Seismicity rate and wave-velocity variations as consequences of rainfall: the case of the catastrophic storm of September 2002 in the Nîmes Fault region (Gard, France)

Abstract

The western Provence in southern France is an intraplate low deforming region, cut by large sinistral strike-slip faults as the Nimes Fault. The deformation rate of this fault was estimated at 0.1 mm yr−1 from geological and morphological observations. Nevertheless, some large earthquakes occurred in this area in historical times inducing well-documented partial destructions on the roman monument 'Pont-du-Gard'. In order to investigate if this area is seismically active, we installed a temporary seismological network of 14 stations between the cities of Nimes and Avignon for a 1-yr period (2002 July–2003 June). We recorded and located a total amount of 80 earthquakes of magnitude Ml smaller than 2.8, and of 153 quarry blasts. The seismic event locations are interpreted in terms of regional tectonics as well as of geometry at depth of the Nimes Fault and of possible seismic activity of the Roquemaure Fault. At the beginning of 2002 September, a catastrophic storm occurred in the same area with cumulated precipitations at 600 mm in 28 hr, causing casualties and major inundations and damages. We identified a clear and sudden increase of the seismic activity as an immediate consequence of this exceptional meteorological phenomenon. We interpret this rainfall triggering of earthquakes as the response of the crust to an abrupt vertical loading. The geological context of this zone is characterized by the presence of sedimentary basins and of an important karstic network in a calcareous environment allowing the retention of a third of the total rainwater volume during few weeks after the catastrophic storm. Month by month inversion of the seismic wave velocities for only the first 1-km thick layer of the crust allows investigating the influence of the stored water. A clear decrease of both P- and S-velocities, and an increase of the Vp/Vs ratio are evidenced. The variations obtained are at 2, 6 and 4.5 per cent for P-, S-velocities and Vp/Vs ratio, respectively. These variations are discussed in confront of previous studies and can be directly related to the known water volume.