Author(s)

Siming Chen, Ning Wang

Corresponding Author

Siming Chen

Abstract

In order to accurately predict the spatial distribution of soil organic matter (SOM), a case study was conducted in the south-western Fujian Province, south-east China. A total of 24 environmental factors were extracted by ArcGIS Geostatistical analyst and remote sensing image analysis technique. Then, the stepwise regression model was used to select the best combination of environmental variables. Finally, a hybrid model of geographically weighted regression kriging (GWRK) was adapted to predict the spatial distribution of SOM, and to compare with regression Kriging (RK). The results show that:(1) B5, B6, NDVI, CI, ELE, SPI, TRI and TMEAN as the most important factors affecting the spatial variability of SOM. The determination coefficients(R2) of stepwise regression model between these variables and SOM is 0.34, and the significance probability value show that P < 0.0001; (2) The SOM spatial distribution patterns derive with the RK and GWRK models are quite similar, showing a spatial pattern of “high in the middle, low in the north and south “. The GWRK model is the highest in prediction accuracy, and the prediction results are more consistent with the actual situation, reflecting the detailed information about spatial distribution of SOM. This method can provide a methodological support for the study of spatial distribution of soil organic matter in the same region

Keywords

Soil organic matter, geographically weighted regression kriging, Auxiliary variables, spatial prediction