Spatial Prediction Accuracy Of Nutrients In Arable Soil

The spatial prediction of soil nutrients has been greatly concerned in county scale, with the development of measuring science and science technique. Ordinary Kriging (OK) is the most popular method that is used, but many factors such as parent material, terrain, vegetation, etc were not considered. In the paper, a spatial prediction model based on GIS and Geostatistics was built, and the influence of environmental factors to soil organic matter (SOM) and available potassium (AK) were conducted in Shimen, Hunan province. The main conclusions are as follows:(1) Terrain, vegetation, climate factors were extracted from Digital Elevation Model (DEM), MODIS, TRMM, TM remote sensing data based on GIS technique. Regression Kriging (RK) was used to predict the SOM, and validation points was adopted to do the accuracy assessment. The correlation analysis showed that significant relation was exist among SOM and longitude, slope, compound topographic index, terrain roughness, amplitude of landforms, land surface temperature, elevation, NDVI. The result of SOM prediction based on RK was better than OK, and relative improvement was6.03%, and RK approach could indicate that remotely sensed data had the potential as auxiliary variables for improving the accuracy and reliability of SOM prediction.(2) The spatial distribution of soil available potassium (AK) was predicted with Kriging method that combined with parent material information (PMK). PMK was better than OK, and relative improvement was68.07%. It showed that PMK approach had the potential as auxiliary variables for improving the accuracy of AK prediction, and it was a more convenient, efficient method in the region that has complex parent materials for soil properties.