The visible and near infrared spectroscopy is a fast and inexpensive non-destructive technique for the prediction of concentrations of salts in wastewater. Conventional chemical methods are usually used, which are very accurate, take more time and require special techniques for sampling, storing and pretreatment of wastewater. In this work we studied the spectral characteristics of water and the effect of salts on the perturbations in the water absorption bands. The generation of multiple regression models with principal components was carried out on standard solutions with composition of salts similar to that of wastewater samples taken along the drainage channel network of the Mexico City Metropolitan Area. The spectral signatures were obtained in situ and in the laboratory using a portable high-resolution spectroradiometer (ASD FieldSpec 3). The prediction model generated showed high precision in the estimation of salinity in wastewater, a coefficient of determination of 89.6% and a low root mean square error of 0.12‰. Other compounds, which are not discussed here, cause distortion of the absorption bands of water at wavelengths less than 900 nm or near the visible region, while our results showed distortions in the water spectrum at higher wavelengths (>1,000 nm).