Research papersValidation of GRACE based groundwater storage anomaly using in-situ groundwater level measurements in India

•First time in-situ validation of two GRACE-products, spherical harmonics and mascon.•Comparison of GRACE-based and in-situ data in 12 major river basins in India.•High resolution specific yield database generated for Indian region.•Groundwater storage from GRACE-mascon data show best match with observed data.

In this study, we tried to validate groundwater storage (GWS) anomaly obtained from a combination of GRACE and land-surface model based estimates, for the first time, with GWS anomaly obtained from a dense network of in-situ groundwater observation wells within 12 major river basins in India. We used seasonal data from >15,000 groundwater observation wells between 2005 and 2013, distributed all over the country. Two recently released GRACE products, RL05 spherical harmonics (SH) and RL05 mascon (MS) products are used for comparison with in-situ data. To our knowledge, this is the first study of comparing the performance of two independent GRACE products at a sub-continental scale. Also for the first time, we have created a high resolution (0.10 × 0.10) map of specific yield for the entire country that was used for calculating GWS. Observed GWS anomalies have been computed using water level anomalies and specific yield information for the locale of individual observation wells that are up-scaled to basin-scale in order to compare with GRACE-based estimates. In general GRACE-based estimates match well (on the basis of the statistical analyses performed in the study) with observed estimates in most of the river basins. On comparing with observed GWS anomaly, GRACE-SH estimates match well in terms of RMSE, while GRACE-MS estimates show better association in terms of correlation, while the output of skewness, kurtosis, coefficient of variation (CV) and scatter analyses remain inconclusive for inter-comparison between two GRACE estimates. We used a non-parametric trend estimation approach, the Hodrick-Prescott (HP) filter, to further assess the performance of the two GRACE estimates. GRACE-MS estimates clearly outperform GRACE-SH estimates for reproducing observed GWS anomaly trends with significantly (>95% confidence level) strong association in 10 out of 12 basins for GRACE-MS estimates, on the other hand, GRACE-SH estimates show significantly (>95% confidence level) strong association in 6 out of 12 basins. On the basis of the study output, we recommend using GRACE-MS estimates for groundwater studies over the region and other regions of the globe with similar climatic, hydrogeologic or groundwater withdrawal conditions.