Scaling methods of leakage correction in GRACE mass change estimates revisited for the complex hydro-climatic setting of the Indus Basin

Abstract

Abstract. Total water storage change (TWSC) reflects the balance of all water fluxes in a hydrological system. The Gravity Recovery and Climate Experiment/Follow-On (GRACE/GRACE-FO) monthly gravity field models, distributed as spherical harmonic (SH) coefficients, are the only means of observing this state variable. The well-known correlated noise in these observations requires filtering, which scatters the actual mass changes from their true locations. This effect is known as leakage. This study explores the traditional basin and grid scaling approaches, and develops a novel frequency-dependent scaling for leakage correction of GRACE TWSC in a unique, basin-specific assessment for the Indus Basin. We harness the characteristics of significant heterogeneity in the Indus Basin due to climate and human-induced changes to study the physical nature of these scaling schemes. The most recent WaterGAP (Water Global Assessment and Prognosis) hydrology model (WGHM v2.2d) with its two variants, standard (without glacier mass changes) and Integrated (with glacier mass changes), is used to derive scaling factors. For the first time, we explicitly show the effect of inclusion or exclusion of glacier mass changes in the model on the gridded scaling factors. The inferences were validated in a detailed simulation environment designed using WGHM fields corrupted with GRACE-like errors using full monthly error covariance matrices. We find that frequency-dependent scaling outperforms both basin and grid scaling for the Indus Basin, where mass changes of different frequencies are localized. Grid scaling can resolve trends from glacier mass loss and groundwater loss but fails to recover the small seasonal signals in trunk Indus. Frequency-dependent scaling can provide a robust estimate of the seasonal cycle of TWSC for practical applications such as regional-scale water availability assessments. Apart from these novel developments and insights into the traditional scaling approach, our study encourages the regional scale users to conduct specific assessments for their basin of interest.