Homogeneity of retrieval models for soil organic matter of different soil types in Northeast Plain using hyperspectral data

Organic matter (OM) content is an important index in reflecting soil fertility, and provides important information for soil digital management and precise fertilization if OM can be estimated using hyperspectral technology. In this experiment, NIR-Visible spectral reflectance of soil was measured using ASD FR 2500 hyperspectral meter. The spectral reflectance curve shapes and the OM prediction models of different soil types were analyzed. The results show that, according to the variation of spectral reflectance curves in the Northeast Plain, those soil types can be divided preliminarily into three categories, and their main differences are the curve slopes in 4601350nm. Further, the sensitive bands to SOM were determined as 650750nm by correlation analysis. The SOM regression models of different soil types were established using reflectance in the sensitive bands, and the homogeneity of variance is tested for regression coefficients and intercept. The results indicate that black and chernozem, and black and dark brown soil pass the homogeneity test respectively. Both of the two homogenous categories can be predicted using common models, but the predicting accuracy is slightly lower.