Benchmarking Amazonian GPS stations: an improved way to model hydrological changes

Abstract

<p>Hydrological loading is one of the main contributors into seasonal displacements of the Earth’s crust, as derived from the Global Positioning System (GPS) permanent stations. Recent studies proved that hydrological signatures may be also observed in GPS displacements outside seasonal band. Such estimates may be, however, biased, since (1) total character of GPS displacements is generated by a set of geophysical phenomena combined with GPS-specific signals and errors and (2) the exact sensitivity of GPS for individual components has not yet been properly recognized. In this study, we propose a completely new approach to establish a set of benchmarks of GPS stations, for which sensitivity to geophysical phenomena is identified. We focus on hydrological changes within the Amazon basin, but the same approach could be employed to analyze other phenomena. Analysis is performed for vertical displacements from 63 GPS stations provided by the Nevada Geodetic Laboratory (NGL), collected between 1995 and 2021. Results are compared to data from GRACE (Gravity Recovery and Climate Experiment) and GRACE Follow-On missions (2002-2021), provided by GFZ (GeoForschungsZentrum) as RL06 solution in a form of spherical harmonic coefficients truncated to d/o 96, filtered with DDK3 decorrelation anisotropic filter. We also utilize GLWS (Global Land Water Storage) datatset provided by University of Bonn, as a result of assimilation of GRACE Total Water Storage (TWS) anomalies into WaterGAP Global Hydrological Model (WGHM). We make also use of two hydrological models: pure WGHM and GLDAS (Global Land Data Assimilation System), for which TWS values are provided. Both GRACE and TWS data are converted to vertical displacements of Earth’s crust using load Love numbers, while GPS displacements are reduced for non-tidal atmospheric and oceanic changes. We find the largest values of trends and annual signals for GPS stations proximate to Amazon river. GRACE, GLWS and hydrological models disagree at the level of 8 mm, at maximum. This is mainly caused by the GLDAS model which lacks in the contribution of surface water. Supplementing GLDAS with surface water layer employed from WGHM reduces this difference to 1 mm. Benchmarks of GPS stations are established by using a wavelet decomposition with Meyer’s mother wavelet. We divide both the GPS, GRACE and GLWS displacement time series into 4 decomposition levels, defined by exact periods they contain. Then, we compute correlation coefficients between individual levels of details. We show that the number of 32%, 64%, 97%, 89% and 68% out of 63 GPS stations is significantly correlated to GRACE for periods, respectively, from 2 to 5 months, from 4 to 9 months, from 7 months to 1.4 years, from 1.1 to 3.0 years and from 3.0 years onwards. These numbers change into: 48%, 73%, 100%, 81% and 50% out of 63 GPS stations, when GRACE is replaced with GLWS. 12 or 21 out of 63 GPS stations correlate positively with GRACE or GLWS within entire frequency band, which means that a character of these GPS displacement time series is generated mostly by hydrological changes.</p>