Application of hyperspectral data for soil organic matter estimation based on principle components regression analysis

Soil organic matter (SOM) content is an important indicator of soil fertility. It provides important information for soil digital management and resource evaluation if SOM can be estimated using hyperspectral technology. In this experiment, Visible-NIR spectral reflectance of soil samples was measured by using an ASD2500 device. The sensitive bands for SOM estimation were initially determined by applying correlation analysis. Furthermore, the processings including the selection of significant variables, collinearity diagnostics and data transformation were carried out by using stepwise regression analysis and principle components regression (PCR) analysis. The predicting model for SOM estimation was developed for black soil in northeast China. The validation results showed that the model was effective and practicable (R2=0.840, RSME=0.226)