Objective The aims were to reveal the spatial distribution patterns, controlling factors, and threshold effects of soil organic matter (SOM) in the black soil region of Inner Mongolia, in order to provide scientific basis for the cultivation of SOM and regional carbon management in this area.

Method The black soil area in the eastern part of Inner Mongolia was as the research object, a random forest (RF) model coupled with the Shapley additive explanation (SHAP) approach was employed to quantitatively assess the influence of multi-source environmental and soil attribute variables, with comparisons drawn against the geographically weighted regression (GWR) and multiscale geographically weighted regression (MGWR) models.

Result The results demonstrated that the RF model outperformed traditional and other machine learning models in both predictive accuracy (R² = 0.747, RMSE = 1.561) and stability, effectively capturing the nonlinear relationships between SOM and environmental factors. Consistent rankings from both %IncMSE and mean absolute Shapley value identified total nitrogen, total phosphorus, vegetation index, mean annual temperature, soil moisture, and mean annual precipitation as the primary determinants of SOM variation. The responses of SOM to those factors exhibited pronounced nonlinearity and threshold effects. Soil moisture and annual precipitation showed an inverted "U-shaped" relationship, indicating that moderate moisture conditions favored SOM accumulation. Spatially, SOM displayed a "high in the northeast-low in the southwest" pattern, with hotspots concentrated in eastern Tongliao and the southern margin of Hulunbuir, corresponding to areas of dense vegetation and favorable hydrothermal conditions.

Conclusion These findings highlights that the spatial pattern of SOM in the study area is mainly driven by the synergy of nutrient supply, vegetation coverage and water and heat conditions. Different predictor variables exhibit nonlinear responses and threshold transitions at their respective gradients.