Hydrologic Representation Complexity in Land Surface Models Influences GRACE Signal Restoration in Forward Modeling Approaches

Modern advancements in remote sensing provide a direct estimat e of mesoscale changes in Earth's terrestrial water storage, a major breakthrough in the nascent field of continental hydrology. The Gravity Recovery and Climate Experiment (GRACE) measures fluctuations in Earth's gravity attributable to mass flux, or changes in total water storage (TWS). However, GRACE TWS estimates are not without error and are limited in scale, with mascon products ranging in lateral resolution from 3° to 0.5°. Noise reduction and further filtering required in processing spherical harmonic solutions result in some signal attenuation; this can be a critical problem, especially for seasonal signals and in basins smaller than the GRACE lateral footprint (approximately 200,000 km²). The most common signal restoration method typically involves bias and leakage corrections in combination with synthetic data from land surface models (LSMs), an approach known as the scaling factor method. In this case, a multiplicative factor is found for each mascon by least squares fit between unfiltered and filtered TWS from the LSM, and GRACE estimates are subsequently multiplied by the scaling factor to restore lost signal. Here, we address several major drawbacks of using synthetic data from LSMs to restore the GRACE signal: 1) LSMs rarely include deep groundwater storage changes, capturing primarily soil moisture anomalies, 2) LSMs frequently relax assumptions of subsurface and surface flows, including lateral flux and lower boundary condition, and 3) abstractions are rarely accounted for in creating the scaling factor. We present an improved scaling factor product for the continental United States derived from an integrated groundwater-surface water model coupled to an LSM. The model extends the lower boundary to over 100 m depth, incorporates lateral flow and stream-aquifer exchange, has been validated against thousands of surface water, groundwater, snow, and land energy observations, and can include pumping from major aquifers. Results show that LSM hydrology complexity is a significant control on the scaling factor. Lateral flow, model depth and human influence should certainly be considerations when using synthetic models to restore GRACE signal.