Mapping surface soil organic matter based on multispectral image

Soil organic matter (SOM) is an essential and dynamic variable in terrestrial ecosystem. This study used a method for mapping its spatial distribution by remote sensing technique. An image observed by Landsat 5 was used to estimate the spatial pattern of surface SOM in a town scale. The results showed that the concentration of surface SOM in study Jianshe town had a negative correlation (r =0.51, P<;0.01) with TM 1 image intensity. A nonlinear model was established using this correlation (R²=0.61, P<;0.001) to estimate surface SOM. The validation showed that the model can predict surface SOM reliably (RMSE=0.96). Based on that model, the spatial pattern of surface SOM in study town can be exactly mapped in details, which may provide potential information for nutrients management in that arid area. The accuracy of prediction model have much space to be improved if more such studies can be carried out to refine the image processing and interpreter technology and enhance the correlation between SOM and multispectral soil reflectance.