Interprovincial food trade aggravates China's land scarcity

The study addresses how effectively interprovincial trade in agricultural products mitigates or exacerbates land scarcity—the imbalance between cropland availability and demand—in China. Contextually, global food demand is rising with population growth and dietary shifts, while cropland is limited and increasingly scarce, making land a pivotal resource for achieving SDGs (notably SDGs 2, 11, and 15). Existing land footprint (LF) indicators quantify land use embodied in consumption but overlook local scarcity. The authors propose a scarce land footprint (SLF) that integrates land scarcity into footprinting, enabling spatially explicit assessment of hotspots and the two-way relationship between trade and scarcity. China is chosen due to its large population, limited cropland, substantial domestic crop trade, and growing food security challenges under climate change. The purpose is to map LF/SLF hotspots at high resolution, quantify virtual (scarce) land flows, and compare a land scarcity index (LSI) with a trade-related LSI (TLSI) to evaluate whether trade mitigates or aggravates provincial land scarcity.

Prior work has developed LF to trace land embodied in inter- and intra-national trade using MRIO frameworks, with studies at global, regional, and national scales showing teleconnections between production and consumption. However, traditional LF does not reflect land scarcity: identical land areas can impose different scarcity pressures depending on local availability. Few studies have attempted scarcity-weighted footprints at grid level, and often without incorporating virtual flows embodied in trade. The authors build on advances in high-resolution footprint mapping (e.g., carbon footprints) and scarcity-weighted indicators (analogous to technology- or scarcity-adjusted footprints in water and energy) to integrate land scarcity into MRIO-based accounting. They identify a gap in combining scarcity metrics with high-resolution spatial mapping and virtual flow analysis within a single framework for land, which this study addresses for China.

• Framework: A Chinese multi-regional input–output (MRIO) model with 31 regions (provinces) and 42 sectors captures intersectoral and interprovincial economic linkages. The standard Leontief demand-driven model X = (I − A)⁻¹Y is used.
• Land footprint (LF): Computed by coupling MRIO with direct land-use intensities E = (e_ik) to obtain LF_ij = Σ_k e_ik y_kj^i, yielding land used in region i to satisfy final demand in region j. Virtual land exports/imports and net virtual land (NVL) are derived by summing LF flows across provinces.
• Land scarcity metrics: Land Scarcity Index (LSI_i) = CL_i / SC_i, where CL is cultivated cropland area and SC is suitable cropland area. Scarcity classes: none (LSI < 0.6), moderate (0.6–0.8), severe (0.8–1.0), extremely severe (≥1.0).
• Trade-related LSI (TLSI): TLSI_i = (CL_i − NVL_i)/SC_i measures hypothetical land stress if a province met its net-imported virtual land with its own land (even if exceeding its maximum availability), indicating the effect of trade on scarcity.
• Scarce land use and footprint: SL_i = LSI_i × CL_i and scarce land footprint SLF_ij computed by applying scarcity-weighted land-use intensities (e_ik × LSI_i) within the MRIO to obtain SLF flows and net virtual scarce land (NVSL).
• Spatial downscaling: Provincial LF and SLF are downscaled to 1 km × 1 km grids using population-based extrapolation coefficients β_LF^i = LF^i/Pop^i and β_SLF^i = SLF^i/Pop^i applied to gridded population, ensuring grid sums match provincial totals. Mapping performed in ArcGIS 10.6.
• Hotspot analysis: Getis-Ord G* statistic identifies LF and SLF hotspot clusters (Z > 1.65, p < 0.05). Buffer zones of 40 km from provincial borders delineate cross-province hotspot clusters; proportions across borders are computed (PLF, PSLF).
• Data sources: 2012 Chinese provincial MRIO (Liu et al., 2019); cultivated land area from China Statistical Yearbook (NBSC, 2013); suitable cultivated land from GAEZ v3.0 simulations (Fischer et al., 2012; Deng et al., 2019); gridded population from Landsat/MODIS (2012). Foreign trade is excluded (only ~9% of China’s virtual land imports).

• Spatial concentration: Over 70% of China’s LF and SLF are concentrated in less than 20% of land area. Within-province polarization is pronounced in Qinghai, Tibet, and Xinjiang (≈95–99% of SLF in ~20% of land), whereas Shandong, Henan, and Jiangsu show more even distributions (<30% of SLF in 20% of land).
• Cross-border hotspots: About 38% of LF and SLF hotspot clusters cross provincial borders, indicating the need for interprovincial coordination.
• LF vs SLF difference: In 91.5% of China’s land area, SLF < LF (negative SLF–LF difference), suggesting traditional LF tends to overestimate scarcity pressures nationally and that locally suitable land remains for conversion in many areas. Largest negative differences occur in Guangxi and Guizhou (exporters facing exacerbated scarcity), while Guangdong shows reductions by importing from less-scarce provinces. Minimal differences in land-abundant Heilongjiang, Shandong, and Hebei.
• Hotspot regions: Major LF/SLF hotspots include Jing-Jin-Ji, Yangtze River Delta, and Sichuan Basin; LF shows dense hotspots in North China, Guanzhong, Jianghan/Dongting plains, while SLF hotspots shift to Songnen, Liao, and Huanghuai plains due to scarcity weighting.
• Virtual (scarce) land flows: Net flows predominantly from North/inland and less-developed provinces to more affluent coastal regions. Exporting hotspots (net virtual land and net virtual scarce land) occupy 60.8% and 62.2% of hotspot areas, 1.55× and 1.65× larger than importing hotspots, respectively.
• Provincial roles: Top exporters of scarce land: Heilongjiang, Inner Mongolia, Henan (combined 30.8% of total exports). Top importers: Jiangsu, Zhejiang, Guangdong (combined 30.6% of total imports). Heilongjiang is the largest net exporter; its exported virtual scarce land volume is 61.5% of its exported virtual land. Jiangsu is the largest net importer; its imported virtual scarce land is about two-thirds of imported virtual land.
• Role transitions under scarcity: Anhui shifts from net virtual land exporter to net virtual scarce land importer; Hunan shows the reverse, indicating trade partners’ scarcity conditions alter roles when scarcity is considered.
• Trade’s impact on scarcity (TLSI vs LSI): 13 net-importing provinces (e.g., Beijing, Shanghai, Guangdong) mitigate scarcity (TLSI > LSI), though many remain severely or extremely scarce. Among 18 net-exporting provinces (TLSI < LSI), nine move from no scarcity to moderate/severe (e.g., Jilin, Heilongjiang, Inner Mongolia, Xinjiang), and Guangxi shifts from severe to extremely severe. Quantitatively, trade mitigates importers’ scarcity by 50.8% but aggravates exporters’ scarcity by 119.8%.

The study demonstrates that while interprovincial trade helps alleviate land scarcity in importing provinces, it simultaneously intensifies scarcity in exporting provinces, often located in less-developed inland regions. Incorporating scarcity into footprinting (SLF) reveals spatial and role re-orderings not visible under traditional LF, highlighting how geospatial mismatches between cropland suitability and consumption create uneven burdens. The findings emphasize telecoupling of food systems across China and the need for hotspot-oriented and cross-border governance to align with SDGs (2, 11, 15). The SLF framework complements material and other environmental footprints by integrating local scarcity, offering a more realistic basis for policy instruments (e.g., adjusting agricultural layout, crop rotations, trading structures, efficiency measures) and for developing scarcity-aware boundary indicators at subnational scales. Identifying that ~38% of hotspots cross provincial borders supports the case for interprovincial mechanisms (e.g., requisition-compensation balance, trans-provincial cultivated land management) to coordinate resource flows and poverty alleviation. Overall, integrating scarcity changes the assessment of sustainability and equity of consumption, informing strategies to reduce land-use pressures while maintaining food security.

This work integrates land scarcity into MRIO-based footprinting to produce a scarce land footprint (SLF) and a trade-related land scarcity index (TLSI), mapped at 1 km resolution for China. It finds that LF and SLF are highly concentrated spatially, with ~38% of hotspots crossing provincial borders, and that traditional LF tends to overestimate scarcity relative to SLF for most areas. Net flows run from northern/inland producers to coastal consumers. Crucially, trade mitigates scarcity in importing provinces but disproportionately aggravates it in exporters, with notable role shifts when scarcity is considered (e.g., Anhui and Hunan). The study advances footprint accounting by embedding site-specific scarcity, informing SDG-aligned policies for land resource management, productivity optimization, and interprovincial collaboration. Future research should extend to time-series and global MRIO linkages, refine sectoral detail and scarcity metrics, and explore multi-resource nexus scenarios to support comprehensive sustainable development planning.

• MRIO model uncertainty: Results depend on the chosen Chinese provincial MRIO (three available for 2012). Monte Carlo analysis indicates disagreements <15.6% across databases, but uncertainty remains.
• Temporal coverage: The analysis is cross-sectional (2012) due to data availability; time-series dynamics of LF/SLF and flows are not examined.
• Sectoral aggregation: An aggregated agriculture sector may misestimate effects of efficiency improvements on virtual land flows.
• Trade scope: International trade is excluded (foreign virtual land ≈9% of China’s total), limiting global perspective and cross-border scarcity assessment.
• Scarcity metric simplifications: LSI (CL/SC) and land-use intensity per monetary output are coarse proxies; more explicit, productivity- and intensity-based metrics at grid scale are needed.
• Data constraints: Lack of high-resolution environmental satellite and gridded consumption data limits precision of downscaling and hotspot attribution.
• Nexus perspective: Interactions with other resources (e.g., water, energy) are not modeled; scenario analysis in a resource nexus framework is suggested for future work.