干旱区生态环境知识资源中心

In order to research the indicator for sandy information, this paper conducts a study on soil organic matter (SOM) in sandy land. Taking the Otindag Sandy Land and its surrounding area as a test site, in Xilingol League, Inner Mongolia, the BJ-1 multispectral image as main data, the soil information parameters were analyzed firstly, and their difference between the sandy land and other land was distinguished. Secondly, the correlation between SOM and each band of multispectral image was analyzed, and the best inversion band was determined. Meanwhile, the quantitative retrieval model for SOM was established and validated. Finally, the soil organic matter was inversed quantitatively, and the whole distribution of SOM was obtained in Otindag Sandy Land. As the results showed that, with the development of land desertification, the content of soil organic matter declined obviously. The correlation between three bands of BJ-1 image and SOM was relatively good, correlation coefficient (r) was as high as 0.7. But the predicted accuracy of multiple regression retrieval model for SOM was higher, and it was more stable than the single band linear regression model. The reason is that three bands contain more effective information than a single band, it can reflected the difference of divergent soil types. The model was validated using independent samples, the standard error RMSE was 0.6445 and model accuracy was 62.65%.