High-resolution groundwater storage changes from GRACE/GRACE-FO using assimilation models and hydrogeological observations

Using assimilation models and hydrogeological observations, a team from the Polish Geological Institute and the Space Research Center of the Polish Academy of Sciences has developed a statistical approach to downscaling GRACE/GRACE-FO data, increasing the spatial resolution to a 0.1° grid. The described approach of downscaling TWS data from GRACE/GRACE-FO was successfully applied to Poland for an area of slightly more than 310 000 km².

Dr. hab. Prof. Tatiana Solovey, Rafał Janica, Agnieszka Brzezińska – Polish Geological Institute - National Research Institute / Dr. Justyna Śliwińska-Bronowicz, Space Research Centre of Polish Academy of Sciences (Warsaw, Poland)

The GRACE/GRACE-FO missions provided the first time global insights into cumulative changes in water storage in snow, ice, lakes, rivers, soil moisture, and underground aquifers. This achievement marked a breakthrough in monitoring the terrestrial part of the water cycle. Total water storage change (∆TWS) measured by GRACE/GRACE-FO effectively closes the water balance, which reflects the equilibrium between precipitation and water losses due to evapotranspiration, river runoff, and changes in terrestrial water storage. This last component of the water balance is crucial for understanding the effects of climate change and is recognized by the Global Climate Observing System (GCOS) as a Essential Climate Variable. Currently, only the GRACE/GRACE-FO missions can provide direct information about TWS. However, the coarse spatial resolution of GRACE/GRACE-FO products limits their usefulness for monitoring the water cycle at regional scales.

Inspired by the latest achievements in developing assimilation models that incorporate hydrogeological observations, a team at the Polish Geological Institute and the Space Research Centre of the Polish Academy of Sciences developed a statistical approach to downscale GRACE/GRACE-FO data by enhancing the spatial resolution to a 0.1° grid. The new method is based on a multivariate regression model that integrates several components of the water balance: ∆TWS from GRACE/GRACE-FO, precipitation (P) from a daily gridded observational dataset (E-OBS), evapotranspiration (ET) from Moderate Resolution Imaging Spectroradiometer (MODIS), river runoff (Q) from observational stations, as well as ∆TWS calculated from the water balance approach (i.e., ∆TWS=P‒ET‒Q). The novel downscaling method identifies common spatiotemporal dependencies among the variables used through least squares regression, and subsequently employs these dependencies to redistribute the GRACE/GRACE-FO observed TWS change at a spatial resolution of 0.1°. The described approach of downscaling TWS data from GRACE/GRACE-FO was successfully applied in Poland for an area of just over 310,000 km². The downscaled TWS was then used to extract groundwater storage (GWS) by removing water storage in other reservoirs, utilizing model data from Global Land Data Assimilation System (GLDAS). This method was specifically tested in the transboundary Bug River basin on the border of Poland, Ukraine and Belarus (see banner picture).

Thanks to the extensive network of groundwater monitoring wells (approximately 1,500 stations) in Poland, it was possible to capture spatiotemporal relationships between (i) GWS determined from GRACE and GLDAS, as well as (ii) GWS from in-situ measurements. These relationships correspond to the division of the study area into specific hydrodynamic zones—recharge and discharge areas—which are characterized by different factors controlling groundwater dynamics (Fig. 2).

In recharge zones located mostly in areas near watersheds, the dynamics of GWS is rather stable, shaped by the long-term accumulation of the effects of precipitation and evapotranspiration. Seasonal variability is less pronounced here and shows a significant delay (over 4 months in the study area) in GWS response to weather conditions. Therefore, in this zone, calculating GWS by removing fast seasonal water storage variations from GRACE-based TWS using TWS obtained from GLDAS is justified and supported by a high correlation with GWS estimates based on in-situ groundwater level measurements.

In turn, in the discharge zone, primarily encompassing valley areas, GWS variability exhibits a strong seasonal cycle due to rapid influx from surface runoff and accumulation in the soil. Therefore, a different approach for estimating GWS is applied here, without removing the component of the rapid water influx. In this case, we propose using the Eckhardt filter, commonly utilized in hydrology (Pozdniakov et al., 2022), to separate rapid baseflow (delayed land and soil flow) from the time series of total river discharge. Using this filter and selecting appropriate hydrogeological parameters, the GWS component can be successfully extracted from GRACE-based TWS. The proposed method contributes to a more accurate estimation of GWS, taking into account the complexities of hydrogeological conditions at a spatial resolution of 0.1°. This approach also allows for a higher consistency between in-situ GWS data, and GWS obtained from GRACE data and hydrological models, which was previously unsatisfactory for the area of Poland (Śliwińska et al. 2019).

• Bertrand, G., Goldscheider, N., Gobat, J. M. et al. (2012). Review: From multi-scale conceptualization to a classification system for inland groundwater-dependent ecosystems. Hydrogeol J 20, 5–25. https://doi.org/10.1007/s10040-011-0791-5

• Pozdniakov, S. P., Wang, P., Grinevsky, S. O., Frolova, N. L. (2022). A physically based model of a two-pass digital filter for separating groundwater runoff from streamflow time series. Water Resources Research, 58, e2021WR031333. https://doi.org/10.1029/2021WR031333 

• Śliwińska, J., Birylo, M., Rzepecka, Z., Nastula, J. (2019). Analysis of groundwater and total water storage changes in Poland using GRACE observations, in-situ data, and various assimilation and climate models. Remote Sensing, 11(24). doi.org/10.3390/rs11242949