Soil organic carbon dynamics in Northeast China under water erosion: Spatiotemporal patterns and driving mechanisms

Abstract

Soil organic carbon (SOC) changes driven by soil erosion directly influence the terrestrial carbon cycle. However, the erosion processes are complex. The interaction of multiple influencing factors introduces substantial uncertainty into SOC dynamics in cropland. Northeast China is a vital commercial grain-producing region that is vulnerable to erosion. This makes it urgent to identify patterns of SOC loss in erosion-prone croplands and understand their driving mechanisms. In this study, we mapped the spatiotemporal distribution of SOC in stable croplands with random forest models by applying remote sensing images and multi-source environmental data. We further integrated the China Soil Loss Equation (CSLE) with several machine learning (ML) methods. This allowed us to explore spatial patterns of erosion-prone SOC loss and determine its drivers. The main findings were as follows: (1) The random forest models showed strong performance for SOC mapping by selecting an optimal set of spectral indices and environmental covariates, with an R² of 0.87 for the training set and 0.63 for the validation set. (2) Cropland pixels exhibiting increased water erosion and SOC loss made up 53.92% of all pure cropland pixels, and 39% of these pixels were located in black soil regions, forming a distinct belt. (3) The primary drivers of SOC loss were the interactions between soil types and multiple factors, including rainfall erosivity trends (0.19), temperature trends (0.17), increased fertilizer use (0.16), and planting patterns (0.15). These findings provided valuable insights for promoting the sustainable carbon management of cropland under erosion.