A Physical Method of Estimating Water Fraction by Combining SMAP, Sentinel-1, and Landsat Measurements

Water fractional cover of the land surface is a key variable to study global water cycle and extremes. Several satellite microwave radiometers provide daily to sub-daily global brightness measurements at deca-kilometer scale which provide the opportunity to estimate global water fraction at daily scale. However, using the brightness temperature (T_B) alone, existing estimating methods exhibit large uncertainty coming from empirically determined specific emissivity (the emissivity of land and water) or the assumption of the equal physical temperature of close pixels. In this study, we remove all these assumptions and empirical values, then physically solve for the specific emissivity for each downscaled SMAP pixel at 9 km resolution. Polarized ratio index (PRI) is computed using T_B at both polarizations to cancel the unknown physical temperature and is finally parameterized by water fractions and specific emissivity. The water fraction dynamics in past flood events and nominal situation is derived by upscaling high-resolution water masks including the Sentinel-1 derived inundation maps using the Radar Produced Inundation Diary (RAPID) system and the Landsat derived USGS Dynamic Surface Water Extent (DSWE) product. For the first time, the estimated water fraction shows strong agreement (see the figures) with the Sentinel-1 derived water fraction at daily scale along at the confluence of Illinois and Kentucky Rivers. The proposed approach has the potential to provide global inundation dynamics or to be used as a start point for downscaling.