Scale effects and spatial heterogeneity of mining impacts on farmland soil organic carbon density

Mining operations represent a major anthropogenic disturbance worldwide, yet their multi-scale impacts on farmland soil organic carbon density (SOCD) remain poorly understood. This deficiency in knowledge poses a significant impediment to the refinement of farmland conservation and soil carbon management in mining regions. Leveraging two open-access datasets of farmland SOCD and a multi-method analytical framework incorporating complex network analysis, Random Forest regression, multi-scale geographically weighted regression and structural equation modeling, this study provides a systematic investigation into the multi-scale effects, spatial heterogeneity, and nonlinear threshold characteristics of mining disturbances on farmland SOCD. We identified that (i) 10 km was identified as the key scale for detecting disturbances. At this scale, coal production emerged as the dominant factor driving SOCD dynamics, and the inclusion of mining factors improved the R² of the SOCD prediction model by 15.5% and reduced RMSE by 4.5% compared with the baseline model. (ii) At the 10 km scale, mining factors surpassed natural factors to become the dominant force regulating farmland SOCD, and the effects of coal and non-ferrous metal production on SOCD exhibited nonlinear responses. (iii) Regionally, non-ferrous metal production indirectly reduced SOCD in East China by suppressing agricultural management (r = -0.86, p < 0.01), while coal production enhanced SOCD in Central and North China through direct and indirect pathways (r = 0.42, p < 0.05). The findings demonstrate that mining impacts on farmland SOCD represent a complex trade-off determined by mining type, intensity and regional context, providing a scientific basis for differentiated soil organic carbon management policies in mining-affected agricultural areas.