Improving Real-Time Flood Forecasting: Probabilistic Validation of Assimilated Remotely-Sensed Soil Moisture Data

Precise hydrological modeling is crucial for effective water resource management, forecasting future water supply, and preparing for extreme events. The devastating July 2021 flood event over western Germany and eastern Belgium has raised concerns about the capacity of current models to predict such previously unrecorded extreme events. These concerns are intensified by the increasing frequency of extreme events linked to global warming. Modeling uncertainties highlight challenges in forecasting such events. To address these challenges, prognostic modeling techniques can be improved by incorporating satellite-derived soil moisture (SM) estimations into hydrological models. This has the potential to minimize uncertainty in soil moisture and streamflow simulations. In this study, the integrated hydrological model ParFlow-CLM was enhanced using an Ensemble Kalman Filter (EnKF) data assimilation (DA) technique. We used Sentinel-1-SM and ESA CCI (European Space Agency Climate Change Initiative)-SM data to improve the estimation of soil moisture and streamflow. We conducted the DA experiment over Germany and nearby regions in July 2021, using ParFlow-CLM with high spatial (611 m) and temporal (hourly) resolution. The key results are as follows: CCI-SM DA is a robust approach for representing soil moisture dynamics, outperforming both Sentinel-1-SM DA and open loop simulations without DA in terms of responsiveness and accuracy. This makes CCI-SM DA a better fit for high-resolution studies that depend on event-driven soil water content changes. Discharge from CCI-SM DA simulations provides the closest match to observed peak discharges, Sentinel-1-SM DA tends to slightly underpredict peak discharge at several locations, however still performs better than the open loop simulation. This graphical abstract presents a framework to enhance flood forecasting by assimilating satellite-derived soil moisture (SM) data into a physics-based hydrological ParFlow-CLM model. The methodology is applied to a real flood event in Germany, demonstrating its operational relevance for improving early warning systems. The core of the framework is the ParFlow-CLM integrated hydrologic model, which captures key surface and subsurface processes, including overland flow, evapotranspiration, snow accumulation, and 3D groundwater-surface water interactions. The model is driven by hourly atmospheric forcing data such as precipitation (APCP), temperature (Temp), specific humidity (SPFH), pressure (Press), wind speed (UGRD/VGRD), and longwave/shortwave radiation (DLWR/DSWR), primarily sourced from the ECMWF HRES medium-range forecast products. To improve model accuracy, soil moisture observations from ESA CCI (European Space Agency Climate Change Initiative), and Sentinel-1 are assimilated using the Ensemble Kalman Filter (EnKF). This sequential data assimilation method updates the model state by combining model forecasts with observational data, accounting for uncertainties in both. The EnKF propagates an ensemble of model states to estimate error statistics and correct the state variables based on incoming observations. Model outputs, including soil water content and streamflow, are evaluated against observations using deterministic and probabilistic performance metrics, such as relative error, RMSE, correlation coefficient, and the first order reliability method (FORM). The results show that incorporating SM data via EnKF significantly enhances the model’s ability to predict flood peaks and soil moisture dynamics. This integrated approach is especially beneficial in flat or low-gradient terrains, where antecedent soil moisture conditions strongly influence runoff generation.