NITROGEN DYNAMICS AND GREENHOUSE GAS PRODUCTION IN

YAQUI VALLEY SURFACE DRAINAGE WATERS

A DISSERTATION

SUBMITTED TO THE DEPARTMENT OF GEOLOGICAL AND

ENVIRONMENTAL SCIENCES

AND THE COMMITTEE ON GRADUATE STUDIES

OF STANFORD UNIVERSITY

IN PARTIAL FULFILLMENT OF THE REQUIREMENTS

FOR THE DEGREE OF

DOCTOR OF PHILOSOPHY

John Arthur Harrison

January 2003

ii

iii

Copyright by John Arthur Harrison 2003

All Rights Reserved

iv

I certify that I have read this dissertation and that in my opinion it is fully adequate, in scope and quality, as dissertation for the degree of Doctor of Philosophy.

__________________________________ Pamela Matson, Principal Advisor

I certify that I have read this dissertation and that in my opinion it is fully adequate, in scope and quality, as dissertation for the degree of Doctor of Philosophy.

__________________________________ Robert Dunbar

I certify that I have read this dissertation and that in my opinion it is fully adequate, in scope and quality, as dissertation for the degree of Doctor of Philosophy.

__________________________________ Scott Fendorf

I certify that I have read this dissertation and that in my opinion it is fully adequate, in scope and quality, as dissertation for the degree of Doctor of Philosophy.

__________________________________ Peter Vitousek

Approved for the University Committee on Graduate Studies

__________________________________

v

ABSTRACT

Agricultural runoff is thought to constitute a globally important source of

the greenhouse gas nitrous oxide (N2O), and may also be a significant source of the

greenhouse gases methane (CH4) and carbon dioxide (CO2). However, production of

N2O, CH4, and CO2 in polluted aquatic systems is poorly understood and scarcely

reported, especially in low-latitude (0-30º) regions where rapid agricultural

intensification is occurring. We measured N2O, CH4, and CO2 emissions, dissolved N2O

concentrations, and factors likely to control rates of greenhouse gas production in Yaqui

Valley drainage canals receiving agricultural and mixed agricultural/urban inputs.

Average per-area N2O flux in both purely agricultural and mixed urban/agricultural

drainage systems (16.5 ng N2O-N cm-2 hr-1) was high compared to other fresh water

fluxes, and extreme values ranged up to 244.6 ng N2O-N cm-2 hr-1. These extremely high

N2O fluxes occurred during green algae blooms, when organic carbon, nitrogen, and

oxygen concentrations were high, and only in canals receiving pig-farm and urban inputs,

suggesting an important link between land-use and N2O emissions. N2O concentrations

and fluxes correlated significantly with water column concentrations of nitrate,

particulate organic carbon and nitrogen, ammonium, and chlorophyll a. A multiple linear

regression model including ammonium, dissolved organic carbon, and particulate organic

carbon was the best predictor of [N2O] (r2 = 52%). Despite high per-area N2O fluxes, our

estimate of regional N2O emission from surface drainage (20,869 kg N2O-N yr-1; 0.046%

of N-fertilizer inputs) was low compared to values predicted by algorithms used in global

budgets. All canals where we measured CO2 were net-heterotrophic. CH4 fluxes, though

variable (range: -1372 – 3990 mg CH4-C m-2 d-1), were generally quite high (mean: 3208

ng-C cm-2 hr-1) compared with fluxes from similar systems in other locations. C gas

evasion from streams was several times greater than stream export of organic C via

lateral transport. During algae bloom conditions we observed rapid and complete

oxidation and reduction of an entire drainage canal. This has not previously been

vi

reported and may have important implications for in-stream transformations and

downstream transfer of N, iron, manganese, and sulfur as well as the production of N2O,

CH4, and CO2.

vii

RESUMEN

Se ha creído que las aguas residuales agrícolas son importantes fuentes del

gas de invernadero, oxido nitroso (N2O), y también puede ser importante por la

producción de dióxido de carbón (CO2), y metano (CH4) (también gases invernaderos).

Sin embargo, la producción de N2O, CH4, y CO2 en sistemas acuáticos contaminados no

esta bien comprendido, ni bien reportado, especialmente en zonas de bajas latitudes

tropicales y subtropicales (0-30º), y en zonas de rápido crecimiento agrícola. Medimos

emisiones de N2O, CH4, y CO2, concentraciones de N2O disuelto, y factores que puede

controlar la taza de producción de gases de invernaderos en los drenes agrícolas y

agricolas/urbanas en el Valle del Yaqui. El flujo promedio (por-área) de N2O en ambos

drenes con aportación sólo agrícolas y drenes con influencia de una mezcla de fuentes

agrícolas y urbanas (16.5 ng N2O-N cm-2 hr-1) fue alto en comparición con los flujos de

otros sitios de agua dulce, y los valores extremos fueron hasta 244.6 ng N2O-N cm-2 hr-1.

Estos valores extremos de flujo de N2O pasaron durante floraciones de algas verdes,

mientras concentraciones de carbón orgánico, nitrógeno reactivo, y oxigeno fueron altas.

También sucedieron solamente en drenes, lagunas de oxidación de desechos de granjas

porcicolas y de zonas urbanas. Esto sugiere una conexión importante entre el uso del

suelo y las emisiones de N2O. Concentraciones y flujos de N2O correlacionaron en una

manera significativa con concentraciones (en la columna de agua) de nitrato, carbón y

nitrógeno orgánico particular, amonio, y clorofila. Un modelo de múltiple retroceso lineal

incluyendo amonio, carbón orgánico disuelto, y carbón orgánico particular como

variables independientes fue el mejor pronosticador de [N2O] (r2 = 52%). Aunque flujos

de N2O por área fueron altos, nuestro estimado de emisión regional de N2O de drenaje

superficial (20,869 kg N2O-N yr-1; 0.046% of N-fertilizer inputs) fue bajo comparado a

valores pronosticados por métodos de algoritmos utilizados en estadares globales. Todos

los drenes donde medimos CO2 fueron heterotróficos. Flujos de CH4, aunque variables

(rango: -1372 a 3990 mg CH4-C m-2 d-1), fueron generalmente altos (promedio: 3208 ng-

C cm-2 hr-1) comparado con los flujos de sistemas parecidos en otras localidades. El gas

viii

de C se desprende a la atmósfera desde los drenes y fue un sendero significativo para

proporcionar C del sistema, en varias ocasiones mas grande que la exportación de C

orgánico por transporta lateral. En mediciones que duraron 24 horas de un dren,

observamos cambios químicos muy rápidos que correspondieron a cambios en la

concentración de oxigeno (O2). Durante este periodo, la nitrificación y la producción de

N2O pararon por lo menos en 8 horas, y la desnitrificación se detuvo al menos en 6 horas.

Concentraciones de CH4 y CO2 también mostraron un ciclo diel fuerte. Los cambios

químicos observados afectaran profundamente las transformaciones y la transferencia de

N, fierro, manganeso, azufre, y también la producción de los gases de invernaderos N2O,

CH4, y CO2.

ix

ACKNOWLEDGMENTS

Thanks to my dissertation committee for support and guidance: Pamela Matson, Peter

Vitousek, Robert Dunbar, and Scott Fendorf.

Thanks to members of the Matson and Vitousek Labs and to the Yaqui Valley Research

Group for comradery, support, and guidance throughout my Ph.D.: Kathleen “Kitty”

Lohse, Ted Schuur, Sharon Hall, Michael Beman, Carrie Nielsen, Karen Carney, Barbara

Cade-Menun, Peter Jewett, Jeanne Panek, Bill Riley, Amy Luers, Eve Hinkley, Toby

Ahrens, Lee Addams, Roz Naylor, Wally Falcon, Greg Asner, Carlos Valdes, and others.

Special thanks to Ivan Ortiz-Monasterio for his invaluable advice and logistical support.

Thanks to Gustavo Vasques, Dennis Rogers, Luis Mendez, Manuel Muñoz, Lindley

Zerbe, Juan Bustamante, Don Chayo, Igenio, Jesus Nieblas, and “Los Chicos” (Cuervo,

Gallo, Oso, y El Viejo) for assistance in the field and laboratory as well as for inspiration.

Thanks also to Lori McVay, Elaine Andersen, and Lorraine Araujo for logistical support

on the home front.

I am also grateful to the following agencies and organizations for their financial support

of my thesis research: The National Science Foundation for Dissertation Enhancement

and Predoctoral Fellowship Awards, NASA for an Earth Systems Science Fellowship,

The Packard Foundation, The Bechtel Foundation, NASA’s Land-Use/Land-Cover

Change program, and Stanford University’s McGee Fund.

Finally, many heartfelt thanks to my dear family and friends, whose love and support

have made my thesis possible: Penny, Marvin, and Rebecca Harrison, Katie Prager,

Michael Kiparsky, Jeremy Eddy, Sarah Jane Lapp, Darcy Karakelian, Joern Hoffman,

Ana Porter, Jeff Dukes, Julia Verville, High Speed Buck, the Fine Tuners, and many

others.

x

DEDICATION

To the residents of the Yaqui Valley, and to all those who supported me during this

work…

xi

TABLE OF CONTENTS

Abstract and Abstract in Spanish ....................................................................................... iv

Acknowledgments............................................................................................................viii

Dedication .......................................................................................................................... ix

List of tables ....................................................................................................................xivi

List of figures .................................................................................................................. xvv

INTRODUCTION............................................................................................................... 1

Chapter 1: Patterns and Controls of Nitrous Oxide Emissions from Waters Draining a

Subtropical Agricultural Valley .................................................................................... 6

Introduction ................................................................................................................... 6

Site Description ............................................................................................................. 8

Watershed................................................................................................................ 8

Drainage canals ..................................................................................................... 11

3. Methods......................................................................................................................... 12

Measurement Program .......................................................................................... 12

Gas Collection and Analysis ................................................................................. 13

Determination of Gas Transfer Coefficient........................................................... 14

Other Chemical Analyses...................................................................................... 15

Denitrification and Nitrification Potential Assays ................................................ 17

Intact Core Experiments........................................................................................ 18

Sectioned Core Experiment................................................................................... 19

Statistical Analyses ............................................................................................... 20

Regional Flux Estimates........................................................................................ 21

Results and Discussion................................................................................................ 21

N2O fluxes ............................................................................................................. 21

Potential Controls on N2O Emission..................................................................... 22

xii

Nitrate.................................................................................................................... 28

Organic Carbon ..................................................................................................... 32

Ammonium............................................................................................................ 33

pH ......................................................................................................................... 36

Temperature .......................................................................................................... 37

Discharge and Flow Rate ...................................................................................... 37

Salinity .................................................................................................................. 38

Algae Blooms (Chlorophyll a) and Related Variables.......................................... 39

Multiple Regression Models ................................................................................. 39

Regional Significance of Aquatic N2O Flux and Implications for Global

Estimates ......................................................................................................... 40

N Transfer to Estuaries.................................................................................... 42

Denitrification ................................................................................................. 44

Chapter 2: Rapid-onset Anoxia Affects Nitrogen Transfer and Greenhouse Gas

Production in Mexican Stream.................................................................................. 466

Methods....................................................................................................................... 59

Gas Concentrations ............................................................................................... 59

Other Chemical Analyses...................................................................................... 60

Chapter 3: Patterns and Controls of Methane (CH4) and Carbon Dioxide (CO2)

Evasion from Urban and Agricultural Drainage waters.............................................. 62

Introduction ............................................................................................................... 622

Methods....................................................................................................................... 62

Study Site .............................................................................................................. 62

Drainage canals ..................................................................................................... 63

Experimental Design ............................................................................................. 63

Gas Collection and Analyses................................................................................. 63

Other Analyses ...................................................................................................... 64

Statistical Analyses ............................................................................................... 65

Results and Discussion................................................................................................ 65

xiii

Methane................................................................................................................. 65

Patterns and Magnitudes of Flux..................................................................... 65

Importance of CH4 Evasion at the Landscape Scale ....................................... 67

Carbon Dioxide ..................................................................................................... 68

Simple and Multiple Regressions.................................................................. 689

Patterns and magnitudes of flux ...................................................................... 69

Importance of CO2 evasion at landscape scale................................................ 70

Total Carbon.......................................................................................................... 71

Appendix A: Abbreviations .............................................................................................. 73

Appendix B: Estmation of Gas Transfer Coefficient ........................................................ 74

Appendix C: Poisoned Versus Unpoisoned Septum Vials ............................................... 79

Appendix D: N2:Ar and N2:O2 Measurements Via Membrane Inlet Mass Sectrometer,

Dssolved Gs Aalysis (MIMS-DGA) ........................................................................... 81

Appendix E: Some Important Unanswered Questions...................................................... 82

Appendix F: Some Tips For Future Yaqui Researchers ................................................... 84

xiv

LIST OF TABLES

Number Page Table 1: Current estimates of global N2O sources .......................................................... 7

Table 2: Canal watershed areas and mean annual discharges ....................................... 12

Table 3: Correlations between measured N2O concentration and independent

variables such as temperature, DO, Chl a, dissolved inorganic nitrogen

[DIN], dissolved organic carbon [DOC], particulate organic carbon

[POC], particulate organic nitrogen [PON], [NO3-], [NH4+], [PO43-],

[NO3-], [NH4+], pH, salinity, and turbidity...................................................... 35

Table 4: CH4 fluxes from tropical and subtropical freshwater systems, including the

ones in this study............................................................................................. 66

Table 5: Correlations between measured CO2 flux (ug-CO2-C cm-2 hr2) and

independent variables such as temperature, DO, Chl a, dissolved inorganic

nitrogen [DIN] defined as [NO3- + NO2- + NH4+], dissolved organic carbon

[DOC], particulate organic carbon [POC], particulate organic nitrogen

[PON], [NO3-], [NH4+], [PO43-], [NO3-], [NH4+], pH, salinity, and turbidity.. 70

Table 6: Abbreviations used in text............................................................................... 73

Table 7: ANOVA table for ppb N2O in septum vial headspaces. See Figure 24 for

data used in this ANOVA................................................................................ 79

xv

LIST OF FIGURES

Number Page 1 Recent increases in anthropogenic N fixation in relation to “natural” N

fixation .............................................................................................................. 2

2 Model-predicted global distribution of dissolved inorganic nitrogen

transport............................................................................................................. 3

3 Study site ........................................................................................................... 9

4 N2O flux in 8 Yaqui Valley drainage canals over two winter wheat cycles

(October 1999- September 2001) determined from supersaturation data ....... 24

5 Per-area rates of N2O flux in study drainage canals (ng N cm-2 hr-1) ............. 25

6 [N2O], [NO3-], and [NH4+] in cores collected from PA1 by depth.................. 26

7 Time series plots of A) [NO3-] (mg L-1), B) [DOC] (mg L-1), C)

[NH4+](mg L-1), D) pH, E) [Chl a] (mg L-1), F) [PON](mg L-1), G)

Temperature (ºC), H) discharge (m3 s-1), and I) salinity (g L-1) .................... 27

8 Mean annual [NO3-] vs. mean N2O flux in several large rivers and the

drainage canals in this study............................................................................ 30

9 A & B) Potenital denitrification and N2O production in sediment slurries

as a function of NO3- and Organic C availability............................................ 31

10 Potential for nitrification and denitrification to produce N2O. Inset: NH4+

enrichment versus N2O production ................................................................. 34

11 Annual N2O-N flux from study drainage canals ............................................. 41

12 Study site for diel experiment ......................................................................... 47

13 Diel measurements from July 2-3, 2001 ......................................................... 51

14 Diel measurements from July 2-3, 2001 continued......................................... 53

15 N2O flux and Chl a values over 23 months of bi-weekly sampling................ 58

16 Chamber-estimated CH4 flux (ug-C cm-2 h-1) from canals A1 and PA1......... 67

xvi

17 Chamber-estimated CO2 flux (ug-C cm-2 h-1) approximately bi-weekly

from canals A1 and PA1 ................................................................................. 69

18 Schematic of triple tracer addition experiment ............................................... 74

19 Photo of tracer addition ................................................................................... 74

20 Concentrations of i) SF6, ii) rhodamine WT, and iii) bromine over time at

two different sites in Drainage Canal PA1...................................................... 75

21 Typical sampling setup for floating chamber flux measurements .............76-77

22 Typical N2O chamber flux data....................................................................... 77

23 Schematic of septum vial used for headspace analysis ................................... 78

24 N2O concentrations in poisoned and unpoisoned septum vial headspaces

from a 24-hour period from noon 12/7/00 to noon on 12/8/00. ...................... 79

25 A comparison between YSI-measured O2 and O2 measured via O2:Ar

method using Membrane Inlet Mass Spectrometer......................................... 81

26 Measured and predicted N2:Ar over 24 hours. ................................................ 81

1

INTRODUCTION

Globally, industrial nitrogen (N) fixation for agricultural use has increased from

less than 10 Tg y-1 (Tg = 1012g) in the 1950's to over 80 Tg y-1 in the 1990’s (FAO 1995),

and approximately half the N fertilizer ever produced has been applied during the last

fifteen years (Galloway and Cowling 2002). This use of N fertilizer, together with

approximately 40 Tg y-1 fixed by leguminous crops and 21 Tg y-1 N inadvertently fixed

during fossil fuel combustion, has more than doubled the rate of N input to terrestrial

systems via natural N fixation (Vitousek and Matson 1993; Figure 1). Increased N

fertilizer use has led to massive increases in agricultural yield (food grown per unit area)

and has allowed humans to largely avoid the food shortages historically predicted to

accompany the recent population boom. In this sense, N fertilizer has been an enormous

boon to humans. However, the recent increase in N use has had serious environmental

drawbacks as well. Along with increased anthropogenic N fixation, levels of riverborne

N have increased dramatically (Turner and Rabalais 1991, Howarth et al. 1996), leading

to coastal oxygen depletion, increases in toxic and nuisance algae blooms, sedimentation,

and loss of biodiversity (Turner and Rabalais 1994, Steidinger 1981, Jickells 1998). In

addition to these local and regional effects of river N loading, this loading may affect the

global climate system by altering the balance of greenhouse gases produced and

consumed in freshwater systems (Paerl 1997, Seitzinger and Kroeze 1998).

2

Figure 1. Recent increases in anthropogenic N fixation in relation to “natural” N fixation.

Modified from (Vitousek and Matson 1993).

3

\

A)

B)

C)

Figure 2. A) Synthetic N fertilizer use in 1990. B) Model predicted N fertilizer use in

2050, and C) Model-predicted dissolved inorganic nitrogen export to coastal systems in

1990. All figures modified from (Seitzinger et al. 2002a).

Anthropogenic N loading is increasing particularly rapidly in the developing

tropics and subtropics (Figure 2). For example, consumption of fertilizer by developing

4

nations has grown from 12% of the global total in 1960 to 65% in 2000 (Fertilizer

Institute 2002). Yet, consequences of anthropogenic N loading for aquatic systems

remain almost completely unstudied outside of the developed world. In my thesis I have

investigated the link between agricultural intensification (the largest human source of

reactive nitrogen globally), N dynamics, and greenhouse gas production in the surface

drainage waters of the Yaqui Valley, an intensively farmed, rapidly developing region in

Sonora, Mexico.

In Chapter 1, I characterize and quantify fluxes of nitrous oxide from the drainage

waters of the Yaqui Valley and investigate the mechanisms controlling the production

and emission of N2O. In this chapter, I report some of the highest per-area fluxes

measured in surface fresh-water systems. I also explore relationships between N2O

concentration and several factors thought to control rates of N2O production. Finally, I

present calculations suggesting that despite high per-area fluxes, regional N2O fluxes

from surface drainage waters in the Yaqui Valley are low compared to values predicted

by algorithms used in global estimates.

In Chapter 2, I focus on diel chemical dynamics in one particular drainage canal,

and show for the first time that a flowing, freshwater system can undergo a complete

cycle of chemical reduction and oxidation within 24 hours. I also explore some of the

implications of this rapid change in chemical conditions for downstream N transfer and

greenhouse gas production.

5

In Chapter 3, I characterize and quantify CO2 and CH4 fluxes from Yaqui Valley

surface drainage waters. Yaqui drainage canals appeared to be net producers of both CO2

and CH4. Per-area CH4-C fluxes, though variable, were generally quite high compared to

other aquatic systems, but about an order of magnitude less than CO2-C fluxes. Finally,

C gas efflux was several times greater than lateral export of organic C, suggesting that it

may be an important pathway for C loss from the Yaqui Valley drainage system.

Together, by quantifying rates of key processes and the links between them, this

work adds significantly to our understanding of nutrient dynamics and trace gas

emissions, enhancing our capacity to manage agricultural and coastal resources and

ecosystems sustainably.

6

CHAPTER 1: PATTERNS AND CONTROLS OF NITROUS OXIDE EMISSIONS

FROM WATERS DRAINING A SUBTROPICAL AGRICULTURAL VALLEY

INTRODUCTION

Because it is an important greenhouse gas that also plays a role in the

destruction of stratospheric ozone, nitrous oxide (N2O) has received considerable

attention (Khalil and Rasmussen 1983, Matson and Vitousek 1990). Currently, global

atmospheric N2O concentration is increasing at the rate of 0.2-0.3% per year, and

fertilization of agricultural fields is thought to be the single most important source of the

observed increase ((IPCC 2001), Table 1). After a decade or more of research, we now

have a reasonably good understanding of the relationship between nitrogen (N)

fertilization and N2O flux from soils (de Klein et al. 2001), (Harrison and Webb 2001).

However, much less is known about the gaseous loss of fertilizer N once it has left

agricultural fields in solution or particulate forms. Some researchers have estimated that

N2O emissions from indirect emissions (emissions from surface water and groundwater

not within agricultural fields) are currently as large as direct emissions from fields (Table

1, (Mosier et al. 1998), and others have projected that increases in N loading to rivers will

triple river N2O production by 2050 (Kroeze and Seitzinger 1998). However, uncertainty

surrounding estimates of N2O emissions from indirect sources is large, ranging over one

and a half orders of magnitude and accounting for over 60% of the uncertainty in current

estimates of total anthropogenic N2O emissions (Table 1). This uncertainty is due to a

7

combination of poor understanding regarding the mechanisms controlling N2O

production and a lack of studies of off-site emissions of N2O from agriculture (de Klein

et al. 2001), (Brown et al. 2001).

Tg N yr-1 Range (Tg N yr-1) Anthropogenic: Agriculture indirect

emissions 2.1 (0.23-11.9)

direct emissions 2.1 (0.4-3.8) cattle and feedlots 2.1 (0.6-3.1) biomass burning 0.5 (0.2-1.0) industrial sources 1.3 (0.7-1.8) Subtotal 8.1 (2.13-21.5) Natural: Ocean 3.0 (1.0-5.0) tropical soils 4.0 (2.7-5.7) temperate soils 2.0 (0.6-4.0) Subtotal 9.0 (4.3-14.7) Total sources 17.1 (11.93-36.3)

Table 1. Current estimates of global N2O sources compiled from Intergovernmental Panel on Climate Change (IPCC) (2001), and Mosier et al. (1998).

Previous work with sediments and soils has indicated that N2O is formed

principally as a byproduct of two microbially mediated N transformations: denitrification

and nitrification. Denitrification, the microbial reduction of nitrate (NO3-) to N2O and

dinitrogen (N2) under anaerobic conditions, is thought to be controlled by inorganic N

availability, organic carbon (C) availability, oxygen (O2) concentration, and temperature

(Nishio et al. 1983), (Seitzinger 1988) (Firestone and Davidson 1989, Robertson 1989).

Nitrification, the oxidation of ammonium (NH4+) to NO3- under aerobic conditions (with

N2O as a byproduct), is controlled by NH4+ availability, temperature, and redox

8

conditions (Firestone and Davidson 1989). Previous work has suggested that the

proportion of total denitrification or nitrification emitted as N2O depends on the relative

availability of resources for microbes, as well as ambient redox conditions and

temperature (Seitzinger et al. 1984, Seitzinger 1988) (Joergensen et al. 1984).

Consequently, environmental variables such as N availability, organic C availability,

temperature, and oxidation-reduction conditions are likely to influence the rate at which

N2O is produced.

In this study, we used a combination of field and laboratory approaches to

estimate rates of N2O production in Yaqui Valley drainage canals and to improve our

understanding of the controls on those rates. These approaches included: 1) regression

analysis examining relationships between measured N2O fluxes and factors likely to

influence N2O production, 2) in situ sampling of sediment cores, 3) potential nitrification

and denitrification assays, and 4) intact core experiments.

SITE DESCRIPTION

WATERSHED

The Yaqui Valley is located between 26° 45' and 27° 33' latitude N and 109° 30'

and 110° 37' longitude W (Figure 3). Containing 226,000 ha of intensively managed,

irrigated, wheat-based agriculture, the Yaqui Valley is the birthplace of the Green

Revolution for wheat and one of Mexico’s most productive breadbaskets. This valley is

9

considered to be agro-climatically representative of 43% of wheat production in the

developing world (Meisner et al. 1992).

Figure 3. The Yaqui Valley- Study drainages are outlined in black and drainage

canals are portrayed as dark rectilinear networks. Water in the Valley generally flows

southwest toward the Gulf of California, shown in black. Dark circles (●) represent the

locations of pig farms, and the city of Obregón (Pop. 300,000+) is shown in the upper-

right.

Mean annual temperature in the valley is 22.5 ºC, and mean annual precipitation

is 28.7 cm, with 82% of that precipitation occurring during a “wet” season (July-

October). Over the course of this 23-month study, total rainfall in the region was just

37.5 cm, and this included only 3 rain events with greater than 2 cm rainfall in 24 hours

10

(Arizona Meteorological Network / PIAES (Patronato para la Investigación y

Experimentación Agrícola del Estado de Sonora 2002). Occasional heavy rain events

associated with major tropical storms may play an important role in this system on the

several year time-scale, but this topic is beyond the scope of this study.

In the Yaqui Valley, the use of fertilizer N has increased markedly in the past

three decades of development; between 1968 and 1995, fertilizer application rates for

wheat production increased from 80 to 250 kg-N ha-1 per 6-month wheat crop, and survey

results indicate substantial increases in fertilizer inputs in just the past decade (Naylor et

al. 2000). Today the most common farming practice for wheat production in the valley is

a pre-planting broadcast application of urea or injection of anhydrous ammonium (at the

rate of 150-200 kg-N ha-1), immediately followed by irrigation. Smaller allocations of

fertilizer are commonly added later in the crop cycle along with additional irrigation

water. These valley-wide fertilization/irrigation events occur principally in November,

but continue intermittently throughout the winter and early spring. These events lead to

large N losses to the atmosphere, ground water, and surface waters (Matson et al. 1998,

Panek et al. 2000, Riley et al. 2001)). Summer crops, primarily maize, are grown when

reservoir water storage is high and irrigation water is available; an additional 250-300 kg-

N ha-1 is typically applied to summer maize. However, during our study, summer crops

were not planted due to water shortages.

11

DRAINAGE CANALS

In the Yaqui Valley, surface irrigation runoff, livestock waste, and largely

untreated urban sewage enter the coastal zone directly via a system of 5 principal, and 14

smaller, open waterways (hereafter referred to as "drainage canals" or “canals”) (Figure

3). These perennial constructed and natural waterways have muddy to sandy bottom

sediments and are lined by shrubby vegetation (mainly Tamarix sp. and Cercidium

microphyllum (palo verde)). Sampling sites were free of macrobenthic flora and fauna.

Approximate drainage areas of study canals ranged from 800 to 43,000 ha, and mean

annual discharges ranged from 1,600 m3 (A3) to 59,000 m3 (A1), averaging 17,017 m3

(Table 2). Over the period of this study, flow in these drainage canals was regulated

almost entirely by upland water use (taken from a reservoir) because rain events were

small and infrequent.

12

Canal Area Served (Ha)

Mean Annual Discharge 1987-1996 (1000 m3)

Canal Surface Area (Ha)

PA1 22400 52000 68-134 PA2 3200 NA 10-19 A1 43600 59000 133-262 A2 6400 14500 19-38 A3 800 1600 2-5 A4 2800 5500 9-17 A5 6400 13000 19-38 A6 8800 15500 27-53

Table 2. Canal watershed areas and mean annual discharges.

3. METHODS

MEASUREMENT PROGRAM

We sampled 8 drainage canals biweekly for 23 months (November 1999-

September 2001) (Figure 3). In these canals, we measured fluxes of N2O as well as

several environmental variables likely to influence and co-vary with N2O emission.

These variables included DO, NO3-, NH4+, temperature, DOC, POC, PON, salinity, Chl a,

turbidity (NTUs), wind-speed and direction, water velocity, and pH. Drainages were

chosen to vary widely with respect to the factors likely to influence N-cycling and N2O

flux.

13

GAS COLLECTION AND ANALYSIS

We estimated gas fluxes using two techniques: floating chambers (as in

Livingston and Hutchinson (1995)), and headspace equilibration (modified from

Robinson et al. (1988). We applied the floating chamber technique in sites A1 and PA1,

the two largest drainage canals. In this technique, we fitted 10 cm high, 25-cm-diameter

acrylonitrile-butadiene-styrene (ABS) plastic chambers with floating collars and tied

them to stakes driven into canal bottoms. We sampled four chambers simultaneously at

10-minute intervals, and flux measurements lasted 30 minutes. During each flux

measurement, we measured water flow rates and depth just behind each chamber. We

calculated concentrations of chamber gases using least squares linear regression, and

calculated N2O fluxes by regressing gas concentration within a chamber against sampling

time, correcting for temperature and chamber volume as in Matson et al. (1998).

Minimum detectable flux of N2O was approximately 0.3 ng N2O –N cm-2 h-1.

In addition to the floating chambers, we employed a headspace equilibration

technique at all 8 sites. For this we pre-sealed 60 ml glass Wheaton bottles with gray

butyl stoppers, then evacuated and flushed them with helium. 15 ml aliquots of canal

water were injected into bottles, brought back to the laboratory, and gently shaken for 4

hours at 25 ºC. Headspace gas was then extracted and analyzed for N2O, and original

[N2O] was calculated using the appropriate solubility tables (Weiss and Price 1980).

Ambient air samples were also collected at each site for use in flux calculations.

14

On several occasions, we compared samples poisoned with saturated mercuric

chloride (HgCl) solution to samples lacking HgCl. There was no detectable difference

between poisoned and unpoisoned vials in 24 hours with respect to N2O (Appendix C).

We made all headspace measurements between 4 and 12 hours after sample collection,

and most samples were not poisoned.

We measured N2O using a Shimadzu gas chromatograph configured with electron

capture detector (ECD). The ECD contained 63Ni as the isotope source and an

argon/methane mixture was used as the carrier gas (Matson et al. 1998). Standards

ranged from 500 ppbv to 500 ppmv N2O, and 500 and 900 ppbv standards bracketed

every 15 samples. Coefficients of variation for standards never exceeded 2%.

DETERMINATION OF GAS TRANSFER COEFFICIENT

A gas transfer coefficient was estimated via a triple tracer experiment in PA1 and

was validated by comparing chamber fluxes with simultaneously measured N2O

concentration. In the tracer experiment, a pulse of dissolved SF6 gas (volatile tracer),

rhodamine (visual tracer), and bromine (conservative tracer) was added to one of the

larger canals (PA1) during a period with minimal wind in a manner similar to

(Wanninkhof et al. 1990). The slug of tracer was sampled over time, and the difference

in rate of loss between the volatile tracer (SF6) and conservative tracer (bromine) was

used to calculate a gas transfer coefficient according to Kilpatrick et al. (1989) (Also, see

Appendix A). In the case of PA1, this transfer coefficient was 4.67 cm hr-1 for N2O at a

Schmidt number of 600, assuming a flat water surface according to Wanninkhof (1992).

15

This estimate is in rough agreement with an estimate based on comparisons between N2O

concentration and chamber fluxes (mean = 7.8 cm hr-1 ± 3.5 cm hr-1 (1 S.D.).

Fluxes were calculated according to:

F = k*(Cw – Ceq) (1)

(Liss 1974) where F is gas flux across the air-water boundary, k is the gas-transfer

velocity, Cw is the dissolved gas concentration in the water column, and Ceq is the

dissolved gas concentration in equilibrium with the atmosphere. Because wind-speed can

play an important role in regulating gas transfer velocity over the range observed in our

study (0.1-7.5 m s-1), we accounted for variation in wind-speed using the relationship:

y = 1.91e0.35k (2)

where y is wind speed in m s-1, and k is the gas transfer velocity in cm hr-1 at a constant

Schmidt number (in this case 600) (Raymond and Cole 2001). We used on-site wind data

after June 26, 2001. Prior to this date, we used wind data from a nearby weather station

(

16

reduction technique utilized. All nutrient samples were filtered through sterile 0.45 um

glass fiber filters on-site and frozen until analyzed. Our detection limit for [NO3--N],

[NH4+-N], and [PO43--P] was 40 ug l-1.

We determined DOC as non-purgeable organic carbon with a Shimadzu total

organic carbon (TOC) analyzer. Samples were pre-acidified with ultra-pure HCl, sparged

with O2 to purge inorganic C, and combusted. We measured extracted Chl a and

phaeophytin concentrations with a Turner Designs, model 10-AU fluorometer according

to EPA protocol (Arar and Collins 1997b). One hundred-fifty ml aliquots were filtered

on site, treated with MgCO3 as a buffer, frozen, and then subsequently freeze-dried for

analysis. We measured POC and PON by filtering 150 ml aliquots of canal water onto

pre-ashed, Whatman GF/C filters, and subsequently analyzing them with a Carlo Erba

(now CE Elantech, Inc.) NA1500 Series II elemental analyzer (EA) (High temperature

combustion direct injection technique (Wangersky 1975, 1993)).

We measured DO (± 0.3 mg l-1), salinity (± 0.1 g l-1), and air and water

temperature (± 0.1 °C) with a YSI model 85 oxygen and salinity meter. We measured

turbidity (± 2%) and pH (± 0.04) with a Solomat 803PS multiprobe. We determined

wind-speed and direction with a handheld Kestrel, propeller anemometer, and water

current velocity and depth with a Marsh-McBirney electromagnetic flow meter and

wading staff according to USGS protocol (Rantz and others 1982b).

17

DENITRIFICATION AND NITRIFICATION POTENTIAL ASSAYS

To estimate the relative importance of autotrophic nitrification and denitrification

as sources of N2O, and to determine what factors might limit denitrification and N2O

production in surface sediments, we performed potential nitrification and potential

denitrification assays on sediments from canal A1.

For denitrification potential experiments (similar to those in Rysgaard, Risgaard-

Petersen et al. (1996), Pfenning and McMahon (1997), and many others) one g sub-

samples of hand-sieved (2 mm sieve) sediment were taken from the top 3 cm of 10 cm

diameter cores collected at site A1. Subsamples were placed in 60 ml Wheaton bottles

with 15 ml of filtered ambient water, and amended with an addition of NO3- (+14.0 mg l-1

NO3--N), organic C (+12 mg l-1 as ethyl acetate-C), a mixture of both NO3- and organic C

(+14 mg l-1 NO3--N, +12 mg l-1 C), or nothing (control). We also incubated sediments

with varying C amendments (+0.28, +0.6, +1.2, and +12 mg l-1-ethyl acetate C) to

determine the affect of increasing C availability on N2O production. After amending the

sediments, we sealed, evacuated, and flushed all vials with helium. We then performed

an acetylene block, inhibiting the reduction of N2O to N2 on half the bottles, allowing us

to estimate potenital denitrification in addition to N2O production. For each treatment,

we sampled and analyzed 3 replicate vials at the beginning of the incubation period, and

after 4 hours gently shaking at 25 °C. We calculated potential denitrification and N2O

production from differences between initial and final N2O concentrations in C2H2-treated

18

and untreated vials respectively. These assays were performed on three occasions (July

2000, December 2000, and February 2001), and yielded similar results each time.

We also performed a potential nitrification assay (similar to (Rysgaard et al.

1996) and many others) in order to determine the potential nitrification rate, estimate the

amount of N2O that could be produced via nitrification, and determine the degree to

which nitrification in these systems is influenced by NH4+ availability. Sediments for this

assay were collected in November 2000 and sub-sampled similarly to the sediments used

in denitrification potential assays. Sediments were amended with additions of PO43- (+10

mg PO43--P l-1) and NH4+ (+1.05, 2.1, 4.2, and 8.4 mg NH4+-N l-1). One set of bottles

was sampled immediately while another was left to incubate for approximately 4 hours at

25 ºC. When the incubation was complete, headspace was analyzed for N2O

concentration and water was filtered and frozen for subsequent nutrient analysis.

Nitrification was calculated as the difference between initial and final [NO3-], and N2O

production was determined by difference as well.

INTACT CORE EXPERIMENTS

To determine the impact of [NO3-], [DOC], and [NH4+] on the production of N2O

under more realistic (non-slurry) conditions, we performed several experiments using

intact cores. In these experiments, 10-15 cm of sediment and 25-30 cm of overlying

water were collected by hand from canal A1 in 10 cm diameter X 40 cm long, clear,

polycarbonate cores. These cores were transported to our Yaqui Valley laboratory where

they were allowed to settle for 12-14 hours at their original temperature (± 1 °C) in a

19

well-aerated, gently circulating, temperature controlled water bath consisting of water

also collected at site A1. Once cores equilibrated, they were amended with 0.6 mg l-1

ethyl acetate-C, 1 mg NO3--N l-1 in N2O production experiments, or, in the case of

denitrification experiments, C2H2, 12 mg ethyl acetate-C l-1. Following amendments,

cores were capped. Incubations were run with at least one blank core (water without

sediment), and at least two control cores (water and sediment without amendment).

Cores were sampled 4 times over the course of their incubation period for [NO3-], [NH4+],

[O2,], [N2O], [CO2], and [CH4]. Cores were incubated under ambient indoor light

conditions and timed such that [O2] did not fall below half saturation.

SECTIONED CORE EXPERIMENT

To determine where N2O production was likely to be occurring in the sediment

column, we sectioned and analyzed cores for [NO3-], [NH4+], [DOC], [POC], [PON], and

[N2O] with depth. On two occasions (June 23 and 25, 2001 at A1 and PA1 respectively)

we collected 10 replicate 5 cm diameter X 30 cm cores from PA1 and A1. Cores were

returned to the laboratory within 2 hours for sub-sampling. We subsampled five cores at

1-2 cm intervals for dissolved N2O concentrations. We used a 3 ml syringe to extract

sediment from holes drilled in the side of each core, which had been covered with

waterproof tape until time of sampling. Subsamples were placed in pre-poisoned 60 ml

Wheaton Vials and a 15 ml aliquot of deionized water was added to each sediment

sample. Bottles were capped immediately, shaken, and allowed to equilibrate for 2 hours

at 25ºC. Headspace gas was then extracted and analyzed for N2O content. After gas

20

sampling was completed, the 5 remaining cores were sectioned and analyzed for

porewater [DIN] (defined as NO3- + NO2-, and NH4+), [DOC], [PON], [POC], and bulk

density.

STATISTICAL ANALYSES

Gas flux, gas concentration, and independent variables with log-normal

distributions were log-transformed prior to statistical analysis using Statview 5.0.1

(Abacus Concepts 1992). Treatment and interaction effects in potential nitrification,

potential denitrification, and intact core experiments were evaluated using a factorial

design ANOVA.

Simple and stepwise regression approaches were used to determine the

relationship between [N2O] at all sites, and several factors likely to influence it:

discharge, water column [NO3-], [NH4+], [DOC], temperature, pH, DO, salinity, turbidity,

[POC], and [PON]. Concentration data were used in regression analyses except in the

cases of water velocity and discharge, when chamber flux data were used. For our

estimates of regional N2O flux, both septum vial and chamber flux estimates were used.

When both measurements were available, we used chamber-based flux estimates. In

order to avoid potentially confounding diel effects (Harrison and Matson 2001), only gas

measurements performed between 10:00 a.m. and 2:00 p.m. were used. Unless otherwise

stated, uncertainties are ±1 Standard Deviation.

21

REGIONAL FLUX ESTIMATES

We estimated total annual N2O-N emission to the atmosphere (Fa) from each

drainage canal in three ways. For the first two estimates, we multiplied median and mean

flux estimates for each drainage canal (data in Figure 4) by estimated surface areas for

each canal. For the third estimate, we integrated fluxes using data from figure 4 and

equation 3

(3)

where A is the surface area of each drainage canal, D is the number of days between flux

measurements, Dtot is the total duration (days) of data set, and Fi is the mean calculated

N2O flux for a particular date (kg cm-2 hr-1). The results of the three methods generally

agreed (Figure 5), and when unspecified, integration-determined estimates were used.

RESULTS AND DISCUSSION.

N2O FLUXES

Per-area N2O fluxes from drainage canals were generally high, and sometimes

extremely high (Figure 4). Mean N2O flux from the two largest drainage canals (A1 and

PA1) was 32.2 ng cm-2 hr-1, and mean flux from all canals was 16.5 ng cm-2 hr-1,

22

comparable to fluxes from eutrophic rivers such as the Potomac and the South Platte (~14

ng cm-2 hr-1 (McElroy et al. 1978, McMahon and Dennehy 1999)). At their highest

(244.6 ng cm-2 hr-1), N2O fluxes from these drainage canals were greater than the highest

reported fluxes from rivers ((McMahon and Dennehy 1999, Cole and Caraco 2001)).

These estimates are conservative as they include only fluxes calculated from

concentration data, which were generally slightly lower than floating chamber estimates

(slope = 1.08, R2 = 0.7). The highest N2O fluxes occurred in the canals receiving pig-

farm and agricultural effluents (PA1 and PA2; Figures 4 and 5) and during summer

months when temperatures were high, flows were reduced, and algae blooms were

occurring (Figures 4 and 7).

POTENTIAL CONTROLS ON N2O EMISSION

Because sediment processes and water column concentrations appeared to be

closely coupled in this system (Figure 6 and Chapter 2), we expected that water column

N2O concentrations, and thus N2O fluxes, would correlate with water column

concentrations of factors controlling N2O production. In Yaqui Valley drainage canals,

many of the factors believed to influence dentrification and nitrification, and hence N2O

production, varied across space and time (Figure 7).

One example of spatial variation was the difference between pig-farm influenced

and non pig-farm influenced canals. The drainage canals in this study could be grouped

as two distinct classes with respect to several characteristics. Canals PA1 and PA2

received significant pig-farm and urban inputs whereas watersheds of A1-A6 received

23

primarily agricultural runoff, though A1 and A6 contained a few small pig-farms without

direct input into drainage canals (Figure 3). For the remainder of this paper, drainage

canals PA1 and PA2 are referred to as pig-farm and agriculture-influenced canals (PA

Canals), and canals A1-A6 are referred to as agricultural drainage canals (A Canals).

Pairwise comparisons between canals indicated that sites PA1 and PA2 had higher

concentrations of [NH4+], [POC], [DOC] and [PO43-] than sites A1-A6 (by ANOVA P <

0.001 in all cases). Sites PA1 and PA2 also differed from sites A1-A6 with respect to

[NO3-], salinity, [DO], and [Chl a] (P = 0.05, 0.02,

24

Figure 4. N2O flux in 8 Yaqui Valley drainage canals over two winter wheat cycles

(October 1999- September 2001) determined from supersaturation data. Closed circles

(●) represent fluxes from low [NH4+] canals (A1-6), and open triangles (∆) and squares represent fluxes from PA1 and PA2 respectively. Shaded areas represent winter (ٱ)

wheat seasons when most irrigation and fertilization occurred. Error bars represent ± 1

S.D., and when not visible are smaller than the data point to which they refer. Sample

size (n) = 3 for each point.

25

Figure 5. Per-area rates of N2O flux in study drainage canals (ng N cm-2 hr-1). Flux

estimates based on concentration data are shown on the left, and flux estimates based on

floating chambers are shown (legend boxed) to the right for comparison. Three

estimates, based on median, mean, and integration-based averages respectively are

presented for each canal. Error bars represent the sum of uncertainties due to error in

flux coefficient (± 45%, estimated based on comparison between floating chamber fluxes

and measured N2O super-saturation values) and headspace measurement (± 3.4 %).

26

Figure 6. [N2O] (% saturation) -open circles (○), [NO3-](mg L-1) – filled circles(●), and [NH4+] (mg L-1) - filled squares(■) in cores collected from PA1 by depth in cm (Y-axis).

Sample size (n) = 5 for each point and error bars represent ± 1 S.D.. A similar pattern

was observed in canal A1.

27

28

Figure 7. Time series plots of A) [NO3-] (mg L-1), B) [DOC] (mg L-1), C) [NH4+]

(mg L-1), D) pH, E) [Chl a] (mg L-1), F) [PON](mg L-1), G) Water temperature (ºC), H)

discharge (m3 s-1), and I) salinity (g L-1). All variables were measured approximately bi-

weekly over two winter wheat cycles (October 1999- September 2002). Distinct drainage

canals are represented as follows: PA1 = ∆, PA2 = ٱ, A1 = ●, A2 = ▼, A3 = +, A4 =▲, A5 = ■, and A6 = ♦. Shaded areas represent winter wheat seasons when most irrigation

and fertilization occur.

NITRATE

Low NO3- availability often limits denitrification in anoxic sediments (e.g.

(Pfenning and McMahon 1997, Hinkle et al. 2001)), and N2O is a product of

denitrification, so NO3- supplied from the water column to upper layers of the sediment

has the potential to exert a strong control on [N2O] in these systems. In our study canals,

[NO3-] ranged from undetectable levels to 14.38 mg l-1 NO3--N, and tended to be highest

during the winter and spring when auxiliary irrigation events occurred. The high NO3-

concentrations we observed are very high compared with NO3- levels in unpolluted

systems, and high even in relation to many anthropogenically influenced systems

(Meybeck 1982, Kemp and Dodds 2002, USGS 2002). Increased [NO3-] was likely due

to direct runoff following N-fertilizer application and irrigation or to leaching of NO3-

remaining in soils following November’s principal fertilization/irrigation event (Figure

4). The latter scenario is consistent with an N-leaching model developed for the Yaqui

Valley (Riley et al. 2001).

29

When all drainage canals and seasons were considered, [NO3-] explained only

6.6% of the variability in [N2O] (P = 0.02), and when only PA canals were considered,

there was no significant relationship between [NO3-] and [N2O] (P = 0.47). Mean annual

[NO3-] and mean annual N2O flux did not correlate significantly either, casting doubt on

the generally accepted relationship between [NO3-] and N2O flux (Figure 8). However, in

canals A1-A6, [NO3-] explained 17.8% of the variability in [N2O] (Table 3).

Laboratory experiments supported a link between NO3- availability and [N2O].

Both denitrification potential assays and intact core incubations responded significantly

to NO3- additions with enhanced N2O production (Figure 9). In the case of denitrification

potential assays, NO3- additions increased N2O production by an order of magnitude.

Correlation between [NO3-] and [N2O] was weaker in field data than in laboratory

incubations. This may be because NO3- only limited N2O production via denitrification

when it was in short supply. When [NO3-] was high, other factors such as organic C,

rather than NO3-, may have limited N2O production due to denitrifcation.

30

Figure 8. Mean annual [NO3-] vs. mean N2O flux in several large rivers plotted

on a log-log scale. Modified from Cole and Caraco (2001) with the addition of drainage

canals in this study. Note that canals PA1& 2 and canals A1-6 cluster with respect to

N2O flux and NO3- load. There was no significant relationship between mean annual

[NO3-] and mean annual N2O flux from these systems.

31

Figure 9. A & B) Potential denitrification and N2O production in sediment slurries as a

function of NO3- and Organic C availability. n = at least 3 for each treatment, and lower-

case letters indicate whether treatments are significantly different (P < 0.05) from each

other by ANOVA. + Nitrate treatment consisted of a 14 mg L-1 enrichment of NO3-. +

Organic C treatment consisted of a 12 mg L-1 C enrichment with Ethyl Acetate. Data

shown are from December 2, 2000. C & D) N2O Production and denitrification in intact

cores. Treatments in acetylene block estimate of denitrification were: Control (same as in

N2O incubations) and +12 mg L-1 organic C as ethyl acetate. Treatments in N2O

32

incubations were as follows: blank (water without sediment), Control (water and

sediment without amendment), + 0.6 mg L-1 Ethyl Acetate-C, and +1 mg L-1 NO3- .

ORGANIC CARBON

If organic C limited N2O production via denitrification when NO3- was abundant,

we would expect to see a significant relationship between organic C and N2O in canals

with relatively high concentrations of NO3-. In study canals, [DOC] ranged from 1.5 to

27.3 mg L-1 with a mean concentration of 5.9 mg L-1. This mean DOC concentration is

similar to the mean DOC concentration in rivers worldwide (5.75 mg L-1 according to

Meybeck (1982)). In canals with [NO3-] >1 mg L-1, [DOC], [POC], and [TOC] correlated

significantly with N2O (r2 = 0.05, 0.15, and 0.089, and P = 0.029, P

33

N2O production some of the time. It may also be that nitrification played an important

role in N2O production.

AMMONIUM

In our study canals, water column [NH4+] ranged from undetectable levels to

12.19 mg L-1 NH4+-N (Figure 7). In canals A1-6, [NH4+] followed a similar pattern to

[NO3-], peaking during winter and spring irrigation and fertilization events. However, in

PA canals, [NH4+] remained high year-round. Mean [NH4+] was 1.50 ± 2.32 mg L-1 and

6.18 ± 3.73 mg L-1 NH4+-N for A canals and PA canals respectively.

When all study drains were considered, [NH4+] was significantly and positively

correlated with [N2O] (P < 0.001 and r2 = 0.36), and [DIN] was the best single predictor

of N2O production, explaining 41% of the variability in N2O in all canals (Table 3). This

is consistent with a significant nitrification source of N2O. So is the observed distribution

of N2O in sectioned cores (concentrated in the top few cm—Figure 6) and the fact that

[O2] correlated significantly and positively with [N2O] in PA canals (P < 0.001 and r2 =

0.473).

Laboratory assays also suggested that nitrification could play an important role in

the production of N2O. Potential nitrification assays suggested that nitrification could be

an important source of N2O in drainage canal sediments (Figure 10), and we observed a

significant, positive relationship between [NH4+] and N2O production via nitrification in

potential nitrification experiments (Figure 10-inset).

34

Figure 10. Comparison of potential for nitrification and denitrification to produce N2O.

Error bars represent 1 S.D.. Inset: NH4+ enrichment versus N2O production. n =3 for

each NH4+ concentration.

In addition to acting as a direct source of N2O, nitrification may also contribute to

N2O production by providing the NO3- necessary to fuel denitrification, as has been

shown to occur in many core studies (e.g. (Thompson et al. 2000, An and Joye 2001,

Laursen and Seitzinger 2002)). However, from our data it is impossible to determine the

extent to which coupled nitrification-denitrification contributes to N2O production.

35

Stepwise Multiple Regressions Simple Regressions Independent Variable Step n P r2(Adj.) F to

enter F to

removen Slope Int. P r2(Adj.)

All Freshwater sites Intercept 0 91 --- --- 805.4 --- --- --- --- --- Log [NH4+] 1 91

36

left, and results of individual regressions are shown on the right. Non-significant

regressions are omitted. For the stepwise multiple regression, criteria for inclusion were:

F to enter > 4, F to remove < 3.95. Unless otherwise noted, units are in terms of mg L-1.

PH

By influencing the activities of nitrifying and denitrifying bacteria, pH has the

potential to control N2O production. In pure culture, ammonium oxidizers have been

shown to have affinities for a pH in the range of 5.8-8.5 (Watson et al. 1989) with an

optimum of 7.8 (Hagopian and Riley 1998). Other studies have shown that denitrifiers

have an optimum pH of 7.0-8.0 and that denitrification is positively related to pH

(Knowles 1982). In our study canals, pH was higher in purely agricultural canals (A1-

A6; mean = 7.95) than in mixed-input canals (PA1 and PA2; mean = 7.8), and in canals

A1-A6 more frequently exceeded the optimal pH range of nitrifiers and denitrifiers. This

may explain why we saw a positive correlation between pH and [N2O] in pig-farm

influenced canals, a slightly negative correlation between pH and N2O in purely

agriculture influenced canals, and no effect of pH when all canals were considered (Table

3). The positive correlation between pH and [N2O] in PA canals could also be a side

effect of the inverse correlation between productivity and water column [CO2], which

correlated with pH (r2 = 0.489).

37

TEMPERATURE

By influencing microbial metabolic rates, temperature can influence the rate at

which microbial N transformations occur (Kadlec and Reddy 2001). Temperature has

been observed to affect rates of N2O production in some aquatic systems (Seitzinger

1988), though not in all (Knowles 1982). In our study, water temperatures ranged from

13.5-38 °C, were highest during July and August, and were relatively uniform across all

study canals (Figure 7). Although the effect of temperature was significant in PA canals,

(r2 = 0.206 and P = 0.007) (Table 3), it was not significant when either canals A1-6 or all

canals were considered. The temperature effect in PA canals was most likely due to the

correlation between temperature and Chl a (P < 0.001, r2 = 0.399). Temperature may be

less important in this system than in other systems because temperatures are generally

within the accepted optimum range for nitrifiers and denitrifiers (Kadlec and Reddy

2001), Figure 7). Additionally, these canals experience wide day-night temperature

swings (up to 10 °C), which may preclude narrow temperature-related metabolic maxima

within microbial populations.

DISCHARGE AND FLOW RATE

It is possible that in-stream flow velocity and discharge could affect N2O fluxes

by altering rates of material exchange across the sediment-water interface or across the

water-air interface, as has been suggested by Raymond and Cole (2001) and many others.

We tested this hypothesis by attempting to correlate flow rate and discharge with

chamber-based flux estimates.

38

Canal mid-channel depths ranged from 3 cm during low flow summer months in

the smallest canal to 2 m during peak flow in one of the major canals (PA1). Flow

velocities were generally low (mean: 0.321 m s-1, range: 0.06 - 0.568 m s-1 at sites PA1

and A1). Neither comparison of replicate floating chambers subject to different water

flow velocities nor comparisons across multiple sampling dates with different discharges

and water velocities indicated a significant relationship between water velocity or

discharge and N2O flux. This is consistent with the findings of McMahon and Dennehy

(1999), but also may be due to the relatively low range of flows we observed in our study

canals.

SALINITY

Some have hypothesized that salinity can influence N2O production by altering

DIN availability, or by influencing the presence of H2S, an inhibitor of nitrification and

the final step in denitrification (Joye and Hollibaugh 1995), (Sorensen et al. 1980).

Although the canals in our study were not subject to tidal mixing, they did exhibit a

striking salinity range due to seasonal variations in runoff-water salinity and evaporation.

Salinities ranged from 0.00 to 26.00 g L-1 with a mean value of 3.16 g L-1, and were

inversely correlated with discharge. Highest salinities occurred during low-flow summer

months, and generally declined during irrigation events (Figure 7). In this study, salinity

didn’t significantly correlate with [N2O] (P > 0.05 in all cases). This may be due to the

fact that salinity was almost always sufficient to prevent the formation of large adsorbed

39

N pools. Salinity may exert a stronger control on [N2O] in systems with lower average

salinities.

ALGAE BLOOMS (CHLOROPHYLL A) AND RELATED VARIABLES

Algae blooms have the potential to create extremely favorable conditions for N2O

production. Algae growth can remove the organic C limitation on denitrification, while

rapid daytime photosynthesis can create the enriched dissolved oxygen environment

necessary for rapid nitrification and coupled nitrification-denitrification (An and Joye

2001).

In summer months, when canal water flows decreased, we observed intense algae

blooms (up to 373.6 mg Chl a m-3), particularly in PA canals (Figure 7). In these canals

[Chl a] was the best single predictor of [N2O] (r2 = 0.47, P < 0.001), and other bloom-

related variables such as [O2], [PON], and [POC] also correlated significantly with [N2O]

(r2 = 0.473, 0.436, and 0.420, and P =

40

pig-farm and agricultural inputs were considered, the best regression model for [N2O]

included [Chl a] and [NO3-], and explained 67% of the variation in [N2O].

In contrast to other studies where [DIN] has correlated strongly with [N2O] (e.g.

(McMahon and Dennehy 1999)), in our study multiple factors appeared to control [N2O],

and controls appeared to change with changes in inorganic N availability. Under low

[DIN] conditions, DIN probably controlled [N2O]. However, when DIN was plentiful,

control of [N2O] likely shifted to stream productivity, a notion supported by the strong

relationship between algae-bloom-related variables and [N2O] in PA canals described

above.

REGIONAL SIGNIFICANCE OF AQUATIC N2O FLUX AND IMPLICATIONS FOR GLOBAL

ESTIMATES

Annual N2O-N fluxes from drainage canals in this study ranged from 13 kg N2O-

N yr-1 in canal A3 to 4,167 kg N2O-N yr-1 in canal A1 (Figure 11), and the total annual

flux from all canals together was 8,678 kg N2O-N yr-1. By assuming that the fluxes we

measured in the study canals (42% of the Yaqui Valley’s total canal surface area) were

representative of canal fluxes valley-wide, we were able to calculate a valley-wide annual

N2O-N flux from agricultural drainage canals of 20,869 kg N2O-N yr-1.

41

Figure 11. Annual N2O-N flux from study drainage canals. Flux estimates based on N2O

concentration data are shown on the left, and flux estimates based on floating chambers

are shown to the right with a boxed legend. Three estimates, based on median, mean, and

integration-based averages respectively are presented for each canal. Error bars represent

the sum of uncertainties due to error in flux coefficient (± 45%, estimated based on

comparison between floating chamber fluxes and measured N2O super-saturation values),

headspace measurement (± 3.4 %), and our estimate of drainage surface area (± 50.6%).

Assuming a relatively conservative valley-wide per-hectare fertilizer application

rate of 250 kg N ha-1 wheat or maize crop cycle-1 (Matson et al. 1998, Naylor et al. 2000),

and a conservative estimate of wheat and maize crop-area (170,000 ha (Secretaria de

Agricultura 1994-2000)), we calculate that 0.046% of the N fertilizer applied to Yaqui

Valley fields in 2000 was lost from drainage waters as N2O-N. Thus, N2O-N emissions

from drainage water were about an order of magnitude lower than emissions from

42

agricultural fields (0.205-1.4% of the applied N in the 1995/96 wheat season (Matson et

al. 1998)).

Multiplying this proportional loss (0.046% of applied fertilizer N lost as N2O

from drainage waters), by the global rate of N-fertilizer application (85 Tg N (Galloway

and Cowling 2002)) yielded our estimate that 0.04 Tg of N are lost from world rivers as

N2O. Typically, 30% of the indirect N2O source (which also includes groundwater and

estuarine emissions), or 0.069 - 3.75 Tg N (Table 1) is assumed to be from rivers (Mosier

et al. 1998), so our estimate is on the very low end of what would be expected for a river

system draining an N-intensive agricultural valley. As time of travel and channel

morphology are thought to influence N processing in rivers (Seitzinger et al. 2002b;

Alexander et al. 2000), one reason for our low estimate may be the proximity of the

Yaqui Valley to the coast. This makes some sense because globally, most drainage

systems are further from the coast (Vorosmarty et al. 2000). However, more work will be

necessary to determine whether this is the case.

N TRANSFER TO ESTUARIES

In addition to producing greenhouse gases, Yaqui Valley drainage canals may

also constitute an environmental threat through their capacity to carry reactive N to the

coastal zone. Using discharge and concentration data it is possible to estimate DIN

transferred to the coastal zone (DINcoast) via Yaqui Valley drainage canals using the

equation:

43

DINcoast = [DIN] Χ Qann (4)

where [DIN] is median DIN concentration over the course of our study and Qann is mean

annual discharge for Yaqui Valley drainage canals over the years 1987-1996 (Comision

Nacional del Agua, Sonora, Mexico). Using this equation, we estimate that the drainage

canals in this study transport 709 Mg (1Mg = 106g) of DIN to the coast annually. Scaling

up to the whole valley, we estimate that Yaqui Valley drainage canals carry 954 - 1688

Mg of DIN to the near coastal zone annually. Adding particulate N to this formulation

increases our estimate of N transfer to 1229-2060 Mg-N. We estimate, that of the total

annual DIN transferred to the coast by collector drains, 738 - 756 Mg-N or 43–77% is

NO3-. High and low estimates of N-transfer were achieved by including and excluding

pig and urban-influenced canals from valley-wide estimates of N transfer, with high

estimates resulting from inclusion of pig and urban-influenced canals. In the two

drainage canals where weekly discharge data were available, we calculated N flux by

multiplying weekly discharge by weekly concentrations. This approach resulted in DIN

flux estimates similar to estimates attained via the simple method described above.

Even when we estimated N transfer conservatively, it was 3-5 times higher than

previous estimates of N transfer to the coastal zone (411 Mg), which were based on

simple input/output ratio models (Rice 1995). Our low estimate of NO3- transfer is more

than 8 times greater than previous estimates of NO3- transport to the coastal zone (Rice

1995). Our high-end estimate of annual DIN flux through Yaqui Valley drainage canals

is equal to 4% of the annual rate of fertilizer application in the Yaqui Valley (See

44

previous section for estimate of fertilizer application). DIN flux through drainage canals

appears to be greater than the amount of N lost from agricultural fields as N2O and NO

trace gases (0.4-2.8 % - Matson et al. 1998), and about two orders of magnitude greater

than N lost to the atmosphere as N2O from drainage waters.

DENITRIFICATION

In many systems, denitrification is thought to exert an important influence on the

amount of reactive N reaching coastal ecosystems (Howarth et al. 1996, Seitzinger and

Kroeze 1998). We estimated denitrification rates in Yaqui Valley drainage canals in

three ways: intact core incubations, denitrification potential incubations, and N2O

measurements.

Intact core incubations yielded a denitrification rate estimate of 0.6 ng-N cm-2 h-1

our lowest estimate of denitrification rate. Denitrification potential incubations yielded a

denitrification rate estimate of 491.5 ng-N cm-2 h-1. An estimate of denitrification based

on field measurements of N2O was similar to the estimate based on potentials. By

assuming, as in Seitzinger and Kroeze (1998), that N2O production was equal to 0.3-3%

of N2 production via denitrification, we estimated a median denitrification rate for all

canals of 533.5 – 5373.5 ng-N cm-2 h-1. When we multiplied this rate by the surface area

of Yaqui Valley drainage canals (1332 ha) it translated into a regional annual

denitrification N flux of 622.50 – 6,269.97 Mg-N yr-1. Intact core and potential

denitrification incubation estimates of denitrification yield annual, valley-wide, drainage

canal N2 flux estimates of 0.7 Mg-N yr-1 and 573.756 Mg-N yr-1 respectively. Thus our

45

estimates of denitrification N flux from surface drainage waters were highly imprecise,

ranging from 0.7-6,269.97 Mg-N yr-1. This is equivalent to 0.002-14.8% of per-crop N

fertilizer inputs to the Valley. This estimate of denitrification straddles our estimate of N

advection through our sampling sites (1229-2060 Mg-N), so the importance of

denitrification in controlling N transfer remains an important topic for further research.

46

CHAPTER 2: RAPID-ONSET ANOXIA AFFECTS NITROGEN TRANSFER AND

GREENHOUSE GAS PRODUCTION IN MEXICAN STREAM

Rivers are a primary pathway for transport of materials from land to sea and an

important source of greenhouse gases (Meybeck 1982, Seitzinger and Kroeze 1998,

Richey et al. 2002). Recent evidence shows that rivers and streams can process nutrients

and other chemicals rapidly over surprisingly short distances (Fuller and Davis 1989,

Fisher et al. 1998, Wollheim et al. 1999, Peterson et al. 2001), and that diurnal cycles of

plant productivity (respiration and photosynthesis) can influence transformations and

transfer of trace metals such as arsenic and manganese. (Brick and Moore 1996, Sullivan

et al. 1998) Here we show that day-night redox oscillations in a polluted, Mexican stream

can strongly and rapidly influence in-stream transformations and downstream transfer of

N and other materials, and thus can substantially affect the production of greenhouse

gases nitrous oxide, methane, and carbon dioxide (N2O, CH4, and CO2).

47

Figure 12. Study site (Yaqui Valley, Sonora, Mexico)- Canal Two is shown as a dark,

rectilinear network, and our sampling station is marked with an X. Water in the Valley

generally flows SW toward the Gulf of California. Shaded area represents agricultural

land, dark circles (●) represent the locations of pig farms, and the city of Obregón (Pop.

300,000+) is shown in black in the upper-right.

Mexico’s Yaqui Valley (Figure 12) was among the first developing regions in the world

to experience the suite of changes in agricultural practice known collectively as the

“Green Revolution”(Naylor et al. 200, Meisner et al. 1992). Beginning in the 1950s, the

Green Revolution greatly increased the Yaqui Valley’s agricultural yield, due to

improved wheat varieties and greatly increased application of nitrogenous fertilizers.

This fertilization, together with regional population growth and increases in regional hog

48

production,1 has greatly increased the amount of N flowing through the valley. Similar

patterns of growth and change are common in regions of agricultural intensification

throughout the developing world, and as a region of rapid change and development, the

Yaqui Valley provides a model system that can help us understand, predict, and respond

to trends throughout the developing subtropics (Naylor et al. 200, Meisner et al. 1992,

Matson et al. 1998).

The Yaqui Valley is drained by a network of natural and man-made streams that

carry a mixture of agricultural and urban wastewater to the Sea of Cortez, providing a

useful system in which to examine the chemical dynamics of polluted rivers and streams

(Harrison and Matson Submitted, and Chapter 1). Due to their seasonally high

productivity, some of the Valley’s drainage streams exhibit pronounced day-night O2

fluctuations. We hypothesised that these large diel swings in oxygen concentrations

would exert a strong control on the concentrations of several biologically and

environmentally important compounds.

1 Between 1968 and 1995, fertilizer application rates for wheat production increased

from 80 to 250 kg N h-1 and survey results indicate substantial increases in fertilizer

inputs in just the past decade (Naylor, Falcon et al. 2000). Since the 1950s the Valley's

population has grown from a few thousand to over 300,000 today, and population

continues to grow (Cajeme Statistical Annuals 1982-2000). Hog production in the

valley has increased as well, from 30-60 hogs in 1960 to more current estimates of

370,000 hogs (Pork Association 1993).

49

As conditions become increasingly reducing in flooded soils, microbial reduction

consumes inorganic compounds, generally in the order: O2 > NO3- > MnO2 > Fe(OH)3 >

SO42- >CO2 (McBride 1994). We would expect to see similar dynamics in a stream’s

water column as a stream becomes increasingly reduced. As the aforementioned electron

acceptors are reduced and depleted, we would expect to see increases in concentrations of

reduction products: N2 (and N2O), Mn2+, Fe2+, H2S, and CH4, in that order.

Concentrations of reduced species should decrease as their source ions are depleted or as

conditions become increasingly oxidized with the onset of daytime photosynthesis during

morning hours.

We tested these predictions by monitoring water-column concentrations of several

ions and gases over a 24-hour period in one of the Yaqui Valley’s principal drainage

streams: Canal Two (Figure 12)2, for which we have extensive, long-term data on

hydrologic, nutrient, and trace gas dynamics 3. During our diel experiment, (July 2-3,

2 Referred to as PA1 in other chapters.

3 See Chapter 1 for details of long-term sampling. Canal Two had high dissolved

inorganic nitrogen (DIN) concentrations, primarily as ammonium (NH4+) (mean:7.21

mg-N L-1), and high orthophosphate (PO43-) concentrations (mean:1.30 mg-P L-1), most

likely due to pig and urban point-source inputs. Canal Two also had high

concentrations of dissolved organic carbon (DOC) (mean:8.46 mg-C L-1), and a

substantial seasonal fluctuation in temperature (range: 14.0-35.3 °C, mean: 24.1°C),

with highest temperatures occurring during summer months. We estimated Canal

Two's re-aeration time, the mean residence time of gases in a river or stream, to be 2-4

50

2001), Canal Two was experiencing an algae bloom (mean [Chl a]: 87 mg m-3), and we

observed significant swings in pH, temperature, and salinity (Figure 13). Changes in

salinity don't explain the observed chemical dynamics because the timing of peak salinity

is not reflected in measured ionic concentrations of Fe, Mn, or S (Fig. 14). Changes in

temperature and pH area are also insufficient to explain the chemical shifts we observed.4

Instead, the dominant control on chemical dynamics appears to be thermodynamic. The

chemical dynamics we observed over this period were consistent with thermodynamic

predictions, and chemical shifts occurred very rapidly (Figure 14). In the evening, with

the cessation of photosynthesis we observed a decrease in [O2], followed by [NO3-], and

[SO42-] (Figure 14). Along with the consumption of these electron acceptors, we saw

hours (Appendix B). Canal Two also exhibited seasonally high Chlorophyll a (Chl a)

concentrations (range:2.55-373.6 mg m-3, mean:69.488 mg m-3).

4 Previous studies have suggested that diel changes in pH, temperature, and salinity might

elicit a chemical response similar to that predicted by thermodynamics (Brick, Moore

1996). In our study, neither pH nor temperature could explain the pattern of chemical

dynamics we observed. Mn3+, Fe3+, and HS- are extremely insoluble over the entire

observed pH and temperature range we observed in Canal Two (Solubilities < 0.0001,

0.002, and 0.5 mg L-1 for each species respectively at maximum pH: 8.5 and minimum

temperature: 27 degrees C). Conversely, Mn2+, Fe2+, and SO42- are extremely soluble

over the same range of pH and temp (solubilities >5000, >20 mg L-1, and >10000mg L-

1 for Mn2+, Fe2+, and SO42- respectively) (Gustaffson 2002). Therefore, the observed

changes in pH and temperature (Fig.10) are unlikely to have caused the patterns in

dissolved Mn, Fe, and S that we observed.

51

predictable increases in the concentrations of reduction products. [CO2] rose first,

followed by [N2], [Mn2+], [Fe2+], and finally [CH4] (Figure 14).

Figure 13. Diel measurements from July 2-3, 2001. a) Insolation expressed as Mjoules

m-2, b) Water temperature (°C), c) pH ± 0.01, d) Salinity (g L-1), e) Dissolved organic

carbon (mg-C L-1) ± 1 mg L-1, and f) Chlorophyll a (mg m-3) ± 1 S.D.. All measurements

52

except Chlorophyll a were made every 2 hours. Chlorophyll a was measured every 4

hours.

53

54

Figure 14. Diel measurements from July 2-3, 2001. Except for CO2, gases (O2, N2, and

CH4) expressed in terms of % saturation. CO2 expressed as concentration (mg-C L-1).

Error bars represent ± 1 S.D. of triplicate samples. Dissolved species expressed in terms

of mg L-1. Error bars represent analytical error. Dissolved N species (NH4+ and NO3-)

are expressed as mg-N L-1. Where no error bars are present, they are contained within the

data point.

In the morning, with the onset of photosynthesis, we observed an increase in

[SO42-], followed by an increase in [O2], and finally an increase in [NO3-]. Though the

order of appearance of SO42- and O2 conforms to thermodynamic predictions, the timing

of [NO3-] increase did not. This unexpected lag in [NO3-] may have been due to

phytoplankton uptake or to the importance of reaction kinetics in determining oxidation

rates. Reaction kinetics may also have exerted an important control on the morning

dynamics of reduction products (Figure 14).

A complete diel reduction and oxidation sequence such as the one we observed

has never been reported in a canal, river, or stream setting. It is, perhaps, for this reason

that current conceptual (Fisher et al. 1998) and numerical (Runkel 1998, Seitzinger and

Kroeze 1998, Wollheim et al. 1999) models of river biogeochemistry ignore diel changes

in stream chemical dynamics. Even models that incorporate the ability of rivers to

transform nutrients (Seitzinger and Kroeze 1998, Wollheim et al. 1999), rather than

treating them as si