Improving the Resolution of GRACE Data for Spatio-Temporal Groundwater Storage Assessment

Abstract

Groundwater has a significant contribution to water storage and is considered to be one of the sources for agricultural irrigation; industrial; and domestic water use. The Gravity Recovery and Climate Experiment (GRACE) satellite provides a unique opportunity to evaluate terrestrial water storage (TWS) and groundwater storage (GWS) at a large spatial scale. However; the coarse resolution of GRACE limits its ability to investigate the water storage change at a small scale. It is; therefore; needed to improve the resolution of GRACE data at a spatial scale applicable for regional-level studies. In this study; a machine-learning-based downscaling random forest model (RFM) and artificial neural network (ANN) model were developed to downscale GRACE data (TWS and GWS) from 1° to a higher resolution (0.25°). The spatial maps of downscaled TWS and GWS were generated over the Indus basin irrigation system (IBIS). Variations in TWS of GRACE in combination with geospatial variables; including digital elevation model (DEM), slope; aspect; and hydrological variables; including soil moisture; evapotranspiration; rainfall; surface runoff; canopy water; and temperature; were used. The geospatial and hydrological variables could potentially contribute to; or correlate with; GRACE TWS. The RFM outperformed the ANN model and results show Pearson correlation coefficient (R) (0.97), root mean square error (RMSE) (11.83 mm), mean absolute error (MAE) (7.71 mm), and Nash–Sutcliffe efficiency (NSE) (0.94) while comparing with the training dataset from 2003 to 2016. These results indicate the suitability of RFM to downscale GRACE data at a regional scale. The downscaled GWS data were analyzed; and we observed that the region has lost GWS of about −9.54 ± 1.27 km3 at the rate of −0.68 ± 0.09 km3/year from 2003 to 2016. The validation results showed that R between downscaled GWS and observational wells GWS are 0.67 and 0.77 at seasonal and annual scales with a confidence level of 95%, respectively. It can; therefore; be concluded that the RFM has the potential to downscale GRACE data at a spatial scale suitable to predict GWS at regional scales.

Evaluating downscaling methods of GRACE data: a case study over a fractured crystalline aquifer in South India

“…The data-based downscaling approach consists in (i) deriving a statistical model of TWS from ancillary data available at high resolution (HR), (ii) calibrating it at low resolution (LR), (iii) applying it at HR, and (iv) removing the contribution of surface and soil moisture water stocks to isolate GWS. This data-driven approach rests on the hypothesis that the hydrological/physical processes that link those variables are identical at all resolutions (Ali et al, 2021;Jyolsna et al, 2021;Karunakalage et al, 2021;Sahour et al, 2020;Seyoum and Milewski, 2017;Vishwakarma et al, 2021;Zhang et al, 2021a, b). In the literature, data-driven methods have been used to downscale GRACE data at various scales, either at the watershed scale for a thematic approach as in Seyoum and Milewski (2017) (5,000 km 2 to 20,000 km 2 ), or grid-based.…”

“…In the literature, data-driven methods have been used to downscale GRACE data at various scales, either at the watershed scale for a thematic approach as in Seyoum and Milewski (2017) (5,000 km 2 to 20,000 km 2 ), or grid-based. The downscaling resolution for data-based techniques is often limited by the coarsest resolution among the predictors : rainfall from the Tropical Rainfall Measuring Mission (TRMM) or model outputs from the NASA's Global Land Data Assimilation System (GLDAS) at 0.25 • (Ali et al, 2021;Jyolsna et al, 2021;Ning et al, 2014;Seyoum et al, 2019;Zhang et al, 2021a), the NASA's North American Land Data Assimilation System (NLDAS) model at 0.125 • (Sahour et al, 2020), the Ecological Assimilation of Land and Climate Observations (EALCO) model at 5 km (Zhong et al, 2021) or evapotranspiration from the Moderate Resolution Imaging Spectroradiometer (MODIS) at 2 km (Yin et al, 2018). Even finer resolution can be targeted when using interpolation based methods, up to the kilometer (Zhang et al, 2021a;Zuo et al, 2021).…”

“…Note that the GWS simulated by models was occasionally used as reference (Houborg et al, 2012;Seyoum and Milewski, 2017). In most studies, the quality of the downscaled GWS data set is evaluated by comparing its time series with that of GWL or GWS derived from in situ measurements for each HR unit (spatialized -HR pixel -or localized -observation well) with one or several metrics including coefficient of determination (R 2 ) or Nash-Sutcliffe efficiency coefficient (NSE), Pearson correlation coefficient (R), root mean squared error (RMSE) and mean absolute error (MAE) (Ali et al, 2021;Jyolsna et al, 2021;Karunakalage et al, 2021;Sahour et al, 2020;Yin et al, 2018;Zhang et al, 2021b, a;Zuo et al, 2021). In those studies, the downscaling procedure is considered efficient if those metrics fall within an acceptable range, or if the downscaled product qualitatively restitutes the long-term trends of in situ data.…”

“…In particular, scaling factors are not expected to efficiently downscale GRACE TWS data as neighbouring pixels are highly dependent (Landerer and Swenson, 2012). Yet, many studies directly use GRACE harmonics solutions at 1 • resolution (Ali et al, 2021;Jyolsna et al, 2021;Karunakalage et al, 2021;Ning et al, 2014;Seyoum et al, 2019;Yin et al, 2018;Zhang et al, 2021b, a;Zuo et al, 2021) or mascons solution at 0.5 • resolution (Karunakalage et al, 2021;Nie et al, 2019;Tian et al, 2017) as LR input data, which is way finer than their actual resolution. There is no study evaluating the uncertainty in downscaled GRACE data associated with the above assumption i.e.…”

Accuracy Comparison and Assessment of DSM Derived from GFDM Satellite and GF-7 Satellite Imagery

“…However, as a basic topographic data, the accuracy of DSM affects its popularization and application in various fields [21][22][23][24]. Due to different imaging configurations and data processing methods, DSM generated by GFDM or GF-7 satellite images contains various errors.…”