Reconstructing groundwater storage variations from GRACE observations using a new Gaussian-Han-Fan (GHF) smoothing approach

Abstract

This study focuses upon development of a new filter, referred to as Gaussian-Han-Fan (GHF) filtering, and its application within a comprehensive procedure for estimating groundwater changes. The Canadian Prairies was the study area. Variations in groundwater were estimated by using 15 years of Gravity Recovery And Climate Experiment (GRACE) twin-satellite observations (April 2002 to June 2017). Surface water storage (sum of soil moisture, snow water equivalent, canopy water, and surface water bodies) was subtracted from reconstructed GRACE-based Terrestrial Water Storage (TWS) changes through GHF filtering. To estimate the required hydrological parameters, both the Global Land Data Assimilation System (GLDAS) and Water Global Hydrology Model (WGHM) were used and evaluated. Water level changes for major surface water bodies were estimated using satellite altimetry-based products. The monthly average of GWS variations over the Prairies ranged between −200 mm and +230 mm. A positive trend was found for both TWS and GWS variations, with the highest values in the region surrounding Hudson Bay, particularly in northern Manitoba (about 55 mm/year). Estimated GWS anomalies error was equivalent to about 10% of its absolute value, with a mean of 19 mm. GWS variations results were validated using 116 active in-situ groundwater level measurements in five different river basins (Peace-Athabasca, Churchill, North Saskatchewan, South Saskatchewan, Missouri), which were all located in Alberta (Canada). Good agreement was achieved in each river basin (correlation > r = |0.70|, P < 10−4, RMSE < 55 mm). Regardless of hydrological system (GLDAS or WGHM), better statistical metrics were found when the average of the five basins was considered (r > |0.90|, P < 10−4), with lowest errors (RMSE or UnRMSE < 30 mm).