Calibrating a large-extent high-resolution coupled groundwater-land surface model using soil moisture and discharge data

We explore the possibility of using remotely sensed soil moisture data and in situ discharge observations to calibrate a large-extent hydrological model. The model used is PCR-GLOBWB-MOD, which is a physically based and fully coupled groundwater-land surface model operating at a daily basis and having a resolution of 30 arc sec (about 1 km at the equator). As a test bed, we use the combined Rhine-Meuse basin (total area: about 200,000 km(2)), where there are 4250 point-scale observed groundwater head time series that are used to verify the model results. Calibration is performed by simulating 3045 model runs with varying parameter values affecting groundwater head dynamics. The simulation results of all runs are evaluated against the remotely sensed soil moisture time series of SWI (Soil Water Index) and field discharge data. The former is derived from European Remote Sensing scatterometers and provides estimates of the first meter profile soil moisture content at 30 arc min resolution (50 km at the equator). From the evaluation of these runs, we then introduce a stepwise calibration approach that considers stream discharge first, then soil moisture, and finally verify the resulting simulation to groundwater head observations. Our results indicate that the remotely sensed soil moisture data can be used for the calibration of upper soil hydraulic conductivities determining simulated groundwater recharge of the model. However, discharge data should be included to obtain full calibration of the coupled model, specifically to constrain aquifer transmissivities and runoff-infiltration partitioning processes. The stepwise approach introduced in this study, using both discharge and soil moisture data, can calibrate both discharge and soil moisture, as well as predicting groundwater head dynamics with acceptable accuracy. As our approach to parameterize and calibrate the model uses globally available data sets only, it opens up the possibility to set up large-extent coupled groundwater-land surface models in other basins or even globally.