ABSTRACT
Recent probabilistic models of event magnitude distribution for the seismic hazard analysis are either log-linear or allow for smooth, unimodal nonlinearity. There are, however, evidences that in some cases magnitude distribution can be multimodal. We present a model-free approach to evaluation of the seismic hazard, based on a nonparametric estimator of magnitude distribution. Analyses of simulated catalogs showed that this approach provides hazard parameters with tolerable and limited errors regardless the magnitude sampling distribution follows the Gutenberg-Richter’s relation or is multimodal. On the contrary, the classic parametric approach can result in strong underestimation or overestimation of the seismic hazard when the guess of magnitude distribution model is improper. A practical performance of the model-free hazard estimation is illustrated by the analysis of mining induced data from a Polish copper mine and Southern California earthquake data. It is concluded that the nonparametric approach is a better and safer tool in any conditions of seismicity.
