Declining potential of local food supplies to satisfy urban demand during 1990‒2015 and the resulting implications for urban food security

ABSTRACT This study developed the two new indexes, the distance to the nearest cropland (DTNC) of urban residents and the foodshed radius of urban residents (FRUR), to evaluate the potential of China's local food system to sustainably meet urban food needs. The results found that the per-capita DTNC increased by 53% over 1990−2015, with a total increase of 8.2 × 106 km on a national scale; the FRUR estimated on a “business-as-usual” basis was estimated to be 10.1 km. This study further predicted that stress on local food supplies will continue increasing until the 2040s, particularly in mega-urban areas. Appropriate spatial planning regarding the distribution of cropland and urban areas is necessary for mega-urban regions to cope with crisis situations when the external food supply is restricted. Such planning will ultimately ensure the long-term sustainability of urban areas.


Introduction
'To achieve food security for all' is a phrase highlighted in the United Nations Sustainable Development Goals (United Nations, n.d.b), although a recent report by the Food and Agriculture Organization of the United Nations implies that the world is not on track to achieve 'Zero Hunger' by 2030 (Din et al., 2022).The vulnerability of food transport networks exacerbates food access inequality for urban residents (Langemeyer et al., 2021).For example, at the onset of the coronavirus disease 2019 (COVID-19) pandemic, urban residents rapidly depleted grocery stores in many cities worldwide; this trend was accompanied by soaring food prices (Bande, 2022;Laborde et al., 2020).Therefore, the potential for local food supplies to feed the urban population is becoming increasingly important.Foodshed potential is defined as the fraction of total dietary demand that could be met if all existing food production were repurposed for local consumption (Zumkehr & Campbell, 2015).It represents the maximum potential of a region to sustain food supplies when external inputs are restricted.This potential is a function of the spatial distribution of the human population, cropland area and productivity, and dietary requirements.Although food product diversity in supermarkets has never been greater, there remains a need to examine whether cities maintain sufficient capacity to meet their local food supply potential.This capacity is particularly important when coping with short-term crises (e.g.floods, sanitary crises, and war) (Bande, n.d.) and longterm stresses (e.g.climate change and environmental degradation) (Johannes et al., 2021;Seto & Ramankutty, 2016;Seto et al., 2012).
Cities depend on agricultural surpluses to survive (G.Chen et al., 2020;Mcneill & Engelke, n.d.;Thebo et al., 2014).Many cities originally developed alongside fertile croplands.This 'symbiosis' between cities and cropland guarantees a generally stable food supply for the urban population, strengthens resilience when external food inputs are restricted, and provides job opportunities for 2-2.5 billion smallholder farmers in peri-urban areas (Bren D'Amour et al., 2017;Masters et al., 2013).However, the relative spatial relationship between urban food consumption and food provision areas has been evolving with urbanization since the Great Acceleration (Mcneill & Engelke, n.d.Periurban areas are now prioritized for high-market-value land uses, rather than food production (Johannes et al., 2021).A widely expressed concern regarding the urbanization process is that the expansion of cities can result in continuous urban encroachment onto cropland.This encroachment has occurred in many countries, including the United States (Zumkehr & Campbell, 2015), China, Ghana, and India (Liu et al., 2019).
It is projected that five billion people will live in cities worldwide by 2030, constituting 60% of the total population (United Nations, n.d.a; United Nations, n.d.b).China, the most populous country in the world, has experienced rapid urbanization during the past three decades (Zhou et al., 2022;Zünd & Bettencourt, 2019).This urbanization process may be the greatest resettlement experiment in human history (Bai et al., 2014).Since 1990, the proportion of the total population living in urban areas has grown at an annual rate of 1% in China (United Nations, 2022c).If this trend continues, the urban population could increase by 330-470 million by 2050 (Wang et al., 2021).However, urbanization trajectories may vary among sub-regions.In the northeastern region of China, cities have decreased in size and population outflows have occurred because of economic stagnation (Long & Wu, 2016).Competition for land use between crop production and urban expansion has been predicted to lead to a 5.3% loss of cropland and an 8.7% decrease in crop production in China by 2030 (Bren D'Amour et al., 2017;Wang et al., 2021).These changes present challenges to the local food supply potential and could weaken resilience in terms of coping with sudden crises (Elmqvist et al., 2019).Gradual increases in the external food supply for urban residents could also lead to negative social and ecological externalities, such as increased transit-related fuel consumption (Cowell & Parkinson, 2003), increased fertilizer application to degraded soils (Barthel et al., 2019;Bren D'Amour et al., 2017, and unemployment among smallholder farmers (Barthel et al., 2019;Johannes et al., 2021;Seto & Ramankutty, 2016;Seto et al., 2012).
Among the many indicators of urbanization, population size is most important.Through the urbanization process, the urban population has substantially increased; the demand for food to sustain urban residents has also rapidly increased.In this study, we considered food demand among urban residents to be a quantifiable attribute of urbanization; thus, we calculated the theoretical change in the ability of local food to feed the urban population between 1990 and 2015 in 271 cities.Such change could reflect the evolution of a city's ability to respond to local food demand during an emergency when external food supplies are restricted.Therefore, we introduced the indexes of the distance to the nearest cropland (DTNC) and the foodshed radius of urban residents (FRUR) to analyze 'hidden links' between urbanization and local food supply potential, including where and how to maintain local food supply potential.These respective indexes reflect a city's ability to cope with conditions when external food inputs are blocked, along with the distance over which local food supplies can theoretically meet the needs of urban dwellers.Data for 271 prefecture-level cities over the period of 1990 − 2015 in China were used to identify food supply dynamics.Shifts in the DTNC for urban residents and the FRUR to meet urban food demand in four geographic regions of China (i.e.eastern, mid, western, and northeastern) were explored.This study attempted to fill the current knowledge gap by addressing the following question: how does urban population growth and cropland loss affect local food supply potential?The answer to this question is particularly important in emerging countries where approximately 90% of global urban population growth is predicted to happen in the next 30 years (United Nations, n.d.c).

Materials and methods
In this study, we used two indexes to explore the spatial relationship between urban residents and food supply sources.The first index was the Euclidean distance from urban residents to the nearest cropland surrounding cities (Lu et al., 2011).This distance directly reflected the residents' ability to access crops.Based on the spatial distribution of urban residents and cropland, as well as the gridded population density, the average DTNC for each target city was estimated using the ArcGIS platform for different years.The second index was the local food supply potential, which quantifies the theoretical upper potential of all existing cropland to meet urban food demand through local food networks based on a range of FRUR values (i.e. 10,20,and 30 km).As in previous studies (Zumkehr & Campbell, 2015), this theoretical potential assumes that all existing cropland is allocated to the fulfillment of calorie demand based on the current standard diet for urban residents in China.The assumed diet was designed to provide the equivalent calories in all food groups (grains, vegetable, fruit, and meat).Thus, the estimates presented here provide a theoretical baseline for examining local food systems across broad spatial and temporal scales.A flow chart describing the calculation processes and data sources is provided in Figure 1.

Estimation of Euclidean distance between urban residents and nearest cropland
We first characterized the changes in spatial distance between food consumption and production in China's 271 prefecture-level cities during 1990 − 2015.We calculated the DTNC at 5-year intervals (Lu et al., 2011).A higher DTNC was associated with greater cropland loss in periurban regions and greater difficulty in accessing local foods (Coulibaly & Li, 2020).To estimate the DTNC, we converted the coverage shapefile of the gridded urban population distribution and land use type data into a raster dataset.The Euclidean distance between each urban population grid and the nearest cropland (from a land use dataset) was calculated for China's 271 prefecture-level cities at a spatial resolution of 1 × 1 km in 1990, 1995, 2000, 2005, 2010, and 2015.China was divided into four geographic regions (i.e.eastern, mid, western, and northeastern) based on location and economic development trajectory (Figure 2) (China Bureau of Statistics, n. d.).Because the gridded population dataset did not distinguish between urban or rural areas, we established a density threshold of 1,000 capita per km 2 for urban residents (Zhu et al., 2019).The estimated results were validated by national city-level census data for 2005, 2010, and 2015.The correlation coefficients between gridded population density estimations and census data were>0.7 (Table S1).The estimated distance between urban residents and their nearest cropland was determined via the ArcGIS platform (ArcMap 10.2) using the Near-DIST function.The average DTNC for each city was estimated based on the Euclidean distance of each urban population grid as follows: where DTNC Ave is the average distance to the nearest cropland from urban residents in the city (km); D Euc is the Euclidean distance between each population grid and its nearest cropland (km); P Grid is the urban population for each grid (capita); and P City is the total urban population (capita).Based on the gridded urban population (1 km × 1 km) and land use datasets (1 × 1 km) for different years, the average DTNC for 271 cities across China was estimated at 5-year intervals for the period of 1990 − 2015.

Estimation of changes in local food supply potential in cities
We characterized the percentage change in urban population that could be theoretically fed within a 10-, 20-, and 30-km FRUR for each population center based on the urban dietary demand of 271 cities during 1990 − 2015 (Zumkehr & Campbell, 2015).As in previous studies (Kinnunen et al., 2020), we calculated the theoretical upper potential for all existing food production to meet the demands of urban residents through local food networks based on the urban population distribution,  agricultural region, crop productivity, and their dietary consumption (Kinnunen et al., 2020;Pradhan et al., 2014).Food production data (i.e.grains, meat, vegetables, and eggs) were derived from the annual China Statistical Yearbook for 1990 − 2015 (Figure S1 and Table S2).To assess changes in food production, these data were updated at 5-year intervals.To determine this theoretical potential, we assumed that all existing food production was allocated to produce the food required by urban residents.Food consumption data for urban residents in 1990-2015 were updated at 5-year intervals (Table S3).The food groups selected in the study were: grains (such as rice and wheat), vegetables, aquatic products, livestock and poultry meat, and eggs.The areas measured were paddy fields, dry lands (from land use and land cover change data), canals, lakes, reservoirs and ponds, livestock and poultry distribution areas (from global livestock and poultry spatial distribution data), and egg production areas (corresponding to the poultry distribution area).We reviewed many articles containing information about food waste in China; we used data from those articles to determine proportions of food losses and waste, which were 20% for crops and vegetables, 18% for meat, and 10% for fish.We first calculated the total local food production, then subtracted the waste to obtain a food production value for the consumed portion, which was used to calculate the local food supply potential.Both food supply and urban demand were converted into equivalent calories.The estimated quantity of calories in each food type is provided in Table S4.The estimation did not consider social (e.g.food preference) and economic factors (e.g.food price); therefore, it only provides a theoretical baseline to examine local food systems across spatial and temporal scales (Zumkehr & Campbell, 2015).A raster dataset based on livestock distributions and cropland distributions was then developed for the period of 1990 − 2015 (Tong et al., 2020).This resulted in a 1 × 1-km gridded map incorporating the calorie supply potential of major food supplies for 1990 − 2015 (Figure S1).Based on the gridded urban population and dietary structure, we calculated the shifting calorie demands of 271 prefecture-level cities for 1990-2015.The fraction of the urban population that could be theoretically fed within the specific FRUR was estimated as follows: where P Local is the fraction of the urban population that can theoretically be fed with the specific FRUR (i.e. 10, 20, and 30 km).If P Local >1, the theoretical calorie supply within the geographical radius is adequate for the urban demand.Supply Cal is the sum of calories provided by food production within the FRUR (kcal); Demand Cal is the total calorie demand from urban residents (kcal).P Local for 271 cities was estimated at 5-year intervals for the period of 1990-2015, using each of the three FRUR values, to indicate the impact of rapid urbanization.
The FRUR that was essential to meet calorie demand in urban areas was estimated under the 'business-as-usual' (FRUR BAU ) and 'food-transport-block' (FRUR FTB ) scenarios.Relative differences between the two scenarios were used to determine the food supply potential for individual cities.The FRUR BAU scenario refers to a situation in which normal trade occurs between provinces.The FRUR FTB scenario refers to a situation in which no agricultural trade occurs with other provinces, and only local food production supplies the needs of urban dwellers.The FRUR is defined as the geographic area required for local food supplies to theoretically satisfy the demand of urban residents (Morrison et al., 2011;Peters et al., 2009;Zumkehr & Campbell, 2015).In the FRUR BAU scenario, a specific percentage of urban calorie demand could be supplied by external food inputs; this percentage was estimated using city-level multi-regional input-output from 2015 (Zheng et al., 2021).External food input in a city refers to input from other cities.In the FRUR FTB scenario, all urban calorie demand was assumed to be provided through local food.Relative changes between the two scenarios were used to evaluate the vulnerability of cities to food-transport-blocks.The FRUR under the FRUR FTB scenario was estimated for 1990 and 2015, respectively.Estimations were performed based on the spatial extent of food production and local food supply in each grid cell, as well as the calorie demand of urban residents in each city.We used the location of the local government as the center of population, then estimated the FRUR required to satisfy the urban demand.Estimations were performed via buffer analysis on the ArcGIS platform using the Python scripting language.The buffer radius was initially assumed to be 0.1 km; it increased by 0.1 km for each iterative computation if the food calorie supply within the radius was smaller than the urban demand.Conversely, the calculation ceased if the supply was larger than the demand.The corresponding radius of the buffer was assumed to represent the FRUR for urban residents in the city.Based on regression analysis involving the FRUR and urban population size, along with future projections of urban population growth (Y.Chen et al., 2020), we simulated changes in the FRUR in mega-urban regions in China for the period of 2020-2080.In the population projection dataset, the population was simulated for five shared socioeconomic pathways (SSPs) to cover the range of development pathways (Zeng et al., 2022).The population dataset was displayed as a 1 × 1-km grid for the period of 2020-2080.To ensure consistency with the historical estimation in 1990-2015, a threshold of 1,000 capita per km 2 was used for urban residents.Based on population growth rates under the SSP1-SSP5 scenarios, we predicted future changes in the FRUR via multiple linear regression.First, the multiple linear regression equations of population and FRUR in three regions (Beijing-Tianjin-Hebei region, Yangtze River Delta, and Pearl River Delta) were obtained using SPSS software, in which the independent variable was the region and the dependent variable was the FRUR of each city.Second, the FRUR responses of three mega-urban regions in China under the five shared socioeconomic pathways for 2020 − 2080 were predicted according to the population prediction data collected in the population dataset.In this study, statistical analyses and data processing were performed using Excel 2010 and ArcMap 10.2.2 software.

Urban land expansion and cropland loss in 1990-2015
Based on the gridded land use and population distribution raster datasets, we first determined changes in urban land during 1990-2015.To host the growing population caused by migration (Table S5), the area of urban settlements in China rapidly increased from 73,979 to 150,322 km 2 between 1990 and 2015, representing a net expansion of 103% and an average area expansion of 3,347 km 2 per year (Figure 2 and Table S5).On a national scale, the expansion of urban regions by 2015 primarily occurred in regions that were under cultivation in 1990.In total, 47492 km 2 of cropland, corresponding to two-thirds of the total urban expansion, was lost because of urbanization in 1990-2015.The area of cropland loss area was similar to the area estimated by Liu et al. in 2020 (49,077 km 2 ) (Liu et al., 2020).However, only 28,851 km 2 of urban land expansion occurred on noncrop land, primarily in the southern region of China.Most (68.6%) urban land expansion occurred in eastern China, covering an area of 52,344 km 2 (Figure 2 and Figure S2).Approximately 47.4% of national urban land expansion occurred in the three mega-urban areas: North China Plain (7,564 km 2 ), lower Yangtze River Delta (18,531 km 2 ), and Pearl River Delta (10,116 km 2 ) (Figure 2).In contrast, the least urban land expansion (<1% of the urban area in 1990), and even some shrinkage, occurred in northeastern China (Figure 2).

Increasing distance of urban residents to nearest cropland
Urban land expansion and land use changes could influence the spatial distances between urban residents and areas of agricultural food production (Bren et al., 2017;G. Chen et al., 2020), which would affect potential local food supplies (Barthel et al., 2019).Based on the spatial distribution of urban residents and cropland, we explored changes in the DTNC for 271 prefecture-level cities in four geographic areas of China at 5-year intervals during 1990 − 2015.A high DTNC indicated difficulty for urban residents to access local foods.On a spatial scale, the largest DTNC in 2015 was observed in eastern China (5.3 ± 4.2 km, n = 85); this was significantly higher than the DTNCs in other areas (mid China: 2.9 ± 1.5 km, n = 78; western China: 3.0 ± 1.8 km, n = 75; northeastern China: 3.2 ± 2.1 km, n = 33, p < 0.01, Student's t-test, Figure 3a and Figure S3).From 1990 to 2015, we found that the DTNC substantially increased in most cities (188 of 227 cities), which indicated increasing difficulty in accessing local foods for urban residents; notably, these trends substantially varied among regions (Figure 3b).The national average DTNC increased by~54% from 2.4 ± 2.0 km (n = 230) in 1990 to 3.7 ± 3.0 km (n = 271) in 2015 (p < 0.01, Student's t-test).The largest increase in DTNC occurred in eastern China from only 2.5 ± 1.4 km (n = 80) in 1990 to 5.3 ± 4.2 km (n = 85) in 2015; this was much larger than the increases in mid China (from 2.1 ± 1.0 km in 1990 to 2.9 ± 1.5 km in 2015) and western China (from 2.6 ± 3.5 km in 1990 to 3.0 ± 1.8 km in 2015, p < 0.01).Northeastern China exhibited a distinct DTNC trend over the study period, compared with the other three areas (Figure 3c), which reflected its unique development pathway (China Bureau of Statistics, n.d.).Since 2000, northeastern China has been characterized by population outflows and a stagnant economy.The average DTNC increased from 2.7 ± 1.5 km in 1990 to 3.4 ± 2.2 km in 2005, then gradually decreased (Figure 3a).Few changes in the DTNC were observed throughout the period (p > 0.1), whereas continuous increases in the DTNC were observed in other regions (Figure 3a).

Declining local food supply potential for urban residents
Although the potential impacts of urbanization on cropland losses were identified (Bren et al., 2017;G. Chen et al., 2020), there was little information regarding the national-scale capacity of the local food supply potential, which is more relevant to national policymaking and implementations (Elmqvist et al., 2019).The food supply potential is a function of the spatial distributions of urban population and cropland, as well as cropland productivity and dietary requirements (Zumkehr & Campbell, 2015).We found that even with a 47% increase in crop production at the national scale (Figure S4) for 1990-2015, the local food supply potential in cities declined regardless of the FRUR (i.e. 10, 20, and 30 km) (Figure 4 and Figure S4).In 1990, half of the cities in eastern China could provide sufficient food to meet the urban demand within a 10-km FRUR; in 2015, this proportion had decreased to only 25% (Table 1).Within the study period, there was a decrease in the percentage of urban residents that could be fed within a 10-km FRUR in 177 cities (77%), such that only 53 cities (23%) had an increasing trend.The cities with an increasing food supply potential were generally located in regions surrounded by abundant cropland (Figure 4).Most cities with a large decrease in food supply potential were located in the Pearl River Delta and lower Yangtze River Delta, or they were coastal cities.Relative to eastern China, similar decreases in local food supply potential were also observed in the other three areas, although the magnitudes of those decreases were much smaller (Table 1).

The ability to cope with a food-transport-block has decreased in many Chinese cities
Based on gridded local food supplies, urban population sizes, and dietary structures, we estimated the FRUR required to meet the calorie demands for urban residents (Zumkehr & Campbell, 2015).We established two scenarios to evaluate the vulnerability of cities in response to a food-transport-block.
Based on city-level multi-regional inputs-outputs in 2015 (Zheng et al., 2021), we approximated the contributions of outside food input in China's cities as shown in Figure S5.We calculated the FRUR values under the FRUR BAU and FRUR FTB scenarios (Figure 5).Relative differences between the two scenarios were used to evaluate the vulnerability of individual cities to food-transport-blocks.With external food inputs, the national average FRUR BAU was estimated to be 10.1 (1.3-30.7,95% Confidence interval) km; under a food-transport-block, the FRUR FTB would increase to 16.1 (1.9-52.3,95% CI) km, which constituted a 12.4-139.8%increase relative to the FRUR BAU scenario (Figure 5 and Figure S6).Under the FRUR FTB scenario, the maximum FRUR FTB was observed in eastern China, with a value of 23.0 ± 17.7 km (n = 85); the lowest FRUR FTB was observed in northeastern China, with a value of 14.4 ± 10.9 km (n = 33, P < 0.05).This indicates that there were fewer stresses in terms of satisfying urban calorie demand through the local food supply in northeastern China than in eastern China (Figure 5 and Figure S6).
We found that from 1990 to 2015, the capacity to cope with food-transport-blocks decreased in many regions (Figure S7).From 1990 to 2015, 176 of 227 cities across China experienced substantial increases in the FRUR FTB [range of 0.3-33.5 km (95% CI)], which indicated a decreasing ability to cope with food-transport-blocks.In contrast, only 51 cities displayed decreases in the FRUR FTB .Because of population growth and urban land expansion in the peri-urban region, the potential for local food provision was lost in many cities (G.Chen et al., 2020).At the national scale, the FRUR FTB substantially increased by 3.4 km during 1990-2015.However, in mega-cities such as Beijing, Shanghai, and Guangzhou (Figure S7), the FRUR FTB values increased by 42.9, 37.1, and 31.1 km, respectively, for 1990-2015; these findings indicated unprecedented pressure to satisfy urban demand when food transport was disrupted.
Approximately five billion people are projected to live in cities by 2030, and 70% of the global population is projected to live in urban settlements by 2050.Despite improvements in agricultural techniques, cropland loss cannot be offset by increasing crop productivity (Bren et al., 2017).It would also be difficult to substantially improve crop productivity in the immediate future (G.Chen et al., 2020;Liu et al., 2019).These factors create challenging conditions for the provision of adequate and sustainable food supplies to urban populations.Because high-income ).Under non-crisis circumstances, the decrease in local food supply potential could be offset by global or regional food trades (Clapp, 2017;Laborde et al., 2020).However, these actions are unlikely to be equally distributed across regions; they will also intensify negative environmental and social externalities, such as soil degradation, greenhouse gas emissions, and water eutrophication (Johannes et al., 2021).
Our results suggest that the rapid urbanization since 1990 in China has been accompanied by a decline in potential local food supply for urban residents, which has weakened urban resilience against food-transport-blocks (Table 1).This trend exhibited regional variance with the most rapid decline occurring in eastern China; the changes were less severe in northeastern China than in other regions (Figures 3 and 4).In contrast to the rapid economic developments in eastern and mid China during 1990-2015, the economy of northeastern China has stagnated since 2000 (China Bureau of Statistics, n.d.).Massive population outflow has occurred toward more developed regions, primarily to eastern and southern China, leading to a decline in urban population from 29.5 million in 2010 to 26.9 million in 2015 (China Bureau of Statistics, n.d.. Urban shrinkage would result in a decline in the DTNC and an increase in the food supply potential relative to developing regions (Table 1).However, this unique trend in northeastern China did not represent the overall national conditions (as shown in Figure 3).

Stresses on local food supply potential will continue to increase in mega-urban regions
Almost all socioeconomic projections have predicted that China's urbanization will continue in most regions in the next two decades (United Nations, n.d.c), although the rates of urbanization may vary across regions or cities (Gao & O'Neill, 2020).Based on gridded (1 × 1 km) population projections under the five shared socioeconomic pathways (i.e.SSP1-SSP5, Figure S8) (Y.Chen et al., 2020) and regression analyses of the urban population and FRUR FTB in 2015 (Figure S9), we estimated the future FRUR FTB trends for 2020-2080 in the three mega-urban regions of China: North China Plain, lower Yangtze River, and Pearl River Basin.We predicted that in the near future, stresses on local food supply potential would continue to increase in mega-urban regions for all socioeconomic pathways (Figure 6).Without productivity improvements, their FRUR FTB will increase in the 2040s to 22.7-23.6,20.3-21.0, and 43.8-44.9 km, respectively (i.e. increases of 9.4-13.9%, 2.7-6.1%, and 4.7-7.1% relative to 2015).These findings indicate that stresses on the local food supply will continue increasing in mega-urban regions; therefore, immediate actions are required by policymakers and urban planners in advance of potential future food-transport-blocks.Conversely, China and many other countries are expected to encounter substantial pressure from urban population declines after 2050 (G.Chen et al., 2020).If these declines in the urban population occur, the FRUR FTB would finally decrease (Figure 6).

Countermeasures to make food systems safer
The COVID-19 pandemic exposed vulnerabilities in food transport networks, both in emerging countries with undeveloped transport systems and in large countries with different food production and consumption patterns (FAO, n.d.).Under non-crisis circumstances, the local food supply potential is always neglected because of the abundance of outside food inputs; however, it becomes important during sudden crises, which have occurred with increasing frequency since the 2000s (Bongaarts, 2021;Laborde et al., 2020;Peter & Rosset, 2008).Thus, the issue of urban food security has arisen, and there is an ongoing need to sustain a desired level of food sovereignty in cities.This issue is not new and has periodically become the focus of global research and policymaking in crisis periods (Johannes et al., 2021); nevertheless, productive cropland surrounding cities continues to be replaced by high-market-value land uses (e.g.built-up and recreational areas) (Clapp, 2017;Satterthwaite et al., 2010;Seto et al., 2012).Despite ongoing discussions regarding the potential benefits and role of local agricultural production in strengthening urban resilience against crises (Costello et al., 2021;Kurtz et al., 2020), the vulnerability of urban food security is consistently neglected in land use policymaking (UN environment programme, 2022).
To meet the twin urban development goals of feeding the growing urban population and guaranteeing a consistent and adequate local food supply, it is essential to guide and shape future urban land expansion to a more sustainable form.The governance and proper planning of urban land development locations and procedures are critical for efforts to ensure that agri-food systems are more resilient to stresses.In China, specific strategies to constrain the sizes of mega-cities have been implemented.The construction of Xiong'an New Area, an important component of China's strategic development, is a pilot project in urban sustainability transition (The Central People's Government, n.d.a).Similarly, the construction of five sub-central business districts in Shenzhen was proposed in 2015, with the primary goal of decentralizing the urban population by 2030 (Shenzhen government, n.d.)).A national 'red line' to conserve arable cropland area was proposed in 2013.In this proposal, food sovereignty was regarded as the foundation of national security.The cropland area in China should be maintained at a level of > 120 Mha to secure the nation's food security (The Central People's Government, n.d.b).To achieve this target, the central government has enacted various regulations, including a cropland requisition -compensation balance and an increasing versus decreasing balance of urban -rural built-up land.A possible policy tool under exploration is the ecological compensation mechanism, which redirects new growth from regions to be protected (e.g.productive cropland) to regions where additional development is desired (Bren D'Amour et al., 2017).However, the effectiveness of this urban containment strategy varies among regions, depending on diverse factors such as policymaker willpower and the geographic and institutional contexts (G.Chen et al., 2020;Wang et al., 2021).
In summary, many cities have population densities far higher than their immediate environment can support.Because they cannot exist in isolation from their surroundings, cities require additional access to natural resources beyond their borders.The current sanitary crisis is not the only concern regarding urban food security, considering the projected 100-110% increase in food demand by 2050 (United Nations, n.d.c)A better understanding of urban resilience and vulnerability is needed for more effective land use planning.Decoupling urban food consumption and local food production would interfere with the local ability to overcome sudden food crises; it would also weaken the ability to recycle nutrients in urban areas into cropland and challenge the United Nations Sustainable Development Goals related to Zero Hunger, Sustainable Cities and Communities, and Clear Water and Sanitation.Our results suggest that rapid urbanization will be accompanied by reductions in local food supply potential, which would weaken the abilities of cities to cope with the disruption of local transport networks.Effective governance of urban land area and population size is critical for efforts to secure the local food supply potential of cities.

Limitations and prospects
Because no available spatial data regarding the crop distribution in China met the time period and accuracy requirements of this research, spatial differences among crops were not considered in this study.The prediction of food production in future scenarios is an important endeavor that requires the use of climate change data and crop growth models.This study only focused on population change in the context of future urban development.Additionally, we believe that because a welldeveloped inner-city supply chain can shorten the food transportation time, it can partly compensate for the difficulty of obtaining food through a well-developed transportation system, which will be the focus of future research.

Figure 1 .
Figure 1.A flow chart describing the calculation processes.

Figure 2 .
Figure 2. Urban land expansion for 1990-2015 in the mainland China and cropland loss (three mega-urban regions.

Figure 3 .
Figure 3. Spatial (a) of urban resident distance to the nearest cropland and the number of cities with different trends in 1990-2015 (b) and temporal changes (c) of urban resident distance to the nearest cropland.

Figure 4 .
Figure 4. Percentages of urban populations that could theoretically be fed with a 10-km foodshed by population centers for 1990-2015 (A: the year of 1990; B. the year of 2015; C. relative change in 1990-2015).Note: 1 Red and blue circles refer to the cities with the adequate and inadequate food supplies with the 10-km foodshed radius.Circle sizes refer to percentages of urban populations that can be fed locally. 2 Results for 10-km and 20-km foodshed radiuses were provided in Fig. S3. 3 A relative increase below 0 represents a decrease in the population fed locally between 1990 and 2015.

Figure 5 .
Figure 5. Shifts in foodshed radiuses based on calorie under the business-as-usual scenario (BAU) and food transport block (FTB) scenarios.

Figure 6 .
Figure 6.Responses in food-shed radiuses in three mega-urban regions in China under five shared socioeconomic pathways for 2020-2080 (A: the North China Plain; B: the Lower Yangtze River; C: the Pearl River Basin).