How much of GPS noise refers to hydrology loading? An insight from GRACE-assimilating hydrological modeling

Global Positioning System (GPS) observations are able to resolve Earth’s surface vertical deformation which originates among others from continental hydrological mass changes. Although long-term signals and seasonal changes of hydrology loading are well-captured by GPS observations, it is still unanswered whether GPS detects also the short-term hydrology-related deformations or not. In this presentation, we use predictions of vertical deformations from a GRACE (Gravity Recovery and Climate Experiment)-assimilating land surface model to separate deterministic and stochastic parts of GPS height changes observed by a set of 221 European EPN (EUREF Permanent GNSS Network) stations. This approach is compared to conventional harmonic functions approach, in which deterministic and stochastic parts are separated by pre-defined annual and semi-annual periods. For the stochastic parts associated with two methods, the noise parameters (spectral indices and amplitudes of power-law noise) are estimated using the Maximum Likelihood Estimation (MLE). Comparing original GPS displacements to displacements reduced for hydrological loading, we notice that annual and semi-annual frequencies are significantly explained by the hydrological model, resulting 60% reduction on average in amplitudes. This means that large part of seasonal crustal deformation arises from hydrological loading or unloading of the lithosphere. We find that the annual and semi-annual peaks are greatly reduced (72% on average) once conventional harmonic functions approach is used instead of GRACE-assimilating hydrological model, but no physical interpretation can be made here since it is difficult to identify the magnitude of each individual processes contributing to seasonal changes. The GRACE-assimilated model can remove the effect of high-frequency hydrological deformations, producing residuals with spectrum closer to the white noise process. Many oscillations present in GPS displacements at periods between 15 and 90 days are well-explained by GRACE-assimilating deformation model. We find the greatest improvement in noise parameters for stations located in the eastern and central regions of Europe, encompassing the Rhine, Elbe, Danube and Oder drainage basins where hydrological mass changes are relatively larger comparing to western Europe. Using GRACE-assimilated model as a deterministic part of GPS displacement time series, we provide a totally new estimates of noise parameters for European sites, which has never been presented before. Our results show that GRACE-assimilating water storage re-analyses can provide essential information for obtaining improved unbiased estimates of GPS vertical velocity, including their uncertainty, which are essential for a range of applications such as upcoming reference frame realizations.