Résumé
Hydrological processes redistribute continental water mass and deform the Earth's elastic crust. Space-based gravimetric and geodetic techniques sense these variations in mass and position and can thus be used to infer changes in continental water storage. Although the deformation field caused by direct hydrological loading is generally well characterized, the situation is more complex near large aquifer systems where both direct loading and poroelastic deformation can occur and interfere in geodetic time series. In this study, we propose a new methodology to isolate the two types of deformation in vertical and horizontal GNSS time series. We chose the Ozark Plateaus Aquifer System in central United States as our test region because of its tectonic quiescence and the availability of geodetic, gravimetric and piezometric data sets as well as geohydrological models. We first extract an aquifer-scale groundwater level signal from the spatially heterogeneous data set using an Independent Component Analysis. We then predict the direct loading deformation with a GRACE-based global model, infer the poroelastic deformation from the GNSS-GRACE residuals and validate our results using the vertical poroelastic deformation and groundwater variations of the aquifer to predict poroelastic horizontal deformation with a 2D analytical poroelastic model. We discuss various parameters influencing the significant recovered poroelastic signals and suggest that they should be accounted for in hydrogeodetic inversions, tectonic transients detection and studies of periodic seismicity.