An improvement in accuracy and spatiotemporal continuity of the MODIS precipitable water vapor product based on a data fusion approach

Precipitable water vapor (PWV) is one of the key variables in the water and energy cycles, whereas current PWV products are subject to spatiotemporal discontinuity, low accuracy, and/or coarse resolution. Based on two widely used global PWV products, i.e., satellite-based MODIS and reanalysis-based ERA5 products, here we propose a data fusion approach to generate PWV maps of spatiotemporal continuity and high resolution (0.01°, daily) for the Upper Brahmaputra River (UBR, referred to as the Yarlung Zangbo River in China) basin in the Tibetan Plateau (TP) during the monsoon period (May‒September) from 2007‒2013. Results show that the fused PWV estimates have good agreement with ground-based PWV measurements from eight GPS stations (correlation coefficient = 0.87‒0.97, overall bias = -0.35‒1.78 mm, and root mean square error = 1.17‒2.04 mm), which greatly improve the accuracy of the MODIS PWV product. The high-resolution fused PWV maps provide detailed spatial variations which are generally consistent with those from the MODIS estimates under confident clear conditions and ERA5. During the monsoon period from 2007‒2013, monthly average PWV estimates across the UBR basin vary from ~6 to ~12 mm, and for each month high PWV values are found mainly along the UBR valley and at the basin outlet. The developed data fusion approach maximizes the potential of satellite and reanalysis-based PWV products for monitoring PWV and can be extended to other data available sources and study regions. The generated PWV estimates are highly valuable in understanding the water and energy cycles and retrieving atmospheric and surface variables for the south TP and its downstream areas.