Rescaling L-band Passive Microwave Satellite Soil moisture products for hydrological applications

Soil moisture at relatively high spatial resolution (e.g.1-10 km) provides information on the hydrological state of watershed antecedent to rainfall events and plays a critical role in a host of hydrological and agricultural applications. In contrast to in-situ measurements, which provide soil moisture information at a point scale, L-band passive microwave satellites such as Soil Moisture and Ocean Salinity (SMOS) and Soil Moisture Active Passive (SMAP) offer global estimates of soil moisture at spatial resolution of roughly 40 km. However, this resolution restricts their use for applications such as hydrological forecasting. Previous studies attempted to downscale satellite soil moisture products notably in arid and semi-arid regions, yet further studies are needed for a range of land surface and climatic conditions. In addition, subsequent use of downscaled soil moisture for hydrological applications is often overlooked. In this study, we aim to improve the spatial resolution of SMOS and SMAP soil moisture products and subsequently use to update hydrological models to better understand interactions between soil moisture, land cover and hydrological regime and for enhancing hydrological forecast skills. Existing downscaling algorithms (i.e. times series regression approach (SMAP baseline) and machine learning (random forest)) were refined and applied to two study areas of contrasting land cover and climatic conditions: the Susquehanna River Basin (SRB), USA; the Manicouagan River Basin (MRB), Canada, to downscale SMAP and SMOS soil moisture products using auxiliary information (e.g., backscatter, land surface temperature) obtained from Sentinel-1 and MODIS. Thereafter, a suite of synthetic experiments will be carried out by ingesting the downscaled soil moisture into a distributed hydrological model to evaluate its benefits in enhancing hydrological forecast skills (SRB) and to evaluate the impacts of forest fires on the hydrological regime of a watershed (MRB). Preliminary results suggest the capability of both downscaling approaches in capturing the spatio-temporal variation of soil moisture. Updating the hydrological model with downscaled soil moisture is among the key tasks to be considered in future works. The expected outcome will be beneficial for better water resources management.