Water vapour is highly variable over tropical region and sensitive to weather condition, monsoon onset, green house effect, and pollution level in Ganga River. In the present study, variability in water vapour derived from Global Positioning System (GPS) over Varanasi (25$^{\circ}$20$^{\prime}$N, 82$^{\circ}$59$^{\prime}$E) during the period 2007–2010 has been studied. The GPS-derived water vapour (WV) has been compared with those retrieved from Moderate Resolution Imaging Spectroradiometer (MODIS) and ECMWF. The GPS-WV data concurrent to MODIS and ECMWF timing has been correlated to perform further analysis. To study the accuracy of water vapour retrieved from the MODIS and ECMWF, root mean square error (RMSE), absolute error (AE), correlation and standard deviation in it are computed with respect to GPS-derived water vapour. Analysis shows an annual correlation $R^{2}$ = 86%, RMSE = 9.5 mm and AE (MODIS–GPS) = 7.0 mm in MODIS retrieval and annual correlation R$^{2}$ = 86%, RMSE = 6.1 mm and AE (ECMWF–GPS) = 2.4 mm in ECMWF reanalysis retrieval. Correlation of ECMWF and MODIS datasets with the GPS datasets are found to vary significantly with seasons. The correlation is high during monsoon season and low during spring season. Water vapour is found to be an indicator for the onset of monsoon.

$\bf{Highlights}$

$\bullet$ Accuracy of water vapor (WV) retrieved from the MODIS and ECMWF with respect to GPS WV.

$\bullet$ High Annual correlation of $R^{2}$ = 0.86 between both MODIS–GPS and ECMWF–GPS.

$\bullet$ The correlation is high during monsoon season and low during spring season.

$\bullet$ The performance of ECMWF is found to be better than that of MODIS.