Potential Changes in Greenhouse Gas Emissions from RefrigeratedSupply Chain Introduction in a Developing Food SystemBrent R. Heard* and Shelie A. Miller

Center for Sustainable Systems, School for Environment and Sustainability, University of Michigan, 440 Church Street, Ann Arbor,Michigan 48109, United States

*S Supporting Information

ABSTRACT: Refrigeration transforms developing food systems,changing the dynamics of production and consumption. This studymodels the introduction of an integrated refrigerated supply chain,or “cold chain,” into sub-Saharan Africa and estimates changes inpreretail greenhouse gas (GHG) emissions if the cold chaindevelops similarly to North America or Europe. Refrigerationpresents an important and understudied trade-off: the ability toreduce food losses and their associated environmental impacts, butincreasing energy use and creating GHG emissions. It is estimatedthat postharvest emissions added from cold chain operation arelarger than food loss emissions avoided, by 10% in the NorthAmerican scenario and 2% in the European scenario. The coldchain also enables changes in agricultural production and diets.Connected agricultural production changes decrease emissions,while dietary shifts facilitated by refrigeration may increase emissions. These system-wide changes brought about by the coldchain may increase the embodied emissions of food supplied to retail by 10% or decrease them by 15%, depending on thescenario.

■ COLD CHAIN INTRODUCTION AND THE FOODSUPPLY CHAIN

This study explores the inherent trade-off of reducing foodlosses and their associated embodied greenhouse gas (GHG)emissions by deploying refrigeration, a technology thatincreases GHG emissions through energy consumption andrefrigerant emissions. This analysis first examines only thedirect postharvest trade-offs between increased energy andrefrigerant emissions compared to the GHG savings of reducedfood loss. This study then takes a broader systems-levelexamination of the potential impacts of introduced refriger-ation, including anticipated impacts on agricultural productionwith development and dietary shifts brought about byimproved access to perishable foods.An integrated refrigerated supply chain, or “cold chain,” can

provide benefits for community health, nutrition, and foodsecurity.1,2 Refrigeration increases access to perishable foods,extends the shelf life of food, and has the potential to reducefood losses.3,4 Access to refrigeration is associated withimproved health outcomes, including reduced risk offoodborne illness3 and improved capacity to store antibioticsand vaccines.5 The cold chain has critical connections to theSustainable Development Goals, with target 12.3 seeking areduction in food loss and waste along the food supply chain,6

and Goal 2 seeking to improve food security and nutrition.7

The global cold chain market was valued at $203.14 billion

USD in 2018 and is expected to grow 7.6% per year, driven byincreased demand in emerging markets.8

Despite these benefits, refrigeration is energy-intensive andoften uses refrigerants with high global warming potentials.9

When accounting only for direct energy use and refrigerantleakage, refrigeration is responsible for approximately 1% of theworld’s total carbon dioxide emissions,10 and can represent 3−3.5% of GHG emissions in developed economies such as theUK.11

In addition to energy use and emissions, refrigerationfacilitates increased consumption of more-perishable foods,which tend to be more environmentally intensive.11 Consumerdemand for food determines the agricultural productionsystems required to provide the types and quantities of fooddemanded. Agricultural industrialization may not initially seemto be a result of the cold chain; however, particularly forperishable goods, cold storage enables more industrializedsystems since it expands distribution capacity, facilitating largerproduction.Food loss and waste is an environmental, economic, and

social loss.12−15 Additionally, food losses that occur furtheralong the supply chain are more carbon-intense due to

Received: September 20, 2018Revised: December 5, 2018Accepted: December 7, 2018Published: December 19, 2018

Article

pubs.acs.org/estCite This: Environ. Sci. Technol. 2019, 53, 251−260

© 2018 American Chemical Society 251 DOI: 10.1021/acs.est.8b05322Environ. Sci. Technol. 2019, 53, 251−260

Dow

nloa

ded

via

UN

IV O

F M

ICH

IGA

N A

NN

AR

BO

R o

n Fe

brua

ry 6

, 201

9 at

21:

41:3

8 (U

TC

).

See

http

s://p

ubs.

acs.

org/

shar

ingg

uide

lines

for

opt

ions

on

how

to le

gitim

atel

y sh

are

publ

ishe

d ar

ticle

s.

additional embodied energy.16 Approximately one-third of allfood produced for human consumption is lost or wasted,17 andreducing food losses and waste has been identified as a key goalin improving food security.12−14,18−20 The cold chain has beenidentified as a key means for reducing food loss and waste,providing savings in embodied GHG emissions.4,15,21,22

Therefore, it becomes crucial to first develop a betterunderstanding of whether the emissions savings from reducedfood loss are offset by increased emissions from the cold chainand determine potential improvements to reduce cold chainimpacts while maintaining these societal benefits.The cold chain is a transformative technology which

influences, codevelops, and interacts with a number of foodsystem properties ranging from consumer behavior toupstream production methods.9 The cold chain fundamentallychanges markets and supply chains, necessitating considerationof not only direct, but also indirect and external factorsassociated with this technology when modeling its environ-mental impacts.9,23 Parfitt et al. characterize the level ofpostharvest infrastructure and supply chain technology as itdirectly relates to the overall development of a country,explicitly noting the presence of the cold chain as a hallmark ofindustrialized countries with advanced food system infra-structure.24 Garnett describes cold chain technologies asubiquitous for a modern food system, embedded in everystage of a product’s life cycle.11 It has also been noted thatsupply chains for several goods are now based on the ability tosupply chilled or frozen products.25 As such, cold chainintroduction is fundamental to food system development.

■ STUDY OVERVIEW

This study examines the extent to which the cold chain mayincrease or decrease net GHG emissions when introduced intoa developing food system.

Academic study of the cold chain has been limited andfragmented, with few connections between the technicalresearch on refrigeration technologies and the broader foodsystems literature, presenting notable research gaps.9 Jamesand James present a valuable analysis of the cold chain’srelationship to climate change, detailing mechanisms throughwhich these emissions could be reduced, but warning ofpotential emissions increases should a rise in ambienttemperatures from climate change occur.26 Garnett discussesrefrigeration from a food systems perspective in a compre-hensive working paper, summarizing the literature on theenvironmental impacts of refrigeration systems, and alsodiscussing how refrigeration may prompt dietary shifts andconsumer behavior changes.3

This study first examines a fundamental trade-off ofrefrigeration: the ability to reduce food losses which carryembodied emissions, but increasing energy use and GHGemissions to do so. This study assesses whether the cold chainadds more emissions per food type supplied to retail than itsaves through avoided losses with its introduction. Once thedirect trade-offs are evaluated, a broader system view is taken,first estimating changes in emissions required to supply eachfood type to retail due to improved efficiencies in agriculturalproduction occurring with development, then estimatingpotential emissions changes from dietary shifts enabled byrefrigeration.Greenhouse gas emissions (in CO2e) are estimated for one

kg of food supplied to retail for seven food categories: cereals,roots and tubers, fruits, vegetables, meat, fish and seafood, andmilk. Additional important impacts associated with agriculture,including blue water consumption, land use change, nutrientrunoff, and biodiversity effects are not included due to a lack ofdata.The food supply chain (FSC) is defined as a linear model of

mass flow with five stages in accordance with Gustavsson et

Figure 1. Visual representations of mass flows for food (F), loss rates (R), and losses (L) in the upstream food supply chain to supply 1 kg of foodto retail. R values are loss rates in each FSC stage for fruits and vegetables for sub-Saharan Africa (top) and North America and Oceania (bottom)from Gustavsson et al.17 Each food type has unique food loss rates at each stage; the values for fruits and vegetables are shown here as an example.Emissions in this study are calculated for the preretail portion of the food supply chain (FSC) due to data constraints and the role of consumerbehavior and retailing heterogeneity in the downstream FSC. Further description of the study boundary and these terms available in the Materialsand Methods.

Environmental Science & Technology Article

DOI: 10.1021/acs.est.8b05322Environ. Sci. Technol. 2019, 53, 251−260

252

al.,17 three of which occur upstream (prior to retail). Thisanalysis defines food loss as edible food at one stage of the FSCthat is not supplied to the next stage of the FSC, correspondingwith common use in the literature.17,24 The boundary of thisstudy is the upstream, preretail portion of the FSC. Therefore,total food loss reported throughout this analysis is edible foodnot successfully supplied to retail. The functional unitconsidered is 1 kg of food reflecting a representative dietcomprised of the seven food types studied. A visual depictionof food mass in the model FSC is displayed in Figure 1.The sub-Saharan African (SSA) food system is the baseline

for this model. Sub-Saharan Africa is an ideal case to examinepotential cold chain deployment as it has some of the highestupstream loss rates for food,17 and is characterized by a lack ofcurrent cold chain infrastructure. The United States wasestimated to have 0.37 m3 of refrigerated storage per capita in2014, compared to estimates of 0.015 m3 per capita in urbanareas of South Africa in 2008, and estimates of 0.002 m3 percapita in urban areas of Ethiopia and the United Republic ofTanzania, and 0.0051 m3 per capita in urban areas of Namibiain 201227,28 (see Supporting Information 1).Two scenarios of cold chain introduction and food system

development are considered: one that substitutes NorthAmerican (NA) parameters into the model, and one thatsubstitutes European (Eur.) parameters. Modeling a transitionfrom the sub-Saharan African food system to one with NorthAmerican or European properties is the closest to a total(“zero-to-one”) introduction of the cold chain as can beexamined with available data. The results of this modelingprovide insights into the direct and indirect emissions effectsassociated with the cold chain as have currently been realizedin development.As seen in the comparison of fruit and vegetable loss rates

between sub-Saharan Africa and North America and Oceaniain Figure 1, a greater quantity of food needs to be produced insub-Saharan Africa to supply a similar amount of food to retail,attributable to more-developed food supply chains. Agricul-tural losses tend to be higher in North America and Oceaniadue to increased grading from higher quality standards set byretailers.17 These changes in grading standards are an exampleof how FSC development may influence consumer and retailerpreferences, affecting the efficiency and environmental impactsof food supply chains. In sub-Saharan Africa, the larger share oflosses occurring after agricultural production are attributed tocrop deterioration from climate exposure as well as crop glutsfrom the seasonality of production.17

Four parameters are integral to modeling the FSC for eachsystem: loss rates (% of food loss at FSC stages), demand (kgfood consumed per capita), agricultural emissions factors (kgCO2e/kg food), and cold chain emissions factors (kg CO2e/kgfood). The relationship between these parameters and specificcalculations conducted are detailed in the Methods section.Due to the fairly sparse and nonstandardized nature of data onfood and its environmental impacts, data sources wereharmonized to the extent possible. Harmonization choicesare detailed in Supporting Information 2.

■ MATERIALS AND METHODSThe changes in food supplied and the emissions associatedwith cold chain introduction are determined by adjusting lossrates (R1, R2, R3, R4), demand (F5), agricultural emissionsfactors (EA), and cold chain emissions factors (Ec). F1−4 aredetermined by the mass balance equations below. Emissions

factors characterize food (and food losses) which enter a stageand are subject to its emissions-contributing processes.Emissions are calculated for the preretail portion of the FSC,though demand is defined at the consumer level due to dataconstraints, and back-calculated using loss rates for the entireFSC. These parameters are drawn from the Monte Carlodistribution types described, with specific parameter describedin Supporting Information 3. Parameter distributions areassumed to be independent and 10 000 Monte Carlosimulations are run to produce this study’s results. Sensitivityanalysis for these parameters is detailed in SupportingInformation 4.There are five stages of the food supply chain corresponding

to Gustavsson et al.:17 1. Agricultural Production, 2.Postharvest Handling and Storage, 3. Processing and Pack-aging, 4. Distribution/Retail, and 5. Consumption, wherestages 1−3 are considered to be “upstream” and 4−5 are“downstream.” This analysis only examines emissions for theupstream supply chain. Values of variables which correspondto one of these stages are indicated with numerical subscripts(e.g., a subscript of “2” for a Postharvest Handling and Storagevalue).Every parameter is defined for each of the seven food types

studied: Cereals, Roots and Tubers, Fruits, Vegetables, Meat,Fish and Seafood, and Milk. Therefore, each model parameterhas a value associated with the seven food types (x) and threestudy regions (y). For example, R1,v,SSA denotes the loss rate ofvegetables between Agricultural Production and PostharvestHandling and Storage in SSA.As depicted in Figure 1, the food present at each section of

the supply chain can be represented by

= { }F F F F F F, , , .x y x y x y x y x y1, , 2, , 3, , 4 , 5, ,

Where Fn represents mass (kg) of each food type at each stageof the region’s FSC, x denotes the food type, and y denotes thestudy region. F5 is defined from a truncated normaldistribution (lower bound of zero) defined with “food” valuesfor each region and type from the 2013 FAOSTAT FoodBalance Sheets,29 capturing the food available for humanconsumption in each region within a given year.Between each stage of the FSC is a loss rate:

= { }R R R R R, , ,x y x y x y x y1, , 2, , 3, , 4, ,

Where Rn represents the percentage of food lost (% of kg)between FSCn and FSCn+1 for each of the seven food types (x)in each region (y). Loss rates calculated by Gustavsson et al.17

are used to define triangular Monte Carlo distributions for thisparameter for each food type and region, with specific valuesprovided in Supporting Information 3.The food loss for each type and region in each stage is

defined as Ln,x,y (kg food) and can be calculated as

= ×L F Rn x y n x y n x y, , , , , , (1)

The mass balance of the system can be represented as

= [{[ × − ] × − }

× − ] × −

F F R R

R R

(1 ) (1 )

(1 ) (1 )

x y x y x y x y

x y x y

5, , 1, , 1, , 2, ,

3, , 4, , (2)

Beginning with values obtained from FAOSTAT and usingmass balance, the food available at each upstream FSC stagecan be computed by

Environmental Science & Technology Article

DOI: 10.1021/acs.est.8b05322Environ. Sci. Technol. 2019, 53, 251−260

253

=−−

−F

F

R(1 )n x yn x y

n x y1, ,

, ,

1, , (3)

Direct Trade-Off between Food Savings and ColdChain Emissions. This analysis first evaluates the direct trade-off of additional cold chain emissions with potential savings infood loss throughout the upstream food supply chain. Thedirect trade-off calculation does not take into account anyindirect behavioral or system-wide changes. As such, itcalculates the potential differences in the system before andafter cold chain introduction by holding all elements of thebaseline SSA model constant, with the exception of theportions of the FSC where the cold chain is introduced andinduces changes in the food loss rates (R2, R3) and cold chainemissions factors (Ec), as detailed in eqs 4−6. The cold chaincodevelops and is integrated with related postharvest storage,processing, transportation, and spoilage-reducing supply chainproperties.9,11,24,30 As a result, some GHG emissions andchanges in loss rates attributed to the cold chain are notdirectly due to refrigeration, but cannot be distinguished orseparated from those which are in the data.eq 4 computes GHG emissions added through cold chain

operation when changed to the North American parameters. Asimilar equation is used to calculate the European scenario.

=+ +

−+ +

Δ

i

kjjjjj

y

{zzzzz

i

kjjjjj

y

{zzzzz

E EF L L

F

EF L L

F

C C x NAx x x

x

C xx x x

x

, ,4, ,NA 2, ,NA 3, ,NA

4, ,NA

, ,SSA4, ,SSA 2, ,SSA 3, ,SSA

4, ,SSA (4)

Where EΔC is the change in GHG emissions (kg CO2e/kgfood) added to the upstream FSC from cold chain operation.Since the baseline models a food system with negligible coldchain infrastructure, EC,x,SSA is assumed to be zero.L1 is not included in eq 4 since it pertains to losses from

agriculture and is not exposed to the cold chain. Cold chainemissions (kg CO2e/kg food) by food type are drawn fromlog-normal distributions, with parameters compiled fromaverages by food type using studies from Porter et al.’s meta-analysis18 which contained sufficient postfarm gate data onemissions from the cold chain.eq 5 calculates the difference in postharvest food loss

emissions from cold chain introduction for the North Americascenario, with a similar calculation performed for the Europeanscenario.

=+ +

−+ +

Δ

Ä

Ç

ÅÅÅÅÅÅÅÅÅÅÅ

i

kjjjjjj

y

{zzzzzz

i

kjjjjjj

y

{zzzzzz

É

Ö

ÑÑÑÑÑÑÑÑÑÑÑ

EF L L

F

F L L

F

E L x A xx x x

x

x x x

x

, , ,SSA4, ,SSA 2, ,SSA 3, ,SSA

4, ,SSA

4, ,NA 2, ,NA 3, ,NA

4, ,NA (5)

Where EΔL,x is the change in GHG emissions (kg CO2e/kg)from changes in food loss emissions associated with cold chainintroduction. Because the analysis only includes food lossemissions directly resulting from cold chain introduction,which occurs after agricultural production losses occur, thevalues associated with R1 (and subsequent calculation of L1)do not change.The EA,x values used in the analysis are weighted averages of

agricultural production emissions (kg CO2e/kg food) by foodtype with a cradle-to-farm gate boundary. Values are drawn

from log-normal distributions with parameters defined from ameta-analysis of life cycle assessments by Porter et al.18 Thesevalues include any environmental burdens prior to food leavingits place of agricultural production.The net emissions change comparing cold chain emissions

and food loss emissions is shown as eq 6.

= −Δ ΔE E ED C L (6)

Induced System-Wide Changes. Once the direct coldchain trade-off is calculated, this analysis estimates potentialsystem-wide shifts associated with cold chain introduction,including changes to agricultural production and shifts indietary patterns.Introduction of the cold chain has the potential to change

system logistics and expand agricultural distribution, makingthe parameters governing the SSA baseline case more similar toagricultural systems in either North America or Europe. Tomodel this, R1, L1, and EA, which were held constant whenestimating direct trade-offs, are now assumed to change inaddition to the direct trade-offs calculated in eqs 4−6.Changes in diet are considered as part of the system-wide

changes induced from the cold chain. Food supplied to retail isnormalized to one kilogram of a representative diet, whereeach fraction corresponds to the fraction of each food type inthe diet examined.Per-capita demand is calculated for each region as

=CF

Px yx y

y,

5, ,

(7)

Where Cx,y is the per-capita food consumption of a food type xin region yAnd Py is the population for the region.The shift toward diets similar to Europe and North America

is then calculated as shown in eq 8.

= ×FC

CF x y x

x y

x5, , 5, ,SSA

,

,SSA (8)

Food supply emissions are calculated in eq 9, both when diethas been held constant and when it has been shifted.

=+

∑+

+ +

∑= =

Ä

Ç

ÅÅÅÅÅÅÅÅÅÅÅÅÅ

i

k

jjjjjjjy

{

zzzzzzzi

k

jjjjjjjy

{

zzzzzzz

É

Ö

ÑÑÑÑÑÑÑÑÑÑÑÑÑE E

F L

FE

F L L

FA x yx y x y

x x yC x y

x y x y x y

x x yP , ,

2, , 1, ,

17

4, ,, ,

4, , 2, , 3, ,

17

4, ,

(9)

■ RESULTSTrade-Off Between Added Emissions and Avoided

Food Losses in the Cold Chain. A fundamental question forrefrigerated supply chain sustainability is whether the increasedemissions from cold chain operation will eclipse the avoidedemissions from reduced food spoilage. eqs 4−6 are used tocalculate this trade-off and the results are depicted in Figures 2and 3.In total, the cold chain is found to add more emissions than

it saves through avoided food losses. Without taking intoaccount any other changes to the system, introducingrefrigeration to sub-Saharan Africa would increase net food-related GHG emissions by 10% from the baseline in the NorthAmerican scenario and 2% in the European scenario, despitereducing postharvest food loss quantities by 23% in bothscenarios. The difference in these emissions increases is due tothe recorded North American cold chain emissions being

Environmental Science & Technology Article

DOI: 10.1021/acs.est.8b05322Environ. Sci. Technol. 2019, 53, 251−260

254

larger than those for Europe for five out of seven food types,while avoided food loss emissions are similar for bothscenarios.While total emissions added are larger than loss emissions

avoided, the difference between these vary by food type andscenario. Figure 2 shows the cold chain adding more emissionsthan it avoids on a per kg basis for five of seven food categoriesif North American values are used, and for three of seven foodcategories if European values are used. The largest cold chainemissions are associated with fish and seafood, meat, andvegetables in the North American scenario, and with fish andseafood, meat, and cereals in the European scenario. The foodtypes that have the greatest reductions in food loss are fish andseafood, vegetables, and milk in both scenarios. This studyfinds mixed results for fruit depending on developmentscenario, though an evaluation of kinnow spoilage in Indiafound GHG reductions of 16% from cold chain presence.31

For both scenarios, emissions associated with food lossactually increase for cereals and meat. For cereals, the increasein food losses result from the addition of a specific “packaging”loss rate in the North American and European processing andpackaging stage (R3), which is not present for sub-SaharanAfrica in Gustavsson et al.17 Meat losses increase by 0.3% inNorth American postharvest handling and storage (R2),affecting the MCA distributions for North America andEurope (see Supporting Information 3). The cause for anincreased postharvest meat loss rate in North America is notdiscussed by Gustavsson et al.,17 but may be from meat supplypractices present in North America but not as common in sub-Saharan Africa (such as the transportation, slaughter, andportioning of meat prior to retail rather than slaughteringanimals for meat at market32 or for immediate consumption).Both food loss-related emissions increases are modest in size,but highlight the need to consider cold chain introduction as

inseparable from interconnected changes in the food supplychain.9

The distribution of differences between added cold chainemissions and avoided loss emissions by food type and in totaldietary emissions are displayed in Figure 3. With theexceptions of meat and fish and seafood, the median differencebetween these values is close to zero, indicating eithernegligible changes to food types that are not typicallyrefrigerated or that any increase in cold chain emissions areoffset by a similar amount of embodied emissions within foodsavings. Meat and fish and seafood both show larger emissionsincreases, and also possess larger variances. This indicates thatthe amount of food savings is insufficient to offset increases inemissions introduced by the cold chain.The histograms in Figure 3c and d show the expected

change in GHG emissions due to cold chain introduction,using the weighted averages of each food type in the averagesub-Saharan diet. A larger share of total emissions differencesare greater than zero for the North American scenario than forthe European scenario. The North American scenario addedmore cold chain emissions than loss emissions avoided in99.9% of runs, and the European scenario resulted in moreemissions added than were saved in 89% of runs.

Indirect Effects of Cold Chain Introduction onUpstream Food Supply Emissions. The influence of coldchain introduction on upstream FSC emissions is nowexamined from a broader, systems perspective, incorporatingchanges to agricultural production and demand.Refrigeration enables structural changes in food production

systems. For example, cold storage allows for agriculturesystem industrialization, since farms can supply a greaterquantity of perishable crops due to lower spoilage rates.33 Theindirect effect of cold chain introduction on agriculturalemissions is modeled by changing the parameters for

Figure 2. Comparison of median emissions added from cold chain introduction and emissions associated with avoided food losses. The calculatedvalues pertain to emissions occurring during the postharvest and preretail supply chain (i.e., L2, and L3 in Figure 1). The calculated differenceindicates the direct trade-off between introduced cold chain emissions and avoided food loss emissions for each food type.

Environmental Science & Technology Article

DOI: 10.1021/acs.est.8b05322Environ. Sci. Technol. 2019, 53, 251−260

255

agricultural emissions (EA) and agricultural production lossrates (R1) from their SSA values to the North American andEuropean values. These changes are made in addition to thepostagriculture loss rates and cold chain emission changesreflected in Figures 2 and 3.Access to refrigeration changes food demand. The cold

chain allows for the supply and consumption of perishablefood products in a way not possible without robust refrigeratedsupply chains,9 and has been linked with shifts in diet asnations develop.3,34 The effects of demand changes reflecting aNorth American or European diet facilated by the cold chainare examined. The food demand parameter (F5) is adjustedfrom its baseline value in addition to the values for agriculturalproduction emissions, loss rates, and cold chain emissions.Figure 4 shows changes in the emissions required to supply a

representative kilogram of food to retail, based on a weightedaverage of each food type using median MCA values for eachparameter. Changes are displayed first with cold chainintroduction and changes in agricultural production emissions

but with the baseline diet, then with demand changes fromdietary shifts.When examining the indirect effects of the cold chain on

agricultural production in addition to its direct effects,emissions decrease in both development scenarios: by 46%for the North American scenario and 49% in the Europeanscenario. Emissions decreases are largest for vegetables, milk,and cereals in the North American scenario, and for milk,vegetables, and meat in the European scenario. These resultsalign with a prior study indicating a decrease in food lossGHGs of 38% is possible from supply chain improvementsincluding cold chain introduction.16

Changes in agricultural production emission factors, whichdecrease with development, put a downward pressure onemissions. It must be noted that there are trade-offs associatedwith industrialized agricultural systems which may decrease theemissions per kg of food produced, but may increase otherenvironmental consequences including water pollution, soil

Figure 3. Boxplots and histograms of the difference between added cold chain emissions and avoided loss emissions in the postharvest cold chainfor both introduction scenarios. Panel (A) is a boxplot of the emissions difference per kg of each food type food delivered to retail for the NorthAmerican scenario, with Panel (C) reflecting these emissions for the European scenario. Boxes show the range of values between the 25th and 75thpercentiles generated from Monte Carlo Analysis, with the box’s line indicating the median. The gray tails are data points generated which falloutside of this interquartile range. Panel (B) shows the histogram of total net emissions for the North American scenario’s model runs based on aweighted average of food types, with Panel (D) showing these results for the European scenario.

Environmental Science & Technology Article

DOI: 10.1021/acs.est.8b05322Environ. Sci. Technol. 2019, 53, 251−260

256

depletion, biodiversity loss, and also geographically concen-trate these effects.35

The agricultural production loss rate for roots and tubersincreases in both development scenarios due to increasedgrading standards for produced food (see SupportingInformation 3).17 Fruits and vegetables see similar increasesin their agricultural production loss rate due to grading, butexperience decreases in loss rates in the later upstream stageswhich result in a net decrease in overall upstream loss rates.Increased grading standards may be considered as a way inwhich consumer demand influences FSC parameters, with thevisual appearance of food being a key determinant of foodacceptance and perceived quality by consumers.36,37 However,since fruit and vegetable exposure to refrigeration is typical intheir developed supply chains,38 these losses are recoupedthrough decreased postharvest spoilage with supply chaindevelopment. Roots and tubers, on the other hand, experiencelosses due to grading and are not always subject to refrigerationin developed supply chains, and in some large storehouses maybe cooled with ventilation from outdoor air.39 Reductions inagricultural loss rates put a downward pressure on emissionsfor all other food types.Upstream emissions do not uniformly change when

incorporating demand changes. Food supply emissionsincrease by 10% for the North American scenario but decrease

by 15% for the European scenario. The difference betweenthese outcomes is primarily due to the level of meatconsumption in the North American diet, where the per-capita meat consumption is 37% greater than in the Europeanscenario, corresponding to a meat emissions increase of 96%over the baseline. The North American scenario also seesemissions increases from fruits and fish and seafood whenincorporating demand shifts. The European scenario seesincreases in meat and milk emissions with dietary change, butstill experiences a total decrease in upstream emissions.The demand shifts modeled capture both substitutions

between food types within a diet, but also increases in totalquantities consumed. In this context of sub-Saharan Africa,increases in calorie consumption would improve healthoutcomes for many individuals,40 an effect not measured inthis model. Pradhan et al. characterize diet types by caloriecomposition, and find low-calorie diets to be decreasingworldwide, with general shifts toward higher-calorie observedwith development.41 Increased availability of refrigeration hasbeen connected to increased consumption of perishable fooditems,3 which may also improve nutritional outcomes.42

Pradhan et al. find low calorie diets observed in the developingworld to have similar GHG emissions as higher-calorie diets inthe developed world, attributable to differences in foodproduction efficiency.41 The connection between the cold

Figure 4. Changes in upstream food supply emissions (kg CO2e) required to deliver one kg of a representative diet, based on a weighted average ofeach food type within a typical diet. Percentage differences in emissions are displayed by corresponding food type in the graph.

Environmental Science & Technology Article

DOI: 10.1021/acs.est.8b05322Environ. Sci. Technol. 2019, 53, 251−260

257

chain and economic development related to shifts in fooddemand, supply, and trade should be examined as the subjectof future research, as there are notable aspects of well-beingand health that are not taken into account in this study.The demand shifts modeled illustrate scenarios of dietary

convergence. In an analysis of the GHG implications of dietaryconvergence, Ritchie et al. find modeled diets for the U.S.,Australia, Canada, and Germany exceeding average per capitaemissions budgets for 1.5 °C of global warming by 2050.43

That being said, the dietary shifts examined in this study arenot preordained, merely reflecting two plausible diets in adeveloped food system.Culture and development individual to any given area will be

a critical determinant of diet. If diets develop to correspondwith South Africa’s nationally recommended diet as modeledby Behrens et al.,44 emissions increase 7% or decrease 4% fromthe baseline, depending on whether North American orEuropean values are used for the other model parameters. Thisfinding illustrates how emissions decreases (or more-modestincreases) could accompany health improvements if dietsdevelop in line with a regional nationally recommended diet.Additional details regarding this diet are in SupportingInformation 5.These results indicate the importance of incorporating a

technology’s influence on consumer preferences into anassessment of its environmental outcomes. Despite decreasedagriculture emissions associated with the cold chain,refrigeration may prompt shifts toward more emissions-intensefoods, creating a scenario of increased environmental impacts.

■ DISCUSSIONIn contextualizing the results of this analysis, it should be notedthat this study focuses only on GHG emissions, and does nottake into account societal benefits of the cold chain, whichinclude food security, health outcomes, nutrition, andeconomic development. The purpose of the study is tohighlight the GHG trade-offs of the technology in order toidentify potential areas for improvement as the cold chaincontinues to expand globally.We find that the emissions from cold chain operation will

likely exceed the emissions saved from reductions in foodlosses, if the cold chain is implemented in a way whichresembles its presence in North America or Europe. While theresults for individual food types vary, these net emissionsincreases are larger and more statistically certain to occur inthe North American development scenario than the Europeanscenario. This difference is due to the magnitude of cold chainemissions recorded for each region.This study presents findings relevant to a number of

stakeholders. Manufacturers of refrigeration equipment canmitigate emissions increases by employing efficiency improve-ments, the substitution of refrigerants with low GlobalWarming Potentials, and/or working with firms along theFSC to increase efficiency. The Postharvest EducationFoundation has produced a valuable white paper onconsiderations for the use of the cold chain in developingareas.4 Potential emissions increases from shifts to high-GHGdiets could be mitigated through reducing food losses and theconsumption of particularly emissions-intensive foods such asbeef.45 Shifting diets is a complex topic, which intersects withelements of culture, equity, and nutrition. Garnett provides adiscussion of the best opportunities for mitigating food systemGHGs, highlighting key opportunities and challenges.46

The Kigali Amendment to the Montreal Protocol will haveAfrican nations freeze the use of hydrofluorocarbon (HFC)refrigerants for most countries by 2024.47 These refrigerantscarry high global warming potential values,48 with HFC leakagefrom stationary refrigeration estimated to release 1740 000tonnes of CO2e in 2005,

49 and use in the mobile portion of thecold chain comprising 7% of global HFC consumption.21 Thisamendment presents the opportunity to reduce directenvironmental impacts from refrigeration. The MontrealProtocol has been a remarkably successful example ofinternational environmental governance,50 with past adherenceby signatories and industry cooperation indicating futuresuccesses for the Kigali Amendment. Refrigerators and coldchain technology will also likely experience increases inefficiency over time, which could decrease direct emissions.Dahmus notes that energy efficiency improvements in U.S.residential refrigerators since the 1960s has been enough tomitigate resource consumption increases driven by increasedrefrigerator ownership and size.51 These improvements areattributed to efficiency mandates, further highlighting the roleof governance and regulation in mitigating potential emissionsincreases from technology.As noted by Porter et al.,18 multiple entries in the literature

find that production/prefarm gate emissions comprise themajority (ranging from 50−90%) of emissions associated witha food product. However, postfarm processes includingrefrigeration make both direct and indirect emissionscontributions. When incorporating indirect emissions impacts(such as dietary shifts), the total emissions from postfarmprocesses are larger than just their direct emissions. The coldchain is an integral element of an industrialized food system,with introduction enabling highly integrated systems connect-ing agricultural producers and the postharvest food supplychain.24 These feedbacks necessitate a systems view of the FSCin order to capture the full influence and environmentalimpacts associated with the cold chain.When incorporating the cold chain’s indirect effects,

decreases in agricultural production emissions and upstreamfood losses decrease total upstream emissions in supplyingfood to retail. However, incorporating shifts in diet leads to anincrease in total emissions in the North American scenario anda decrease in the European scenario. This difference isattributable to higher meat consumption in the NorthAmerican diet. The outsized role of meat-intense diets incomprising food system emissions has been quantified for theUnited States’ diet.52 It is possible that dietary shifts enabledby increased access to perishable foods could eclipse GHGadditions from the cold chain, but this depends largely onconsumer choices. Promoting reduced-meat diets requiresengaging with sociocultural norms as well as psychologicalperceptions, and may require different strategies to be effectivefor different groups of people.53

The influence of behavioral choices and diet on food systememissions has been noted in the literature.45,46 Whileanticipated shifts in diets are modeled and addressed in thesustainability literature, they are infrequently integrated withmore-technically oriented models of the FSC. Similarly,differences in food production systems are often not accountedfor in studies of sustainable diets.34 Without includingbehavioral and production system differences in modelingthe FSC, important influences on environmental outcomesmay not be captured.

Environmental Science & Technology Article

DOI: 10.1021/acs.est.8b05322Environ. Sci. Technol. 2019, 53, 251−260

258

Data on food losses and waste are limited anduncertain,24,54,55 presenting distinct challenges in creatinginformed models. There is similarly limited data on the coldchain, particularly in the developing world.30 These dataquality issues affect this study, which draws on limited anduncertain data for all major model parameters. While therehave been means proposed to better-optimize data collectionfrom food life cycle assessments (studying the environmentalimpacts of a product throughout its lifespan),56 differentreporting formats, functional units, and system boundariespose challenges in data collection and standardization.Improving the quantity and quality of estimates for food lossand waste rates, and the environmental impacts from foodproduction and supply are critical research needs.Sub-Saharan Africa is not a uniform region, and contains

notable heterogeneity and differences within it. Theaggregation of this region as a baseline case is a limitation ofthis study which can be improved upon by future work. Inaddition to differences in cold chain penetration, diet, andagricultural production, sub-Saharan Africa differs from NorthAmerica and Europe in local ambient temperature. This willaffect elements of the food system ranging from agriculturalproduction57 to the efficiency and emissions of cold chainoperation.26

Development does not occur smoothly, and is oftenasymmetric in ways which are difficult to capture in a model.Assumptions including the matching of food demand withsupply and reliable provision of energy from the electricity gridmay differ from an observed development process. Thisanalysis assumes no improvements in cold chain technologyupon introduction; however, James and James suggest that thecold chain can be extended without an increase in global CO2,or possibly even with a decrease, if the most energy efficientrefrigeration technologies are used.58 The deployment ofrenewable and alternative energy technologies such decentral-ized solar power in areas of Africa59,60 could also provideimportant emissions reductions within the food systemstudied, and have been identified as a key means of reducingpostfarm food system emissions.46

Refrigerated supply chains transform food systems. Examin-ing the introduction of the cold chain requires modeling morethan the technology itself: incorporating the behavioral andbroader systemic changes which accompany it. This systemsview allows for greater insights into environmental trade-offsand changes in food system sustainability.

■ ASSOCIATED CONTENT*S Supporting InformationThe Supporting Information is available free of charge on theACS Publications website at DOI: 10.1021/acs.est.8b05322.

Five further-detailed descriptions of methods, nine tablesof model parameters, one figure displaying results ofsensitivity analysis, and one figure displaying detailedresults from modeling the nationally recommended dietscenario (PDF)

■ AUTHOR INFORMATIONCorresponding Author*E-mail: brheard@umich.edu.ORCIDBrent R. Heard: 0000-0001-9761-2043Shelie A. Miller: 0000-0003-0379-3993

NotesThe authors declare no competing financial interest.

■ ACKNOWLEDGMENTSWe thank Martin C. Heller for his advice and suggestions ondata harmonization for the cold chain emissions factors.

■ REFERENCES(1) Aung, M. M.; Chang, Y. S. Temperature Management for theQuality Assurance of a Perishable Food Supply Chain. Food Control2014, 40, 198−207.(2) Sahin, E.; Zied Babaï, M.; Dallery, Y.; Vaillant, R. EnsuringSupply Chain Safety through Time Temperature Integrators. Int. J.Logist. Manag. 2007, 18 (1), 102−124.(3) Garnett, T. Food Refrigeration: What Is the Contribution toGreenhouse Gas Emissions and How Might Emissions Be Reduced?;FCRN Working Paper; April 2007; 2007.(4) Kitinoja, L. Use of Cold Chains for Reducing Food Losses inDeveloping Countries; 13−03; 2013.(5) Zhang, J.; Pritchard, E.; Hu, X.; Valentin, T.; Panilaitis, B.;Omenetto, F. G.; Kaplan, D. L. Stabilization of Vaccines andAntibiotics in Silk and Eliminating the Cold Chain. Proc. Natl. Acad.Sci. U. S. A. 2012, 109 (30), 11981−11986.(6) SDG Target 12.3 on Food Loss and Waste: 2016 Progress Report;Food and Agriculture Organization of the United Nations, 2016.(7) UN Resolution 2020 Agenda for Sustainable Development. InSeventieth Session Agenda Items 15 and 116; United Nations, 2015.https://doi.org/10.1007/s13398-014-0173-7.2.(8) Markets and Markets. Cold Chain Market worth 234.49 BillionUSD by 2020 http://www.marketsandmarkets.com/PressReleases/cold-chain.asp.(9) Heard, B. R.; Miller, S. A. Critical Research Needed to Examinethe Environmental Impacts of Expanded Refrigeration on the FoodSystem. Environ. Sci. Technol. 2016, 50 (22), 12050−12071.(10) James, S. J.; James, C. The Food Cold-Chain and ClimateChange. Food Res. Int. 2010, 43, 1944−1956.(11) Garnett, T. Food Refrigeration: What Is the Contribution toGreenhouse Gas Emissions and How Might Emissions Be Reduced?, 2007.(12) Food Wastage Footprint: Impacts on Natural Resources; Food andAgriculture Organization of the United Nations, 2013. ISBN 978-92-5-107752-8.(13) Papargyropoulou, E.; Lozano, R.; Steinberger, K. J.; Wright, N.;Ujang, Z. B. The Food Waste Hierarchy as a Framework for theManagement of Food Surplus and Food Waste. J. Cleaner Prod. 2014,76, 106−115.(14) Food Loss and Waste Accounting and Reporting Standard; WorldResources Institute, 2016.(15) Food and Agriculture Organization of the United Nations. SaveFood for a Better Climate: Converting the Food Loss and WasteChallenge into Climate Action; Rome, 2017.(16) Food and Agriculture Organization of the United Nations. FoodWastage Footprint & Climate Change; 2011.(17) Gustavsson, J.; Cederberg, C.; Sonesson, U. Global Food Lossesand Food Waste: Extent, Causes, and Prevention; 2011.(18) Porter, S. D.; Reay, D. S.; Higgins, P.; Bomberg, E. A Half-Century of Production-Phase Greenhouse Gas Emissions from FoodLoss & Waste in the Global Food Supply Chain. Sci. Total Environ.2016, 571, 721−729.(19) Hic, C.; Pradhan, P.; Rybski, D.; Kropp, J. P. Food Surplus andIts Climate Burdens. Environ. Sci. Technol. 2016, 50, 4269−4277.(20) U.S. Government Global Food Security Strategy; United StatesAgency for International Development, 2016.(21) Global Food Cold Chain Council; BIO Intelligence Service.Assessing the Potential of the Cold Chain Sector to Reduce GHGEmissions through Food Loss and Waste Reduction; 2015.(22) Carrier Transicold. India Pilot Study Shows How the Cold ChainCan. Help Reduce Food Loss and Carbon Emissions https://www.carrier.com/carrier/en/us/news/news-article/india_pilot_study_

Environmental Science & Technology Article

DOI: 10.1021/acs.est.8b05322Environ. Sci. Technol. 2019, 53, 251−260

259

shows_how_the_cold_chain_can_help_reduce_food_loss_and_carbon_emissions.aspx.(23) Miller, S. A.; Keoleian, G. A. Framework for AnalyzingTransformative Technologies in Life Cycle Assessment. Environ. Sci.Technol. 2015, 49, 3067−3075.(24) Parfitt, J.; Barthel, M.; Macnaughton, S. Food Waste withinFood Supply Chains: Quantification and Potential for Change to2050. Philos. Trans. R. Soc., B 2010, 365 (1554), 3065−3081.(25) Zanoni, S.; Zavanella, L. Chilled or Frozen? Decision Strategiesfor Sustainable Food Supply Chains. Int. J. Prod. Econ. 2012, 140,731−736.(26) James, S. J.; James, C. The Food Cold-Chain and ClimateChange. Food Res. Int. 2010, 43 (7), 1944−1956.(27) AGRO Merchants Group. Worldwide Cold Storage CapacityEstimated at 552 Mi Cubic Meters; 2018.(28) International Institute of Refrigeration. Developing the ColdChain in the Agrifood Sector in Sub-Saharan Africa; Food andAgriculture Organization of the United Nations, 2016.(29) Food and Agriculture Organization of the United Nations.FAOSTAT.(30) Yahia, E. M. Cold Chain Development and Challenges in theDeveloping World. In 6th International Postharvest Symposium; Erkan,M., Aksoy, U., Eds.; 2010; pp 127−132.(31) Cold Chain Development for Fruits & Vegetables in India: KnnowCold Chain Study; Carrier & United Technologies, 2016.(32) Grace, D.; Roesel, K. What’s Eating Sub-Saharan Africa?; AlJazeera January 27, 2015; pp 1−5.(33) Reddy, G. P.; Murthy, M. R. K.; Meena, P. C. Value Chains andRetailing of Fresh Vegetables and Fruits. Andhra Pradesh. Agric. Econ.Res. Rev. 2010, 23 (July), 455−460.(34) Garnett, T. Plating up Solutions. Science 2016, 353 (6305),1202−1204.(35) Horrigan, L.; Lawrence, R. S.; Walker, P. How SustainableAgriculture Can Address the Environmental and Human HealthHarms of Industrial Agriculture. Environ. Health Perspect. 2002, 110(5), 445−456.(36) Wadhera, D.; Capaldi-Phillips, E. D. A Review of Visual CuesAssociated with Food on Food Acceptance and Consumption. Eat.Behav. 2014, 15 (1), 132−143.(37) Aschemann-Witzel, J.; de Hooge, I.; Amani, P.; Bech-Larsen,T.; Oostindjer, M. Consumer-Related Food Waste: Causes andPotential for Action. Sustainability 2015, 7 (6), 6457−6477.(38) Paull, R. E. Effect of Temperature and Relative Humidity onFresh Commodity Quality. Postharvest Biol. Technol. 1999, 15, 263−277.(39) Gottschalk, K. Mathematical Modelling Of The ThermalBehaviour Of Stored Potatoes & Developing Of Fuzzy ControlAlgorithms To Optimise The Climate In Storehouses. Acta Hortic.1996, No. 406, 331−340.(40) Abrahams, Z.; McHiza, Z.; Steyn, N. P. Diet and MortalityRates in Sub-Saharan Africa: Stages in the Nutrition Transition. BMCPublic Health 2011, 11, 801. https://doi.org/10.1186/1471-2458-11-801. DOI: 10.1186/1471-2458-11-801(41) Pradhan, P.; Reusser, D. E.; Kropp, J. P. Embodied GreenhouseGas Emissions in Diets. PLoS One 2013, 8 (5), 1−8.(42) International Organization for the Development of Refriger-ation. 5th Informatory Note on Refrigeration and Food: The Role ofRefrigeration in Worldwide Nutrition; Paris, 2009.(43) Ritchie, H.; Reay, D. S.; Higgins, P. The Impact of GlobalDietary Guidelines on Climate Change. Glob. Environ. Chang. 2018,49 (February), 46−55.(44) Behrens, P.; Kiefte-de Jong, J. C.; Bosker, T.; Rodrigues, J. F.D.; de Koning, A.; Tukker, A. Evaluating the Environmental Impactsof Dietary Recommendations. Proc. Natl. Acad. Sci. U. S. A. 2017, 114(51), 13412−13417.(45) Heller, M. C.; Keoleian, G. A. Greenhouse Gas EmissionEstimates of U.S. Dietary Choices and Food Loss. J. Ind. Ecol. 2015,19 (3), 391−401.

(46) Garnett, T. Where Are the Best Opportunities for ReducingGreenhouse Gas Emissions in the Food System (Including the FoodChain)? Food Policy 2011, 36, S23−S32.(47) United Nations Environment Programme. Kigali Amendment tothe Montreal Protocol: Another Global Commitment to Stop ClimateChange https://www.unenvironment.org/news-and-stories/news/kigali-amendment-montreal-protocol-another-global-commitment-stop-climate.(48) United Nations Environment Programme. The Kigali Amend-ment to the Montreal Protocol: HFC Phase-down, OzonAction Fact Sheet,2016.(49) Emissions and Projections of HFCs, PFCs and SF6 for the UK andConstituent Countries; AEA Technology Environment, 2004.(50) DeSombre, E. R. The Experience of the Montreal Protocol:Particularly Remarkable and Remarkably Particular. UCLA J. Environ.Law Policy 2000, 19 (1), 49−81.(51) Dahmus, J. B. Can Efficiency Improvements Reduce ResourceConsumption? J. Ind. Ecol. 2014, 18 (6), 883−897.(52) Heller, M. C.; Willits-Smith, A.; Meyer, R.; Keoleian, G. A.;Rose, D. Greenhouse Gas Emissions and Energy Use Associated withProduction of Individual Self-Selected US Diets. Environ. Res. Lett.2018, 13, No. e044004.(53) Uta, S. S.; Schmidt, J. Reducing Meat Consumption inDeveloped and Transition Countries to Counter Climate Change andBiodiversity Loss: A Review of Influence Factors. Reg. Environ. Chang.2017. .171261(54) Reutter, B.; Lant, P. A.; Lane, J. L. The Challenge ofCharacterising Food Waste at a National LevelAn AustralianExample. Environ. Sci. Policy 2017, 78 (September), 157−166.(55) Xue, L.; Liu, G.; Parfitt, J.; Liu, X.; Van Herpen, E.; Stenmarck,A.; O’Connor, C.; Ostergren, K.; Cheng, S. Missing Food, MissingData? A Critical Review of Global Food Losses and Food Waste Data.Environ. Sci. Technol. 2017, 51 (12), 6618−6633.(56) Pernollet, F.; Coelho, C. R. V; van der Werf, H. M. G. Methodsto Simplify Diet and Food Life Cycle Inventories: Accuracy versusData-Collection Resources. J. Cleaner Prod. 2017, 140, 410−420.(57) Rosenzweig, C.; Elliott, J.; Deryng, D.; Ruane, A. C.; Muller, C.;Arneth, A.; Boote, K. J.; Folberth, C.; Glotter, M.; Khabarov, N.; et al.Assessing Agricultural Risks of Climate Change in the 21st Century ina Global Gridded Crop Model Intercomparison. Proc. Natl. Acad. Sci.U. S. A. 2014, 111 (9), 3268−3273.(58) James, S. J.; James, C. Sustainable Cold Chain. In SustainableFood Processing; Tiwari Brijesh, K., Norton, T., Holden, N. M., Eds.;John Wiley & Sons, Ltd, 2013; pp 463−496. https://doi.org/10.1002/9781118634301.ch19.(59) Ulsrud, K.; Winther, T.; Palit, D.; Rohracher, H. Village-LevelSolar Power in Africa: Accelerating Access to Electricity Servicesthrough a Socio-Technical Design in Kenya. Energy Res. Soc. Sci. 2015,5, 34−44.(60) Szabo, S.; Bo dis, K.; Huld, T.; Moner-Girona, M. EnergySolutions in Rural Africa: Mapping Electrification Costs ofDistributed Solar and Diesel Generation versus Grid Extension.Environ. Res. Lett. 2011, 6, 1−9.

Environmental Science & Technology Article

DOI: 10.1021/acs.est.8b05322Environ. Sci. Technol. 2019, 53, 251−260

260