[MIS09-12] 日本におけるb値の時空間変化の統計的検定と地震予測
キーワード：b値、地震検知力、赤池情報量基準
The Gutenberg-Richter (GR) law: log10NM=a-bM denotes the relationship between the frequency-magnitude distribution (FMD), in which N is the cumulative number of earthquakes with magnitude equal to or larger than M. Constant b measures difference in the relative proportion of small and large earthquakes and shows the variation of seismicity activity and stress evolution before earthquake. Many studies show that b-value decreased prior to great earthquakes. Since the b value depends very much on the magnitude of completeness (Mc), we investigated the spatial variation of Mc in Japan, combining the maximum curvature (MAXC) technique and the bootstrap approaches. Then, for the area that has approximately equal Mc value, we check the temporal variation of Mc and apply its maximum to calculate b values. This process helps the computation of b value in this area and its uncertainty can be computed more properly.
We used the bootstrap approach to generate reference b values and applied the Akaike Information Criterion (AIC) to assess the spatial variation of b values of the area statistically. Here, we define the P(ΔAIC≧2) value, where P(ΔAIC≧2) means that the percentage of ΔAIC between the b value at the target space and time and that of the reference is larger or equals to 2. This gives the significance level of difference between target and the reference periods. In the area, we narrowed a target area depending on P(ΔAIC≧2), after that we quantified the temporal variation of the target area and discuss the precursor of the large earthquake occurred in this area. Also, the spatial distribution of ΔAIC between reference and the period prior to the main shock will also be investigated. In this paper, we will demonstrate capability of the proposed method. A case study of the 2016 Kumamoto Earthquake (M7.3) shows a statistically clear decrease in b value and large increase of P(ΔAIC≧2) value few months before the main shock. It indicates that the b value with statistical assessment may emerge positive impact for forecasting a future large earthquake.
We used the bootstrap approach to generate reference b values and applied the Akaike Information Criterion (AIC) to assess the spatial variation of b values of the area statistically. Here, we define the P(ΔAIC≧2) value, where P(ΔAIC≧2) means that the percentage of ΔAIC between the b value at the target space and time and that of the reference is larger or equals to 2. This gives the significance level of difference between target and the reference periods. In the area, we narrowed a target area depending on P(ΔAIC≧2), after that we quantified the temporal variation of the target area and discuss the precursor of the large earthquake occurred in this area. Also, the spatial distribution of ΔAIC between reference and the period prior to the main shock will also be investigated. In this paper, we will demonstrate capability of the proposed method. A case study of the 2016 Kumamoto Earthquake (M7.3) shows a statistically clear decrease in b value and large increase of P(ΔAIC≧2) value few months before the main shock. It indicates that the b value with statistical assessment may emerge positive impact for forecasting a future large earthquake.