Non-grain production of cropland (NGPCL) is a common consequence of rapid urbanization, but excessive NGPCL threatens food security and sustainable cropland use. However, the evolutionary processes and mechanisms of different NGPCL types remain largely unknown, compromising the scientific basis for NGPCL management. Thus, taking rapidly urbanized Deqing as the study area, this research constructed an NGPCL theoretical framework from the perspective of agricultural location, classified NGPCL types using Landsat images, random forest algorithm and Google Earth Engine, and revealed their spatiotemporal changes and different influencing factors through the Multinomial Logit Regression model, and provided targeted zoning and categorized policy suggestions. Our constructed theoretical framework modeled NGPCL by the complex interactions among geographical location, natural conditions, socioeconomic levels, and policy forces. From 2000 to 2022, NGPCL in Deqing County had a high degree and significantly increased from 68.65% to 89.19%. Pond farming and horticultural crops increased by 4046.46 ha (38.0%) and 1307 ha (27.1%), while tree planting slightly decreased by 54 ha (-2.5%). However, these NGPCL expansions resulted in a drastic decrease of 5299 ha (-65.5%) in grain crops plantation. Different NGPCL types showed varied spatial distribution characteristics and differentiated driving mechanism. Horticultural crops depended on geographical location and socioeconomic factors and were widely planted near towns and roads with high population density and economic strength. Pond farming mainly extended along the eastern flat rivers and relied on resource endowment. Tree planting was closely related to the natural condition and was concentrated in the sparsely populated mountainous areas. The NGPCL is the result of balancing various factors, and under conditions of extremely scarce cropland resources, the Chinese government needs to strike a reasonable threshold for non-grain production and propose targeted policies of zoning classification to prevent excessive NGPCL from threatening grain yield security.
Keywords: Agricultural location; Influencing factors; Multinomial logit regression; NGPCL type classification; Non-grain production of cropland; Remote sensing.
Copyright © 2025 Elsevier Ltd. All rights reserved.