Assimilation of High-Resolution Soil Moisture Data Into an Integrated Terrestrial Model for a Small-Scale Head-Water Catchment

Land surface-subsurface modeling combined with data assimilation was applied on the Rollesbroich hillslope (Germany). Dense information from a soil moisture sensor network was assimilated with the ensemble Kalman filter applying different scenarios including the update of model states with or without updating of saturated soil hydraulic conductivity on an ensemble size of 128 (or 256) realizations with 3-D heterogeneous fields of Mualem-van Genuchten parameters. Simulations were also carried out with a synthetic test case mimicking the Rollesbroich site, to get more insight in the role of model structural errors. The combination of joint updating of model states and hydraulic conductivity was more efficient in updating the soil water content than state updating alone for the real-world case. On average, the root-mean-square error at the sensor locations was reduced by 14% if states and parameters were updated jointly, but discharge estimation was not improved significantly. Synthetic simulations showed much better results with an overall root-mean-square error reduction by 55% at independent verification locations in case of daily soil water content data assimilation including parameter estimation. Individual synthetic data assimilation scenarios with parameter estimation showed an increase of the Nash-Sutcliffe-Efficiency for discharge from −0.04 for the open loop run to 0.61. This shows that data assimilation in combination with high-resolution physically based models can strongly improve soil moisture and discharge estimation at the hillslope scale. Large performance differences between synthetic and real-world experiments indicated the limits of such an approach associated with model structural errors like errors in the prior geostatistical parameters.