Assessment of the seismic hazard, comparison of probabilistics and deterministic methods with application to the moderate North European context.
Céline BEAUVAL, PhD defended on December 16, 2003 in Grenoble (France)
This thesis proposes a new methodology that allows to pinpoint the key parameters that control probabilistic seismic hazard assessment (PSHA) and at the same time to quantify the impact of these parameters uncertainties on hazard estimates. Probabilistic seismic hazard assessment consists in computing return periods of ground motion exceedances. The Cornell-McGuire’s method is used here. First, the dependence of the b-value (slope of the recurrence curve) with the magnitude range used is highlighted. The high dependence found in the southeastern part of France implies that, for conservative modeling, earthquakes recurrence should be established only on magnitudes above 3.5. Secondly, uncertainties on magnitude and location determinations are modeled and quantified: resulting variability on hazard estimates ranges between 5% and 25% (=COV, coefficient of variation), depending on the site and the return period.
An impact study is then performed, in order to determine the hierarchy between the impacts of the choices of four other parameters: intensity-magnitude correlation, minimum and maximum magnitudes contributing to the hazard and the truncation of the attenuation relationship. The results at 34 Hz (PGA) indicate that the maximum magnitude is the less influent parameter (from 100 to 10000 years); whereas the intensity-magnitude correlation and the truncation of ground motion predictions (>2s) are the controlling parameters at all return periods (producing up to 30% decrease each at 10000 years). An increase in the minimum magnitude contributing to the hazard, from 3.5 to 4.5, can also produce non-negligible impacts at small return periods, with up to 20% decrease of hazard results at 475 years.
Finally, the overall variability on hazard estimates due to the combined choices of the four parameters can reach up to 30% (COV, at 34Hz). For lower frequencies (<5Hz), the overall variability increases and maximum magnitude becomes a controlling parameter. Therefore, variability of estimates due to catalog uncertainties and to the choices of these four parameters must be taken into account in all probabilistic seismic hazard studies in France. To reduce variability in hazard estimates, future research should concentrate on the elaboration of an appropriate intensity-magnitude correlation, as well as on a more realistic way of taking into account ground motion dispersion