Mapping site effects at large scale
Résumé
One of the strongest historical earthquakes in France metropolitan territory occurred in 1909, in
Provence, south of France. In the eighties, a scenario study predicted that a similar earthquake may
lead to more than the 46 deaths of 1909 and a tremendous economical cost caused by increasing
urbanisation in this area. The 1909 maximal intensity was estimated at IX. But a lot of municipalities
exhibited strong variations in damage distribution. For some of them, like Rognes and Vernègues, the
historical perched village suffered more damage than constructions built on the flat part of the
territories.
While seismologists realised site effect importance in earthquakes, this 1909 damage distribution
became the most famous french illustration of topographic site effect. But if ray theory explains that
relief can indubitably focus waves and amplify seismic signal for specific wavelength according to the
location on the slope, some doubts remain about the real impact of topographic effects in 1909
damage distribution. It may also be related to the fact that the different types of building were not
uniformly spread on the territories and/or that the old structures were more vulnerable than new ones.
Finally, was the seismic signal really different along the relief during 1909 earthquake ? Trying to solve
this question, several field campaigns were conducted on the village of Rognes. The first one
consisted in measuring microtremors on several points and computing H/V ratios (Nogoshi, 1970,
Nakamura, 1989). The H/V curves on flat part of the territory do not exhibit any clear peak except for
one site on the north where a high frequency peak should be relative to a superficial and thin soft
layer. On the contrary, the H/V curves obtained on the top of the relief show a high peak around 1
Hertz. We then decided to install 9 seismic stations to record continuously seismicity at key-points of
the relief. The seismicity rate is very low in this region, but the 2 years of operation allow to compare
recordings of local, regional and teleseismic events both in time and spectral domains. The first
analyse confirms the importance of signal amplification on the top of the relief.
Now, numerical simulations are conducted to confirm this amplification and the concerned frequency
range. It should precise the repartition of the amplification along a 2D profile. These simulations are
conducted simultaneously with different techniques (Finite Elements, Fast Multipole method,
Discontinuous Galerkin Finite Elements). These numerical results, confirmed by experimental results,
should latter drive to canonical models of seismic response for different relief. They may be directly
used to predict ground motion along hills, which is an essential task for seismic risk mitigation,
particularly in south France.
Origine : Fichiers produits par l'(les) auteur(s)