Spatio-temporal clustering of successive earthquakes: analyses of global CMT and F-net catalogs

概要

Spatio-temporal clustering of seismicity is an interesting and important phenomenon that is relevant for earthquake generation process and operational earthquake forecasting. It is well known that earthquakes sometimes closely occur in space and time beyond aftershock area. They are, for example, an 𝑀w 6.5 earthquake hit at Hinagu active fault in Kyushu, Japan, on April 14, 2016, was followed by a large earthquake with 𝑀w 7.3 that took place at Futagawa active fault on April 16 (Goda et al., 2016) and a few successive large earthquakes occurred at Sumatra Island in Indonesia after the 2004 𝑀w 9.2 Sumatra earthquake. Contrary to many studies working on the main shock and aftershock sequences and triggering of smaller earthquakes by large earthquake, few studies have paid attentions on the successive occurrences of large earthquakes. In the present study, therefore, we systematically investigate how often earthquakes successively occur in space and time using some of the most reliable global and regional centroid data catalogs. We aim to find some significant features in the occurrence of successive earthquakes by analyzing the data covering a wide magnitude range. We further examine the results by using Epidemic-Type Aftershock Sequence (ETAS) model and the Coulomb stress change (ΔCFF) hypothesis for understanding their generation mechanism.

In Chapter 1, we first introduce basic and important statistical laws recognized in seismicity. Then, we review previous studies that examine earthquake clustering and, successive occurrence of earthquakes at various regions. We also introduce several successive earthquakes observed around the world. Finally, we present the purpose of this study.

In Chapter 2, we present an algorithm to identify successive earthquakes which is developed in the present study. Also, to evaluate the successive occurrence in space and time, we introduce a method to measure the triggering distance that is defined as a distance from the centroid of a source event. The method we select successive earthquakes consists of the two parts. Firstly, we remove aftershocks that occur in and around the fault of earthquakes with magnitude larger than the target magnitude ranges we investigate. We do not use the earthquakes occurring within a distance of three times of the square root of the slip area and within 180 days to 1825 days from the occurrence of the larger events. Secondly, we search the earthquakes that occur within a horizontal distance (D) and a lapse time (𝑇𝑎) from a source event and group them as a cluster. The source event is selected from the beginning of the data catalog, and the same procedure is repeated for the other earthquakes. By setting various D and 𝑇𝑎, we count the number of the clusters. To examine whether or not successive earthquakes randomly occur, we compare the results with simulations in which the earthquakes in data catalog are set to occur randomly in time but to keep the locations same with the centroids reported in the catalog. We then define the intersection point where the number of clusters obtained from real data merges with that obtained from simulated data as “triggering distance”.

In Chapter 3, we apply the method to the global Centroid Moment Tensor catalog provided by Columbia University. Shallow earthquakes (depth less than 70 km) with a moment magnitude, 𝑀w, of larger than or equal to 5 for the period from 1976 to 2016 are analyzed.

We divide the earthquake catalog into 5 sub-groups: 5.0 ≤ 𝑀w < 5.5, 5.5 ≤ 𝑀w < 6.0, 6.0 ≤ 𝑀w < 6.5, 6.5 ≤ 𝑀w < 7.0and 𝑀w ≥ 7.0. For each magnitude range, we group the earthquakes that occur within a horizontal distance from the centroid of source event and a lapse time from the source event. The results show that the number of clusters increase with increasing the horizontal distance, merging with those obtained for simulated data at a short distance, that is triggering distance. Systematic analyses for the wide magnitude ranges show that the triggering distance increases with increasing the magnitude of source event and decreases with the lapse time. For instance, large earthquakes with magnitude 𝑀w ≥ 7 are triggered up to 550 km, 440 km and 220 km within the lapse time of 60 days, 180 days and 365 days, respectively. This implies that the seismic activity turns to the normal condition in which the occurrence time intervals of large earthquakes obey a Poisson distribution. The triggering distance increases with being almost proportional to the seismic moment of source earthquake with an exponent of about 1/6 to 1/5. The results also show that successive earthquakes are distributed along all types of plate boundaries, and the percentage of successive earthquakes to the total number of earthquake distributes from approximately 8% to 20% for the earthquakes with 𝑀w ≥ 5.0.

In Chapter 4, in order to examine whether or not similar features of successive earthquakes could be retrieved for smaller earthquakes, we use Full Range Seismograph Network of Japan (F-net) catalog provided by NIED, Japan, and analyze the earthquakes occurring around Japan islands with magnitudes of less than 5.5. We divide the earthquake catalog for the period from 2001 to 2010 into 4 sub-groups: 3.5 ≤ 𝑀w < 4.0, 4.0 ≤ 𝑀w < 4.5, 4.5 ≤ 𝑀w < 5.0 and 5.0 ≤ 𝑀w < 5.5. Following the same procedure as in chapter 3, we group the earthquakes that occur within the horizontal distance ranges and lapse times. The results obtained for 3.5≤ 𝑀w< 5.5 also show similar characteristics with those for the global CMT catalog. The triggering distance increases with being proportional to the seismic moment of source event with an exponent of about 1/5 to 1/3 and almost exponentially decreases with the lapse time increasing. The triggering distances are, for instance, 25 km, 17 km and 12 within a lapse time of 60 days, 180 days
and 365 days, respectively, for 3.5 ≤ 𝑀w < 4.0 . Within the Japan islands, successive earthquakes account for approximately 2% to 14% for all the earthquakes with magnitude 3.5≤ 𝑀w < 5.5.

In Chapter 5, we first show that the seismic moment dependency of triggering distance obtained from the analysis of global CMT catalog is well matched with that from F-net catalog. This strongly suggests that the successive earthquakes are generated by a same process over a wide magnitude range. Then, we compare the results obtained from F-net catalog with those from the data that follow ETAS model, and find that the observed and simulated ones by ETAS model are well matched with each other. We further derive empirical scaling relations between the seismic moment and triggering distance from the equations in ETAS model, and the observed exponent of 1/5 to 1/4 are well predicted from the estimated ETAS parameters at various regions around the world. These consistencies show that successive occurrence of earthquakes is well explained by ETAS model. Subsequently, we examine ΔCFF of a source event as triggering mechanisms. The results show that the number of earthquakes occurring in the region with positive ΔCFF are more than about 60% of the total number of the successive earthquakes. This suggests that static stress change introduced by a source event is one of the triggering mechanism of successive earthquakes, although the other mechanisms such as strong motion and visco-elastic response of the structure may play a role. We also estimate the triggering distance at several regions using global CMT catalog to know how the triggering distance changes with regions and tectonic settings.

In this thesis, we introduced a new method to detect and investigate earthquakes that closely occurr in a space and time. The developed method is applicable for the data catalog with a small numbers of earthquakes. Our results highlight that the observed systematic spatio-temporal characteristics are well explained by the ETAS model. The obtained results can be useful to prepare and evaluate successive earthquakes consisting of moderate and/or large earthquakes that sometimes cause the earthquake disaster.