Spatiotemporal variation in soil available potassium in Central-Western China

The status of soil available potassium (AK) and their spatiotemporal changes are essential for nutrient management and food security. However, high-resolution spatial mapping of soil AK at the regional scale in combination with multiple environmental variables remains limited. Using data from 3,318 sites in the 1980s and 2010s and a variety of machine learning methods, this study systematically assessed the spatial distribution and stock changes of available potassium density (AKD) in central-western China. The results showed that the random forest model achieved moderate predictive performance in both periods (RMSE: 0.55-0.59 g m⁻², R²: 0.36-0.43). From the 1980s to the 2010s, the total soil AK stock in central-western China decreased by 12.65%. AK stock in natural ecosystems declined by 16.01%, likely reflecting insufficient sustained replenishment, increasing water limitation, ecological degradation, and grazing disturbance, whereas AK stock in agroecosystems increased by 12.71% under enhanced potassium fertilization and management practices such as straw incorporation. Spatially, AKD generally exhibited a pattern of higher values in the west and lower values in the central region. AK stock on the Tibetan Plateau declined markedly, whereas regions such as the Chengdu Plain and the Loess Plateau showed evident AK accumulation. SHAP analysis indicated that topographic and climatic factors were the dominant factors affecting the spatial differentiation of AKD. Slope, SAGA topographic wetness index (TWI), annual evaporation (Evap), sunshine duration (Solar) and elevation (DEM) showed relatively high contributions, while land use and vegetation played important regulatory roles. This study highlights the necessity of high-resolution assessment of soil potassium dynamics and provides a scientific basis for precision nutrient management, cultivated land quality improvement, and regional agricultural development and ecological restoration.