DETECTING MOISTURE CONTENT OF SOUTHERN SOIL IN CHINAUSING NEAR-INFRARED SPECTROSCOPY

摘要

Three types of Southern soil (yellow-brown soil, paddy soil and alluvial soil) were studied based on NIR spectra. Soil moisture analysis model of pretreated soil samples was established by PLS, with R2=0.9946 and 0.9919 and RMSECV=0.801%, RMSEP=0.912% for validation and prediction, respectively. 12 unpretreated soil samples were predicted by re-established PLS model based on Kennard-Stone algorithm,with average absolute error reducing from 0.78% to 0.42%, which showed that the soil sample space expansion method can improve model prediction accuracy.