Hyperspectral predicting model for Black soil moisture at different depth

摘要

This study aims to quantify the relationship between field Black soil hyperspectral reflectance and soil moisture at different depth. The stepwise regression predicting model of soil moisture at different depths was built with the spectral derivatives as independent variable, and the models were evaluated with root mean square error (RMSE) and detemining coefficient. The results are as follows: (1) the overall correlation coefficients between spectral reflectance and soil moisture at different depth vary considerably, and the largest correlation coefficient is between soil spectral reflectance and soil moisture at the depth of 10–20 cm, maximum at around 960 nm. (2) This article identifies the best model of soil water content at 0–2cm, 2–10cm and 10–20cm, pre-determination coefficients R2 were 0.48, 0.51, 0.78, root mean square error (RMSE) were 0.4, 0.48 and 0.24, respectively. (3) Among different mathematical transform of soil spectral reflectance, of the correlation between the first derivative and the soil water content becomes much more remarkable, the number of selected band in water content model at different depth also increases, and the models become more stable.