Abstract: Monitoring and evaluation of soil salinity and water content quickly and timely is important for agricultural production and land restoration in saline areas. Based on data of measured soil water and salt content, and hyperspectral reflectance and contemporaneous Landsat-8 OLI image in Pingluo, northern Yinchuan of Ningxia, we used resampling technology to match the measured hyperspectral data with OLI image bands, screened out sensitive bands and 11 salinity indices by 11 linear and nonlinear multiple function models. We established the regression models between soil salt content (SSC) and water content (SMC) with multiple linear regression (MLR), partial least squares regression (PLSR), and support vector machine regression (SVM). The results showed that the reflectance of resampling measured hyperspectral bands had a significant correlation with that of the OLI image bands under different soil salinity and water content. The R² of SSC inversion models was 0.626, 0.510 and 0.829 for MLR, PLSR and SVM, respectively, and was 0.455, 0.204 and 0.731 in SMC inversion models. Results from the accuracy evaluation of the models through the validation set showed that the SVM model had better performance in predicting SSC and SMC than MLR and PLSR. Our results provide reference for the prediction of soil salinity and water content in different periods in the study region and similar regions.

Key words: Phyllostachys heterocycla, soil infiltration characteristics, mulching, root length density, soil fauna.