EXTREME VALUE RISK MODEL ADAPTED TO SPARSE DATA IN SOUTHEASTERN CHINA OF LOW SEISMICITY
适用于弱震的中国东南地区地震资料稀疏情况的极值灾害模型

The Gumbel extreme model, especially the third distribution G(M)=exp-(ω-m/ω-μ)1/λ] is used usually to earthquake risk research because of its prominent advantages, such as the upper bound to magnitude ω etc. Due to the complications of nonlinearity and the limitation of sparse observations, however, its importance has been largely underestimated. The methodology developed here concentrates on two aspects: exploring widening and adaptable use of the observations and finding the proper starting parameters to guarantee the convergence in the nonlinear fitting. Moreover,this new method makes it possible to study the occurrence pattern of large earthquakes in low seismicity regions. In order to expound the distinctive advantage of this method, two data sets from different seismotectonic backgrounds, a low seismicity region from southeastern China and a high seismicity region from western Greece, are analysed. Both of these results are good and stable, as with low fitting errors.