On the basis of effectively handling the gross errors of landslide monitoring data and fully considering the characteristics of landslide monitoring data, we develop a deep learning landslide displacement prediction model combining time series decomposition and similar component reorganization. First, we deal with the significant gross errors of landslide time series monitoring data using the isolation forest algorithm, and then comprehensively analyze its smoothness, autocorrelation, and normality to determine the optimal length of input feature sequence. Second, the non-stationary landslide monitoring data are decomposed into multiple smooth time series using the ensemble empirical mode decomposition(EEMD) method, which is then classified into three types combining the sample entropy and K-means algorithm, namely high, medium, and low frequency. Finally, comparing the prediction accuracy of different neural networks, the prediction models suitable for three types of time components are constructed respectively, and then the prediction results are superimposed to realize the high-precision prediction of landslide displacement. The testing results of Beidou/GNSS monitoring data of typical landslide body in experimental area show that the combination prediction model proposed in this paper has a better applicability to the landslide monitoring data containing significant gross errors, and can significantly improve the prediction accuracy of landslide displacement compared with single and existing combination models.

We comprehensively use InSAR and machine learning technology to evaluate landslide susceptibility in Jinyang county, Sichuan province. The data set is updated by interpreting landslides. Based on 12 evaluation factors, we use three models including random forest(RF), support vector machine(SVM) and extreme gradient boosting(XGBoost) for training in Python environment to complete landslide susceptibility mapping. We use ROC curve to verify prediction performance. We optimize the negative samples, obtain the landslide susceptibility evaluation results after sample optimization using machine learning and update the landslide susceptibility results using surface LOS deformation rate. The results show that the three machine learning models have good zoning effect, and the mapping effect of XGBoost model is best. The accuracy of XGBoost model after sample optimization is highest, and the AUC value reaches 0.95. The surface deformation rate obtained by SBAS-InSAR can reduce zoning errors and give timeliness to landslide susceptibility evaluation.

Based on the mobile Beidou deformation monitoring system at Xiluodu hydropower station, we per- form the horizontal deformation simulation experiment. The static differential positioning, post-processed kinematic positioning and precise point positioning methods are employed to compute actual displacements and evaluate accuracy. The results indicate that the accuracy of single BDS and GPS+BDS dual systems is consistent. The static differential method has the highest accuracy, with the external coincidence accuracy of dual systems and single BDS better than 0. 5 mm and 0. 7 mm, respectively. The baseline differential method can effectively eliminate interference from dam area water vapor, ionospheric and multipath errors through 24 hours long-term observations. The maximum errors of PPK for dual systems and single BDS are 0.5 mm and 1 mm, respectively, while under dynamic PPP are 1. 7 mm and 1 mm, respectively. This shows thatPPK and PPP can also achieve deformation monitoring results close to 1 mm, demonstrating significant application potential.

We combine the latest horizontal and vertical GPS velocity fields in the southeastern margin of Qinghai-Xizang plateau, based on the strain solving method of spherical wavelet multi-scale analysis, consider the weight relationship between regional seismic moment and strain distribution, obtain the three-dimensional crustal strain characteristics of the southeastern margin of Qinghai-Xizang plateau, and analyze the relationship between regional deformation characteristics and seismicity. The results show that: 1) There is a consistency between surface strain and vertical strain in the study area. The high surface strain and low vertical strain are mainly concentrated in the central Qinghai-Xizang plateau and northern Jinshajiang regions, while the low surface strain and high vertical strain are mainly concentrated in the Eastern Himalayan Syntaxis and Longmenshan regions. The spatial inverse distribution relationship between surface strain and vertical strain shows that the horizontal velocity and vertical velocity obtained by GPS observation in this region are coupled, which indicates that vertical motion is also an indispensable consideration for the study of crustal deformation and seismic risk in this region. There is a good agreement between surface strain and vertical strain in terms of magnitude, which further indicates that the crustal deformation in this region can be approximated as a continuous and homogeneous elastic behavior. 2) There is a consistency between the type of strain rate and focal mechanism solution. The strain types obtained from GPS data in the southeastern margin of Qinghai-Xizang plateau can be well matched with the focal mechanism solution in this region. The source mechanism of normal fault is mainly distributed in Jinshajiang region, thrust fault is mainly concentrated in Longmenshan and Eastern Himalayan Syntaxis regions, and strike-slip fault is mainly distributed near the fault zone with significant strike-slip shear such as Xianshuihe-Xiaojiang fault, indicating that the strain characteristics obtained from GPS observation can better describe the tectonic stress field characteristics in this region. Combined with the spatial distribution characteristics of stress and strain in this region and seismic tomography results, we believe that the Longmenshan fault and Eastern Himalayan Syntaxis are in compressive strain state.

We investigate the crustal stress field in Inner Mongolia using 728 focal mechanism solutions and 14 460 P-wave polarity data derived from 2 793 earthquakes based on three different strategies. The Inner Mongolia area is divided into 1°×1° grids, and three methods are applied: a composite focal mechanism solution method based on first-motion polarity data, a Bayesian right trihedra method(BRTM) based on focal mechanisms, and a weighted fusion method of polarity and focal mechanism data. The results from polarity-based and focal mechanism-based methods show good consistency. The abundance of polarity data, compared to limited focal mechanism data, enables stress inversion in a wider spatial extent. After weighted fusion of polarity-based results with focal mechanism data, the BRTM successfully obtains stress results at 125 grid points, expanding the coverage and reducing uncertainties. The study reveals significant spatial variations in the stress field across Inner Mongolia area. The eastern regions are dominated by shear and extension, with shearing concentrated along the Daxinganling and extension characterizing the basins on both sides. Localized compression is observed near Chifeng and Tongliao in the southeast. The northern margin of Ordos block exhibits extension characteristics, while the Alxa block displays shear features. The orientations of maximum principal stress exhibit considerable differences, while the minimum principal stress directions remain relatively uniform.

Based on 963 records of intensity meters and 410 records of strong motion seismographs of 37 earthquakes of M_L3.2 to 5.9 in Beijing-Tianjin-Hebei and its surrounding areas from 2019 to 2022, we use the P_d method to establish 36 magnitude estimation models between P_d amplitude within 2 to 7 seconds after P-wave arrival time of vertical waveform and magnitude. By fitting the standard deviation, the optimal filtering frequency bands of intensity meters and strong motion seismographs are determined to be 1 to 3 Hz and 0.5 to 3 Hz respectively. Meanwhile, we establish the magnitude estimation model of intensity meters and strong motion seismographs within 3 seconds after P-wave arrival time using P_d method. The parameter fitting results show that fitting slope and correlation coefficient are high, the standard deviation is small, and the overall reliability is high. The results of reverse and new data verification indicate that the magnitude deviation of P_d method is mainly within ±0.5, which shows that the two models established in this paper are reasonable and can be well applied in Beijing-Tianjin-Hebei region.

Based on the catalogue data from Fujian digital seismic network, a total of 3 720 earthquakes that occurred in Fujian from 2010 to 2023 are selected to study the local magnitude (M_L) deviations. The magnitude residual method is used to obtain a new calibration function (R₃(Δ)), station correction values (S_j) and medium propagation direction correction values (S_θ) which are applicable to Fujian, enhancing the accuracy of local magnitude measurement. The results indicate that standard deviation of magnitude deviations decreases from 0.29 to 0.19 after correction. In order to improve the accuracy of earthquake warning magnitude measurement, we apply the first station's magnitude, resulting in the standard deviation of magnitude deviations decreases from 0.31 to 0.25. It is evident that the new calibration function, station correction values and medium propagation direction correction values exhibit strong practicality in Fujian.

We use T-shaped, linear, regular pentagonal, and spiral arrays at the same site for micromotion detection comparison experiments to analyze the effects of cross-correlation time window length, geometry of observation array, and their spreading on the quality of frequency-Bessel (F-J) dispersion energy imaging. The results indicate that, for the same observation time, increasing the cross-correlation time window length can effectively distinguish the fundamental and higher-mode dispersion curves, but it will reduce the number of superposition segments, resulting in lower dispersion spectrum resolution.The F-J method does not have special requirements for the arrangement of observation array and is suitable for random and linear distributions. When the array aperture is the same, the quality and resolution of dispersion imaging for an observation array with a greater number of array spacing combinations will be higher. The F-J method has certain requirements for the deployment range of observation array. For example, when the deployment range is too small, the quality of dispersion imaging using the F-J method is poor and it is unable to separate higher-mode dispersion curves.

The dynamic evolution characteristics of regional gravity field at different time scales before the Wushi M_S7.1 earthquake are obtained by using the mobile gravity observation data of southern Tianshan area from 2020 to 2023, and the approximate field source depth corresponding to each order wavelet gravity detail is obtained by power spectrum analysis method. The results show that: 1)Before the Wushi M_S7.1 earthquake, the gravity changes in Wuqia-Bachu and Aksu areas show an obvious four-quadrant distribution, and the epicenter is located at the edge of four quadrants and near the zero line. 2)The wavelet transform results of gravity field in southern Tianshan area from 2020 to 2023 show that before the Wushi M_S7.1 earthquake, the gravity changes in Wuqia-Bachu area shows obvious four-quadrant distribution, and the epicenter is located at the edge of four quadrants and near the zero line.

We use the precise clock offset data released by IGS to analyze the key performance indicators of six GPS-Ⅲ satellite clocks that have been launched and operated in orbit, including frequency accuracy, frequency drift rate, and frequency stability, and compare them with the previous GPS clocks. The results show that rubidium clocks carried by GPS-Ⅲ are generally stable in operation and have good data quality, which is related to the improvement of anti-interference ability. There is no significant change or difference in time frequency performance, which is basically consistent with the previous generation Block ⅡF rubidium clocks. The frequency stability of GPS-Ⅲ rubidium clocks are better, which helps to improve the accuracy of PNT services. The research can provide reference and guidance for the upgrading and transformation of other GNSS system satellites.

The solar pressure perturbation, as the largest non-conservative force experienced by in-orbit navigation satellites, is an important error source for precise orbit determination. Currently, most solar radiation pressure models are established for global positioning system(GPS) satellites, and there is relatively little adaptability analysis of solar radiation pressure models for precise orbit determination accuracy of Beidou-3(BDS-3) satellites. We use observation data from the multi-GNSS experiment(MGEX) to conduct orbit determination experiments based on five solar radiation pressure models, including ECOM1-9, ECOM1-7, ECOM1-5, ECOM2, ECOMC. The results show that the applicability of solar radiation pressure model varies for satellites developed by different manufacturers. For satellites developed by China Academy of Space Technology(CAST), the ECOMC model exhibits optimal applicability in radial, tangential, and normal directions. For satellites developed by Shanghai Engineering Center for Microsatellites(SECM), the radial, tangential, and normal directions have the best accuracy for ECOM1-5, ECOMC, and ECOMC models, respectively. For 3D entirety, the ECOMC model is recommended for orbit determination of both CAST and SECM satellites. From the 24-hour prediction results, for CAST type satellites, the accuracy of ECOM1-9, ECOM1-7, ECOM1-5, ECOM2, and ECOMC models is 23.3 cm, 20.6 cm, 17.2 cm, 21.8 cm, and 10.4 cm, respectively. For SECM type satellites, the accuracy is 25.7 cm, 19.1 cm, 15.9 cm, 12.9 cm, and 11.9 cm, respectively.

We propose an adaptive stochastic model for low-cost receiver precise point positioning(PPP) in complex environments. The model improves PPP performance by adaptively adjusting the weight ratio of pseudorange observations and carrier phase observations of all GNSS satellites in a single epoch. We take the Hexinxingtong UM982 low-cost GNSS receiver to conduct PPP tests in three complex environments, including tree shade, high buildings and glass walls. The static PPP results show that the positioning accuracy of adaptive stochastic model is improved by 24%, 45% and 50% in three complex environments, respectively, and the convergence time is shortened by 49%, 27% and 24%, respectively, compared with the traditional empirical weighted stochastic model. The kinematic PPP results show that the positioning accuracy of adaptive stochastic model is improved by 35% in complex environment.

To reduce the positioning errors of global navigation satellite system(GNSS) in complex environments, we propose a method that combines high-precision points with distance intersection to accurately estimate the coordinates of an undetermined point. The observation equation is constructed as a nonlinear Gauss-Helmert model. To address the nonlinearity within this model, we introduce a back-propagation(BP) neural network for auxiliary processing. Compared with the traditional linearization methods, the BP neural network can effectively fit complex nonlinear functional relationships. The simulation experiments and actual measurement results show that this method can significantly reduce the impact of complex environments on positioning accuracy, and improve the positioning accuracy in E, N and U directions by 78.1%, 72.8%, and 79.2%, respectively.

The elevation datum is an important component of the surveying and mapping datum, providing data support for maior national proiects, Addressing the issue of accurately monitoring elevation datum affected by factors such as surface deformation, we present a study to analyze surface deformation based on time-series InSAR data and leveling time-series data. The study conducts periodic re measurement of elevation datum in deformation area, achieving dynamic monitoring primarily through updating elevation datum, point usability analysis, and point subsidence analysis. Taking Shandong province as an example, we carry out an experiment, The results show that the method proposed in this paper can effectively reduce monitoring costs and difficulties, yield reliable results, significantly enhance the supportability and service quality of surveying and mapping datum, and provide a new dynamic monitoring approach for regional elevation datum.

We focus on the complex characteristics of groundwater level data, including nonlinear trends, seasonal fluctuations, and random disturbances, and introduce the Prophet time series prediction model developed by Facebook. The aim is to use its nonlinear trend capture, seasonal fluctuation analysis, and flexible response ability to outliers and data missing to significantly improve the accuracy of groundwater level anomaly identification. Through observation data from Beilin district seismic station in Suihua city, Heilongjiang province, it is shown that the Prophet model performs well in capturing dynamic characteristics of time series data and can effectively identify anomalies. The high fitting accuracy and predictive ability of adjusted model have been confirmed, with low prediction error and high determination coefficient. In addition, the model identifies water level anomalies related to earthquakes in earthquake prediction, providing a new perspective for earthquake precursor research. This study demonstrates the effectiveness of Prophet model in processing complex time series data, providing a new tool for earthquake prediction.

GNSS technology is widely used in landslide monitoring, but it is difficult to install monitoring equipment in complex high-risk landslide environment. UAV-dropped is expected to realize unmanned deployment of monitoring equipment. In view of the problems of UAV-dropped endurance and terrain threat in complex environment, the path planning of UAV-dropped is particularly critical as the basic mission. In this paper, the path planning problem of UAV-dropped in complex mountainous area is studied in three aspects: 3D real map construction of geological disaster area, cost function design and flight path planning algorithm. In addition, whale optimization algorithm based on adaptive weight and Levy flight strategy is applied to the flight path planning of UAV-dropped. Taking Hongyanzi landslide in Hanyuan county, Sichuan province as the study area, the path planning of UAV-dropped GNSS monitoring equipment was realized.

We focus on the apparent resistivity method in earthquake precursor observation, develop a detection device suitable for its observation system, and conduct theoretical simulation and on-site testing on the device. First, we analyze the apparent resistivity measurement system, and point out the shortcomings of existing detection systems. Second, according to the demand analysis of observation system, the corresponding detection index and circuit are designed, and the detection accuracy of instrument is designed to 0.001 level to improve the performance. Finally, the developed device is theoretically simulated using Simulink, and the impact of interference on observation results is compared and studied. Random earthquake observation stations in Gansu region are selected for on-site testing. The results show that the developed device can not only visually display the working status of apparent resistivity measurement system, but also accurately evaluate the real-time superiority and inferiority of apparent resistivity observation system, and fill the gap in detection device of apparent resistivity observation system and provide great convenience for frontline earthquake workers.