Soil samples with various salt and moisture contents were artificially prepared by adding different amount of NaCl solutions to costal saline soil to simulate the evaporation process. During the evaporation process, soil moisture contents and soil spectra were regularly collected, and then analyzed using spectral indices and partial least squares regression (PLSR) to quantify soil moisture content. The results showed that the differential moisture index derived from the reflectance value of 2 027 nm and 1 878 nm was the best index to predict soil moisture content, and the indexes obtained from NDMI _2027,1878 in both calibration and validation process were slightly better than these from PLSR, with the determined coefficient (R²) of the prediction as high as 0.99. The root mean square error (RMSE) was only 21.84 g/kg, and not affected by the salinity grades. It could be concluded that soil moisture content can be accurately predicted by NDMI _2027,1878 .