Introduction
A conversation has been evolving in North America and Europe about the role local and regional food systems can and should play in creating more sustainable societies. This discussion has ebbed and flowed for many years, and the origins of this debate within the US have been traced to a literature that was developed in the 1970sReference Feenstra1. While, these earlier writings tended to postulate why modern society should obtain more of its food from sources closer to the point of consumption, current debate focuses on two key questions. First, do ‘local foods’ offer real ecological, economic and social benefits? Second, can ‘local food’ move beyond niche markets and supply a significant share of total food demand? No consensus has been reached on either question, but the existing literature provides insight on where further work is needed.
Several appeals have been made to avoid characterizing locally produced food as inherently more sustainable than food distributed through national and international channelsReference Hinrichs2–Reference Schönhart, Penker and Schmid4. These warnings are warranted. Martinez et al.Reference Martinez, Hand, Da Pra, Pollack, Ralston, Smith, Vogel, Clark, Lohr, Low and Newman5 reviewed the literature and found few studies that evaluated the economic, health and environmental benefits of local food systems. They concluded that the issue constitutes an important data gap. Similarly, Edwards-Jones et al.Reference Edwards-Jones, Milá i Canals, Hounsome, Truniger, Koerber, Hounsome, Cross, York, Hospido, Plassman, Harris, Edwards, Day, Tomos, Cowell and Jones6 evaluated the evidence regarding the energy use and greenhouse gas emissions benefits of local food and found existing information to be insufficient to evaluate these impacts for any food system. Even if sufficient data were available, adequate comparisons of local and conventional food supply chains may be misleading because, as Schönhart et al.Reference Schönhart, Penker and Schmid4 point out, existing local food systems may have significant potential for improved efficiencies.
In spite of these data limitations, public interest in and consumption of local foods fuels discussion about the potential for local food to supply a greater share of food needs. Two articles in a recent issue of Choices magazine, a publication of the American Agricultural Economics Association, address this issue. King et al.Reference King, Gómez and DiGiacomo7 outline the challenges of ‘re-localizing’ foods in mainstream supermarkets. While significant hurdles exist, the authors see growth potential for the local food sector, similar to that already seen in the organic sector. Clancy and RuhfReference Clancy and Ruhf8 argue that local food systems will not be sufficient to provide food in large volumes and at reasonable cost to large populations. Rather, they claim that development of regional food systems will be better able to deliver the economic, environmental and social benefits often attributed to local food systems. The extent to which local and regional food systems can expand remains to be seen.
Attempts to measure the capacity for foods to be supplied from ‘local’ sources have been made periodically since the early 1980s for a variety of locationsReference Messing9–Reference Peters, Bills, Lembo, Wilkins and Fick16. These studies vary in methodology, so it is difficult to reach general conclusions about local food in the US. However, multiple studiesReference Messing9, Reference Peters, Wilkins and Fick13, Reference Timmons, Wang and Lass15, Reference Peters, Bills, Lembo, Wilkins and Fick16 have estimated New York State's (NYS's) capacity to supply local food, and all have found that capacity is just a fraction of total food needs (from 22 to 34%). Given its high population density and relatively modest endowment of agricultural land, New York can be seen as a microcosm of the Northeast US. In this part of the country, and perhaps in other heavily urbanized areas, food will continue to be sourced from regions that have lower population densities. This raises the question, ‘If all food cannot or will not be produced locally or regionally, which food should be?’
The answer to this normative question depends on which societal goals local and regional food systems are supposed to serve. This paper will examine the question based on an economic goal, namely, maximizing the value of the land based on its agricultural returns. While economic rents may not be the most important reason to promote local and regional food systems, all agree that the empirical evidence shows that economic returns influence land use decisions. According to the theory of highest and best use, land tends to be used for the purpose that generates the greatest return, given the available marketReference Barlowe17. A classic example of this principle is the conversion of farmland to residential or other urban uses. Similarly, patterns of agricultural land utilization in metropolitan areas demonstrate how the principal of highest and best use influences decisions on production of farm commodities. For example, vegetable production, which has a high-use value per acre relative to other agricultural enterprises, often persists in metropolitan areas after lower-value enterprises have been supplanted by developmentReference Heimlich and Anderson18.
The principle of highest and best use is also embedded in federal and state land use policies. At the federal level, land retirement programs such as the Conservation Reserve Program (CRP) and Wetlands Reserve Program (WRP) and farmland protection programs compensate land owners not to convert land to certain uses, such as cropping in the case of CRP or WRP. In these programs, rental payments (or easement prices) are intended to cover the opportunity cost of giving up the right to use land for agricultural purposes (i.e., cropping or grazing)Reference Wiebe and Tegene19. These payments appear to be effective. Lubowski et al.Reference Lubowski, Bucholtz, Claasen, Roberts, Cooper, Gueorguieva and Johansson20 attribute the reduction in the area of active cropland that occurred between 1982 and 1997 largely to the institution of CRPReference Lubowski, Bucholtz, Claasen, Roberts, Cooper, Gueorguieva and Johansson20. As a result, the United States Department of Agriculture (USDA) employs models based on expected returns to land to estimate the land use changes that would result from elimination of payments through CRPReference Sullivan, Hellerstein, Hansen, Johansson, Koenig, Lubowski, McBride, McGranahan, Roberts, Vogel and Bucholtz21 and WRPReference Heimlich, Wiebe, Claasen, Gadsby and House22. Thus, the value of land can be a compelling lens through which to explore the question of which foods should be grown locally.
The primary goal of this paper is to evaluate the capacity for NYS to meet its food needs while prioritizing which foods should be produced locally and regionally. This issue is explored using a ‘foodshed analysis’Reference Peters, Bills, Wilkins and Fick23 framework. Throughout the paper, the term ‘foodshed’ refers to a geographic area that could potentially supply a population center with food. The approach builds on earlier analysis of potential, local foodsheds in NYSReference Peters, Bills, Lembo, Wilkins and Fick16, and adds to a small but growing body of research using spatial analysis techniques to measure the potential for supplying food needs locally. While our previous work attempted to map theoretical foodsheds by minimizing food miles, this study mapped potential local foodsheds that maximize returns to land.
Methods
A hybrid spatial-analysis-optimization modeling approach was used to evaluate the capacity for NYS population centers to meet food needs within the state's boundaries and thereby map potential local foodsheds. The model characterizes the food production potential of the state's cropland and pasture and the food needs of its population centers within a geographic information system (GIS). The GIS provided input for an optimization model that allocated NYS food production capacity to meet the food needs of NYS population centers while maximizing economic returns to land. The details of each step are outlined in the subsequent sections.
Outline of model design
A broad range of data were processed to provide the input data for the foodshed optimization model (Fig. 1). Spatial data on land cover and soils provided estimates of the location of agricultural land and the relative productivity of the underlying soils for growing key indicator crops. Urban area delineations and spatially referenced population data were used to define the location of population centers and to estimate the number of people that reside in or near them. Data from prior research on the land requirements of the human dietReference Peters, Wilkins and Fick13 were used to derive estimates of per capita food need and average NYS food yields.
Figure 1. Generalized spatial flow diagram for the foodsheds model.
The GIS provided the platform needed to integrate these data and produce input for the optimization model. Estimates of food production potential were calculated for each food producing unit (production zone) using the information on agricultural land cover, soil productivity and average food yields. Estimates of food needs were calculated for the population living in or around major urban centers (consumption zones). Distances between all production zones and population centers were calculated within the GIS. Data on the potential productivity of the agricultural soils found within production zones were integrated with additional data on transport costs and economic returns to land. These expected returns, when capitalized, were used to estimate land use values (LUVs) for various agricultural enterprises. All data matrices were incorporated into an optimization model that allocated available production to meet the needs of population centers such that the aggregate agricultural LUV of all production zones was maximized. Output from the model was used to map and characterize potential foodsheds. All mapping and geospatial operations were performed using ArcGIS 9.1 and Manifold® System 6.5.
Diet and land requirements
Data from previous research were used to calculate the agricultural land area required to meet the annual food needs of the average person in NYSReference Peters, Wilkins and Fick13. Of the 42 diets examined in the earlier study, one diet was selected to represent food need in the foodshed model. The diet provides 2300 kcal per day per person, meets the USDA Food Guide Pyramid (FGP)24 recommendations for all food groups and relies on meat and eggs as the primary protein sources. Diets were based on representative food commodities from each food group, but, with the exception of sugar crops, only included commodities that are produced in NYS. Per capita intake of each food commodity was estimated based collectively on the quantities of meat and fat included in each diet, the food group recommendations of the FGP, the USDA established serving size for the commodity and the relative preferences for that commodity indicated by national food consumption survey data. Total edible food need was estimated and converted into quantities of agricultural commodities (crops and livestock) by using conversion factors to account for the removal of inedible portions and transformations that occur in food processing. The amount of feed crops necessary to supply the food coming from livestock and poultry products was combined with the food crop estimates and crop yield data to determine the average land requirements of the diet. These estimates assume that cattle, swine and poultry are raised under conventional management.
Agricultural classes and LUV
In the model, farm commodities are divided into six classes: grains, vegetables, fruits, dairy, eggs and meat. These classes group individual crop and animal enterprises based on expected gross returns per acre and some general agronomic similarities. The ‘grains’ class includes all field crops used for direct human consumption while the livestock classes include the feed crops required to support the animals. Eggs and meat were placed in separate agricultural classes because egg production generates very different returns to land than do meat production enterprises.
Partial enterprise budgets, showing expected costs and returns for these commodities, were assembled and adjusted to reflect agricultural practices on New York's farms (see Appendix 6 of PetersReference Peters25). The budgets represent conventional crop and livestock production practices. Annual net returns were calculated and averaged for the 8-year period, 1997–2004. Because of some data gaps, certain fruit, grain and vegetable crops included in the calculation of dietary land requirements were omitted from the analysis of LUV, but each agricultural class was represented by least three component enterprises.
The calculation of LUV used an approach similar to the method used by the NYS Office of Real Property Services (ORPS) to administer an agricultural tax assessment programReference Dunne and Lynk26. New York law mandates that the agricultural value of farmland be determined using 8-year average net returns to land from crops grown on New York farms27.
This method was adapted to estimate LUV for each agricultural class (Eqn 1).
The LUV of an agricultural class (i) is equal to the weighted average use value of all representative enterprises in an agricultural class. For each enterprise (j), LUV is equal to the net returns to land (average net income (I) plus the average land charge (L) minus a management charge (M)) divided by the capitalization rate (C). Each enterprise is then weighted (w) based on its relative prevalence in New York agriculture. Crop classes are weighted based on the area of land they occupied during the 8-year period. Livestock classes are weighted based on the number of animals in the state's livestock inventory.
Estimating potential food production
As in earlier workReference Peters, Bills, Lembo, Wilkins and Fick16, the productive potential of NYS soils was determined using two sources of soils data. Spatial data on the distribution of soils were obtained from the State Soil Geographic (STATSGO) database28. Information on expected yields and recommended crop rotations for individual soils were obtained from the Master Soils List (MSL), a resource maintained by the Department of Crop and Soil Sciences at Cornell University and published annually by the NYS Department of Agriculture and Markets29. Data from STATSGO represent soils in a generalized manner appropriate for statewide and regional analyses30. Each STATSGO map unit contains multiple soil types each of which corresponds to a soil-mapping unit from the county-level Soil Survey Geographic (SSURGO) database30. The MSL database provides information for all NYS soils, including all the mapping units in the SSURGO database. Thus, the data were integrated by matching each soil component in STATSGO to its corresponding soil in the MSL. The two databases were joined together and expected yields and recommended rotations for good conservation management were calculated for each STATSGO map unit (see PetersReference Peters25 for more details). Together, these data provided estimates of the spatial variability in expected yield of two indicator crops, corn (Zea mays) silage and hay (multiple species). These forage crops are widely grown throughout NYS to support livestock and milk production. The MSL also provides a corn–hay rotation prescribed for good conservation management. Together, the corn silage yields, hay yields and recommended rotations are used to calculate an index of soil productivity based on the total digestible nutrients (TDN) produced per acre per year in a corn silage and hay rotation.
The location of agricultural land in NYS was determined using the most recent land cover data available at the time the study was conducted, the 1992 National Land Cover Dataset (NLCD)31. The 1992 NLCD data were processed to reduce inherent error as recommended by the US Geological Survey32. This involved using the data at the most general level of land-cover classification (agriculture, barren, forest, urban, water and wetland) and performing some spatial aggregation (see Appendix 3 of PetersReference Peters25 for details).
The combined spatial layers of soil and land cover data displayed potential productivity at a very fine resolution; each pixel represented a 30 m by 30 m land area. In order to avoid the computational issues at this level of resolution, a set of data was created that represented the land area of NYS in 2 km by 2 km grid cells called production zones. Each of these zones was treated as a potential food producing unit. The relative productive potential and land management limitations of each zone were estimated based on a spatial overlay of the production zone boundaries with the layers of combined soils and land cover data. Data from this overlay were used to derive estimates of the area suitable for annual and perennial crops (Eqn 2) and the expected food yields of each agricultural class (Eqn 3).
The potential areas (P) suitable for annual and perennial crops are equal to the area of agricultural land (A) in the production zone (j) times the recommended rotation factor (R) for the type of crop (i). For annual crops, R is equal to the number of years corn can be grown in a 10-year rotation divided by 10. For perennial crops, R is equal to the number of years that hay crops should be grown divided by 10. An average R value was calculated for each production zone.
The expected food yield (Y) of a given agricultural class (i) within each production zone (j) was the product of the TDN index for the production zone and the average yield (Y) of food for the agricultural class in NYS divided by the land area weighted average TDN index (
) of the indicator crop across NYS. The TDN values were calculated from the spatial data layers. Term Y was derived from the earlier study of the land requirements of dietReference Peters, Wilkins and Fick13 and is expressed in Mg ha−1.
Potential fruit production was handled in a different manner. Winter low temperatures and spring frosts pose threats to perennial fruit crops in temperate climates like New York. To account for the fact that the best locations for fruit production are clustered in certain areas of NYS, spatial data on the location of tax parcels coded as fruit farms were used to approximate suitable locations for fruit production. Spatial tax parcel data were obtained from the NYS ORPS (statewide data for 2004 obtained from L. Morin, personal communication, 2006).
Estimating food needs of population centers
The US Census Bureau's Urbanized Areas and Urban Clusters (UAUC) data33 were used to identify the locations of population centers. These data delineate clusters of contiguous, highly developed land throughout the US, and thus represent population centers as discrete geographic entities. One hundred and thirty-two of these entities exist in NYS. To constrain the size of the problem while maintaining the resolution of the production zones, we reduced the number of population centers contained in the model by dividing the state's population into eight zones covering the largest urban areas in the state: Albany, Binghamton, Buffalo, New York City, Poughkeepsie-Newburgh, Rochester, Syracuse and Utica (Fig. 2). These zones were created by assigning all census block groups34 to the nearest major city. The populations contained within each urbanized area and within each consumption zone are shown in Table 135.
Figure 2. Map of consumption zones for foodshed analysis. Dots show the geographic center of each major city. Lines demarcate the boundaries of the consumption zones based on the constituent Census Block Groups.
Table 1. Population of consumption zones.
1 Includes only persons living within each individual urbanized area as defined by the US Census Bureau35.
2 Includes all persons within the zones demarcated in Fig. 2.
3 ‘New York–Newark’ is a Census Bureau designation for the contiguous urbanized are that contains both New York City, NY and Newark, NJ. The population values reported here include only residents from NYS. To avoid confusion about whether or not the New Jersey population has been included in the analysis, this consumption zone is referred to as ‘New York City’ elsewhere in the text.
Adjusting LUVs for soil productivity and transport costs
The production zones differ in terms of their soil productivity and their proximity to population centers. In order to estimate the LUV associated with a given agricultural class for a given production zone, these factors must be included (Eqn 4).
Adjusted land use value (ALUV) was calculated for each unique combination of agricultural class (i), production zone (j) and population center (k). For each production zone, the LUV was multiplied by the ratio of the productivity (TDN) of the soils within the zone and the average productivity of agricultural soils found in NYS. This product was then discounted for transport costs by subtracting the product of the distance (D) between the production zone and population center times the transport cost per km (T) associated with a given agricultural class.
Euclidian distances were calculated from the geographic center of each production zone to the geographic center of each population center. Costs assume that all products are moved by truck. Estimates of truck transport costs were obtained from the USDA Agriculture Marketing Service for refrigerated agricultural products and bulk grains36, 37.
Optimization: allocating the available production potential
The optimization problem solved in the foodshed model maximizes the LUV derived from supplying the food needs of NYS population centers. The optimization problem was specified as a matrix of producers and consumers in which each production zone is a potential source of land for meeting the food needs of any consumption zone. The optimization software solved for the matrix of values that maximized the total ALUV throughout the study area while not exceeding the food needs of consumption zones. This function is meant to mimic the allocation of land to its highest and best use. Total ALUV (Eqn 5) was defined as the sum of the products of the area of agricultural land (A) allocated from each production zone (j) to produce food from each agricultural class (i) for each consumption zone (k) and the associated ALUVs. The value reached by the model was a global maximum for the entire state.
This optimization problem was subject to two categories of constraints. The first set of constraints (Eqn 6a) required that the quantity of food (F) supplied to each consumption zone (k) be less than or equal to the need (C) of the consumption zone for a given type of food (i). These constraints prevent any consumption zone from receiving more food than it needs, while allowing the model to leave the needs of some consumption zones unmet. Note that the quantity of food supplied to a given consumption zone is the sum of the products of the area allocated by each production zone and its corresponding yield. The second set of constraints (Eqn 6b) required that the area of land (A) allocated by each production zone (j) respect the land use limitations implied by the recommended crop rotations from Equation 2. These constraints ensure that the area of land allocated to agricultural classes derived primarily from a given type of land (i) be less than or equal to that zone's potential agricultural area available for such crops (P).
Because certain agricultural classes contained both annual and perennial crops, a heuristic approach was used to represent these constraints. Namely, the estimates of land availability for annual and perennial crop production were converted into limits of land availability for ‘eggs, grains and vegetables’ (all annual crops) and ‘dairy, fruit and meat’ (primarily perennial crops). The area of land available limited to production of ‘dairy, fruit and meat’ was inflated to account for the fact that dairy and meat production are supported by a mix of annual and perennial crops.
Optimization was performed using the Premium Solver Platform™ v6.5 and the Extended Large-Scale LP Solver Engine v6.5 from Frontline Systems38, 39. These programs are upgrades of the basic solver in Microsoft Excel® and are accessed through the solver dialogue box in Excel®40, 41.
Scenarios
Eleven scenarios were examined within this study, each representing a different level of receptivity to meeting dietary needs with locally sourced foods. This ‘willingness’ to eat a local diet refers to the acceptability of substituting a product which can be grown in NYS, such as an apple, for one which cannot be grown in the state's climate, such as a banana. In addition, ‘willingness’ includes consumer openness to eating foods within their windows of seasonal availability, rather than expecting year-round access. This definition of ‘willingness’ differs from the economic concept of ‘willingness to pay’ and is useful for thinking about how much dietary change would be culturally acceptable, irrespective of differences in price.
Each scenario also accounts for the fact that some proportion of current consumption is already supplied by production on New York farms. The actual ‘baseline’ level of local or regional food consumption is not well quantified because information on trade patterns and marketing channels for individual farm commodities is fragmentary and often non-existent. Thus, a baseline was calculated under the admittedly inaccurate, background assumption that current crop and livestock production in New York is not exported but used instead to meet in-state food needs. While this assumption does not reflect the actual flow of food, it does reflect the fact that a non-trivial amount of the state's food is already supplied from NYS sources, irrespective of consumer interest in local and regional food or deliberate policies to promote such interest.
According to this baseline estimate, NYS already produces a sufficient quantity of fluid milk to meet the food needs of its population, but falls short in the other agricultural classes (Table 2)42. Each of the other 10 scenarios represents a 10% increase in the ‘willingness’ of the population to consume a deliberately local diet. Total estimated NYS food needs and the estimated baseline consumption of NYS foods were used to determine the incremental shift in ‘willingness’ to consume local foods for each 10% shift toward a local diet. This could be interpreted as an additional 10% of the population completely following a local diet or the entire population following a local diet an additional 10% of the calendar year. As ‘willingness’ increases, a larger share of statewide food needs may be supplied by ‘in-state’ production, if adequate capacity is available, and if additional production increases the LUV.
Table 2. Baseline ‘willingness’ to consume local food and incremental changes with each scenarioFootnote 1.
1 Scenarios vary based on the ‘willingness’ of consumers to consume a diet composed of entirely local foods. The baseline represents the background level of consumption of foods produced in-state. Other scenarios assume incremental shifts toward local diets. See text for more details.
2 Baseline ‘local’ food consumption was estimated by dividing the estimated quantity of agricultural commodities required to meet in-state food needs by statewide agricultural production. For this calculation, food needs were expressed in quantities of primary agricultural commodities required at the farmgate. Estimates of per capita farmgate level food needs for each agricultural class came from the model described by Peters et al.Reference Peters, Wilkins and Fick13 Agricultural production for each class was estimated from production reported in New York Agricultural Statistics: 2006–07 Annual Bulletin 42. Grain production equals sum of the harvested production of wheat, rye, soybeans and dry beans. Vegetable production equals the sum of the harvested weight of fresh vegetables, processed vegetables and potatoes. Meat production equals the sum of broiler, duck and turkey production plus marketings of cattle, calves, hogs and pigs, and sheep and lambs. Summary statistics for dairy, egg and fruit production were used directly.
3 Total food needs of NYS are equal to the total NYS population reported in Table 1 multiplied by the per capita food needs from Table 3. Values are expressed in weight of primary food commodities.
4 The maximum quantity of food which New Yorkers are ‘willing’ to consume from in-state agricultural production in the baseline scenario.
5 The incremental changes in the quantity of food New Yorkers are ‘willing’ to consume from in-state sources with each 10% shift from a baseline diet to a deliberately local diet.
Results
Food needs, food yields and LUV by agricultural class
Based on the dietary assumptions used in the model, each New Yorker requires 727 kg of primary food commodities annually (Table 3). Plant foods constitute approximately two-thirds of the diet (on a weight basis), while livestock products account for the remaining third. Land requirements for producing each agricultural class vary widely, spanning nearly two orders of magnitude. Vegetables and fruits have high yields of primary food commodity per hectare. Grains, dairy and eggs have intermediate yields. Meat products have the lowest yields.
Table 3. Per capita food needs and average yield of food products used in the foodshed model, summarized by agricultural class.
1 Agricultural classes aggregate broadly similar crop and livestock enterprises.
2 Values for food needs represent the quantity of primary food products required to provide nutritional needs accounting for losses and inedible portions.
3 Values shown for food yields represent the quantity of primary food products that could be produced on ‘average’ NYS agricultural land accounting for crop yields, livestock feed requirements and weight changes that occur in processing.
4 The ‘grains’ category includes all agronomic crops used for direct human consumption: cereals, oilseeds, pulses and sugar crops. Agronomists tend to refer to these plants as ‘field crops’.
Returns to land also differ greatly between the agricultural classes (Table 4). Vegetables were found to be the most lucrative class, with net returns and LUV is almost four times higher than the next most valuable class. Eggs, fruit and dairy classes offered more modest returns relative to vegetables. Grains and meat provided the lowest returns. Indeed, returns for the meat class were negative on land of average quality. All classes are quite sensitive to shifts in yield, with changes in LUV relatively greater than a 1% change in yield.
Table 4. Estimated annual net returns to land and capitalized LUV for each agricultural class. Based on NYS prices, 1997–2004Footnote 1.
1 Complete details on the calculation of net returns and LUV for all component enterprises are reported in PetersReference Peters25.
2 Agricultural classes aggregate broadly similar groups of enterprises. Details on the individual enterprises included within each class and their respective net annual returns and LUV are reported in PetersReference Peters25.
3 Note that the dairy, eggs and meat classes include both livestock enterprises and the crop enterprises required to support the animals.
4 Net annual returns to land represent the residual value after subtracting variable costs, fixed costs and the value of the operator's time and management from gross returns.
5 LUV equals net returns to land capitalized at a 10% discount rate.
6 Sensitivity to changes in yield assumes that gross returns are a direct function of yield whereas only some costs vary across yield levels.
7 Data were not available on variation of fruit yields with soil type. Thus, sensitivity of returns and LUV to yield changes were not calculated.
Allocating production potential to maximize LUV
The area of agricultural land allocated to meet the food needs of the in-state population increased as ‘willingness’ to eat a local food diet increased (Figs. 3a and b). Interesting differences were observed between the land available for enterprises based on ‘annual’ cropping (eggs, grains and vegetables) and the land available for enterprises based on ‘perennial’ cropping (dairy, fruit and meat). Land area devoted to ‘annual’ cropping increased for all classes up to the 30% scenario (Fig. 3a). However, as willingness to consume local food increased beyond 30%, the area of land devoted to grains decreased relative to eggs and vegetables and ultimately declined to zero. In addition, the majority of the 0.46 million ha of land available for grains, vegetables and eggs was used in most scenarios. In contrast, the area devoted to agricultural classes supported by perennial crop production increased gradually as willingness to eat local food increased, but never came close to reaching the total 2.7 million ha of land available for perennial-based agriculture (Fig. 3b). Dairy occupied the largest area of any single agricultural class in every scenario.
Figure 3 . (a, b). Land allocated to meet in-state food needs by agricultural class and model scenario for systems based primarily on annual cropping (a) and those based primarily on perennial cropping (b).
The quantity of food supplied by in-state agricultural land increased as ‘willingness’ to eat local increased (Fig. 4). Values ranged from 4.5 to 9.5 teragrams (one teragram=one million metric tons) of food, as consumed on a primary food commodity basis. Dairy, eggs, fruit and vegetables were supplied in all scenarios. However, grains were supplied only in scenarios where the ‘willingness’ to eat local was 80% or less. No meat was supplied in any scenario.
Figure 4. Food supplied by in-state production by agricultural class and model scenario.
Similar to food production, the maximum total LUV attained in each model run increased as the allowable in-state consumption increased (Fig. 5). The total LUV ranged from 5.0 to 8.9 billion dollars. Dairy accounted for the vast majority of the LUV in the baseline scenario, but its share of LUV decreased relative to eggs, fruit and vegetables as the scenarios allowed for a greater share of consumption to be ‘local’. Grains, where present, constituted a very minor share of total LUV.
Figure 5. LUV attained from allocating agricultural to meet in-state food needs, by agricultural class and scenario.
Mapping foodsheds by agricultural class
Potential local foodsheds were mapped by agricultural class (Figs. 6a–d). Given the large number of scenarios, results are reported only for the 100% local-food scenario. This scenario was selected because it can be compared to earlier work that also mapped foodsheds for 100% local dietsReference Peters, Bills, Lembo, Wilkins and Fick16. While the spatial extent of the area devoted to each agricultural class is different, the spatial dispersion of each foodshed is similar across agricultural classes. Consumption zones near the map edge (Buffalo, Rochester and Syracuse) generally received food from nearby land. However, zones closer to New York City (Albany, Binghamton, Poughkeepsie-Newburgh and Utica) display a ‘ray-like’ pattern. The foodshed for New York City extends throughout the state.
Figure 6. (a–d) Production zones allocated to meeting in-state food needs for the 100% local scenario, by consumption zone and agricultural class: (a) vegetables, (b) eggs, (c) dairy and (d) fruit.
Potential food production was distributed almost directly in proportion to population (Table 5). New York City received a slightly lower share of total food needs than the seven consumption zones located in ‘upstate’ New York. In contrast, the ‘food distance traveled’ for the New York City foodshed was much greater than those of the other consumption zones. Overall, 69% of total statewide food needs were met in the 100% local scenario, while the average food distance traveled was 238 km.
Table 5. Food needs satisfied and food distance traveled in 100% local scenario.
1 Food distance equals the quantity of food supplied from a production zone to a consumption zone multiplied by the distance between those zones. Total food distance is the sum of the food distances for all production zones serving a consumption zone (or group of zones).
2 Equals total food distance divided by food supplied.
Discussion
The primary goal of this study was to evaluate and map the potential to supply food locally by food groups. Since agriculture tends toward specialization, the model requires a method for deciding which foods should be produced locally. LUV was chosen because it has been proven empirically to influence land use decisions. Although the findings are foremost a proof of concept, they illustrate some key principles worthy of consideration in the broader discussion of local and regional food systems. In addition, the results have implications for the development of local and regional food systems in NYS.
Key principles
One principle illustrated by the findings is that of highest and best use. This was most clearly demonstrated by the allocation of agricultural land. The reduction in land devoted to ‘grain’ at scenarios greater than 30% willingness to eat local food was the result of displacement by higher value eggs and vegetables. Indeed, land was only dedicated to the lower value grains class if the ‘demands’ for all higher value classes were already satisfied and suitable land was still available. Thus, while the potential markets for locally or regionally sourced foods is not known with any certainty, the results are a reminder that land tends to be devoted to the most valuable use possible.
A second principle illustrated by the model is that of marginal land. In all scenarios, large areas of land were not used by the model. Given the objective function, this result means that this land could not produce an economically viable (i.e., positive) return without reducing the total LUV. This land may have had uses for which returns were positive. Indeed, vegetable production was estimated to produce positive returns on many types of land. However, the better (more productive) land was used first, leaving marginal land behind. This is consistent with the land use history of NYS, in which large areas of land were abandoned from farming over the course of the 20th centuryReference Bills, Stanton, Hirschl and Heaton43. However, the model makes this determination based on a single factor (LUV), whereas farmland use decisions in reality are multi-factorial. They incorporate, for example, attitudes toward risk, availability of agricultural support services and non-economic land ownership goals. Nonetheless, the model illustrates an important point: not all technically available land will be used for food production.
Implications for NYS
Beyond these general principles, the findings raise several issues regarding the capacity for increased reliance on local and regional food in NYS.
First, local and regional food systems will specialize in production of certain foods. The model illustrates that in NYS, competition for the highest and best agricultural use of land would favor production of dairy, eggs, fruit and vegetables relative to grains and meat. This is consistent with what one would expect based on the LUV estimates, and in a general sense, it reflects actual land use patterns in New York. Fruits and vegetables combined constituted 56% of the 1.75 billion dollars of crop production generated in 2007, and dairy constituted 86% of the 2.76 billion dollars of livestock and poultry production for the same year (authors’ calculations from New York Agricultural Statistics, 2008-2009 Annual Bulletin)44. Thus, increasing local food generally means growing more of the things the state already produces.
Second, specialization could enable local and regional food systems to supply a large share of the state's food needs. The model shows that up to two-thirds of statewide food needs, on a fresh weight basis, could be supplied by agricultural land in the state. In contrast, an earlier study of NYS that examined the capacity to provide complete diets found that the state could supply approximately one-third of its food needsReference Peters, Bills, Lembo, Wilkins and Fick16. This finding results directly from the fact that the LUV-based model favors high-value enterprises which are, coincidentally, also water rich. Therefore, a larger share of the mass of food could be supplied than when foods were produced in proportion to nutritional needs. This suggests that food miles might be reduced strategically by focusing on the water-rich foods with relatively high yields per acre, such as most fruits and vegetables.
Third, the model suggests that local and regional food systems would focus on supplying certain foods rather than certain cities. Our earlier work minimized food miles, and the model allocated food preferentially to cities in upstate New York at the expense of New York City. In contrast, the productive capacity of New York's agricultural land was more equitably distributed in the differentiated food group model. While the food miles based model can provide an estimate of the theoretical minimum distance food would need to travel, the LUV-based model is probably more realistic.
Directions for further research
One modification that would enhance the value of this model is to allow land to be allocated to food ‘exports.’ In such a scenario, the NYS population would ‘compete’ with the larger US population for foods produced on in-state agricultural land. For example, the use of land for production of fluid milk to supply the New York City foodshed might compete with demand from New Jersey. Such an approach would help to shed light on the question of where local and regional food systems fit in a system where national and international distribution of food predominates.
Conclusions
Models represent a simplified version of reality, but they can be valuable for gaining insight into how systems function. The LUV-based foodshed model shows that an optimal local and regional food system in NYS would specialize on high-value crops and livestock. Since these products can be produced in a relatively large volume on a relatively small acreage, a focus on fruits, vegetables, dairy and eggs could enable the state to supply a large share of total food needs (on a fresh-weight basis). These findings suggest local and regional food systems have great potential for expansion and supplying a larger share of nutritional needs. However, it remains unclear as to how much of this potential will be realized.
The principles of highest and best use and marginality of land are long lived in the literature, and LUV serves as a reasonable optimizing parameter for understanding allocation of land. However, a much broader spectrum of societal concerns motivates public interest in local and regional food systems. Future work should examine how well a strategy of focusing on high-value crops and livestock serves other important goals for the food system, such as supporting small- and medium-sized farms, reducing greenhouse gas emissions and improving access to foods lacking in the American diet. Such efforts will help to create an informed and comprehensive vision of the ideal local food system.
Acknowledgements
This research was supported in part by the National Research Initiative of the USDA Cooperative State Research, Education, and Extension Service, grant number 2005-55618-15640 and by the W.K. Kellogg Foundation, grant number P3008987.
