Impact of Earth Model Selection on Terrestrial Water Storage Inversion from GNSS Vertical Displacements

In existing GNSS-based terrestrial water storage (TWS) inversion studies, the PREM model is commonly adopted, and crustal structural heterogeneity is often neglected. Here, we conduct a comprehensive assessment of how different Earth models affect inversion results using both checkerboard-model experiments and continuous smooth-model experiments. The results show that, under realistic hydrological loading conditions within the study region (100°E–115°E, 25°N–40°N), inversion differences among global 1-D reference Earth models are below 2%, whereas the differences between global 1-D reference models and regional crustal models are ~11%; meanwhile, discrepancies between the two regional crustal models remain below 4%. Application to observed GNSS coordinate time series in Yunnan indicates that the spatial pattern of the annual equivalent water height (EWH) amplitude derived from GNSS is broadly consistent with that from the GLDAS hydrological model; however, the choice of Earth model can still substantially alter the magnitude of the inferred amplitude and its spatial distribution. Correlation analyses further suggest that Earth-model dependence is weak for large-scale inversions, but becomes non-negligible at smaller spatial scales. For a representative small-scale subregion (101.75°E–102°E, 22.75°N–23°N), we therefore recommend using the AK135F model to construct Green’s functions. Overall, our findings demonstrate that Earth-model selection is a key source of uncertainty in GNSS-based TWS inversion, and provide practical guidance for choosing appropriate Earth models to improve inversion accuracy.