244

www.frontiersinecology.org © The Ecological Society of America

Consumption of local food has rapidly escalated inrecent years, in parallel with a broad debate on the

potential benefits of transitioning to a food system thatinvolves much smaller distances between areas of produc-tion and consumption (Halweil 2002; Hinrichs 2003; Roseet al. 2008; Peters et al. 2009; Morrison et al. 2011). Localfood systems are characterized by “foodsheds” – geographicareas in which food is both produced and consumed. Whilereducing “food miles” may generally lead to only modestreductions in atmospheric greenhouse-gas (GHG) emis-sions relative to those possible through reduced consump-tion of animal products, there remains an unprecedentedimpetus to design a food system that accommodates localfood production, because of the potential for synergies withagroecosystem sustainability, rural economies, and foodsecurity (Kloppenburg et al. 1996; Allen et al. 2003; DuPuisand Goodman 2005; Born and Purcell 2006; Weber andMatthews 2008; Martinez et al. 2010; DeLind 2011).

First, there are specific cases in which reducing foodmiles will result in large GHG emission reductions; forexample, highly perishable produce and frozen productsoften require air freight and refrigeration during transit(Pelletier et al. 2011). Second, greater dependence onlocal food shortens the distances required for economicand energy-efficient recycling of waste streams betweenfarms and cities (Dawson and Hilton 2010). For instance,the recycling of urban organic waste as a compost fertil-izer can reduce solid waste accumulation in landfills andsoil degradation, but this would require cities and farms tobe in close proximity because of the high shipping costsfor bulky and low-value compost material (Kan et al.2010). Finally, local food systems may increase commu-

nity involvement in food production issues, potentiallyleading to improved environmental constraints on land-use practices (Gussow 1999; Magdoff 2007). The transi-tion to local food production may accelerate as globaldemand for transportation fuel intensifies and as newpolicies – such as the 2014 US Farm Bill – provideincreasing support for local and regional food systems.

While most research to date has focused on evaluatingthe potential benefits of local food systems, an emergingpolicy-relevant question is whether local food systemscan potentially scale beyond niche markets and replacean appreciable fraction of conventional food systems(Herrin and Gussow 1989; Cowell and Parkinson 2003;Peters et al. 2007, 2009; Thompson et al. 2008; Timmonset al. 2008; Hu et al. 2011). We define the foodshedpotential as the fraction of total dietary needs that couldbe met if all existing croplands were repurposed for localfood consumption. This foodshed potential is a functionof the spatial distribution of human populations, croplandareas, cropland productivities, and dietary requirements.It may also be an upper estimate of capacity because itdoes not account for economic and social factors such asfood preferences and prices that may constrain theexpansion of local food systems. One study focusing onNew York State found that repurposing all existing crop-lands as local food systems would theoretically meet only34% of the total statewide food requirements (Peters et al.2009). This case study reveals considerable spatial vari-ability in foodshed potential throughout the state and, inparticular, the low potential for large coastal cities.

Although the possible role of localized food provision hasbeen demonstrated in select regions (Peters et al. 2007, 2009;Hu et al. 2011), little is known about the theoreticalnational capacity of local food systems, which is more rele-vant to national policy. We also lack an understanding of

RESEARCH COMMUNICATIONS RESEARCH COMMUNICATIONS

The potential for local croplands to meetUS food demand Andrew Zumkehr and J Elliott Campbell*

Local food systems may facilitate agroecological practices that conserve nutrient, energy, and water resources.However, little is known about the potential for local food systems to scale beyond niche markets and meet asubstantial fraction of total food demand. Here we estimate the upper potential for all existing US croplands tomeet total US food demand through local food networks. Our spatially explicit approach simulates the years1850 through 2000 and accounts for a wide range of diets, food waste, population distributions, cropland areas,and crop yields. Although we find that local food potential has declined over time, particularly in some coastalcities, our results also demonstrate an unexpectedly large current potential for meeting as much as 90% of thenational food demand. This decline in potential is associated with demographic and agronomic trends, result-ing in extreme pressures on agroecological systems that, if left unchecked, could severely undermine recentnational policies focused on food localization. Nevertheless, these results provide a spatially explicit foundationfor exploring the many dimensions of agroecosystem sustainability.

Front Ecol Environ 2015; 13(5): 244–248, doi:10.1890/140246

Sierra Nevada Research Institute, University of California, Merced,Merced, CA *(ecampbell3@ucmerced.edu)

A Zumkehr and JE Campbell Local croplands and US food demand

245

© The Ecological Society of America www.frontiersinecology.org

the changes in foodshed potential that may have occurredover time and that may be related to historical trends in cropyields, suburban sprawl, urbanization, agricultural abandon-ment, and population growth (Figure 1, a and b). The his-torical increase in crop yields may enhance the foodshedpotential. Alternatively, foodshed potential may havedecreased due to competing trends, including stabilization ofcropland area, agricultural abandonment, and populationgrowth (NASS 2013; Zumkehr and Campbell 2013).

n Methods

We approximated the theoretical upper potential for allexisting US croplands to meet total US food demandthrough local food networks, based on a range of esti-mates for foodshed radius, diet, food waste, populationdistribution, cropland area, and crop productivity. As inprevious work, this theoretical potential assumes that allexisting cropland is allocated to produce the food groupsrequired for a standard US diet. While the local food sys-

tem will not provide exactly the same foods that are cur-rently consumed, the assumed diet is designed to providethe equivalent calories in all food groups (grains, vegeta-bles, fruit, dairy, eggs, and meat) as are currently con-sumed. Such a transition to local food would require sea-sonal storage and changes in crop types, marketing, andinfrastructure – shifts that would be further constrainedby social (eg food preferences) and economic (eg foodprices) factors. Thus, the estimates presented here pro-vide a needed theoretical baseline for examining localfood systems across broad spatial and temporal scales.

Our foodshed analysis consisted of three sequential steps:(1) estimating the per-capita food demand, (2) creatingmaps of the number of people that could be fed based onthe amount of cropland present in each map grid cell, and(3) selecting which cells should be allocated to whichcities in order to maximize the percentage of the total USpopulation that can be fed with locally produced food.

The food demand for the first step of our analysis wasobtained from a previously published diet model (Peters et

Figure 1. Historical US demographic trends, agronomic trends, and fraction of the population that can be fed by local food. Trendsare for the years 1850 to 2000 in the US, including (a) population and (b) cropland area and corn yields. Percentage of thepopulation that could theoretically be fed within a local foodshed in the conterminous US for vegetarian, standard (ie typical US diet),and meat-intensive diets for (c) a 50-mile foodshed radius and (d) a 100-mile foodshed radius.

(a) (b)

1860 1880 1900 1920 1940 1960 1980 2000

Year(c) (d)

1860 1880 1900 1920 1940 1960 1980 2000

Year

1860 1880 1900 1920 1940 1960 1980 2000

Year

US

po

pul

atio

n fe

d lo

cally

(%)

US

po

pul

atio

n fe

d lo

cally

(%)

Po

pul

atio

n (m

illio

ns)

Cro

pla

nd a

rea

(mill

ion

hect

ares

)Total population

US population Cropland area and corn yield

Cropland areaCorn yield

Co

rn y

ield

(met

ric

tons

ha–1

)

VegetarianStandardMeat intensive

50-mile foodshed

VegetarianStandardMeat intensive

100-mile foodshed

1860 1880 1900 1920 1940 1960 1980 2000

Year

Local croplands and US food demand A Zumkehr and JE Campbell

246

www.frontiersinecology.org © The Ecological Society of America

al. 2007, 2009), which we summarize here. The diet modelis based on per-capita food consumption for a wide rangeof foods, calculated through the food group recommenda-tions of the US Department of Agriculture (USDA) andthe personal food preferences reported in national foodconsumption survey data. The diet model converts per-capita food demand to per-capita requirement for agricul-tural commodities (crops and livestock) by using conver-sion factors to account for removal of inedible portions,transformations that occur as a result of food processing,and the feed and forage quantities needed for animal prod-ucts (ie efficiency of calorie conversion from forage andfeed to animal products). We used three diets from thediet model: a standard American diet that relies on meatand eggs, rather than pulses (edible seeds of plants in thelegume family, such as peas, beans and lentils), as themajor protein sources; a more meat-intensive diet (381 gof meat consumed per capita per day); and a vegetariandiet. The meat-intensive diet requires roughly three timesas much land for production as does the vegetarian diet.This model is based on conventional livestock practicesand does not include purely grass-fed livestock, which isan important scenario to explore in future work.

The second step was to create maps of the number ofpeople that could be fed based on the cropland present ineach map grid cell. These maps were based on the spatialextent and productivity of cropland in each grid cell, aswell as the per-capita demand for agricultural commoditiesfrom the previous step. The cropland areas were obtainedfrom a recently developed gridded database of historicalUS cropland areas, derived from USDA county-level crop-land area data (Zumkehr and Campbell 2013). The pro-ductivity of the land was determined through the use ofUSDA county-level crop yield data, which reflect the vari-ability in soil quality, climate, irrigation, and other man-agement factors (NASS 2013). As in previous work(Peters et al. 2009), we used corn silage yield data as a spa-tial and temporal proxy for the productivity of annualcommodity crops, and hay yield data as a proxy for peren-nial forage crops. We assumed that the food loss from “farmto table” was 40% of production, but we also considered ascenario with no food losses (ERS 2012).

The third step involved creating a foodshed allocationalgorithm that selected which cropland map grid cellsshould be allocated to which cities in order to maximizethe percentage of the total US population that could be fedwith local food. As a baseline, we set a 50-mile maximumdistance between each city and its allocated cropland, butwe also considered a range of scenarios in which the maxi-mum distance was set to as little as 30 miles and to as muchas 300 miles. In all scenarios, surplus cropland was presentin some foodsheds. This surplus is consistent with the factthat the US is currently a net food exporter (ie domesticfood production is greater than food imports).

We examined the sensitivity of our foodshed potentialresults with respect to the size of the foodshed radius, typeof diet, and uncertainty in the driver data (eg population

sizes, crop yields, and food losses). Additional details onthese methods and the results can be found in WebPanel 1,WebFigures 1–3, and WebTable 1.

n Results

We found that for the years 1850 to 1900, nearly all of theUS population could be fed entirely from food grown infoodsheds with a maximum radius of 50 miles to 100 milesand with diets ranging from vegetarian to meat-intensive(Figure 1, c and d). This is consistent with expectations forthis time period, which predates the long-distance supplychains that currently dominate our food supply systems.However, as population size increased and cropland areastabilized and shifted to the Midwest US region, local foodpotential became overstretched. In the year 2000, thepotential of the 50-mile foodshed to feed its inhabitantsdeclined to 85%, 82%, and 80% for the vegetarian, stan-dard, and meat-intensive diets, respectively. The 100-milefoodshed reflected a similar decline in potential, to 93%,90%, and 88% for the vegetarian, standard, and meat-intensive diets, respectively. Although future projections indemographics and agricultural trends are highly uncertain,any continuation of these trends could greatly underminefood localization efforts. Nevertheless, our results suggest anunexpectedly large foodshed potential at a national scalethat is more than twice the potential reported for the NewYork State case study by Peters et al. (2009).

Most cities can feed 100% of the population with 50-mile foodsheds, with a standard US diet, and in any his-torical time period, but large cities with relatively fewlocal cropland resources have a declining foodshedpotential over the historical time period (Figure 2,WebFigure 2). New York City and Los Angeles, for exam-ple, can support only about 10% of their populations witha 50-mile foodshed radius but as much as 30–50% with a100-mile foodshed radius (WebTable 1). Large cities thatare located near major cropland resources, such as SanFrancisco and Chicago, currently have the potential tofeed 100% of their populations with 50-mile foodsheds.In earlier years, some small cities within otherwise rela-tively rural areas exhibited low foodshed potentials; thismay be because our analysis was based on croplands butnot pasturelands (Figure 2a). Including pasturelands infuture work would likely lead to only modest increases inthe national foodshed potential because the large popula-tion centers that cannot be completely fed with localfood are located far from pastureland areas.

We considered the sensitivity of these results withrespect to a number of factors. Several cities are particu-larly sensitive to changes in the foodshed radius. Seattle,for instance, has a food localization potential of 32% and100% for a 50-mile and 100-mile radius, respectively(WebTable 1). At a national scale, we find that 100% ofthe population can be fed within a 200-mile radius.Increasing the foodshed radius has diminishing returns onfoodshed potential (Figure 3), reflecting the fact that a

A Zumkehr and JE Campbell Local croplands and US food demand

247

© The Ecological Society of America www.frontiersinecology.org

larger radius may extend fartherinto unsuitable areas (eg oceans)and will increase competition withneighboring cities for local food.

We also considered the sensitiv-ity of these results to diets and foodloss rates. Our baseline estimateof an 82% national potentialincludes a 40% food loss fromproduction to consumption. Iffood loss is neglected (ie 0% foodloss), then the national poten-tial increases only by a modestamount, to 87%. Similarly smallsensitivities are observed whenshifting dietary food requirementsfrom a standard diet to a vegetar-ian diet. The national foodshedpotential is not highly sensitive tofood waste and diet assumptionsbecause most US cities have a100% potential in the baselinescenario for the year 2000 (50-milefoodshed radius and a standarddiet). Reducing food losses orswitching to more efficient dietswill not increase food potentialfor cities that already have a 100%potential. Although national poten-tial is not particularly sensitive tothese drivers, individual cities canbe highly sensitive to these drivers(eg San Diego, which can feedonly 30% of its population on ameat-intensive diet but 50% on avegetarian diet).

Similarly, the foodshed poten-tial at a national scale is nothighly sensitive to crop yield.Switching between food loss sim-ulations of 40% loss and 0% losswould be equivalent to simula-tions of a very large yield increaseof 66%, with the same low sensi-tivity (5%). Furthermore, afterconsidering a very wide range ofcrop types for crop yield data, Huet al. (2011) found a foodshed potential of 100% in theUS Midwest. We detected the same foodshed potentialin the Midwest region, but our method relied on cropyield data for only two crop types as indicators of landproductivity.

We also examined the sensitivity of our foodshedpotential results with respect to the population data andthe cropland allocation approach. Previous work forNew York State was based on similar diet assumptionsbut used different population data and a different crop-

land allocation algorithm (Peters et al. 2009). That studypredicted that 34% of the state’s population could be fedby production within the state. Likewise, we found that31% of the population in New York State could be fedfrom within simulated foodsheds (1) with a maximumradius of 50 miles and (2) that may extend outside thestate but also must compete with cities in adjacent states.These comparable results suggest that our analysis isrobust with respect to uncertainty in the population dataand the foodshed allocation method.

Figure 2. Percentage of the population that could theoretically be fed within a 50-milefoodshed by population center, based on a typical US diet in (a) 1900, (b) 1950, and (c)2000. Each panel uses a different classification of graduated circles to represent populationmagnitudes, to show more clearly the relative importance of population centers.

(a) 1900

(b) 1950

(c) 2000

US population fed locally(%)

80–100

60–80

40–60

20–40

0–20

Total population<125 000

125 000–250 000250 000–500 000500 000–1 000 000

1 000 000–2 000 000

>2 000 000

Local croplands and US food demand A Zumkehr and JE Campbell

248

www.frontiersinecology.org © The Ecological Society of America

n Conclusions

Agroecological principles reinforce the foodshed con-cept but the potential scale of the underlying agro-ecosystems is largely unknown. Here, we evaluated thepotential for food localization by repurposing all exist-ing cropland for use in local foodsheds. Our results wererobust with respect to a wide range of input parameters,which could be further explored in future research thatexamines alternative nutritional models and proxycrops. An emerging question is whether this theoreticalpotential is large enough to allow local food systems toscale beyond niche markets and replace a large fractionof conventional food systems. Although we found thatfoodshed potential has declined over time as a result ofdemographic and agronomic trends, our baseline mapssuggest that the current foodshed potential couldnonetheless still satisfy the vast majority of US fooddemand, based on standard US caloric consumption offood groups. The decline in foodshed potentialoccurred in parallel with demographic and agronomictrends such as suburban sprawl, abandonment of agri-cultural land, and population increases, all of whichmay also need to be considered in light of recent effortsin the 2014 Farm Bill to support food localization.Nevertheless, our results indicate that the current food-shed potential of most US cities is not limited by cur-rent agronomic capacity or demographics to any greatextent, and that the critical barriers to this transitionwill be social and economic.

n Acknowledgements

We thank L Kueppers, Q Guo, N Gurwich, and BHalweil for their helpful comments on this study. Thiswork was supported by NSF/CBET #0955141.

n ReferencesAllen P, FitzSimmons M, Goodman M, and Warner K. 2003.

Shifting plates in the agrifood landscape: the tectonics of alter-native agrifood initiatives in California. J Rural Stud 19: 61–75.

Born B and Purcell M. 2006. Avoiding the local trap. J Plan EducRes 26: 195–207.

Cowell SJ and Parkinson S. 2003. Localisation of UK food produc-tion: an analysis using land area and energy as indicators. AgrEcosyst Environ 94: 221–36.

Dawson CJ and Hilton J. 2010. Fertiliser availability in a resource-limited world: production and recycling of nitrogen and phos-phorus. Food Policy 36: S14–S22.

DeLind L. 2011. Are local food and the local food movement tak-ing us where we want to go? Or are we hitching our wagons tothe wrong stars? Agr Hum Values 28: 273–83.

DuPuis EM and Goodman D. 2005. Should we go “home” to eat?:toward a reflexive politics of localism. J Rural Stud 21: 359–71.

ERS (Economic Research Service). 2012. Food availability (percapita) data system. Washington, DC: USDA.

Gussow JD. 1999. Dietary guidelines for sustainability: twelve yearslater. J Nutr Educ 31: 194–200.

Halweil B. 2002. Home grown: the case for local food in a globalmarket. Washington, DC: Worldwatch Institute.

Herrin M and Gussow JD. 1989. Designing a sustainable regionaldiet. J Nutr Educ 21: 270–75.

Hinrichs CC. 2003. The practice and politics of food system local-ization. J Rural Stud 19: 33–45.

Hu G, Wang L, Arendt S, and Boeckenstedt R. 2011. An optimiza-tion approach to assessing the self-sustainability potential offood demand in the midwestern United States. JAFSCD 21:195–207.

Kan I, Ayalon O, and Federman R. 2010. On the efficiency of com-posting organic wastes. Agr Econ 41: 151–63.

Kloppenburg JJ, Hendrickson J, and Stevenson GW. 1996. Cominginto the foodshed. Agr Hum Values 13: 33–42.

Magdoff F. 2007. Ecological agriculture: principles, practices, andconstraints. Renew Agr Food Syst 22: 109–17.

Martinez S, Hand MS, Da Pra M, et al. 2010. Local food systems:concepts, impacts and issues. Washington, DC: USDA ERS.ERR 97 (volume 9).

Morrison KT, Nelson TA, and Ostry AS. 2011. Methods for map-ping local food production capacity from agricultural statistics.Agr Syst 104: 491–99.

NASS (National Agricultural Statistics Service). 2013. Quick stats2.0. Washington, DC: USDA.

Pelletier N, Audsley E, Brodt S, et al. 2011. Energy intensity of agri-culture and food systems. Annu Rev Env Resour 36: 223–46.

Peters CJ, Bills NL, Lembo AJ, et al. 2009. Mapping potential food-sheds in New York State: a spatial model for evaluating thecapacity to localize food production. Renew Agr Food Syst 24:72–84.

Peters CJ, Wilkins JL, and Fick GW. 2007. Testing a complete-dietmodel for estimating the land resource requirements of foodconsumption and agricultural carrying capacity: the New YorkState example. Renew Agr Food Syst 22: 145–53.

Rose N, Serrano E, Hosig K, et al. 2008. The 100-mile diet: a com-munity approach to promote sustainable food systems impactsdietary quality. J Hunger Environ Nutr 3: 270–85.

Thompson EJ, Harper AE, and Kraus S. 2008. Think globally – eatlocally: San Francisco foodshed assessment. Berkeley, CA:Autumn Press.

Timmons D, Wang Q, and Lass D. 2008. Local foods: estimatingcapacity. J Extension 46: 5FEA7.

Weber CL and Matthews HS. 2008. Food-miles and the relativeclimate impacts of food choices in the United States. Envir SciTech Lib 42: 3508–13.

Zumkehr A and Campbell JE. 2013. Historical US cropland areasand the potential for bioenergy production on abandoned crop-lands. Envir Sci Tech Lib 47: 3840–47.

Figure 3. Sensitivity of US local foodshed potential to changesin maximum foodshed radius. Simulations are based on theassumption of a US standard diet for the year 2000. Points areindividual simulations and the line is fit to data.

40 80 120 160 200 240 280 320

Foodshed radius (miles)

US

po

pul

atio

n fe

d lo

cally

(%)

100

90

80

70

60

50

40