%0 Thesis
%A Li, Fang
%T Assimilation of groundwater level and cosmic-ray neutron sensor soil moisture measurements into integrated terrestrial system models for better predictionst
%V 650
%I RWTH Aachen University
%V Dissertation
%C Jülich
%M FZJ-2024-06980
%@ 978-3-95806-796-7
%B Reihe Energie & Umwelt / Energy & Environment
%P xvii, 172
%D 2024
%Z Dissertation, RWTH Aachen University, 2024
%X Groundwater and soil moisture (SM) play a crucial role in the hydrological cycle, and therefore the dynamics of these two variables need to be accurately quantified on spatial and temporal scales. In situ observation networks can provide direct and accurate information on groundwater level (GWL) and SM. However, observations from observation networks are not sufficient to fully represent the Earth’s hydrological system without the help of models. Integrated models such as the Terrestrial System Modelling Platform (TSMP) can simulate the hydrological system from the subsurface to the atmosphere and accurately capture the full terrestrial hydrological cycle. Current model estimates of GWL and SM are highly uncertain due to data limitations and model uncertainties. The main sources of uncertainty are related to atmospheric forcings, model structural errors, and uncertain parameterization. Data assimilation (DA) can merge numerical models with observations, resulting in a correction of hydrological states and fluxes and improved parameter estimates. Different sources of uncertainty may lead to unsatisfactory simulations of groundwater hydrodynamics with hydrological models. The goal of first study is to investigate the impact of assimilating groundwater data into TSMP for improving  hydrological modelling in a real-world case. Daily groundwater table depth (WTD) measurements from the year 2018 for the Rur catchment in Germany were assimilated by the Localized Ensemble Kalman Filter (LEnKF) into TSMP. The LEnKF is used with a localization radius so that the assimilated measurements only update model states in a limited radius around the measurements, in order to avoid  unphysical updates related to spurious correlations. Due to the mismatch between groundwater measurements and the coarse model resolution (500 m), the measurements need careful screening before DA. Based on the spatial autocorrelation of the WTD deduced from the measurements, three different filter localization radii (2.5 km, 5 km and 10 km) were evaluated for assimilation. The bias in the simulated water table and the root mean square error (RMSE) are reduced after DA, compared with runs without DA (i.e., open loop (OL) runs). The best results at the assimilated locations are obtained for a localization radius of 10km, with an 81% reduction of RMSE at the measurement locations, and slightly smaller RMSE reductions for the 5 km and 2.5 km radius. The validation with independent WTD data showed the best results for a localization radius of 10 km, but groundwater table characterization could only be improved for sites less than 2.5 km from measurement locations. In case of a localization radius of 10km the RMSE-reduction was 30% for those nearby sites. Simulated soil moisture was validated against soil moisture measured by cosmic-ray neutron sensors (CRNS), cut no RMSE reduction was observed for DA-runs compared to OL-run. However, in both cases, the correlation between measured and simulated soil moisture content was high (between 0.70 and 0.89, except for the Wüstebach site). CRNS fill the gap between locally measured in situ SM and remotely sensed (RS) SM by providing accurate SM estimation at the field scale. This is promising for improving hydrologic model predictions, as CRNS can provide valuable information on SM in the root zone at the typical scale of a model grid cell. In a second study of this PhD-work, SM measurements from a network of 12 CRNS in the Rur catchment (Germany) were assimilated into TSMP to investigate its potential for improving SM, evapotranspiration (ET) and river discharge characterization and estimating soil hydraulic parameters at the larger catchment scale. DA experiments (with and without parameter estimation) were conducted in both a wet year (2016) and a dry year (2018) with the Ensemble Kalman Filter (EnKF), and later verified with an independent year (2017) without DA. The results show that SM characterization was significantly improved at measurement locations (with up to 60% RMSE reduction), and that joint state-parameter estimation improved SM simulation more than state estimation  one (more than 15% additional RMSE reduction). Jackknife experiments showed that SM at verification locations had lower and different improvements in the wet and dry years (an RMSE 2 reduction of 40% in 2016 and 16% in 2018). In addition, SM assimilation was found  to improve ET characterization to a much lesser extent, with a 15% RMSE reduction of monthly ET in the wet year and 9% in the dry year. In a third study, we tested different approaches for joint DA of observed SM data from CRNS and GWL data into the TSMP model. A experiments (with and without parameter estimation) were conducted with LEnKF for the Rur catchment in Germany for the years 2016-2018, followed by cross-validation (if parameters were estimated) in independent years. Univariate SM assimilation reduced the RMSE of SM over the assimilation locations by more than 50%. Univariate GWL assimilation reduced the monthly RMSE of GWL at assimilation locations by 70%. Within 5 km of the assimilated sites, GWL estimation was still obviously improved, with RMSE reductions 2-50%. However, the univariate assimilation of GWL degraded the characterization of SM, and the univariate assimilation of SM also diminished the simulation of GWL. A new multivariate DA approach that assimilates GWL and SM separately is proposed. GWL data are assimilated and used to estimate the interface between the unsaturated and saturated zones, and update the states (and possibly parameters) of the saturated zone. SM measurements are assimilated to update states of the unsaturated zone. In addition, observation specific localization is proposed. With multivariate DA, at the assimilation locations the estimates of variables (GWL, SM, and ET) are close to those in univariate assimilation. However, there were more than 15% RMSE reductions for GWL at 2.5~5 km validation locations compared to univariate assimilation. In addition, only SM assimilation (univariate or multivariate) improves very slightly ET estimates, with an overall RMSE reduction of 3%. Parameter updating reduced the RMSE of variable estimates by up to 17% compared to updating states alone. This work was carried out for the Rur catchment (2354 km²), which has a well-established monitoring infrastructure and considerable regional diversity in climate, soil types, and land use. In contrast to previously reported small-domain tests, which were primarily conducted in synthetic experiments or oversimplified real-world cases, the assimilation of real-world data at the larger catchment scale faces additional complexities and challenges. The effectiveness of DA can be limited by the uneven distribution of monitoring stations, coarse model resolution, and model structure errors. This thesis, using EnKF and its variants, proposes specific strategies for the assimilation of GWL and SM (from CRNS) separately or jointly in the integrated terrestrial model TSMP. Overall, the results of this thesis provide insights for improving the characterization of multiple variables and parameters (GWL, SM, ET, and saturated hydraulic conductivity (Ks)) by DA. Possible promising approaches for future improvement of DA performance in coupled models are: (i) improving the accuracy of terrestrial system modeling, including the addition of an atmospheric model, the inclusion of more detailed agro-ecological processes into land surface models and increasing model resolution; (ii) attempting to assimilate data from more diverse sources such as RS, unmanned aerial vehicles (UAVs), and small satellites to address the sparse distribution of in situ observations; and (iii) exploring advanced DA algorithms, potentially EnKF variants or hybrid methods with machine learning (ML) integration, to address the issues and challenges of multivariate assimilation at large scales.
%F PUB:(DE-HGF)3 ; PUB:(DE-HGF)11
%9 BookDissertation / PhD Thesis
%R 10.34734/FZJ-2024-06980
%U https://juser.fz-juelich.de/record/1034169