The spatial variability of temporal changes in soil organic carbon and its drivers in a mountainous agricultural region of China

Abstract

The mountainous agricultural region of China (MARC) is characterized by complex natural conditions, fragmented farmland landscapes, and rapid socio-economic development. The relative contribution of these factors to the spatiotemporal variability of soil organic carbon (SOC) in MARC remains unclear. In this work, a total of 5121 topsoil (0–20 cm) samples (2,883 in 2012 and 2,238 in 2021) were collected from a typical mountainous area (43,700 km²) of MARC. Descriptive statistics, semivariance analysis, random forest (RF), and partial dependence plot (PDP) analyses were applied to investigate the spatiotemporal variability of SOC and its relationships with natural factors (climate, topography, and lithology), landscape pattern indices, and socio-economic factors. The average SOC content in 2021 (13.86 g kg⁻¹) was significantly lower than that in 2012 (15.08 g kg⁻¹). SOC exhibited moderate spatial autocorrelation in both years, with nugget/sill ratios of 58.98 % in 2012 and 64.16 % in 2021, respectively. The RF model explained 51 % of the spatial variation in SOC changes. Mean annual precipitation (MAP), GDP changes, elevation, population density changes, and landscape contagion index were identified as the main factors affecting the spatial variability of SOC changes. PDP analyses revealed that SOC decline was more pronounced at higher elevation and MAP, but this trend slowed down in areas experiencing faster economic growth and population outflow. Moreover, SOC decline was more severe in highly connected landscapes. These findings highlighted the influence of landscape pattern and socio-economic factors in the spatiotemporal variability of SOC, providing valuable insights for developing effective SOC management strategies for mountain agriculture.