The most significant feature of active faults on remote sensing images is fault lineament. How to identify and extract fault lineament is an important content of active fault research. The rapid development of remote sensing technology has provided people with extremely rich remote sensing data, and has also created the problem of how to choose suitable data for fault interpretation. In the traditional fault interpretation, people pay more attention to high-resolution optical images and high-resolution DEM, but optical remote sensing images are greatly affected by factors such as weather condition, vegetation and human impacts, and the time and economic costs for obtaining high-resolution DEM are relatively high. Due to the low resolution, the medium-resolution DEM(such as Aster GDEM, SRTM1, SRTM3, etc.)is generally used to automatically extract structural lineament, and then analyze the overall regional structural features, but it is rarely used to visually interpret active faults. ALOS-PALSAR DEM is generated from SAR images acquired by the phased array L-band synthetic aperture radar mission sensor of the Japanese ALOS satellite. It is currently a free DEM with the highest resolution(resolution of 12.5m)and the widest coverage. Based on ALOS-PALSAR DEM and ArcGIS 10.4 software, this paper generates a hillshade map and visually interprets the fault lineaments in the West Qinling Mountains. When generating a hillshade map, we set the light azimuths to be oblique or orthogonal to the overall trend of the linear structures, the light azimuths to be consistent with the slope direction of the hillslope, and the light dips to be a medium incident angle. Based on the hillshade map generated from ALOS-PALSAR DEM, this paper summarizes the typical performance and interpretation markers of fault lineaments on the hillshade map(generated by DEM), and visually interprets the V-shaped fault system in West Qinling Mountains where the research on fault geometry is limited based on the interpretation markers. The results of the research are as follows: First, this study discovers a number of fault lineament zones, including the fault lineament located between the Lintan-Dangchang Fault and the Guanggaishan-Dieshan Fault, the NE-directed fault lineament zone between the Lixian-Luojiapu Fault and the Liangdang-Jiangluo Fault, and the arc-shaped dense fault lineament zones distributed south of the Hanan-Daoqizi Fault and the Wudu-Kangxian Fault; Second, this study completes the geometric distribution images of the known active faults, such as the western and eastern sections of the Lintan-Dangchang Fault, the western and eastern sections of the Liangdang-Jiangluo Fault; Third, fault lineaments in the West Qinling Mountains exhibit a “V” shape, with two groups of fault lineaments trending NW and NE, whose tectonic transformation mainly consists of two kinds: mutual cutting and arc transition. The Lintan-Dangchang Fault cuts the Lixian-Luojiapu Fault, the Lintan-Dangchang Fault and the Guanggaishan-Dieshan Fault are connected with the Liangdang-Jiangluo Fault in arc shape, and the Tazang Fault is connected with the Hanan-Daoqizi Fault in arc shape. The research results show that ALOS-PALSAR DEM has an outstanding capability to display fault lineaments due to its topographic attributes and strong surface penetration. In circumstances when the surface is artificially modified strongly, the spectrum of ground objects is complex and the vegetation is dense, the ALOS-PALSAR DEM can display fault lineament that cannot be displayed on optical remote sensing images, indicating that the medium-resolution DEM is an effective supplement to high-resolution optical remote sensing images in the fault lineament interpretation. The research results are of great significance for improving the geometric image of the V-shaped fault system in the West Qinling Mountains. It is also the basis for further research on fault geometry, kinematics, regional geodynamics and seismic hazard.

The Sanweishan Fault is located in the front of the northwest growth of the northern margin of Tibetan plateau, a branch fault of the Altyn Tagh Fault which extends to the northwest. The latest seismic activity of the Sanweishan Fault reflects the tectonic deformation characteristics of the northern plateau. Meanwhile, it is of great significance for the seismic risk assessment of Dunhuang and its adjacent areas to understand the characteristics of earthquake recurrence. The Sanweishan Fault runs along the western piedmont of the Sanwei Shan, with a total length of 175km. The fault is characterized by left-lateral strike-slip and reverse faulting, with local normal fault features. Based on the geometry, the fault can be divided into three segments, i.e. the Shuangta-Shigongkouzi, the Shigongkouzi-Shugouzi and the Shugouzi-Xishuigou segment from east to west. Previous studies about the paleoearthquakes on the Sanweishan Fault mainly focus on the middle and east segments of the fault, while the west segment of the fault has been less studied. Meanwhile, the available research does not involve the recurrence characteristics and possible magnitude of the paleoearthquakes. Based on high-resolution satellite images, we found that the main fault has grown toward the basin and formed fault scarps in the western segment of the Sanweishan Fault. We have carried out a detailed study on these fault scarps. Based on trench excavation and chronological study on the latest fault scarps, this paper determines the sequence of the paleoseismic events on the fault and discusses the recurrence characteristics and possible magnitude of earthquake for the Sanweishan Fault.
In the western segment of the fault, through satellite image interpretation and field investigation, we found new fault scarps distributed on the alluvial fan in front of the mountain near Gedajing. We called it Dunhuang segment of the Sanweishan Fault. The activity characteristics of the fault scarps may reflect the latest seismic events in the western part of the Sanweishan Fault. Different from the sinistral strike slip of the main Sanweishan Fault, this fault segment shows the characteristics of thrust with low angle. According to the differential GPS survey, the height of the fault scarp is approximately 2.2m. The paleoseismic trench was excavated across the fault scarp. Based on the analysis of paleoseismological trenching and optical stimulated luminescence dating, two paleoseismic events are determined. Event E1 occurred at approximately(35.1±3.7)~(36.7±4.1)ka; event E2 occurred at approximately(76.5±8.8)~(76.7±8.3)ka. According to the strata offset and corresponding age, the vertical slip rate of the Sanweishan Fault is determined to be(0.03±0.01)mm/a, with a corresponding shortening rate of(0.09±0.01)mm/a. Together with the previous results, we consider that the Sanweishan Fault is characterized by segmentation. The middle and east segments may have the ability of independent rupture, and also the characteristics of cascading rupture with the Dunhuang segment. According to the existing results, we conclude that the recurrence interval for cascading rupture behavior on the Sanweishan Fault is approximately 40ka, which shows a characteristic of low slip rate and long-term recurrence. The best estimated magnitude is inferred to be in the range between M_W7.1 and M_W7.5 based on the empirical relationships between moment magnitude and rupture length.

Outlier detection is a key step in satellite gravity data preprocessing. As the theory and practice of GOCE satellite gravity gradient measurement get more and more sophisticated, the spatial resolution of satellite gravity data can reach the order of 1mgal and the accuracy of 1~2cm. However, due to the interference of various uncertain factors and the characteristics of massive observation, the satellite gravity gradient data often have some outliers. Simulation studies have shown that outliers will adversely affect the interpretation of various physical phenomena. In addition, the existing outlier detection methods have the disadvantages of high time consumption and low accuracy, which reduces the reliability of data analysis and affects the accuracy of the results. Therefore, outliers need to be eliminated. In recent years, with the in-depth development of artificial intelligence technology in earth science research and applications, many new methods and achievements in geoscience have been obtained at home and abroad. Inspired by the fact that long short-term memory networks can capture long-term or short-term information in data sequences, in this paper, a long short-term memory(LSTM)network for outlier detection of gravity gradient data is proposed. This network is a special type of cyclic neural network that can avoid long-term dependence. It adopts the special gate structure of LSTM network, trains the sample characteristics through the calculation of forgetting gate, input gate and output gate, and the LSTM network selectively updates or discards the neuron vector so as to preserve the long-term state of neurons and make LSTM network perform better on long-time series. In order to prove the reliability of extracting outliers by long short-term memory neural network method, the simulated satellite gravity data can be used for the analysis. Firstly, through the 300-order EMG96 model, the normal ellipsoid GRS80 simulates the gravity gradient data with a sampling rate of 5s and a length of 1 day, and by selecting the function whose expected value is equal to 0 and standard deviation is 0.01σ, a white noise sequence is generated, which is randomly added to the gravity gradient data, then adding outlier to the gravity gradient data sequence with a proportion of 1% and a value of 2σ, the gravity gradient data set containing white noise and outlier is obtained; Secondly, the data is normalized to the standard interval by data preprocessing, which is conducive to obtain the optimal solution. Then the gravity gradient data set is divided into training set and testing set according to the proportion of 8:2. After the data are grouped, the network structure is trained to avoid over-fitting and enhance the adaptability of the model to the samples. Through the sliding time window, the data are processed, and the neural network is easier to learn from the data set. Then, the LSTM network constructs the training module, and through the data input layer, the forward and back propagation of training parameters, changes the neuron information. After many iterative processes, the loss function of the LSTM network tends to be stable. Finally, the model is tested through the test set to obtain the final recognition result. In order to obtain higher accuracy, based on the characteristics of neural network, the LSTM continuously updates parameters, increases the complexity and depth of the network, calculates the output value at the current time, and effectively identifies the position of outlier. Compared with the traditional cyclic neural network method, the unit in LSTM records all historical cumulative information and can capture the dependence between gravity gradient time step distance and large data. On this basis, considering that the number and distribution of outliers in the measured satellite gravity gradient data are unknown, two indicators of success rate and failure rate are introduced to evaluate the effect of outlier detection and verify the effectiveness and accuracy of outlier detection method. The method in this paper realizes the outlier detection ability of long-time series observation data. The calculation results show that after the LSTM training model is applied to the test set, the prediction accuracy reaches 99.4%, and it only takes 4.26 seconds, the processing time is short, without manual intervention. In the prediction process, increasing the training data or increasing the number of LSTM neurons can improve the prediction effect, and the loss function, learning rate, number of iterations, etc. are the main model parameters affecting the prediction effect. The experimental results of outlier recognition show that LSTM model can realize feature extraction and effectively solve the problem of outlier recognition. The complexity of the original time-consuming outlier recognition technology is reduced, and the network can be supplemented with new synthetic data for training to identify new features. It has good adaptability to anomaly removal, and provides a new method to remove all kinds of anomaly interference from the actual observation data of satellite gravity.

On 27 and 30 Mar. 2013, an M_S4.5 and M_S4.7 earthquake occurred at Zigui County of Hubei Province, which were two larger earthquakes in this region since the impoundment of the There Gorges Reservoir in 2003. Characteristics of the focal and seismogenic tectonic of the two earthquakes are analysed and discussed in this paper. Moment tensor solutions of two earthquakes are inverted by the Kiwi method which uses the wideband waveform of 13 stations from Hubei and Chongqing seismic network and a six-layer crustal velocity structure model. The inverted results show that there is a good fit between the observed waveform spectrum and theoretical waveform spectrum, and the non-fitting errors of two earthquakes are less than 0. 57. These indicate that the inversion results are reliable. Rupture modes of the earthquakes are both strike-slip with a small amount of thrust component, but the former earthquake is of sinistral strike-slip and the latter is the dextral one, and the fact that there is less DC component and more ISO component in their moment tensor solutions is likely the manifestation of the effect of the reservoir water on physical properties of crustal strata. More than 500 earthquakes from Mar. 27 to Apr. 27 in this sequence are relocated by Double-Difference Method using the waveform recording of 15 sub-stations from the Three Gorges Seismic Network, and the results show that aftershocks are distributed along NNW and NE, but mainly along NE direction. Two depth profiles along NNW and NE direction show that focal depths of the sequence are from 4.5 to 10km. Two significant planes in the deep strata are formed along NE direction by these aftershocks, which are in accord with the occurrence of the NE-striking nodal plane of the focal mechanisms. The field intensity survey points out that the isoseismal line of the intensity Ⅴ meizoseismal area is an ellipse, the major axis is along NNW direction and the minor axis is along NE direction. Combining the results of field intensity survey with the tectonic setting of the epicenter area, we deduce that the rupture plane along NE direction at the north of Xiannüshan Fault is the seismogenic fault plane of the two earthquakes, the distribution of aftershocks along NE and NNW, the shape of the seismogenic fault plane and the characteristic of the depth profiles all indicate that these earthquakes are controlled and influenced by Xiannüshan Fault and Jiuwanxi Fault.

On 13, April, 2010, a great earthquake of M_W7.0 occurred in Yushu County, Qinghai Province, which is another big one in China since 2008 Wenchuan earthquake. And with seismic wave data, InSAR data and field investigations, many researchers studied the focal mechanism and source rupture process of this earthquake and many valuable results were obtained. However, there are some arguments on the rupture velocity. Some think that this earthquake is a super-shear rupture event, and some insist on opposite opinion. In order to explore whether it is a super-shear rupture event or not, this study chooses the teleseismic wave data recorded by 33 seismic stations with epicentral distances between 30～90 degrees, good azimuth coverage and high signal-noise ratio to reexamine the rupture process using Yagi's program. By comparison of different given rupture velocities in the range of 2.5～5.5km/s, it is found that rupture velocity of 4.7km/s yields the smallest normalized misfit between the observed and synthetic waveforms. And the inversion result is more in accordance with field observation. The relationship between subfault dimension, rise time and rupture velocity is discussed, which shows that the rupture velocity is not so dependent on the two parameters. And by teleseisemic analyses using an envelope deconvolution method with an empirical Green's function, the location and timing of the high-frequency event also show a rupture velocity of 4.7 to 5.8km/s, which is apparently greater than the shear wave velocity in this region. By comprehensive analyses, it can be concluded that the super-shear rupture exists in this earthquake. According to our inversion result, the strike, dip, and rake angle of this earthquake separately is 300, 88 and 4. Beach ball shows the seismogenic fault is of strike-slip type, which is consistent with the Ganzi-Yushu Fault. And the rupture extended to the surface on the northwest and southeast segments of the Yushu Fault with the length of 19km and 31km. Due to the existence of pull-apart Longbao Basin, the central part where the epicenter is did not rupture. By comprehensive analysis, super shear rupture is one of the main reasons that caused serious damage to Yushu County.

Ruichang-Yangxin earthquake is another moderate-size earthquake in the Yangxin-Jiujiang area since the M5.7 Jiujiang-Ruichang earthquake in 2005.In order to have a better understanding of the seismic activities in this area,we investigate the moment tensor solution and the seismogenic structure of Ruichang-Yangxin earthquake. Precise earthquake relocation shows that the main shock occurred on the southwestern part of the NE-trending fault and aftershocks are distributed along both NNE and NW directions. By comprehensive analysis of the earthquake distribution,isoseismal features,focal mechanism,and regional structure characteristics,it is inferred that the this earthquake is caused by the southern segment of the NNE-trending Tanlu Fault(TLF).In addition,it has close relationship with the conjugated NW-trending fault as well.

Accurate earthquake location is of primary importance in studying the seismicity of reservoir area. The well known program Velest was used to inverse the minimum 1-D velocity,hypocenter location and station correction simultaneously.We selected 529 earthquakes,which were recorded at least by 10 stations each and whose max azimuth gap was less than 180?,to re-determine the velocity model and station corrections with three different initial models used by different researchers before.Using the 1-D minimum velocity model and station correction,we relocated 2138 earthquakes in the Three Gorges area.The average focal depth changes remarkably in different areas,as 5.6km in the area near Xiannvshan Fault,2.9km in the Xietan area and 2.6km in the Badong area.Divided by the depth of 5km,the b values are different,as 1.07,1.16 and 0.90 for all,shallower and deeper events respectively. The frequency of shallower or deeper events per 15 days increases with impounding.But that of the shallower ones drops quickly and the deeper attenuates slowly when water lever keeps stable.

Based on field investigation in the reservoir head area of the Yangtze Three Gorges,combining with its seismogeological background and past research achievements,the reservoir head area is divided into 31 predictive units,and together with 8 impact factors,the possibility and magnitude of reservoir induced seismicity(RIS)are predicted using statistical forecasting model.The results show as follows:(1)it is quite possible that M_L=3.0～4.5 earthquakes will be triggered along the Jiuwanxi-Lukouzi Fault and Xiannvshan Fault in the reservoir area;(2)According to the analysis of earthquakes at home and abroad,the RIS takes place mainly in carbonatite and igneous rocks,and concentrates in karst developing segment,yet it is little possible that earthquake happens in clastic rock area.From the predictive results,it is possible to trigger M_L 4.5～6.0 earthquake in two places,the limestone area of southern Badong and the limestone area on the Gaoqiao Fault;(3)Around the Gaoqiao Fault,tectonic reservoir-induced earthquake is quite likely to occur.Geological investigation shows that the M_L 5.1 Longhuiguan earthquake in 1979 was possibly related to the Gaoqiao Fault which had a certain activity during the early period of reservoir impounding.The maximum magnitude of earthquake happening around the Gaoqiao Fault reaches to M3.3.The bigger earthquakes might be induced near the reservoir segment of Gaoqiao Fault along with the water storing to the design level.

This paper presents the results of analysis of observed data of water levels from 8 observation wells, water temperature from 4 wells and soil radon from 4 well sites obtained during the year of 2001. Based on the work, the normal behaviors of each observation item are established and their responsibility to earthquake event are tested. The results indicate that all the groundwater levels from 8 wells are able to reflect the solid tide. The sensitivity of the response of groundwater to volumetric strain of the earth crust in eight wells ranges from 10^-9 to 10^-10 volumetric strain/mm. Groundwater levels in most wells within the well network have barometric effects and are influenced to some extent by the rainfall. The water temperature in 4 wells is relatively stable, the daily change of which is ranging within several degrees a thousandth. The soil radon in 4 well sites fluctuates dramatically, and the annual change ranges from ten to several tens Bq/L. The anomalies of water levels in 6 wells, water temperature in one well, and soil radon in one well site were observed before the M_L3.6 earthquake occurred at Meijiahe, Zigui County on October 16, 2001. This may indicate that the Three Gorge observation well network has a relatively high capability to respond earthquake event.

The Three-Gorges well network is the first professional network in China established for the purpose of monitoring the groundwater precursors of reservoir-induced earthquake. It is composed of 8 observation wells. In addition to observe the change of water level, 4 wells are enabled to measure water temperature and soil Radon, 3 wells are enabled to measure rainfall, atmospheric pressure and temperature, and 1 well can be used to observe the water level of the reservoir. The observation items are totaled 26. The present paper introduces in detail the technical system of the observations used in the Three-Gorges well network, including the technical system of the whole network and each well, the technical index, installation and adjustment of every sort of instruments and equipments, as well as the corresponding supporting facilities such as the power supply, communication and lighting arrester. In addition, the running of the technical system of these observations and its preliminary results are also introduced briefly.