The use of soil moisture - remote sensing products for large-scale groundwater modeling and assessment

Abstract

In this thesis, the possibilities of using spaceborne remote sensing for large-scale groundwater modeling are explored. We focus on a soil moisture product called European Remote Sensing Soil Water Index (ERS SWI, Wagner et al., 1999) - representing the upper profile soil moisture. As a test-bed, we used the Rhine-Meuse basin, covering about 200000 km2 and having more than four thousand in-situ groundwater head observations. The thesis explores the potential of using SWI in an empirical transfer function-noise (TFN) model and in a physically-based model PCR-GLOBWB-MOD. We first show that there is correlation between groundwater head and SWI dynamics, which is apparent mostly for shallow groundwater areas. For deep groundwater areas, the correlation may become apparent if delay time is considered. Given such correlation, head prediction based on SWI time series should be feasible. Hence, we performed two exercises in which SWI time series were used as TFN model input. For the first exercise - focusing on temporal forecasting, the parameters were calibrated based on head measurements in the period 1995-2000. Then, the forecasts were validated in the period 2004-2007. In the second exercise - focusing on spatio-temporal prediction, model parameters were predicted by using a digital elevation map. Using these estimated parameters, spatio-temporal head prediction was created. Both exercises show that observed head dynamics can be well simulated, especially for shallow groundwater where head fluctuations are dominated by meteorological forcing, which is reflected in soil moisture dynamics. In this thesis, we also introduce a physically-based and coupled groundwater-land surface model PCR-GLOBWB-MOD (1 km resolution), built by using only globally available datasets. We started building it by modifying PCR-GLOBWB land surface model (van Beek et al., 2011) and then performing its daily simulation to estimate groundwater recharge and river discharge. Subsequently, a MODFLOW groundwater model (McDonald and Harbaugh, 1988) was created and forced by the recharge and water levels calculated by the land surface model. Results are promising despite the fact that an offline coupling procedure was still used (i.e. both models were separately and sequentially simulated). The model simulation can reproduce the observed discharge and groundwater head reasonably well. We also introduce the online-coupled version of PCR-GLOBWB-MOD including a two-way feedback between surface water and groundwater dynamics and between groundwater and upper soil stores, enabling groundwater to sustain upper soil moisture states and fulfil evaporation demand (during dry conditions). For this online coupled model, we explored the possibility of using SWI to calibrate it by performing more than three thousand runs with various parameter sets and evaluating their results against discharge, SWI and head measurements. From these runs, we conclude that SWI can be used for calibrating upper soil saturated hydraulic conductivity, affecting groundwater recharge. However, it is difficult to calibrate the model by using SWI only. Discharge data should be included to resolve equifinality problems of fitting soil moisture dynamics and to constrain aquifer transmissivities and runoff-infiltration partitioning. Moreover, head measurements are important to capture finer resolution heterogeneity, which cannot be captured by current resolution of spaceborne soil moisture products (50 km)