Modeling soil organic carbon spatial distribution for a complex terrain based on geographically weighted regression in the eastern Qinghai-Tibetan Plateau

Permafrost regions store a large amount soil organic carbon (SOC), and the decomposition of these carbon pools can release greenhouse gases and further strength climate warming. An explicit spatial distribution of SOC is one of the basic databases for Earth System Models. However, efficient approaches for obtaining the spatial distribution of SOC remain challenging, especially in mountainous areas which are characterized by complex terrains. Here, we modeled the spatial SOC distribution using the geographically weighted regression (GWR) approach in an area on the eastern part of the Qinghai-Tibetan Plateau (QTP). We analyzed multiple environmental variables and soil profile data (n = 73) to find the best prediction models for the SOC density (SOCD) for the 0-50 cm layers. The results showed that normalized difference vegetation index (NDVI), elevation, and slope gradient are the significant predictors for the SOCD. For the upper 50 cm soil layers, the SOCD ranged from 1.08 to 18.32 kgm(-2), with higher values in mountain slopes but lower values in mountain valleys and basins. The GWR model had a higher prediction accuracy in the modeling SOCD in comparison with other models such as ordinary kriging (OK) interpolation, multiple linear regression (MLR) model. Our results showed that GWR model is a useful tool for modeling of SOC distribution and potentially can be integrated into Earth system models in areas of complex terrains.