Prediction of soil organic carbon in mining areas

Coal mining activities have destroyed natural landscapes and surface vegetation, resulting in a large loss of soil organic carbon (SOC) storage in farmlands. Therefore, farmland SOC mapping in mining areas is of great significance to low-carbon reclamation and ecological compensation. Natural factors such as topography, climate, vegetation, and soil properties are key variables for SOC mapping. Human activities, including crop rotation and farmland damage caused by coal mining, also severely interfere with farmland SOC in mining areas. Here, we use take mining areas of the Changhe watershed as a case study to explore the effects of these activities on farmland SOC storage. Farmland damage was quantified using mining area geological hazard data, and four random forest (RF) models were constructed to predict SOC content based on natural variables alone (RF-N), natural variables + crop rotation (RF-C), natural variables + farmland damage (RF-F), and natural variables + crop rotation + farmland damage (RF-A). We found that: (1) the RF-A model most accurately captured SOC spatial differentiation information (followed by RF-C, RF-F, and RF-N). (2) There were significant differences in the SOC of different crop rotation areas and between low and high damage farmlands; specifically, the SOC content of wheat-corn rotations (one-year two cropping system) was higher than wheat-soybean-corn rotations (two-year three cropping system) and wheat (one-year one cropping system), and the SOC of high damage farmlands was lower than that of low damage farmlands. (3) Variable importance analysis showed that crop rotation and farmland damage were the two most important variables for predicting SOC in mining areas—demonstrating the effectiveness of common farming practices in predicting SOC in mining areas. Our results provide a scientific basis for land reclamation, soil carbon input, and land remediation planning in mining areas.