, such as FACE experiments for CO2

enrichment. Far more research is needed on agricul-tural pests and their response to climate change,particularly in the development of models thatcould project changing ranges and incidences ofpests. There remains a need for economic analysisto better study the dynamics of adjustment tochanging conditions.

Climate-agriculture-environment interactions may beone of the more important vulnerabilities, but exist-ing research is extremely limited. Soil, water quality,and air quality should be included in a comprehen-sive study of interactions.

Perhaps one of the most serious and potentiallymost important is the need to consider changes invariability. This has proved difficult both because itdemands improved projections from climate modelsand detailed and carefully validated models of cropgrowth. To move forward, agricultural modelingmust be more closely integrated with climate model-ing so that modelers can develop better techniquesfor assessing the impacts of climate variability. Thiswork requires significant advances in climate predic-tions to better represent changes in variability, aswell as assessment of and improvements in the per-formance of crop models under extreme conditions.

Coping Options

The ultimate question for US agriculture over thenext several decades is,“Can agriculture becomemore resilient and adaptable given the many forcesthat will reshape the sector—of which climatechange is only one?” US agriculture has, in fact, beenvery adaptable and resilient along many dimensions;to stay ahead in a competitive world, however, wecan always ask: “Can it do still better?” The individ-ual farmer, agribusiness company, agronomist, orfarm-dependent community is not concerned withwhether prices are low because of climate change,technological change, or a market collapse in Asia.Similarly, if commodity prices rise sharply because ofdemand pressures, production failure in Russia, orworsening climatic conditions across the world, theimpact on farmers and resources will be similar.Each of these scenarios represents a change in therelative economic conditions across regions. Thesetypes of events and forces create short-term variabil-ity and shape long-term trends. They presentchanged conditions that are potential opportunitiesfor those that act quickly (and in the right direction)and threats to those who are slow to respond. Of

120


course, there can be real losses and real gains to dif-ferent regions; in this assessment, we have tried toillustrate such changes resulting from climatechange. The challenge for adaptation is to do as wellas possible with what the world presents. Limitingclimate change is another option for avoiding nega-tive impacts involved with climate change, but thatapproach involves much more than what happensto US agriculture. Coping options for climatechange can be divided into two broad categories:the market and policy environment for agricultureand technological response options.

The Market and Policy Environment for Agriculture

Over the past half-century, federal farm policy hasaimed to boost farm and rural incomes, smooth outthe ups and downs of commodity prices, insurefarmers against the inevitable disasters of droughtsand floods, feed the poor, improve productivity, pro-tect natural resources, and come to the aid of thesmall farmer. There have been great successes: since1950, US agricultural productivity has doubled; realworld food prices have fallen by two-thirds, so feed-ing the world is cheaper; and the average US farmhousehold is now wealthier than the average non-farm household. There also have been contradictoryand costly policies such as supply control with pro-duction-based payments and “conservation” pro-grams that idled land with only minimal environ-mental benefits.

In this new century, we must be realistic aboutinevitable market and global forces that are simplytoo powerful to change and avoid the policy pitfallsof the past half-century. As our assessment shows,the probability that climate change will increaseagricultural productivity in the United States is atleast as good as—if not better than—the probabilitythat climate change will decrease productivity.Although improved productivity is good for US con-sumers, it generally reduces income and wealthamong farmers and agricultural landholders.

Given the current structure of agriculture (chapter1) and the forces that are likely to shape agricultureover the next several decades, four broad considera-tions with regard to the market and policy environ-ment for agriculture will affect its ability to copewith climate change.

First, successful adaptation to climate change willrequire successful R&D. Traditional public R&D ispart of the research portfolio, but the engine ofinvention now is in private firms. Basic researchremains the province of the public sector. Theimportant element for the future is how to encour-age and direct the power of the private researchengine to improve environmental performance.Science-based environmental targets implementedwith market-based mechanisms can provide soundincentives for innovations that improve environmen-tal performance. Designing market-based mecha-nisms to deal with nonpoint pollution has proveddifficult; more attention is needed to assure thatwhatever mechanisms are chosen, they provideincentives for the private sector to develop andcommercialize agricultural technologies and prac-tices with improved environmental performance.

Second, the lesson from the last 50 years of agricul-tural policy is that use of broad-based commoditypolicy to fight rural poverty is an extremely bluntinstrument. These payments often end up dispro-portionately in the hands of the wealthiest farmers.Fifty years ago, when the farm population was muchpoorer than the general population, the regressiveaspects of these policies were minimal, but that isno longer true today. A goal could be to targetincome assistance far more carefully to disadvan-taged people in rural areas—many of whom are notactually farmers on any significant scale. Tying aidto the business of farming also tends merely toinflate the value of assets (mainly land) tied to farm-ing. Ultimately, the next generation of farmers paysa higher price for the land and faces a higher coststructure than if the payments had not been inplace. This situation sets the stage for anotherincome crisis when inevitable commodity price

121


variability leads to a downturn in prices. The 1996farm legislation eliminated most of these elements,replacing them with payments that ultimately wereto be phased out after seven years. Farm sectoreuphoria over the program when prices were highturned to disenchantment when prices fell. This dis-enchantment risks a drift back to programs that paypeople to produce product that depresses prices,forcing government to buy it up to prop up prices,dump stocks on the market and depress prices, andpay people not to produce. These lessons are broad-ly relevant to disadvantages and dislocations thatmay result from climate change.

Third, climate variability and its potential increasenecessarily focus attention on risk-managementstrategies. Contract production, vertical integration,forward markets, private savings, household employ-ment decisions, and weather derivatives are marketresponses to risk. These strategies are likely toevolve further, and farmers who are not adept atusing them will have to become so. Farmers canadopt technological solutions to risk—such as irriga-tion as insurance against drought or shorter matur-ing varieties against frost. If farmers adopt thesesolutions primarily to reduce variability in income,however, these strategies can increase costs andmake the farm uncompetitive with other farms thathave accepted the risk and pooled income variabili-ty through savings, contract production, or othermarket mechanisms.

Crop insurance is another response, for which thefederal government now takes some responsibility.Federal crop insurance contains a devilish publicpolicy dilemma. One aspect of insurance is what isknown in economics as “moral hazard.” The exis-tence of insurance reduces the incentive to under-take technological solutions to risks. A secondaspect of insurance is that under a pure insuranceprogram, the enrollee pays insurance premiumseach year but over several years should expect toget back in loss payments no more than he or shepaid. If the farmer can expect more, the insuranceprogram also is a subsidy program. This situationmay involve cross-subsidization among enrollees;

the subsidizers then tend to drop out, however, or—where federally managed—the entire program canrun a deficit with tax dollar support. There is a risk,then, that the desire to create a federal insuranceprogram that enrolls a large proportion of farmerswill end up as largely a subsidy program. If climatechange causes a drift toward more frequent disastersin an area, the premiums for farmers in the areawould have to be adjusted upward to maintain theprogram as a pure insurance program. Failure toadjust premiums ultimately could mean that insur-ance is paying out almost every year. A federalprogram would have difficulty, however, raising pre-miums substantially on areas that have sufferedrepeated disaster years. Ultimately, crop insuranceor a broader form of producer insurance cannotoffer much protection if an area is drifting towardreduced viability.

Fourth, environmental and resource policies need tobe realistic, tough, and market-based and adapt asconditions change and put the ultimate objectives ofthe programs at risk. These situations can be “win-win.” In the climate scenarios we examinedincreased yields and lower prices led to a reductionin resource use. In the past, acreage-reduction pro-grams took vast tracts of land out of production toboost prices. In the same way, environmentally tar-geted programs that reduce production—throughland retirement or through other types of con-straints on production practices—can offset climate-induced productivity increases, raise commodityprices, and restore income levels. These programsalso can be beneficial for the United States overall ifthe programs are targeted to generate substantialand real environmental gains. If—as projected inour analysis—use of water and land resourcesdeclines because of climate change, reallocatingresources to environmental and conservation goalsmay be more feasible. Keep in mind, however, thatwe project reduced resource use compared with areference. If far greater demand for resourcesoccurs for other reasons (e.g., demand growthabroad), we will not see these reductions comparedto current levels. Thus, again, climate change is justone of the factors that needs to be considered.

122


Technological Response Options

In addition we can think about some specific techno-logical response options that could be useful inadapting to climate change. Considerable caution isneeded in recommending specific technological solu-tions or directions for agricultural research, however,because of the remaining high degree of uncertaintyin climate projections. A decade ago, the main fear ofclimate change was drought; in the scenarios weexamined, however, precipitation over much of thecountry increased, reducing the number of irrigatedacres and the demand for water. Flooding and exces-sively wet field conditions may pose a greater threat,at least as currently projected. Rather than bet onone scenario or another, a distributed portfolio ofresearch, representing a variety of perspectives onhow the future might evolve, is needed.

The surprising finding in our analysis is that theimpact of climate change on agriculture may well bebeneficial to the US economy through the next cen-tury. It will, however, create winners and losers andcontribute to dislocation and disruption that impos-es costs on localities. Our case studies of theChesapeake Bay drainage area and the Edward’sAquifer region in Texas illustrated that local andregional effects and issues can differ substantially.Agriculture—or some types of agriculture—maywell become nonviable in some areas under climatechange. The truly difficult aspect of adaptation andadjustment is to decide when to make furtherinvestments in a particular farming practice or farm-ing region and when conditions have become soadverse that the sensible strategy is to find anotherline of work.

Although identifying specific technological responsesto climate change is difficult in view of the level ofuncertainty in predictions of climate at a local andregional level, we found that adaptations such aschanging planting dates and choosing longer seasonvarieties offset losses or further increase yields.Adaptive measures are likely to be particularly criti-cal for the Southeast because of large reductions inyields projected for some crops under the moresevere climate scenarios (chapter 3). Breeding for

response to CO2 probably will be necessary toachieve the strong fertilization effect assumed in thecrop studies. This technology is an unexploitedopportunity; the prospects for selecting for CO2

response are good. Attempts to breed for a singlecharacteristic often are unsuccessful, however, unlessother traits and interactions are considered.Breeding for tolerance to climatic stress already hasbeen heavily exploited, and varieties that do bestunder ideal conditions usually outperform other vari-eties under stress conditions as well (chapter 4).Breeding specific varieties for specific conditions ofclimate stress therefore is less likely to be successful.

Some adaptations to climate change and its impactscan have negative secondary effects. For example,an examination of use of water from the Edward’saquifer region found increased pressure on ground-water resources that would threaten endangeredspecies that rely on spring flows supported by theaquifer. Another example relates to agriculturalchemical use. An increase in the use of pesticides isone adaptation to increased insects, weeds, and dis-eases that could be associated with warming.Runoff of these chemicals into prairie wetlands,groundwater, and rivers and lakes could threatendrinking water supplies, coastal waters, recreationareas, and waterfowl habitat.

The wide uncertainties in climate scenarios; regionalvariation in climate effects; and interactions of envi-ronment, economics, and farm policy suggest thatthere are no simple and widely applicable adapta-tion prescriptions. Farmers will have to adaptbroadly to changing conditions in agriculture—ofwhich changing climate is only one factor. Somepotential adaptations that are more directly relatedto climate include the following:

• Sowing dates and other seasonalchanges. Planting two crops instead of one ora spring and fall crop with a short fallow periodto avoid excessive heat and drought in mid-sum-mer. For already warm growing areas, wintercropping could possibly become more produc-tive than summer cropping.

123


• New crop varieties. The genetic base is verybroad for many crops, and biotechnology offersnew potential for introducing salt tolerance,pest resistance, and general improvements incrop yield and quality.

• Water supply, irrigation, and drainagesystems. Technologies and management meth-ods exist to increase irrigation efficiency andreduce problems of soil degradation. In manyareas, however, economic incentives to reducewasteful practices do not exist. Increased pre-cipitation and more-intense precipitation proba-bly will mean that some areas will have toincrease their use of drainage systems to avoidflooding and waterlogging of soils.

In Summary

Climate and weather are intimately connected withagriculture. We have only scratched the surface ofunderstanding how climate might change and howthose changes would affect agriculture. While wefound generally positive effects on the country as awhole even these changes will require adjustmentand change. Southern portions of the US could suf-fer substantial losses if some of the more severe cli-mate changes projected actually occur even asNorthern regions benefit. Nearly all of theresearchers involved in this effort have been study-ing climate and agriculture interactions intensely for15 years or more. Even after this period of study,research, and analysis, we recognize that our abilityto foresee the future with great resolution is limited.

125

References

Abler, D. G. 2000. Climate change and agriculture inthe mid-Atlantic region. Climate Research 14:185–94.

———, and J. S. Shortle. 1995.Technology as an agri-cultural pollution control policy. AmericanJournal of Agricultural Economics 77: 20–32.

———, J. S. Shortle, and J. Carmichael. 2000. Climatechange, agriculture, and water quality in theU.S. Chesapeake Bay region. Available athttp://www.nacc.usgcrp.gov/sectors/agriculture/working-papers.html.

Acock, B., and M.C.Acock. 1993. Modeling approach-es for predicting crop ecosystem responses toclimate change. In International Crop ScienceI, edited by D.R. Buxton, R. Shibles, R.A.Forsberg, B.L. Blad, K.H.Asay, G.M. Paulsen, andR.F.Wilson. Madison: Crop Science Society ofAmerica, pp. 299-306.

Adams, R.M., B.H. Hurd, and J. Reilly. 1999a.Agriculture and global climate change: Areview of impacts to U.S. agriculturalresources. Arlington,VA.: PEW Center onGlobal Change.

———, B.A. McCarl, K. Segerson, C. Rosenzweig, K.J.Bryant, B.L. Dixon, R. Connor, R.E. Evenson, andD. Ojima. 1999b. The economic effects of cli-mate change on US agriculture. In Impact ofClimate Change on the United StatesEconomy, edited by R. Mendelsohn and J. E.Neumann. Cambridge; Cambridge UniversityPress.

———, C. Rosenzweig, R.M. Peart, J.T. Richie, B.A.McCarl, J.D. Glyer, R.B. Curry, J.W. Jones, K.J.Boote and L.H.Allen. 1990. Global climatechange and U.S. agriculture. Nature 345(6272):219–24.

———, R.A. Fleming, C.C. Chang, B.A. McCarl, and C.Rosenzweig. 1995. A reassessment of the eco-nomic effects of global climate change on U.S.agriculture. Climatic Change 30: 147–67.

Agriculture Sector Assessment: Report of aStakeholder/Sector Assessment Team Meeting.1999. Available at http://www.nacc.usgcrp.gov/sectors/agriculture/workshop-report.pdf.

Ahearn, M., J.Yee, E. Ball, and R. Nehring. 1999.Agricultural Productivity in the UnitedStates. Resource Economics Division,Economic Research Service, US Department ofAgriculture. Agriculture Information BulletinNo. 740. Available at http://www.econ.ag.gov/epubs/pdf/aib740/.

Alcamo, J., E. Kreileman, M. Krol, R. Leemans, J.Bollen, J.Van Minnen, M. Schaeffer, S.Toet, andB. De Vries. 1998. Global modelling of environ-mental change: An overview of Image 2.1. InGlobal Change Scenarios of the 21st Century:Results from the IMAGE 2.1 Model, edited byJ.Alcamo, R. Leemans, and E. Kreileman.Oxford: Pergamon, 3–94.

Alig, R.J., R.M.Adams, B.A. McCarl, and others. 1997.Assessing effects of mitigation strategies forglobal climate change with an intertemporalmodel of the U.S. forest and agriculture sec-tors. Environmental and ResourceEconomics 9: 259–74.

Berbecel, O., and M. Eftimescu. 1973. Effect of agrom-eteorological conditions on maize growth anddevelopment (English translation). Bucharest,Romania: Institute of Meteorology andHydrology, 10–31.

Bockstael, N.E., and K. Bell. 1998. Land use patternsand water quality: The effect of differentialland management controls. In Conflict andCooperation on Transboundary WaterResources, edited by R. Just and S. Netanyahu.Norwell, Mass.: Kluwer, 169–91.

Boote, K.J., J.W. Jones, G. Hoogenboom, and N.B.Pickering. 1998.The CROPGRO model forgrain legumes. In Understanding Options forAgricultural Production, edited by G.Y.Tsuji,G. Hoogenboom and P.K.Thornton. Dordrecht,The Netherlands: Kluwer Academic Publishers,99–128

126


Bradshaw,T.K., and B. Muller. 1998. Impact of rapidurban growth on farmland conversion:Application of new regional land use policymodels and geographic information systems.Rural Sociology 63 :1–25.

Brown, R.A., and N.J. Rosenberg. 1999. Climatechange impacts on the potential productivityof corn and winter wheat in their primaryUnited States growing regions. ClimaticChange 41: 73–107.

Bruce, J. P., M. Frome, E. Haites, H. Janzen, R. Lal, andK. Paustian. 1998. Carbon sequestration in soils.Paper presented at Soil and Water ConservationSociety’s Carbon Sequestration in SoilsWorkshop, Calgary, Alberta, Canada, May 1998.

Bouwman,A.F., and R. Leemans. 1995. The role offorest soils in the global carbon cycle. InCarbon forms and functions in forest soils,edited by W.W. McFee and J. M. Kelly. Madison,Wisc.: Soil Science Society of America.503–526.

Bunton, Lucius D. III. 1996. Order on the SierraClub’s Motion for Preliminary Injunction,Sierra Club, et al. v. Bruce Babbitt et al.MO-91-CA-069, United States District Court,Western District of Texas, Midland-OdessaDivision. August 23.

Census Bureau. 2000. Data accessed via Census Website at http://www.census.gov/population/censusdata/urpop0090.txt.

Chang, H., B.M. Evans, and D.R. Easterling. 2000.Theeffects of climate variability and change onnutrient loads in Pennsylvania watersheds.Paper presented at Association of AmericanGeographers Annual Meeting, Pittsburgh.

Chen, C.C., and B.A. McCarl. 2000. Pesticide usage asinfluenced by climate:A statistical investiga-tion.Available at http://www.nacc.usgcrp.gov/sectors/agriculture/working-papers.html.

Chen, C.C., B.A. McCarl, and R.M.Adams. 1999a.Economic implications of potential climatechange induced ENSO frequency and strengthshifts. Available at http://www.nacc.usgcrp.gov/sectors/agriculture/working-papers.html.

Chen, C.C., B.A. McCarl, and D.E. Schimmelpfennig.1999b. Yield variability as influenced by cli-mate: A statistical investigation. Available athttp://www.nacc.usgcrp.gov/sectors/agriculture/working-papers.html.

Chen, C.C., D. Gillig, and B.A. McCarl. 2000. Effects ofclimatic change on a water dependent regionaleconomy: A study of the Texas EdwardsAquifer. Available at http://www.nacc.usgcrp.gov/sectors/agriculture/working-papers.html.

Chesapeake Bay Program. 1997. The State of theChesapeake Bay, 1995. Available at http://www.chesapeakebay.net/pubs/state95/state.htm(accessed April 1999).

———. 1999. The State of the Chesapeake Bay. EPA903-R99-013 and CBP/TRS 222/108. Availableat http://www.chesapeakebay.net/pubs/sob/index.html (accessed December 1999).

Collinge, R., P. Emerson, R.C. Griffin, B.A. McCarl, andJ. Merrifield. 1993. The Edwards Aquifer:AnEconomic Perspective.TR-159. College Station:Texas Water Resources Institute,Texas A&MUniversity.

Council for Agricultural Science and Technology(CAST). 1992. Preparing U.S. agriculture forglobal climate change. Task Force Report No.119. Ames, Iowa: CAST.

Dalrymple, Dana. 1980. Spread of Semi-DwarfVarieties of Wheat and Rice in the UnitedStates: An International Perspective.AER-425.Economic Research Service, US Department ofAgriculture, June.

Darwin, R. F. 1997.World agriculture and climatechange: Current questions. World ResourceReview 9(1): 17–31.

———. 1999.A FARMer’s view of the Ricardianapproach to measuring agricultural effects ofclimate change. Climatic Change 41: 371–411.

127

References

———, M.Tsigas, J. Lewandrowski, and A. Raneses.1995. World Agriculture and Climate Change:Economic Adaptations.Agricultural EconomicReport No. 703. Washington, D.C.: NaturalResources and Environmental Division,Economic Research Service, U.S. Departmentof Agriculture.

Delécolle, R., D. Ripoche, F. Ruget, and G. Gosse,1994: Possible effects of increasing CO2 con-centration on wheat and maize crops in northand southeast France. In Implications ofClimate Change for International Agricul-ture: Crop Modeling Study, edited by C.Rosenzweig and A. Iglesias. Washington, D.C.:Environmental Protection Agency.

Dinar,A., R. Mendelsohn, R. Evenson, J. Parikh,A.Sanghi, K. Kumar, J. McKinsey, and S. Lonergan.1998. Measuring the impact of climate changeon Indian agriculture. World Bank TechnicalPaper No. 402. Washington, D.C.:World Bank.

Doorenbos, J., and W.O. Pruitt. 1977. Guidelines forPredicting Crop Water Requirements. FAOIrrigation and Drainage Paper 33. Rome: UnitedNations Food and Agriculture Organization.

Easterling, D. Recent changes in frost days and thefrost-free season in the United States. In press.Bulletin of the American MeteorologicalSociety.

Easterling,W.E. III, N.J. Rosenberg, M.S. McKenney, C.A. Jones, P.T. Dyke and J.R.Williams. 1992.Preparing the erosion productivity impact cal-culator (EPIC) model to simulate cropresponse to climate change and the directeffects of CO2. Agricultural and ForestMeteorology 59: 17–34.

———, P.R. Crosson, N.J. Rosenberg, M. McKenney,L.A. Katz, and K. Lemon. 1993. Agriculturalimpacts of and responses to climate change inthe Missouri-Iowa-Nebraska-Kansas (MINK)region. Climatic Change 24: 23–61.

Economic Research Service, USDA. 2000. FamilyFarm Report. Summary and details on orderingthe report are available at http://www.econ.ag.gov/epubs/htmlsum/aib735.htm.

Evans, L.T. 1993. Crop, Evolution and Yield.Cambridge: Cambridge University Press.

Fan, S., M. Gloor, J. Mahlman, S. Pacala, J. Sarmiento,T.Takahaski, and P. Tans. 1998. A large terrestrialcarbon sink in North America implied by atmo-spheric and oceanic carbon dioxide data andmodels. Science 282: 442–46.

Fuglie, K., N. Ballenger, K. Day, C. Klotz, M. Ollinger, J.Reilly, U.Vasavada, and J.Yee. 1996. AgriculturalResearch and Development: Public andPrivate Investments Under AlternativeMarkets and Institutions. Economic ResearchService, US Department of Agriculture.Agricultural Economic Report No. 735.Available at http://www.econ.ag.gov/epubs/pdf/aer735/.

Gleick, P.H., et al. 2000. Water: The PotentialConsequences of Climate Variability andChange. A Report of the National WaterAssessment Group for the US Global ChangeResearch Program. [Pacific Institute, USGeological Survey, US Department of Interior],Oakland, California. 151 pp.

Griffin, R.C., and C. Chang. 1991. Seasonality in com-munity water demand. Western Journal ofAgricultural Economics 16: 207–217.

Haith, D.A., R. Mandel, and R.S.Wu. 1992. GWLF—Generalized Water Loading Function: Version2.0 User’s Manual. Ithaca, N.Y.: Department ofAgricultural and Biological Engineering,Cornell University.

Hammer, G.L., D.P. Holzworth, and R. Stone. 1996.The value of skill in seasonal climate forecast-ing to wheat crop management in a regionwith high climatic variability. AustralianJournal of Agricultural Research 47(5):717–37.

128


Hardie, I.W., and P. J. Parks. 1997. Land use in a regionwith heterogeneous land quality: An applica-tion of an area base model. American Journalof Agricultural Economics, 79: 299–310.

Heimes, F. J., and R.R. Luckey. 1983. Estimating1980 Groundwater Pumpage for Irrigationon the High Plains in Parts of Colorado,Kansas, Nebraska, New Mexico, Oklahoma,South Dakota, Texas, and Wyoming. WaterResources Investigation Report 83-4123.Denver: US Geological Survey.

Holland, H.D. 1978. The Chemistry of theAtmosphere and Oceans. New York: JohnWiley & Sons.

Hyami, F., and V. Ruttan. 1985. AgriculturalDevelopment: An International Perspective.Baltimore: Johns Hopkins University Press.

Iglesias,A. and M.I. Minguez. 1994. Perspectives formaize production in Spain under climatechange. In Climate Change and Agriculture,edited by L. Harper, S. Hollinger, J. Jones and C.Rosenzweig. Madison,Wisc.: American Societyof Agronomy.

Intergovernmental Panel on Climate Change (IPCC).1996. Climate Change 1995: The Science ofClimate Change. Cambridge: CambridgeUniversity Press.

———. 1999. Available at http://www/ipcc.ch.

Irland, L.C., D.Adams, R.Alig, C.J. Betz, C.C. Chen, M.Mutchins, B.A. McCarl, K. Skog, and B.L.Sohngen. 2001. Assessing socioeconomicimpacts of climate change on U.S. forests,wood product markets, and forest recreation.In press. BioScience.

Izaurralde, C.R., R.A. Brown, and N.J. Rosenberg.1999. US regional agricultural production in2030 and 2095: Response to CO2 fertilizationand Hadley climate model (HadCM2) projec-tions of greenhouse-forced climatic change.PNNL-12252. Richland,Wash.: PacificNorthwest National Laboratory.

Johnson, D. Gale. 1983.The world grain economyand climate change to the year 2000:Implications for policy. Washington, D.C.:National Defense University Press.

Joyce, L.A., J.Aber, S.G. McNulty,V.H. Dale,A.J.Hansen, L.C. Irland, R.P. Neilson, and K. Skog.2001. Chapter 17: Potential consequences ofclimate variability and change for the forests ofthe United States. pp. 489-522 in ClimateChange Impacts on the United States: ThePotential Consequences of Climate Variabilityand Change Foundation Report, CambridgeUniversity Press, Cambridge UK, 612 pp.

Just, R., and R.D. Pope. 1978. Stochastic specificationof production function and economic implica-tions. Journal of Econometrics 7: 67–86.

Kaiser, H.M., S.J. Riha, D.S.Wilkes, and R. Sampath.1993a.Adaptation to global climate change atthe farm level. In Agricultural Dimensions ofGlobal Climate Change, edited by H. M.Kaiser and T. E. Drennen. Delray Beach, Fla.: St.Lucie Press, 136–52.

———, S.J. Riha, D.S.Wilkes, D.G. Rossier, and R.Sampath. 1993b. A farm-level analysis of eco-nomic and agronomic impacts of gradualwarming. American Journal of AgriculturalEconomics 75: 387–98.

Kane, S., J. Reilly, and J.Tobey. 1992. An empiricalstudy of the economic effects of climatechange on world agriculture. Climatic Change,21: 17–35.

———, J. Reilly, and J.Tobey. 1991. Climate change:Economic implications for world agriculture.AER No. 647. Washington DC: EconomicResearch Service.

Karl,T., and R. Knight. 1998. Secular trends in precip-itation amount, frequency, and intensity in theUnited States. Bulletin of the AmericanMeteorological Society 79: 231–41.

———, R. Knight, D. Easterling, and R. Quayle. 1996.Indices of climate change for the United States,Bulletin of the American MeteorologicalSociety 77: 279–92.

129

References

———, R. Knight, and N. Plummer. 1995. Trends inhigh-frequency climate variability in the twen-tieth century. Nature 377: 217–20.

Katz, R.W., and B.G. Brown. 1992: Extreme events ina changing climate: Variability is more impor-tant than averages. Climatic Change 21:289–302.

Kellogg, R.L., S.Wallace, K.Alt, and D.W. Goss. 1997.Potential priority watersheds for protection ofwater quality from nonpoint sources related toagriculture. Paper presented at 52nd AnnualSoil and Water Conservation SocietyConference,Toronto, Ontario.Available athttp://www.nhq.nrcs.usda.gov/land/pubs/wqpost2.html.

Kenny, G.J., P.A. Harrison, J.E. Olesen, and M.L. Parry.1993. The effects of climate change on landsuitability of grain maize, winter wheat andcauliflower in Europe. European Journal ofAgronomy 2: 325–38.

Kimball, B.A., P.J. Pinter, Jr., R.L. Garcia, R.L. LaMorte,G.W.Wall, D.J. Hunsaker, G.Wechsung, F.Wechsung, and T. Kartschall. 1995. Productivityand water use of wheat under free-air CO2enrichment. Global Change Biology 1:429–42.

Lal, R., J.M. Kimble, R.F. Follett, and C.V. Cole. 1998.The Potential of U.S. Cropland to SequesterCarbon and Mitigate the Greenhouse Effect.Chelsea, Mich.: Ann Arbor Press.

Leemans, R., and A.M. Solomon. 1993. Modeling thepotential change in yield and distribution ofthe earth’s crops under a warmed climate.Climate Research 3: 79–96.

Lewandrowski, J., and D. Schimmelpfennig. 1999.Agricultural adaptation to climate change:Issues of long-run sustainability. LandEconomics 75(1): 39–57.

———, and R. Brazee. 1993. Farm programs and cli-mate change. Climatic Change 23: 1–20.

Longley, G. 1992. The subterranean aquatic ecosys-tem of the Balcones fault zone Edwards aquiferin Texas—threats from overpumping. InProceedings of the First InternationalConference on Ground Water Ecology, editedby J.A. Stanford and J. J. Simons.Tampa, Fla.:,291–300.

McCarl B.A. 2000. Results from the national andNCAR agricultural climate change effectsassessments draft. Department of AgriculturalEconomics,Texas A&M University. Available athttp://www.nacc.usgcrp.gov/sectors/agriculture/working-papers.html.

———, K.O. Keplinger, C. Dillon, and R.L.Williams.1998. Limiting pumping from the Edwardsaquifer:An economic investigation of propos-als, water markets and springflow guarantees.Water Resources Research 35(4): 1257–68.

Mearns, L.O. 1996.The effect of change in daily andinterannual variability on CERES-wheat:A sensi-tivity study. Climatic Change 32: 257–92.

———. 1997. Mean and variance change in climatescenarios: Methods, agricultural applications,and measures of uncertainty. Climatic Change35: 367–96.

———, C. Rosenzweig, and R. Goldberg. 1992.Theeffect of changes in interannual climatic vari-ability on CERES-wheat yields: Sensitivity and2xCO2 studies, Journal of Agricultural andForest Meteorology 62: 159–89.

———, R.W. Katz, and S.H. Schneider. 1984. Extremehigh temperature events: Changes in theirprobabilities with changes in mean tempera-ture. Journal of Climate and AppliedMeteorology 23(12): 1601–1613.

———,T. Mavromatis, E.Tsvetsinskaya, C. Hays, andW. Easterling. 1999. Comparative responses ofEPIC and CERES crop models to high and lowresolution climate change scenarios. Journal ofGeophysical Research (special issue on NewDevelopments and Applications with theNCAR Regional Climate Model (RegCM))104(D6): 6623–46.

130


Mederski, H.J. 1983. Effects of water and tempera-ture stress on soybean plant growth and yieldin humid, temperature climates. In CropReactions to Water and Temperature Stressesin Humid, Temperature Climates, edited by C.D. Raper, Jr., and P. U. Kramer. Boulder, Colo.:Westview Press, 35–48.

Meehl, G.A. 1996.Vulnerability of fresh waterresources to climate change in the tropicalPacific region. Journal of Water, Air and SoilPollution 92: 203–13.

———, F. Zwiers, J. Evans,T. Knutson, L.O. Mearns,and P.Whetton. 2000.Trends in extreme weath-er and climate events: Issues related to model-ing extremes in projections of future climatechange. Bulletin of the AmericanMeteorological Society, 81(12): 427-436.

Meinke, H., and G.L. Hammer. 1997. Forecastingregional crop production using SOI phases:Anexample for the Australian peanut industry.Australian Journal of Agricultural Research48(6): 789–93.

Mendelsohn, R. 1999.The effect of climate variationon agriculture. In The Impacts of ClimateChange on the US Economy, edited by R.Mendelsohn and J. Neumann. Cambridge:Cambridge University Press.

———,W.D. Nordhaus, and D. Shaw. 1994.Theimpact of global warming on agriculture:ARicardian analysis. American EconomicReview 84(4): 753–71.

Midwestern Regional Climate Center. 1999.Midwestern Climate Information System.Available at http://mcc.sws.uiuc.edu/.

Mínguez, M.I., and A. Iglesias. 1994. Perspectives offuture crop water requirements in Spain: Thecase of maize as a reference crop. InDiachronic Climatic Changes: Impacts onWater Resources, edited by A.Angelakis. NewYork: Springer-Verlag.

Morris, S.J. 2001. Spatial distribution of fungal andbacterial biomass in southern Ohio hardwoodforest soils; fine scale variability and microscalepatterns. Soil Biology & Biochemistry. Inpress.

National Oceanic and Atmospheric Administration(NOAA). 1999. United States HistoricalClimatology Network (HCN) SerialTemperature and Precipitation Data: 1999.Available at http://ftp.ncdc.noaa.gov/pub/ushcn.

National Research Council (NRC). 1992. Chapter 34in Policy Implications of GreenhouseWarming: Mitigation, Adaptation and theScience Base. Washington, D.C.: NationalAcademy Press, 541–66.

———. 1997. Precision Agriculture in the 21stCentury.Washington, D.C.: National AcademyPress.

Nordt, L.C., L.C.Wilding, and L.R. Drees. 2000.Pedogenic carbonate transformations in leach-ing soil systems: Implications for globalchange. In Global Climate Change andPedogenic Carbonates, edited by R. Lal, J.W.Kimble, H. Eswaren, and B.A. Stewart. BocaRaton, Fla.: Lewis Publishers, 43–64.

Office of Chief Economist, Global Change ProgramOffice, USDA. 1999. Economic Analysis of U.S.Agriculture and the Kyoto Protocol (May1999). Available at http://www.usda.gov/oce/gcpo/gcponews.htm.

Office of Technology Assessment (OTA), USCongress. 1993. Preparing for an UncertainClimate—Volume I, OTA-O-567.Washington,DC: US Government Printing Office (October).

Olesen, J.E., and K. Grevsen. 1993. Simulated effectsof climate change on summer cauliflower pro-duction in Europe. European Journal ofAgronomy 2: 313–23.

Parker, D.,T. Legg, and C. Folland. 1994. Interdecadalchanges of surface temperatures since the late19th century. Journal of GeophysicalResearch 99: 14373–99.

131

References

Parry, M.L. 1990a.Agriculture and forestry. In ClimateChange:The IPCC Impacts Assessment, editedby W.J. McG.Tegarrt, G.W. Sheldon, and D.C.Griffiths. Canberra:Australian GovernmentPrinting Office.

———. 1990b. Climate change and world agricul-ture. London: Earthscan Publications.

Parton,W. J., D.S. Schimel, and D.S. Ojima. 1994.Environmental change in grasslands:Assessment using models, Climatic Change28(1-2): 111–41.

Patterson, D.T., J.K.Westbrook, R.J.V. Joyce, P.D.Lingren, and J. Rogasik. 1999.Weeds, insectsand diseases. Climatic Change 43(4): 711–27.

Paul, E.A., and F.E. Clark. 1996. Soil Microbiologyand Biochemistry. 2nd ed. San Diego:Academic Press.

———, and J. Kimble. 2000. Global climate change:interactions with soil properties.Available athttp://www.nacc.usgcrp.gov/sectors/agriculture/working-papers.html.

Paustian, K., D. Ojima, R. Kelly, J. Lackett, F.Tubiello, R.Brown, C. Izaurralde, S. Jagtap, and C. Li. 2000.Crop model comparisons: Report of workshopat Natural Resource Ecology Laboratory,September 1999. Available at http://www.nacc.usgcrp.gov/sectors/agriculture/working-papers.html.

Phillips, J., B. Rajagopalan, and C. Rosenzweig. 1999.The role of ENSO in determining climate andmaize yield variability in the U.S. cornbelt.International Journal of Climatology 19(8):877–88.

Plucknett, D.L., and D.L.Winkelmann. 1996.Technology for sustainable agriculture. In KeyTechnologies for the 21st Century. New York:Freeman, 133–38.

Polsky, C., J.Allard, R. Crane, and B.Yarnal. 2001. Themid-Atlantic region and its climate: Past, present,and future. Climate Research, in press.

Reilly, J., Hohmann, N., and S. Kane. 1994. Climatechange and agricultural trade: who benefits,who loses? Global Environmental Change:Human and Policy Dimensions 4(1): 24–36.

———, et al. 1996. Agriculture in a changing cli-mate: Impacts and adaptations. In ClimateChange 1995: Impacts, Adaptations andMitigation of Climate Change: Scientific-Technical Analyses, edited by R.T.Watson, M.C. Zinyowera, and R.H. Moss. Cambridge:Cambridge University Press, 427–69.

Reilly, J., (ed). 1999. Climate change: impacts on agri-culture (special issue). Climatic Change 43(4):645-793.

———, and D. Schimmelpfennig. 1999. Agriculturalimpact assessment, vulnerability, and the scopefor adaptation. Climatic Change, 43(4):745–88.

———, and K. Fuglie. 1998. Future yield growth infield crops: What evidence exists. Soil andTillage Research 47: 275–90.

———, N. Hohmann, and S. Kane. 1993. Climatechange and agriculture: Global and regionaleffects using an economic model of interna-tional trade. MIT-CEEPR 93-012WP. Cambridge,Mass.: Center for Energy and EnvironmentalPolicy Research, Massachusetts Institute ofTechnology.

Riha, S.J., D.S.Wilks, and P. Simoens, 1996. Impact oftemperature and precipitation variability oncrop model predictions. Climatic Change 32:293–311.

Rosenberg, N.J. (ed.). 1993. Towards an IntegratedAssessment of Climate Change:The MINKStudy. Boston: Kluwer Academic Publishers.

———, M.S. McKenney,W.E. Easterling III, and K.L.Lemon. 1992.Validation of EPIC model simula-tions of crop response to current climate andCO2 conditions: Comparisons with census,expert judgement and experimental plot data.Climatic Change 36(2): 35–51.

132


Rosenzweig, C., and A. Iglesias (eds.). 1994:Implications of Climate Change forInternational Agriculture: Crop ModelingStudy. EPA 230-B-94-003.Washington, D.C.: USEnvironmental Protection Agency.

———, and D. Hillel. 1998. Climate Change and theGlobal Harvest: Potential Impacts of theGreenhouse Effect on Agriculture. New York:Oxford University Press, 324 pp.

———. and F.N.Tubiello. 1997. Impacts of FutureClimate Change on Mediterranean Agriculture:Current Methodologies and Future Directions.Mitigation and Adaptation Strategies forClimate Change 1: 219–32.

———, L.H.Allen, Jr., L.A. Harper, S.E. Hollinger, andJ.W. Jones. (Eds.). 1995. Climate Change andAgriculture:Analysis of Potential Interna-tional Impacts. American Society of AgronomySpecial Publication No. 59. Madison,WI, 382 pp.

———, and M.L. Parry. 1994. Potential impact of cli-mate change on world food supply. Nature367 (6459): 133–38.

———, M.L. Parry, and G. Fischer. 1995.World foodsupply. In As Climate Changes: InternationalImpacts and Implications, edited by K.M.Strzepek and J.B. Smith. Cambridge and NewYork: Cambridge University Press, 27–56.

Rötter, R., and C.A. van Diepen. 1994. Rhine BasinStudy.Vol.2, Climate Change Impact on CropYield Potentials and Water Use. SC-DLOReport 85.2. Wageningen and Lelystad, theNetherlands.

———, and S.C. van de Geijn. 1999. Climate changeeffects on plant growth, crop yield, and live-stock. Climatic Change 43(4): 651–81.

Sage, R.F. 1995.Was low atmospheric CO2 during thepleistocene a limiting factor for the origin ofagriculture? Global Change Biology 1: 93–106.

Sanghi,A., and R. Mendelsohn. 1999. The impact ofglobal warming on Brazilian and Indian agricul-ture. New Haven, Conn.:Yale FES.

———, R. Mendelsohn, and A. Dinar. 1998. The cli-mate sensitivity of Indian agriculture. InMeasuring the Impact of Climate Change onIndian Agriculture, edited by A. Dinar, R.Mendelsohn, R. Evenson, J. Parikh,A. Sanghi, K.Kumar, J. McKinsey, and S. Lonergan.TechnicalPaper No. 402.Washington, D.C.:World Bank.

Schimmelpfennig, D., J. Lewandrowski, J. Reilly, M.Tsigas, and I. Parry. 1996. AgriculturalAdaptation to Climate Change: Issues ofLongrun Sustainability. Agricultural EconomicsReport No. 740. Washington, D.C.: EconomicResearch Service, US Department of Agriculture.

Segerson, K., and B. Dixon. 1999. Climate change andUS agriculture: The role of farmer adaptation.In The Economic Impacts of Climate Changeon the Economy of the United States, editedby R. Mendelsohn and J. Neumann. Cambridge:Cambridge University Press.

Semenov, M.A., and E. Barrow. 1997. Use of a stochas-tic weather generator in the development ofclimate change scenarios. Climatic Change 35:397–414.

———, J.R. Porter, and R. Delecolle. 1993a. Climatechange and the growth and development ofwheat in the UK and France. EuropeanJournal of Agronomy 2: 293–304.

———. 1993b. Simulation of the effects of climatechange on growth and development of wheatin the UK and France. In The Effect of ClimateChange on Agricultural and HorticulturalPotential in Europe, edited by G.J. Kenny, P.A.Harrison, and M.L. Parry. Oxford: Environmen-tal Change Unit, University of Oxford, 121–36.

Shaw, R.H. 1983. Estimates of yield reductions incorn caused by water and temperature stress.In Crop Reactions to Water and TemperatureStresses in Humid, Temperature Climates,edited by C.D. Raper, Jr., and P.U. Kramer.Boulder, Colo.:Westview Press, 49–66.

133

References

Shortle, J.S., and D.G.Abler. 1997. Nonpoint pollu-tion. In The International Yearbook ofEnvironmental and Resource Economics1997/1998, edited by H. Folmer and T.Tietenberg. London: Edward Elgar.

———, D.G.Abler, and A. Fisher. 1999. Developingsocioeconomic scenarios: Mid-Atlantic case.Acclimations 7: 7–8. Available at http://www.nacc.usgcrp.gov/newsletter/1999.08/issue7.pdf (accessed September 1999).

Shriner, D.S., and R.B. Street. 1998. North America. InThe Regional Impacts of Climate Change:AnAssessment of Vulnerability, edited by R.T.Watson, M. C. Zinyowera, and R. H. Moss.Cambridge: Cambridge University Press,253–330.

Smith, J.B., and D.Tirpak (eds.). 1989. The PotentialEffects of Global Climate Change on theUnited States. EPA-230-05-89-050.Washington,D.C.: US Environmental Protection Agency.

Solow,A.R., R. F.Adams, K.J. Bryant, D.M. Legler, J.J.O’Brien, B.A. McCarl, and W.I. Nayda. 1998. Thevalue of improved ENSO prediction to U.S.agriculture. Climatic Change 39: 47–60.

Southerland, J., J.C. Randolph, M. Hableck, O.Doering, R.A. Pfeifer, G. Rao, and J. Johnston.1999. Consequences of future change andchanging climate variability on corn yields inthe Midwest US. In press, Agriculture,Ecosystems, and Environment.

Stockle, C.O., J.R.Williams, N.J. Rosenberg ,and C.A.Jones. 1992a. Estimation of the effects ofCO2-induced climate change on growth andyield of crops. I. Modification of the EPICmodel for climate change analysis. AgriculturalSystems 38: 225–38.

———. 1992b. Estimation of the effects ofCO2-induced climate change on growth andyield of crops. II. Assessing the impacts onmaize, wheat and soybean in the midwesternUSA.Agricultural Systems 38: 239–56.

Timmermann,A., J. Oberhuber,A. Bacher, M. Each, M.Latif, and E. Roeckner. 1999. Increased El Niñofrequency in a climate model by future green-house warming. Nature, 398(6729): 694–97

Tobey, J., J. Reilly, and S. Kane. 1992. Economic impli-cations of global climate change for worldagriculture. Journal of Agricultural andResource Economics 17(1): 195–204.

Trimble, S.W. 1999. Decreased rates of alluvial sedi-ment storage in Coon Creek Basin,WI,1975–93.Science 285: 1244–46.

Tsuji, G.Y., G. Uehara, and S. Balas (Eds.). 1994. DSSATv3. University of Hawaii, Honolulu, Hawaii.

Tubiello, F.N., C. Rosenzweig, B.A. Kimball, P.J. Pinter,Jr., G.W.Wall, D.J. Hunsaker, R.L. Lamorte, and R.L. Garcia. 1999. Testing CERES-Wheat withFACE data: CO2 and water interactions.Agronomy Journal 91: 1856–65.

———, S. Jagtap, C. Rosenzweig, J. Jones, D. Ojima,and K. Paustian. 2000. US national assessment:Effects of climate change on agricultural pro-duction of wheat, maize, soybean, sorghum,rice, potato, tomato, hay, and citrus.Available athttp://www.nacc.usgcrp.gov/sectors/agriculture/working-papers.html.

US Department of Agriculture (USDA). 1991–1997,Agricultural Statistics.Washington D.C.: USDA.

———. 1997. Agricultural Resources andEnvironmental Indicators.Washington D.C.:Economic Research Service.

US Department of Agriculture (USDA). 1999a.Agricultural Chemical Usage. PCU-BB.Washington, D.C.: National AgriculturalStatistics Service,Agricultural Statistics Board,Economic Research Service, US Department ofAgriculture.Available at http://usda.mannlib.cornell.edu/reports/nassr/other/pcu-bb/#top.

———. 1999b. 1997 Census of Agriculture.Washington, D.C.: National AgriculturalStatistics Service.Available at http://www.nass.usda.gov/census/ (accessed April 1999).

134

US Department of Agriculture (USDA), NationalAgricultural Statistical Service. 1999c. 1999Agricultural Statistics. Available at http://www.usda.gov/nass/pubs/agr99/acro99.htm.

US Department of Agriculture, National AgriculturalStatistical Service. 1999d. Historical Data.Available at http://www.usda.gov/nass/pubs/histdata.htm.

US Geological Survey (USGS). 1997. Recharge toand Discharge from the Edwards Aquifer inthe San Antonio Area, Texas.Available athttp://tx.usgs.gov/reports/district/98/01/index.html.

Wang,Y.P., J.R. Handoko, and G.M. Rimmington. 1992.Sensitivity of wheat growth to increased airtemperature for different scenarios of ambientCO2 concentration and rainfall in Victoria,Australia—a simulation study. ClimaticResearch 2: 131–149.

Wetherald, R.T., and S. Manabe, 1999. Detectability ofsummer dryness caused by greenhouse warm-ing. Climatic Change, 43(3): 495-511.

Williams, J.R., C.A. Jones, and P.T. Dyke. 1989. EPIC-Erosion/Productivity Impact Calculator, TheEPIC Model.Temple,Tex.: U.S. Department ofAgriculture-Agricultural Research Service.

Winters, P., R. Murgai,A. de Janvry, E. Sadoulet, and G.Frisvold. 1999. Climate change and agriculture:Effects on developing countries. In GlobalEnvironmental Change and Agriculture:Assessing the Impacts, edited by G. Frisvoldand B. Kuhn. Northampton, Mass.: EdwardElgar Publishing.

Wolf, J., 1993a: Effects of climate change on wheatand maize production in the EC. In The Effectof Climate Change on Agricultural andHorticultural Potential in Europe, edited byG. J. Kenny, P.A. Harrison, and M. L. Parry. Ox-ford: Environmental Change Unit, University ofOxford, 93–119.

Wolf, J. 1993b: Effects of climate change on wheatproduction potential in the EuropeanCommunity. European Journal of Agronomy2: 281–92.

Woodward, F.I. 1993. Leaf responses to the environ-ment and extrapolation to larger scales. InVegetation Dynamics and Global Change,edited by A.M. Solomon,A.M. and H.H. Shugart.New York: Chapman and Hall, 71–100.

Yarnal, B. 1999. Mid-Atlantic region’s climate. In Mid-Atlantic Regional Assessment (MARA) DraftPreliminary Report on Impacts from ClimateChange. University Park: Pennsylvania StateUniversity, 27–33.

Zwiers, F.W., and V.V. Kharin. 1998. Changes in theextremes of the climate simulated by CCCGCM2 under CO2 doubling. Journal ofClimate 11: 2200–2222.


135

Appendix A: Agriculture SectorAssessment Working Papers

This volume contains the principal findings of the agricultural assessment. Detailed reports of results and meth-ods are reported in the following working paper reports. All of these are available at http://www.nacc.usgcrp.gov/sectors/agriculture/workshop-report.pdf. Most of the substantive new modeling and statistical analysis inthese reports has now been accepted for publication or is in review in peer-reviewed scientific journals.1999. Agricultural Sector Assessment: Report of a Stakeholder/Sector Assessment Team Meeting.

Abler, D., J. Shortle, and J. Carmichael. 2000. ClimateChange,Agriculture, and Water Quality in the U.S.Chesapeake Bay Region.Working Paper.

Chen, C., Gillig, Dhazn, and McCarl, B.A. 2000.Effects of Climatic Change on a Water DependentRegional Economy:A Study of the Texas EdwardsAquifer.

Chen, C. and McCarl, B.A. 2000. Pesticide Usage asInfluenced by Climate: A Statistical Investigation.

Chen, C. and McCarl, B.A. 2000. EconomicImplications of Potential Climate ChangeInduced ENSO Frequency and Strength Shifts.

Chen, C., McCarl, B.A. and Schimmelpfennig, D.2000. Yield Variability as Influenced by Climate: AStatistical Investigation.

Izaurralde, R. C., R.A. Brown, and N. J. Rosenberg.1999. U.S. regional agricultural production in2030 and 2095: response to CO2 fertilization andHadley Climate Model (HADCM2) projections ofgreenhouse-forced climatic change. Rep. No.PNNL-12252. Pacific Northwest NationalLaboratories, Richland,WA. 42 pp.

McCarl, B.A. 2000. Results from the National andNCAR Agricultural Climate Change EffectsAssessments.

Paul, E.A. and J. Kimble, 2000. Global ClimateChange: Interactions with Soil Properties.

Paustian, K., D. Ojima, R. Kelly, J. Lackett, F.Tubiello, R.Brown, C. Izaurralde, S. Jagtap, and C. Li, 2000.Crop Model Analysis of Climate and CO2 Effects,Workshop Report - Preliminary Draft.

Tubiello, F. N., C. Rosenzweig, R.A. Goldberg, S.Jagtap, and J.W. Jones, 2000. U.S. NationalAssessment Technical Report Effects of ClimateChange on U.S. Crop Production Part I:Wheat,Potato, Corn, and Citrus.

Appendix B: Abbreviations

AOGCM atmosphere-ocean general circulation model

ASM Agriculture Sector Model

AUM animal unit month

BC Blaney-Criddle procedure

CAST Council on Agricultural Science and Technology

CCC Canadian Climate Center

CRP Conservation Reserve Program

CV Coefficient of Variation

DOC dissolved organic carbon

DOE US Department of Energy

DSSAT Decision Support Systems for Agrotechnology Transfer

EA Edwards Aquifer

EDSIM Edwards Aquifer Simulation Model

EFS environmentally friendly, smalleragriculture

ENSO El Niño Southern Oscillation

EPA US Environmental Protection Agency

EPIC erosion productivity impact calculator

EPRI Electric Power Research Institute

EQIP Environmental Quality Incentives Program

ERS Economic Research Service

FAIR Federal Agricultural Improvement Reform Act of 1996

GCM general circulation model

GCRA Global Change Research Act

GDP gross domestic product

GHG greenhouse gas

GISS Goddard Institute for Space Studies

GMO genetically modified organism

GPS global positioning system

GWLF Generalized Watershed Loading Functions

GWP Global Warming Potential

IPCC Intergovernmental Panel on Climate Change

IPM integrated pest management

M & I municipal and industrial

MINK Missouri, Iowa, Nebraska, Kansas

MMT million metric tons

NOAA US National Oceanographic and Atmospheric Administration

NRC US National Research Council

NREL Natural Resource Ecology Laboratory

OTA US Office of Technology Assessment

Pg C Petagrams of carbon

PNNL Pacific Northwest National Laboratory

R&D research and development

SQ status quo

t/ha tonnes per hectare

UKMO United Kingdom Meteorology Office

USDA US Department of Agriculture

USGCRP US Global Change Research Program

USGS US Geological Survey

WRP Wetland Reserve Program

WUE water use efficiency

136

Color Plates

Figure 3.2: Simulated yield changes from baseline for dryland corn grown in (a) 2030 and (b) 2095 under climate scenarios projected withthe HadCM2 general circulation model.

Yield diff. (Mg ha-1)Had2095, CO2 = 560 ppm



Corn yield (Mg ha-1)Baseline, CO2 = 365 ppm

(a)

(b)

Figure 3.3: Simulated yield changes from baseline for winter wheat grown in (a) 2030 and (b) 2095 under climate scenarios projectedwith the HadCM2 general circulation model.



Wheat yield (Mg ha-1)Baseline, CO2 = 365 ppm


(a)

(b)

Documents

ii PUBLISHED BY THE PRESS SYNDICATE OF THE UNIVERSITY OF CAMBRIDGE The Pitt Building,Trumpington Street,Cambridge,United Kingdom CAMBRIDGE UNIVERSITY PRESS The Edinburgh Building,