The purpose of this study was to investigate the possibility of the GPS Precipitable Water Vapor (GPS-PWV) to forecast reservoir intake. The correlations were examined between monthly average GPS-PWV and monthly intake of two reservoirs as well as they were inspected mathematical relationship by the polynomial neural networks algorithm. The monthly reservoir intake is proportional to the monthly GPS-PWV trend; furthermore, the peak of reservoir intake often away from the peak of average GPS-PWV 1-2 months. The mathematical model between monthly average GPS-PWV and monthly intake of UBONRATANA reservoir has the model fit coefficient of determination (R2) of 0.97 and mean absolute error value (MAE) of 54.19x106 m3; in addition, the validation test has R2 of 0.93 and MAE value of 61.36x106 m3. Moreover, the mathematical model of LUMPOW reservoir has the model fit R2 of 0.98 and MAE value of 25.65x106 m3; in addition, the validation test has R2 of 0.72 and MAE value of 46.64x106 m3. Indeed, both models has correlation coefficient of 0.99; therefore, it may be possible to use GPS-PWV predict reservoir intake.