Objectives The time-varying gravity field models derived from the gravity recovery and climate experiment (GRACE) offer a novel approach for studying terrestrial water storage. However, GRACE data resolve the total water storage changes at grid points, encompassing surface water, soil moisture, groundwater, and vegetation water, without the ability to differentiate the vertical distribution of water components.

Methods We employ wavelet decomposition to process GRACE signals, from which the surface water component, as modeled by the global land data assimilation system (GLDSA) hydrological model, has been subtracted. The resulting wavelet sub-functions are then combined with groundwater measurements from wells across the USA to perform a regression analysis on the groundwater component. By employing two-dimensional surface interpolation, we obtain regression coefficients for various wavelet sub-functions throughout the USA, enabling the reconstruction of long-term continuous sequences of groundwater storage changes.

Results The results show that over 61.84% of the test sites have a correlation coefficient greater than 0.4, and 62.90% of the sites exhibit root mean square (RMS) residuals below 1.0 m.

Conclusions This methodology successfully captures the spatiotemporal characteristics of groundwater distribution, providing valuable data support for the utilization and research of groundwater resources.