Examining the spatial and temporal distribution of seismic activity holds significant importance for seismic risk assessment, particularly in regions prone to frequent and intense earthquakes such as the Sichuan-Yunnan region in China. It is widely recognized that earthquakes exhibit non-random patterns in both spatial and temporal dimensions.

Early scientists endeavored to predict earthquakes using statistical principles, leading to the development of various forecasting methods. Among these, the Relative Intensity(RI)and Pattern Informatics(PI)methods emerged as statistical approaches to earthquake prediction modeling. Essentially, both methods fall under the category of smoothing seismic activity models. They employ techniques to quantify temporal changes in seismic activity graphs, generating maps that highlight areas(hot spots)where earthquakes may occur during specific future periods. While the RI algorithm’s theory is straightforward, its forecasting efficacy is robust, particularly notable in predicting major earthquakes, demonstrating similar advantages to the PI algorithm. Widely adopted globally for proactive predictions across diverse tectonic systems, it has shown commendable results in seismic forecasting practices both domestically and internationally. Over years of development, its predictive performance has gained prominence. However, further research is needed to assess its suitability for predicting minor seismic events in low-seismicity zones. Additionally, its successful application hinges on background seismic activity and the selection of target magnitudes.

To aid seismic activity prediction in the Sichuan-Yunnan region and identify potential future seismic source areas, a comprehensive parameter analysis was conducted using the Relative Intensity(RI)algorithm with the parameter traversal test(PTT). The RI algorithm operates on the premise that the predicted intensity of future earthquakes in a given region closely mirrors the intensity of past earthquakes. While it may not explicitly consider the “active” and “quiet” characteristics of seismic activity, as a fundamental prediction algorithm, it often yields improved prediction outcomes when applied to assess seismic probability in regions with high seismic activity, such as the Sichuan-Yunnan region.

In this study, the statistical-based Relative Intensity(RI)algorithm is employed to calculate the relative intensity of earthquakes based on quantitative earthquake characteristics. The study involves gridding the investigated area and statistically analyzing historical earthquake occurrences within each grid unit under specific magnitude conditions to inform predictions of future earthquake frequencies. The research focuses on evaluating the influence of four key model parameters: grid size, length of the anomalous learning window, starting point of the prediction window, and length of the prediction window, on the algorithm’s prediction efficiency. Furthermore, the study investigates the applicability of the RI algorithm to the Sichuan-Yunnan regions in China. The results yield two significant findings:

(1)The integration of the Relative Intensity(RI)algorithm with the Parameter Traversal Test(PTT)yielded significantly improved results compared to random guessing, primarily due to its optimized parameter selections. These parameters include the grid size, length of the anomalous learning time window, starting time of the prediction time window, and length of the prediction time window.

(2)The parameters of the prediction model exhibit a degree of stability and demonstrate predictive capability for seismic activity in the Sichuan-Yunnan region over the next 1-5 years. The study revealed specific rules and effective parameter intervals applicable to earthquake-prone areas in Sichuan-Yunnan.

The findings suggest that the integration of the Relative Intensity(RI)algorithm with the Parameter Traversal Test(PTT)holds promise for predicting seismic activities in the Sichuan-Yunnan region. This approach enhances the pool of references available for predicting earthquake trends in regions prone to frequent and intense earthquakes. Further research on the RI algorithm is anticipated to yield a more refined numerical model for earthquake trend prediction, contributing to enhanced forecasting accuracy and preparedness in earthquake-prone areas.

In order to realize the rapid and efficient identification of earthquakes, blasting and collapse events, this paper applies the Convolutional Neural Network(CNN)in deep learning technology to design a deep learning training module based on single station waveform recording of single event and a real-time test module based on multiple stations waveform recording of single event.
On the basis of ensuring that the data is comprehensive, objective and original, the three-component waveforms of the first five stations that recorded the P-wave arrival time of each event are input, and the current mainstream convolutional neural network structures are used for learning test. The four main convolutional neural network structures of AlexNet, VGG16, VGG19 and GoogLeNet are used for learning training, and the learning effects of different network structures are compared and analyzed. The results show that in the training process of various convolutional neural network structures, the accuracy rate and the cost function curve of the training set and the test set of each network are basically the same. The accuracy rate increases gradually with the increase of the training times and exceeds 90%, and finally stabilizes around a certain value. The cost function curve decreases rapidly with the increase of the training times, and eventually the stability does not change near a relatively small value. At the same time, over-fitting occurred in all convolutional neural network structures during training, except for AlexNet. In the end, the cost function of each type of structural training set and test set is finally lower than 0.194, and the recognition accuracy of each type of structure for training sets and test sets is over 93%. Among them, the recognition accuracy of AlexNet network structure is the highest, the accuracy of the training set of AlexNet network structure is as high as 100%, the test set is 98.51%, and no overfitting occurred; the accuracy of VGG16 and VGG19 network structure comes second, and the recognition accuracy of GoogLeNet network structure is relatively low, and the trend curves of the accuracy and cost function in training and test set of each network in the training process are basically the same. Subsequently, in order to test the event discrimination efficiency of the CNN in deep learning in the real-time operation of the digital seismic network, we select the trained AlexNet convolutional neural network to perform event type determination test based on the waveform recording of multiple stations of a single event. The final result shows that the types of a total of 89 events are accurately identified in the 110 events with M ≥0.7 recorded by Shandong seismic network, and the accuracy rate is about 80.9%. Among them, the accuracy rate of natural earthquake is about 74.6%, that of explosion is about 90.9%, and that of collapse is 100%. The recognition accuracy of collapse and explosion events is relatively high, and it basically reaches or exceeds the recognition accuracy of manual determination in the daily work of the seismic network. The accuracy of natural earthquake identification is relatively low. Among the 18 misidentified natural earthquakes, up to 13 events were judged as blasting or difficult to identify due to distortion of waveforms recorded by some stations(They are determined to be explosion and earthquake each by the records of two of the five stations). If sloughing off the recognition type error events caused by waveform distortion due to the background noise interference that overwhelms the real event waveform or waveform drift, the recognition accuracy of earthquake will become 91.4%, and the recognition accuracy of all events will increase from 80.9%to 91.7%, which is basically equivalent to the recognition accuracy of manual judgment in the daily work of the seismic network. This indicates that deep learning can quickly and efficiently realize the type identification of earthquake, blasting and collapse events.

Since the earthquake of M_L3.8 occurring on October 1, 2013 in Ruishan, Weihai City, Shandong Province, the sequence has lasted for about 4 years(Aug. 31, 2017). Seismicity is enhanced or weakened and fluctuated continuously. More than 13250 aftershocks have been recorded in Shandong Seismic Network. During this period, the significant earthquake events were magnitude 4.2(M_L4.7)on January 7, 4.0(M_L4.5)on April 4, M3.6(M_L 4.1)on September 16 in 2014 and M4.6(M_L5.0)on May 22, 2015. The earthquake of M_L5.0 was the largest one in the Rushan sequence so far. In order to strengthen the monitoring of aftershocks, 18 temporary stations were set up near the epicenter at the end of April, 2014(official recording began on May 7)by Shandong Earthquake Agency, which constitutes an intensified network in Rushan that surrounds the four quadrants of the small earthquake concentration area together with 12 fixed stations nearby, and provides an effective data foundation for the refinement of Rushan earthquake sequence.
The velocity structure offers important information related to earthquake location and the focal medium, providing an important basis for understanding the background and mechanism of the earthquake. In this paper, double-difference tomography method is used to relocate the seismic events recorded by more than six stations of Rushan array from May 7, 2014 to December 31, 2016, and the inversion on the P-wave velocity structure of the focal area is conducted. The Hyposat positioning method is used to relocate the absolute position. Only the stations with the first wave arrival time less than 0.1 second are involved in the location. A total of 14165 seismic records are obtained, which is much larger than that recorded by Shandong Seismic Network during the same period with 7708 earthquakes and 2048 localizable ones. A total of 1410 earthquakes with M_L ≥ 1.0 were selected to participate in the inversion. Precise relocation of 1376 earthquakes is obtained by using double-difference tomography, in which, there are 14318 absolute traveltime P waves and 63162 relative travel time P waves. The epicenters are located in distribution along NWW-SEE toward SEE and tend to WS, forming a seismic belt with the length about 3km and width about 1km. The focal depths are mainly concentrated between 4km and 9km, occurring mainly at the edge of the high velocity body, and gradually dispersing with time. It has obvious temporal and spatial cluster characteristics. Compared with the precise relocation of Shandong network, the accuracy of the positioning of Rushan array is higher. The main reason is that the epicenter of Rushan earthquake swarm is near the seaside, and the fixed stations of Shandong Seismic Network are located on the one side of the epicenter. The nearest three stations(RSH, HAY, WED)from the epicenter are Rushan station with epicentral distance about 13km, the Haiyang station with epicentral distance about 33km, and Wendeng station with epicentral distance about 42km. The epicentral distance of the rest stations are more than 75km. In addition, the magnitude of most earthquakes in Rushan sequence is small. The accuracy of phase identification is relatively limited due to the slightly larger epicentral distance of the station HAY and station WED in Shandong Seismic Network. Furthermore, the one-dimensional velocity model used in network location is simple with only the depth and velocity of Moho surface and Conrad surface. The epicentral distances of the 18 temporary stations in Rushan are less than 10km, and the initial phase is clear. The island station set up on the southeast side and the Haiyangsuo station on the southwest side form a comprehensive package for the epicenter. Compared with the double-difference algorithm method, the double-difference tomography method used in this paper is more accurate for the velocity structure, thus can obtain the optimal relocation result and velocity structure.
the velocity structure shows that there are three distinct regions with different velocities in the vicinity of the focal area. The earthquakes mainly occur in the intersection of the three regions and on the side of the high velocity body. With the increase of depth, P wave velocity increases gradually and there are two distinct velocity changes. The aftershock activities basically occur near the dividing line to the high velocity side. The south side is low velocity abnormal body and the north side is high velocity abnormal body. High velocity body becomes shallower from south to north, which coincides with the tectonic conditions of Rushan. Considering the spatial relationships between the epicenter distribution and the high-low velocity body and different lithology of geological structure, and other factors, it is inferred that the location of the epicenter should be the boundary of two different rock bodies, and there may be a hidden fault in the transition zone between high velocity abnormal body and low velocity abnormal body. The interface position of the high-low velocity body, the concentrating area of the aftershocks, is often the stress concentration zone, the medium is relatively weak, and the intensity is low. There is almost no earthquake in the high velocity abnormal body, and the energy accumulated in the high velocity body is released at the peripheral positions. It can be seen that the existence of the high-low velocity body has a certain control effect on the distribution of the aftershocks.

Many small earthquakes occurred intensively and continuously and formed an earthquake sequence after the M_L3.8 earthquake happened at Rushan County, Shandong Province on October 1, 2013. Up to March, 2017, more than 13 000 events have been recorded, with 3 429 locatable shocks, of which 31 events with M_L ≥ 3.0. This sequence is rarely seen in East China for its extraordinary long duration and the extremely high frequency of aftershocks. To track the developing tendency of the earthquake sequence accurately, 20 temporary seismometers were arranged to monitor the sequence activities around the epicenter of the sequence since May 6, 2014. Firstly, this paper adopts double difference method to relocate the 1 418 earthquakes of M_L ≥ 1.0 recorded by temporary seismometers in the Rushan earthquake sequence (May 7, 2014 to December 31, 2016), the result shows that the Rushan earthquake sequence mainly extends along NWW-SEE and forms a rectangular activity belt of about 4km long and 3km wide. In addition, the seismogenic fault of Rushan earthquake sequence stretches along NWW-SEE with nearly vertical strike-slip movement and a small amount of thrust component. Then we apply the P-wave initial motion and CAP to invert the focal mechanism of earthquakes with M_L ≥ 1.5 in the study area. The earthquakes can be divided into several categories, including 3 normal fault earthquakes (0.9%), 3 normal-slip earthquakes (0.9%), 229 strike-slip earthquakes (65.8%), 18 thrust fault earthquakes (5.2%), 37 thrust-slip earthquakes (10.6%)and 58 undefined (16.6%). Most earthquakes had a strike-slip mechanism in Rushan (65.8%), which is one of the intrinsic characteristics of the stress field. According to the focal mechanism solutions, we further utilized the LSIB method (Linear stress inversion bootstrap)to invert the stress tensor of Rushan area. The result shows that the azimuth and plunge of three principal stress (σ₁, σ₂, σ₃) axes are 25°, 10°; 286°, 45°; 125°, 43°, respectively. Based on the stress field inversion results, we calculated the focal mechanism solutions consistency parameter (θ)and the angle (θ₁)between σ₁ and P axis. The trend lines of θ and θ₁ were relatively stable with small fluctuation near the average line over time. Furthermore, the earthquake sequence can be divided into three stages based on θ and θ₁ values. The first stage is before September 16, 2014, and the variation of the θ and θ₁ values is relatively smooth with short period. All focal mechanism solutions of the three M_L ≥ 3.0 earthquakes exhibited consistence. The second stage started from September 16, 2014 to July 1, 2015, the fluctuation range of θ and θ₁ values is larger than that of the first stage with a relative longer period. The last stage is after July 1, 2015, values of θ and θ₁ gradually changed to a periodic change, three out of the four M_L ≥ 3.0 earthquakes (strike-slip type)displayed a good consistency. Spatially, earthquakes occurred mainly in green, yellow-red regions, and the focal mechanism parameters consistency θ was dominant near the green region (around the average value), which presents a steady state, and the spatial locations are concordant with the distribution of θ value. Moreover, all of M_L ≥ 3.0 earthquakes are located in the transitional region from the mean value to lower value area or region below the mean value area, which also indicates the centralized stress field of the region.

A notable swarm occurred in Rushan, Shandong Peninsula and its activities continue since Oct. 2013 till now. Up to Sept. 30, 2014, more than 7 000 events have been recorded, in which locatable shocks exceed 2000, and 18 events with M_L≥3.0. The swarm is rarely seen in East China for its extraordinary duration time and surprising high frequency of aftershocks. 18 temporary seismometers have been deployed around the swarm since May 6, 2014, and composed a seismic array for monitoring the swarm activities. Based on data from permanent networks and temporary array, we relocated the earthquake sequence by using hypoDD method. It has been shown that, there is obvious difference between permanent network results and temporary array results. The permanent network of Shandong has a relative large coverage gap(more than 200°)for this swarm. Its location results therefore should not be reliable. There are maybe other errors in the permanent network result due to some problems in the raw data, such as too few stations for most locatable events(3 stations), and relative lower proportion of located events in final result(74.3%, while 95.1% in temporary array result). It can be found by comparing location results from permanent network and temporary array that, using temporary array's data can improve the location accuracy significantly. The results of temporary array are: aftershocks distribution of Rushan swarm is in NWW direction, the dip-direction of fitted fault plane is SW, and the strike and dip angle agree with focal mechanism of the mainshock. Focal depths of aftershocks are at 4.5～8km; the swarm is restricted in a small area about 3km×3km×1km, and has some characteristics such as clustering, staged activities, and etc; the aftershock activities are in accord with crack growth behavior pattern, hence we deduced that there may be fluid intrusion in source area. Finally, we discussed the seismogenic structures and active mechanisms of this swarm combined with relative geologic knowledge. We draw some conclusions as follows: 1)Rushan swarm probably occurred at the boundary of rock bodies of Duogu Mountain and Haiyangsuo super-unit; 2)The seismogenic structure is a blind fault, which should be a part of adjacent Heishankuang-Jilincun Fault, or might be a new fault at rock body boundaries; 3)Rushan swarm might be an evidence for the existence of the disputed Shidao Fault.

After a large earthquake, more seismic activities are observed in the focal region and its adjacent area. The obvious increased earthquakes are called the aftershocks. Generally speaking, aftershock sequence gradually weakens and sometimes has ups and downs. The time when the aftershock activity begins to be confused with background seismic activity is known as the aftershock activity duration. Aftershock sequence is one of the enduring research fields in seismology. Aftershocks accord with two important statistical relationships, one is the G-R relationship describing the relation between the magnitude and frequency, the other is the modified Omori formula describing the characteristics of aftershock decay with time. On this basis, a number of studies from different angles explain the mechanism of aftershock activity. From the perspective of the medium heterogeneity, it is universally accepted that aftershock is a result of further rupture of residual asperities. From the perspective of stress, these models, e.g. rate-state dependence, subcritical crack growth, creep or afterslip and so on, think that the fault stress change caused by mainshock is the main cause for aftershock. But other researchers, by studying real aftershock observations, think that the fault stress change caused by mainshock is not the main cause or has very weak control over the aftershocks. Pore pressure diffusion caused by mainshock fault slip is also considered as an important incentive for aftershocks. There is a relationship between the frequency of aftershocks and pore pressure changes. Dry rock pressurized in physical experiment can produce acoustic emission sequence similar to mainshock-aftershock sequence type earthquake. Though fluid plays an important role in aftershock activities, it is not the essential element for aftershock. Overall, there is no single model which can fully explain the phenomenon of aftershock activity.
Assuming the rupture of the residual asperities inside the mainshock rupture plane randomly leads to the aftershocks, the size of the residual asperities conforms to fractal distribution, and the rupture or instability strength of the residual asperities accords with the lognormal distribution. Taking the postseismic stress relaxation as the mechanical load, the loading stress attenuates according to negative exponential law. Taking the Coulomb failure as the judgment criterion of the instability, combining the mechanical interactions among the residual asperities, the artificial aftershock sequence, including occurring time, location and magnitude, is simulated under different conditions. The agreement between output and the actual statistical characteristics of aftershock activities is detected by G-R relationship and modified Omori formula as a basis for further adjustments to the model parameters. On this basis, the influences of the medium viscosity properties on aftershock activities have been discussed.
The results show that viscosity coefficient of rheological properties of the lower part of the lithosphere has an important effect on the duration of aftershock activity. The viscosity coefficient of the lower part of the lithosphere controls the duration of the aftershock activity, the lower the viscosity coefficient, the sooner the stress relaxation of the lower lithosphere, and the faster the loading rate to the upper part of the lithosphere, the shorter the duration of the aftershock activity. On the contrary, the higher the viscosity coefficient, the slower the loading rate to the upper part of the lithosphere, and the longer the duration of the aftershock activity. This simulation conclusion is consistent with the observed result. The viscosity coefficient as one of the important lithosphere physical parameters controls the decay rate of aftershock activity. Under this model conditions, p value, the decay rate of modified Omori law, changes with the viscosity coefficients in a negative exponential function. The relationship that the viscosity coefficient is lower and the decay of aftershock sequence is faster provides a reference for the study of the main influence factors of aftershock decay. The relationship corresponds to the observation that the decay rate of the aftershock sequence shows a good positive correlation. The b value of the G-R relationship of aftershock sequence characterizes the ratio relationship of large to small earthquakes. The modeling studies suggest that the G-R relationship of the aftershock sequence is irrelevant with the viscosity coefficient, but mainly controlled by the size distribution of the residual asperities. In another word, it is mostly correlative to the heterogeneity of tectonics and medium.