A Water Storage Reanalysis over the European Continent: Assimilating GRACE Data into a High-Resolution Hydrological Model
Abstract
Hydrological models are useful in the assessment of e.g. water resources and long-term climate variability. Because models are built on a simplified representation of reality, simulation results are subject to large uncertainties leading to limited predicting skills. A common approach for reducing these uncertainties is the assimilation of observational data from remote sensing and in-situ measurements. In this study, we assimilated GRACE total water storage observations into the Community Land Model version 3.5 (CLM3.5) over the European CORDEX domain, utilizing the Parallel Data Assimilation Framework (PDAF), in order to create an improved high-resolution regional reanalysis of terrestrial water storage. In the assimilation, the disparate spatial scales of the hydrological model and GRACE constitute a major challenge. While our CLM implementation simulates the evolution of water storage and flux for up to 27 compartments at hourly time steps, on a 12.5 km grid, GRACE monitors monthly and vertically aggregated TWS at spatial scales of a few hundred kilometers. In addition, TWS estimates from GRACE are contaminated with correlated noise, which cannot be neglected in the context of data assimilation.We study the effects from assimilating GRACE data into CLM at different spatial scales, and evaluate the experiments with independent data sets including soil moisture, discharge, and vertical deformation from continuously-operating GPS sites.The validation shows that the number of in-situ soil moisture stations with correlations larger than 0.3 increased from 7% for the open-loop run to 35% for the assimilation run. An even larger improvement was found for discharge from 21 gauging stations. While correlation coefficients of the open-loop run were distributed between -0.3 and 0.3, for the assimilation run 50% of the correlation coefficients were above 0.5. Moreover, the root mean square scatter of daily time series of vertical land motion showed an increased reduction ≥ 10% when subtracting the assimilated model instead of GRACE observations or the open-loop run.In the next step GRACE data will be assimilated into the fully coupled Terrestrial System Modeling Platform, TerrSysMP, which includes CLM as a land surface model and closes the terrestrial water and energy cycles from groundwater into the atmosphere.
- Publication:
-
AGU Fall Meeting Abstracts
- Pub Date:
- December 2018
- Bibcode:
- 2018AGUFM.H51M1463K
- Keywords:
-
- 1836 Hydrological cycles and budgets;
- HYDROLOGYDE: 1847 Modeling;
- HYDROLOGYDE: 1855 Remote sensing;
- HYDROLOGYDE: 1873 Uncertainty assessment;
- HYDROLOGY