Rapid and accurate monitoring of soil organic matter (SOM) content is of great significance for precision fertilization of farmland. However, the SOM retrieval models are mainly established by statistical methods, which have limited application scope and incomplete theoretical foundation. Moreover, the accuracy of the SOM retrieval models remains raised. In this paper, for the first time, a semi-empirical SOM content retrieval model is constructed, which has certain theoretical basis, strong

more »

... licability and higher accuracy than before. Based on the Kubelka-Munk (KM) theory, the SOM retrieval model with the absorption coefficient k and scattering coefficient s related to SOM (r = k/s) is derived. The validity and reliability of the model are confirmed with validation set (n = 26) including three sorts of soils. Results show that the model can estimate SOM content in different sorts of soils with high prediction accuracy and good prediction ability (root mean square errors of prediction (RMSEP), coefficients of determination (R 2 ) and relative percentage deviation (RPD) values of 0.18%, 89.9% and 3.2, respectively ) in the range of 552-950nm. The model provides an innovative method for predicting SOM content. INDEX TERMS Soil organic matter retrieval, reflectance, semi-empirical model, KM.