Beyond GRACE: Benchmarking Global Hydrological Models Back to the 1990s with Satellite Laser Ranging

Abstract

GRACE and GRACE Follow-On (GRACE-FO) provide the primary satellite gravimetry observations of terrestrial water storage (TWS) variations from 2002 onwards. However, continuously improving global hydrological models often span substantially longer periods, while their validation in the pre-GRACE era remains challenging and requires complementary observations or long-term gravity field reconstructions. The validation can be conducted by combining information from GRACE with other geodetic techniques, such as Satellite Laser Ranging (SLR) and Doppler Orbitography and Radiopositioning Integrated by Satellite (DORIS). An important question, however, is to what extent hydrological model assessments depend on the choice of long-term gravity field reconstruction used as a benchmark.To address this question, we use our recently developed long-term SLR+GRACE gravity field reconstruction, which extends GRACE information back to 1995 by combining the complementary strengths of SLR and GRACE. The lowest-degree spherical harmonics are constrained directly by SLR, while intermediate and higher degrees incorporate GRACE-derived temporal and spatial information to provide gravity field solutions up to degree and order 60 also for the pre-GRACE period. We compare this reconstruction with an independently developed SLR+DORIS gravity field solution over the common period 1995–2022. Their applicability as long-term hydrological benchmarks is evaluated for the 100 largest river basins using TWS estimates from OS LISFLOOD and the Land Surface Discharge Model (LSDM), which has been widely used in geodetic applications for nearly two decades. Model benchmark agreement is quantified using the centered Kling–Gupta Efficiency (cKGE) and Nash–Sutcliffe Efficiency (NSE). In addition, differences in NSE between OS LISFLOOD and LSDM (ΔNSE) are compared between the SLR+GRACE and SLR+DORIS benchmarks to assess whether both gravity field reconstructions lead to the same qualitative ranking of the hydrological models.The results show that OS LISFLOOD is generally more consistent with the gravimetric benchmarks than LSDM, particularly in large tropical and subtropical basins. For the SLR+GRACE reconstruction, median cKGE values are 0.59 for OS LISFLOOD and 0.50 for LSDM for detrended TWS, and 0.41 and 0.23, respectively, for interannual variability. More importantly, SLR+GRACE and SLR+DORIS yield the same sign of ΔNSE in 70.7% of cases for detrended TWS and 64.3% for interannual variability. The median absolute difference in ΔNSE between the two benchmarks increases from 0.12 to 0.20, indicating that model assessment is largely robust to benchmark choice, but becomes more sensitive at interannual timescales.In addition, degree-1 coefficients and the corresponding geocenter motion were reconstructed from the SLR+GRACE solution for 1995–2023. This extends the reconstructed geocenter series into the pre-GRACE era, fills the gap between GRACE and GRACE-FO, and complements our long-term gravity field reconstruction with degree-1 information.