Cropland abandonment in mountainous China: Patterns and determinants at multiple scales and policy implications

Understanding the formation and causes of cropland abandonment is crucial for developing effective strategies to mitigate this process. However, studies on cropland abandonment often look at one spatial scale, although the relationship between abandonment and evaluated factors may change depending on the specific scale, which limits our understanding of the process of cropland abandonment. Taking mountainous Southern Sichuan of China as an example, we utilized satellite-derived cropland abandonment maps to assess the evolving patterns of cropland abandonment from 2003 to 2018 across various scales. We then carried out a multi-scale (parcel, local and regional) assessment of the relationships between selected 17 variables representing natural, location and socioeconomic conditions and cropland abandonment. Results showed, the cropland abandonment rate increased from 2 % in 2003 to 15 % by 2018, and cropland abandonment always showed a higher concentration in the central part of the study area. At different scales, a combination of factors, which determined cropland abandonment, differed. We also found the distance to the nearest county center was scale-dependent factor and statistically significant factor of cropland abandonment only at the local scale. Among the factors with large contributions, topography mattered at all three scales; cropland per capita, aging and agricultural labor per unit of cropland mattered at the parcel and regional scales. Cropland fragmentation, air temperature and distance to the nearest settlement mattered at the local scale; the distance to the nearest water system, the distance to nearest county center and agricultural mechanization mattered at the regional scale. This multi-scale assessment framework embodied with multi-variable and multi-temporal could provide a new perspective to reveal the complex nonlinear relationship between the determinants and cropland abandonment, and further help to predict future cropland abandonment hotspots and identify relevant land-use policies.