.png){kind=icon}
.png){kind=icon}
Return
Soil Organic Matter in Rocky Mountainous Area of Northern China: Spatial Distribution, Drivers and Mechanisms
.png){kind=icon}
PRE
AI
Y
Z
Z
J
DOI：10.1002/ldr.70320.png)
.png){kind=icon}
Original request for help.png){kind=icon}
Share.png){kind=icon}
Save
Abstract
En 中文
Soil organic matter (SOM) is a key indicator of soil fertility and ecosystem resilience, yet its spatial and vertical variation mechanisms remain poorly understood in fragile mountainous landscapes. This study integrates machine learning and interpretable causal modeling to elucidate the multiscale drivers of SOM distribution in the Yihe River Basin (YRB), a typical erosion-prone region in northern China. The Random Forest (RF) model was used to predict the spatial distribution of SOM content, and the relative contributions and pathways of natural and anthropogenic factors were identified using the Piecewise structural equation (PSEM) model and Shapley additive explanations (SHAP). Results showed that the RF model successfully predicted SOM content with coefficients of determination (R2) of 0.695, 0.671, 0.626, and 0.730 for the four soil horizons, respectively. Horizontally, SOM content was higher in the eastern and northern regions of the basin. Vertically, the SOM content showed a decreasing trend with increasing soil depth. Among the factors influencing SOM distribution, soil properties, particularly total nitrogen content, were identified as the most significant determinants. Moreover, human activities primarily influence SOM indirectly by modifying soil physicochemical properties and climatic conditions. Their effects are more pronounced in surface soils, with total impact effects ranging from 0.94 to 1.04. These findings provide new insights into SOM dynamics in fragile landscapes and offer scientific guidance for targeted soil management and ecological restoration in erosion-prone regions.
Keywords:
erosion-prone regions
influencing factors
piecewise structural equation modeling
random forest
soil organic matter
spatial prediction
Journal
IF:
3.7
Papers:
4.3K
Citations:
1.3W
Organization
Cited Papers