Abstract:Taking farmland of oasis in Xinjiang Manas as the example, in order to timely, accurately and dynamically monitor water and salinity of saline soils,the partial least squares regression (PLSR) for model was applied to model the moisture and salt content of different moistures and salt soils based on hyperspectral analysis technique, the stability and predictive ability of the model was validated. The results show that the prediction precision of soil salinity and moisture were effectively improved through continuum removal (CR) and the logarithm of first order differential (lgR)′ in 12 kinds of data transformation. The prediction models were better when soil salt content was less than 819dS/m, R2cal were greater than 079, R2val were greater than 064, with no significant difference between RMSEP. The prediction precision was poor when soil salt content was greater than 10dS/m with R2val less than 045 in the moisture prediction models. The better prediction accuracy when the moisture is less than 15%, R2cal were greater than 077, R2val were greater than 064,with RMSEP less than 46. The model prediction accuracy was poor when soil moisture greater than 15%. It was concluded that the large soil moisture, salt content will have a significant impact on salt water prediction model.