Statistical Downscaling of GRACE‐Derived Groundwater Storage Using ET Data in the North China Plain

Abstract

Due to coarse spatial resolution, the application of Gravity Recovery and Climate Experiment (GRACE) data in local groundwater resource management has been limited. To overcome this issue, a downscaling approach is presented to improve the spatial resolution of GRACE‐derived groundwater storage anomalies using evapotranspiration (ET) data. The statistical downscaling method is only applied in areas where there is a strong relationship between GRACE‐derived groundwater storage (GWS) and ET, and the relationship can be established under different spatial resolutions. In this study, GWS anomalies are downscaled from 110 km to 2 km in the North China Plain, capturing subgrid heterogeneity in groundwater storage changes. The downscaled results are further validated using data from 111 observation wells from 2003 to 2014 in the Beijing Plain. The results show that the in situ groundwater levels agree with the Gravity Recovery and Climate Experiment (GRACE) downscaled results for the Beijing Plain in both interannual and monthly scales, with the correlation coefficient of 0.94 and 0.80, respectively. The correlation coefficients vary from 0.63 to 0.78, in 10 of the 13 Beijing Plain administrative regions. The regional downscaling approach employed in this study may be readily integrated into local water resources planning activities.

An Improved GPS-Inferred Seasonal Terrestrial Water Storage Using Terrain-Corrected Vertical Crustal Displacements Constrained by GRACE

“…Since different regions around the globe exhibit different hydro-climatic and topographic conditions, further experiments in regions with relatively small seasonal TWS variations and various terrain types should be conducted to confirm the feasibility of the presented methodology. Furthermore, we anticipate that the presented methodology can be ultimately applied as a model-based downscaling tool for a higher spatial resolution of TWS, while comparing it to recently employed statistical downscaling [68,69].…”

Estimating High-Resolution Groundwater Storage from GRACE: A Random Forest Approach

“…The performance of the RF model mostly depends on selecting appropriate biophysical variables that are highly correlated with GRACE. Precipitation, temperature, DEM, Natural Resources Conservation Service soil data [2,49], evapotranspiration [50], and runoff [51] were used for downscaling GRACE data. In this work, different predictors, i.e., aspect, BFGR, DEM, PCSW, ET, HF, precipitation, RZSM, slope, SM, SSR, SWE, SP, temperature, and WS, were used to develop RF, as those are strongly co-related with groundwater.…”

“…Although Seyoum and Milewski [55] have considered the lag effect of predicting variables, the ANN model shows poor performance in terms of its Pearson correlation and exhibits more variability than observed GWSA. On the other hand, the simple statistical approach [50,51] showed comparatively better results. The output of the currently proposed RF model shows an error of less than 8 mm even in extreme hydrologic scenarios.…”

Reconstructing Terrestrial Water Storage Variations from 1980 to 2015 in the Beishan Area of China

“…As a critical state variable in the hydrological cycle, terrestrial water storage (TWS) integrates surface water storage (SWS, including canopy interception, reservoirs, wetlands and lakes, rivers, and snow water equivalent), soil moisture storage (SMS), and groundwater storage (GWS) [1]. TWS changes also reflect changes in accumulated precipitation, evapotranspiration, and surface and subsurface runoff within a given area or basin [2].…”