Data Assimilation of Remotely Sensed Soil Moisture in Hydrological Modeling to Improve Flood Forecasting

A project is being proposed which will investigate multiple issues that arise when assimilating remotely sensed soil moisture data into land surface models for flood forecasting. The first issue is estimating subsurface soil moisture from satellite derived surface soil moisture measurements. Although root zone soil moisture is more important for streamflow modeling, satellites can estimate soil moisture only for a thin uppermost soil layer. This project will compare between different methods of estimating subsurface soil moisture from satellite derived surface soil moisture measurements and investigate how these differences in subsurface soil moisture generation methods may impact streamflow modeling performance. The second issue is the spatial interpolation/extrapolation of satellite soil moisture data over an entire watershed. Soil moisture data from a satellite may not be available over the whole of a watershed of interest during one select pass over that watershed, due of reasons like dense vegetation, radio frequency interference or simply parts of that watershed being out of the satellite’s swath during that pass, causing spatial data discontinuity. This project will investigate whether it is better to interpolate/extrapolate satellite derived soil moisture data over a study watershed before assimilating the interpolated/extrapolated soil moisture in a flood forecasting model. Also to be investigated is which method of spatial interpolation is most suited for remotely sensed soil moisture. The additional benefits of using artificial intelligence-based methods will be assessed for the abovementioned tasks of subsurface soil moisture estimation and spatial interpolation from satellite derived soil moisture data. Finally, data assimilation experiments will be performed using different satellite-based soil moisture products both individually and jointly. The study area for this project will be the Au Saumon watershed in Quebec and the Susquehanna watershed New York-Pennsylvania-Maryland. The contrasting nature of these two watersheds will help address different challenges faced during data assimilation in watersheds having differing physical characteristics.