Résumé
The increasing number of hydro-gravimetry studies proves the rising interest of the hydrology community towards this monitoring method. The accuracy of Superconducting Gravimeters (SG) potentially allows the retrieval of small Water Storage Changes (WSC) down to a few millimeters of equivalent water thickness However, the importance of these corrections applied to SG data to achieve such a precision in gravity residuals should be recalled. The Djougou permanent gravity station presented in this paper and located in northern Benin, West-Africa, provides a good opportunity to review these considerations. This station is equipped since July 2010 with the superconducting gravimeter SG-060 aimed at deriving WSC at different time-scales, daily to inter-annual. In this area, WSC are 1) part of the control system for evapotranspiration (ET) process, a key variable of the West-African monsoon cycle and 2) the state variable for resource management, a critical issue in storage-poor hard rock basement contexts such as in northern Benin. The potential for deriving WSC from time-lapse gravity data partly depends on environmental features such as topography and the instrument shelter. Therefore, this issue is addressed first, with the background idea that such sensitivity analysis should be undertaken before setting up any new instrument. In Djougou, local topography is quite flat leading to a theoretical straightforward relationship between gravity changes and WSC, close to the standard Bouguer value. However, the shelter plays a significant masking role, which is the principal limitation to the retrieval of fast hydrological processes such as ET following a rain event. Several issues concerning classical gravity corrections are also addressed in the paper. These include gap-filling procedures during rain-events and drift estimates for short time series. Special attention is provided to atmospheric corrections, and different approaches are tested: a simple scalar admittance, a filtered scalar admittance, a frequency-dependent admittance and direct atmospheric loading calculations. It is shown that only the physically-based approach of direct loading calculations performs better in both residual minimization and ET retrieval. Moreover, non-local hydrological effects are investigated and account for about 20% of the gravity residuals. Finally, gravity residuals are briefly analyzed at two distinct time scales: rapid (up to a few days) and seasonal. At the rapid time-scale, it is shown that ET retrieval is hardly achievable given shelter size and state-of-the-art atmospheric corrections. Still, mean values retrieved from this study are in accordance with known values of potential ET and constant lateral flow. Direct comparison of gravity changes with hydrological data (neutron probe monitoring and water table levels) show some discrepancies, particularly for the hydrological year of 2011, for which all hydrological data show a deficit, but SG and FG5 data do not. This preliminary analysis both provides a basis and call for further hydro-gravity modeling, to comprehensively investigate the water-cycle at the Djougou station.