EARTHQUAKE PREDICTION: STUDIES AND EXPERIMENTS（1999-2002）
ZHANG Xiaodong, FU Zhengxiang,
Center for Analysis and Prediction, China Seismological Bureau, Beijing 100036, China
In China the earthquake prediction is usually divided into four kinds: long-term (few years to tens years), intermediate-term (few months to few years ), short-term (few weeks to few months ) and imminent-term (few days to few weeks ). The present report will give some research results on earthquake prediction during the period from 1999 to 2002. The methods of earthquake prediction are based on mainly the precursory observation of seismicity, active fault, crust deformation, electromagnetism, underground water, and so on.
I. RESEARCHES ON LONG-TERM EARTHQUAKE PREDICTION
The long-term earthquake prediction amounts to determining the recurrence time of earthquakes on a certain fault segment or a certain seismic zone and predicting the approximate time of the next earthquake from the known time of the previous one. Long-term earthquake prediction has extensively used earthquake rates for non-precursory information to estimate the rate of earthquakes and the probability of future event occurrence from seismological, geological and geodetic information. At the same time, the long-term precursory seismicity patterns (for example “seismic gap”) are also used in the long-term earthquake prediction.
(1) The estimation of annual seismic moment rate for a seismic zone or seismic area is necessary in the time-magnitude predictable model (Papazachos, 1989). Shao et al. (1999) proposed a method of estimating annual seismic moment rate by using different seismic data including different precision and completeness, even the paleo-earthquake; at the same time, Jin et al. (1999) used fault slip information to estimate of annual seismic moment rate. And their research results have applied to the time-magnitude predictable model for long-term earthquake prediction of North and Southwest China seismic area. Yi and Wen (2000) have analyzed the recurrence behavior of historical strong events for 39 seismogenic sources in North-South seismic zone of Western China and preliminarily constructed a statistically time-magnitude predictable model. The result shows that within the coming 30 years, 5 of the 39 seismogenic sources have higher conditional probabilities (≥0.4) for strong earthquake recurrence.
Fu et al. (2001) analyzed the time process of historic and recent seismicity and the empirical Gutenberg-Richter relation of magnitude-frequency along the Zhangjiakou-Penglai (Z-P) fault zone in the northern part of North China, and calculated the conditional probabilities of strong earthquake occurrence for eastern and western segment of the Z-P fault zone from the time of the last earthquake to 2010 using the Poison model.
(2) Based on the original stress release model of seismicity proposed by Vere-Jone (1978), Liu et al. (1999) have developed a stochastic coupled stress release model of time-dependent seismicity, which considers the earthquake interaction and stress transfer between different seismic subregions and is applied to a statistical analysis of the historical earthquake catalog with magnitude M≥6.0 during the period from 1480 to 1996 in North China. According to the Akaike information criterion (AIC), the results show that the coupled stress release model is better than the original model, which demonstrates the existence of long-range correlation between different seismic subregions.
(3) Huang and Yu (2000) have investigated the changes in time-space on seismic apparent strain (Wyss, 1970) of middle events, before and after 31 strong earthquake occurred in China since 1955. The result shows that there is a rather well corresponding relationship between the anomaly region of seismic apparent strain and the zone of strong event occurrence within the time range of one to about five years. Huang and Wu (2000) studied the relation between occurrence time and cumulative frequency of strong earthquakes on five subregions of South-North seismic zone of China, which obey an exponential distribution or power function and the time interval of strong earthquake occurrence had changes ranging from few years to tens years.
(4) To investigate the recurrence behaviors of segment-rupturing earthquakes on active faults of the Chinese mainland, Wen (1999a, b) analyzes quantitatively earthquake history of 19 fault segments based on earthquake data segments having behaviors of quasi-periodic earthquake recurrence at previous location and 12 segments with characteristics of time-predictable earthquake recurrence. And the empirical distributions for earthquake recurrence interval for the above-mentioned two earthquake recurrence behaviors have been established, both of them obey the normalized recurrence interval very well, which are no significant difference to that on the Circum-Pacific plate boundaries (NB model). Wen's researches might be important for long-term probabilistic earthquake prediction of active faults of Chinese mainland.
(5) Guo et al. (2001) illustrate the gradient of vertical deformation rates, seismotectonic and seimicity in China continent, as well as their relationship, and get the gradients of vertical deformation rates on a GIS platform based on the map of vertical deformation rate of China continent (1951-1990). The variation of the gradients may reflect the magnitude of vertical crustal shear strain. In the period from 1990 to 1997, some of the strong earthquakes occurred along some high gradient zone, such as 1990 Gonghe Earthquake (M=7.1), 1996 Lijiang earthquake (M=7.0), 1996 Baotou earthquake (M=6.4) and 1998 Zhangbei earthquake (M=6.2). So the analysis of relation between the gradient of vertical crustal deformation rates will apply to the long- term earthquake prediction (Guo and Xie, 2001).
(6) Some investigators (e.g., Utsu, 1977); Aki, 1981; Kenji and Akio, 1990) have investigated the temporal probabilistic synthesis models of earthquake prediction. Wang et al. (2000) investigated a spatial and temporal synthesized probability gain model for the long-term earthquake prediction, which considers not only the relation between strong earthquake occurrence and abnormal temporal process of long-term precursor (seismicity , crust deformation and so on ), but also the abnormal geological structure, active fault behaviors and geophysical fields etc in space. The model has applied to the long-term earthquake prediction for North China.
II. RESERCHES ON MEDIUM-TERM AND SHORT-TERM EARTHQUAKE PREDICTION
1. Seismicity and Seismicity Triggering
Liu (1999) and Wang et al. (2001) use the seismicity data of moderate or small earthquake to space scanning of seismic inhomogenous degree (GL-value) before some strong earthquakes in North China. The result shows that there are obviously anomaly zones of GL-value in the neighbourhood of epicenter in general during the mid-term of 1-3 years before moderately strong earthquakes in China.
Yang and Qu (1999) calculated b value in the relation between magnitude and frequency by using robust regression of M estimation and choosing function as normal density function, and discussed the difference between the results obtained by the methods of robust.
Zhou and Zhu (1999) deduced a method for estimating strong earthquake recurrence period by the fractal based on the dynamic equation of the seismic energy accumulation.
Peng et al. (1999) deduced a formula of effective shear stress for the slip fault, the thrust fault and the strike fault based on the characteristics of regional fault activity and the result of tectonic stress field, and proposed a new Loading/Unloading distinction criterion. Chang et al. (1999) Studied the variation characteristics of the Loading/Unloading response ratio with time before the Baotou Ms6.4 earthquake. The result shows that there is obvious anomaly before the 1996 Baotou earthquake.
Shao et al. (2000) used the Morlet wavelet translation method to analyze the dynamic period of seismic energy in some area of North China. The results show that there exist both comparatively stable periods and some variable periods with time.
Jiao and Ding (2000) analyzed 24 earthquake cases with M≥6 occurred in China. The research results show that about 80% of the cases, earthquake activity increased within the epicentral area and its surroundings before the mainshock.
Zhao et al. (2001) applied the Morishita index Iδ to analyze the features of the day-night and seasonal distribution of earthquakes in China. The statistical results indicated that the occurrence time of strong earthquakes in China's inland is mainly concentrated at night.
Pei et al. (2002) introduced the semi-variogram analysis to study the seismic relativity, and detected the distinct difference of relativity between foreshocks and aftershocks of Songpan earthquake.
Cai et al. (2002) calculated the multi-fractal spectrums of the temporal distribution of earthquakes occurred in China from 1960 to 1982 and analyzed aftershock sequences of the four great earthquakes. The result may be useful for earthquake prediction.
Zhou et al. (1999) provided a method to distinguish the pre-and after-shocks of a strong shock from earthquake catalogue and then to form a sequence based on the analysis of SLC (Single- Link- Cluster).
Zhou et al. (2000) studied the dynamic changes of stress field of small earthquakes occurred before four moderate-strong earthquakes in North China by using the data of comprehensive mechanism solution of small earthquakes and using the initial P wave recorded by seismic recording. They also generalized preliminarily the dynamic evolution characteristics of stress field in the focal region and its adjacent area before moderate and strong earthquakes.
To understand the statistical characteristics of the time process of earthquakeoccurrence, Kagan and Jackson (1991) have defined a coefficient of variation (Cv) of earthquake interoccurrence time (T) as a ratio of the standard deviation (s) to the average time (Ta): Cv = s / Ta. Cao et al. (2000) analyzed the variation of Cv before the 6 strong earthquakes along North-south earthquake belt in China from 1970. The result shows that several months or one year before the main-shock, the coefficient of variation drops slowly, ranging from 0.5-1.3. Among six strong earthquakes, the variation of Cv drops in long and intermediate-term before four earthquakes. The result indicates the coefficient of Cv reflect some complexity of seismic precursors before strong earthquakes.
Li et al. (2001) investigated the co-relationship between seismisity in the Taiwan area and its vicinity, finding that optimal angles for frequency over double standard deviations are about 50° of offset after quadratures and full moon. Li (2000) analyzed the triggering of the seasonal changes in earth rotation rate and tide for strong earthquakes occurring in North China, and suggested that earth rotation and tide force can trigger strong earthquakes. Zhang et al. (2001) analyzed the relationship between the Ms≥7.0 events in the Chinese mainland and the modulation and triggering from the earth tides.
Fu et al. (1999) investigated the interaction between parallel strike-slip faults and its effect on seismicity in visco-elastic model. The slipping along a strike-slip fault would lead to shear stress reduction on parallel fault planes, and delay the slip occurrence along the parallel faults.
2. Crustal DeformationOn the crust movement and the stress-strain filed of China mainland
Based on the regional GPS observations, the principal movement trend of China main blocks have been given out. Liu and Shao (1999), Zhou and Guo (2000), and Zhou and Wu (2001) have explained the characteristics of crust movement in Fujian coast which may be related to the Taiwan 7.8 earthquake of 21 Dec, 1999; Jiang and Zhang (2000a, 2000b, 2001) have studied the recent movement of Qing-Zang block and western China, Wang and Ding (2000) have explored the crust deformation in Xinjiang region.
The GPS data from China Crust Movement Observation Network have attracted many authors to study the characteristics of China mainland. Sui and Wu (1999), Gu (2001) have draw the images of movement of China mainland from 1998 to 1999, and explained the characteristics of crust movement of China mainland.
Except of the movement of block based on the GPS, the study on the stress-strain field is also an important research subject. Huang and Song (1999) have given the mechanism of horizontal deformation based on the GPS observation in northern China, He and Niu (2000) have simulated the stress field change in Northern China based on the regional tectonic data, Chen and Nie (2001) have simulated the distribution of deformation and stress field for a elastic body model, Chen and Zhang (2001) have studied the evolution of stress status and fault activity in Northern China, Jiang and Zhang (1999, 2000a, 2000b, 2001) and Chen and Jiang (2000) have discussed the method for solving the strain and calculated the stress status and block movement by using the non-continuous deformation method.