Can ASCAT-derived soil wetness indices reduce predictive uncertainty in well-gauged areas? A comparison with in situ observed soil moisture in an assimilation application

Although soil moisture is a key variable controlling the hydrological response of a catchment to rainfall events, the utility of Earth Observation products for soil moisture monitoring in hydrological applications remains controversial. It is not clear under which hydrological modeling scheme remote sensing may have a positive impact on the runoff forecasts and to what degree the practical utility of these data suffers from limitations related to their uncertainty, as well as to their spatial and temporal resolution. More specifically, there is limited understanding of whether remotely sensed soil moisture data can improve forecasts in well gauged catchments, or if their utility is restricted to poorly gauged areas. This paper contrasts the use of space-based and in situ based soil moisture monitoring in a data assimilation exercise in the Bibeschbach experimental catchment in Luxembourg. Bi-daily soil wetness indices obtained from ASCAT METOP-A satellite data are used as proxies of soil water storage and have been employed to periodically update the water budget of a lumped conceptual hydrological model. The hydrologic model was specifically developed and calibrated to represent catchment characteristics in terms of observed run-off and soil moisture conditions. Nevertheless, the assimilation of in situ soil moisture measurements using a Particle Filter-based data assimilation approach even further improved both discharge and soil wetness forecasts, indicating that continuously recorded in situ measurements, even if taken only over a few points within the catchment, are useful for updating model states. On the other hand, the assimilation of the remotely sensed soil moisture data resulted in a negative or only small positive impact. This suggests that for small and well-instrumented catchments, where well-calibrated “à-la-carte” models are available, coarse-resolution remotely sensed soil moisture data add little or no extra value for runoff prediction. It remains an open research question if this result can mainly be attributed to errors in the ASCAT-based profile soil moisture estimates, or if it is mainly related to the stringent hydrological modeling scheme as used in this study. We further illustrate that the efficiency of the approach varies seasonally, with soil moisture recordings being particularly useful for improved flood predictions during transition periods from wet to dry in early spring and from dry to wet in early autumn.

► We introduce a conceptual framework for integrating soil moisture into a hydrological model.
► The data assimilation makes use of the particle filtering technique.
► In well-gauged catchments remotely sensed soil moisture hardly impact runoff prediction.
► The assimilation of in situ observations improves soil moisture and discharge forecasts.
► Assimilation is most efficient in transition periods between wet and dry conditions.