Using interpolation cut method, the expanded gravity anomalies of ideal model on the surface at different scales is separated into several gravity anomalies caused by strata at different depths. Based on separated results, the effects of expanding edge length on gravity anomalies separation are studied from two aspects: the difference of gravity anomalies at different scales and depths, and the mean square deviation between the expanding edge scale and the gravity anomalies separation. The results show that 1) The region boundary effect of ideal model data is related to the expanding edge length, the depth of cutting separation, location, scale and intensity of geological abnormal bodies. 2) In selecting the expanding edge scale, accuracy and computational efficiency should be taken into consideration. The expanded edge size should not be too large, and the expanded edge size and maximum cutting depth for gravity anomalies separation should at least meet the requirements of 2∶1; this significantly improves the boundary effect. When cutting and separating the boundary with large gravity anomalies,it is necessary to further expand the boundary on the basis of twice the scale of expansion. 3) The spatial distribution of the boundary effect is not evenly distributed from the boundary of the data area to the central area, but is related to the scale and spatial location of the anomalous body.

Based on the massive long-term observation data from existing geomagnetic stations, this paper disregards the formation of geomagnetic field and the complex mechanisms of many influencing factors, and analyzes the temporal and spatial correlation of the variable geomagnetic field, mining the regular information contained in the data. It further constructs the prediction model of geomagnetic field based on the BP neural network. The forecasting data of the established model is validated by the observation data of geomagnetic stations, and the result shows that root mean squared error of the 100 arbitrarily chosen validation data is 4.8 nT, and this solution accuracy can meet the need of general scientific research.

We adopt the gravity observation data of the southwestern margin of Ordos to train the long short-term memory recurrent neural network (LSTM). The results show that LSTM can obtain good estimation results based on limited data. Through comparing and analyzing the estimated results of LSTM and ordinary Kriging method based on free-air gravity anomaly data, we find that the estimation ability of neural network is better than ordinary Kriging method, but Kriging method performs better in terms of computing efficiency. Using the free-air gravity anomaly data to estimate the entire area, we show that LSTM method is significantly better than Kriging method. Adding elevation data as a constraint can effectively improve the accuracy of free-air gravity anomaly field estimated by LSTM method.

Based on the continuous observation data of seven gPhone gravimeters in Fujian gravity network, we analyze the coseismic response characteristics of earthquakes with different epicenter distances and magnitudes and the maximum epicenter distance of seismic response capability. The results show that: 1) The seismic response capability of Fujian gravity network are affected by magnitude and epicenter distance. In terms of seismic response ability, Quanzhou station is superior to Putian station, Xiamen station is superior to Fuzhou Station, Zhangzhou station and Longyan station are superior to Nanping station. 2) Fujian gravity network instrument has good stability and high sensitivity. Compared with CMG-3ESP 120S seismometer, it has a wide seismic frequency band, and its duration of coseismic response is highly positively correlated with wave amplitude. 3) The maximum epicenter distance of seismic response capability(Δ) and magnitude (M_S) of Fujian gravity network has a corresponding formula: M_S=-0.218(logΔ)3+2.056(logΔ)2-4.594logΔ+6.459.

This paper uses the observation data for the past eight years recorded by fixed and mobile stations of the Xinjiang seismic network in the south Tianshan area. We carry out local earthquake travel time tomography technology to obtain the three-dimensional crustal velocity structure, and use the three-dimensional velocity structure to relocate local earthquake events. After relocation, the root mean square of the travel time residual decreases from 1.29 s to 0.64 s, and the uncertainty of the source location is about 0.1 km. The three-dimensional velocity structure shows that the velocity structure in the region has obvious lateral and vertical inhomogeneities. At 5 km, it is a high-speed anomaly north of the Artux anticline and the Keping fault, a low-speed anomaly in the south. The velocity ratio is dominated by high values on the whole. At the depths of 15 km and 20 km, the P-wave velocity in the study area is dominated by high-speed anomalies. At the depth of 25-30 km, the P-wave high-speed anomalies in the area turn into obvious near north-east with low-speed anomalies, the velocity ratio is also dominated by the low value anomalies in the NNE direction. The seismic activity in the area has a strong correspondence with the crustal velocity structure. The source is mainly concentrated in the high speed region at the junction of high and low P wave velocity. The velocity ratio also tends to the side of the high velocity ratio. There is a low-speed band under the focal area, which provides a strong environmental foundation for the accumulation and release of strain energy.

We obtain the seismic-source spectral parameters of seismic energy, corner frequency, seismic moment and apparent stress by spectral analysis of small and medium earthquakes in Shizhu and Wulong areas. We comparatively analyze the differences of seismic-source spectrum parameters of the Shizhu M4.5 earthquake and the Wulong M5.0 earthquake. The results show: 1) Corner frequency and apparent stress in Shizhu area are lower than that in Wulong area under the same magnitude conditions, which may be related to the closer distance from Shizhu to the Three Gorges Reservoir area. 2) The apparent stress in Shizhu area presents two years rising process, decreasing slightly before the main shock, and corner frequency also increasing before the earthquake. 3) The apparent stress in Wulong area presents abnormal changes of “rise-descent-rise” before the Shizhu M4.5 and the Wulong M5.0 earthquakes, and corner frequency also increases before the two earthquakes. 4) The apparent stress of the whole study area has undergone an ascending change process before the Shizhu M4.5 and the Wulong M5.0 earthquakes, and the stress in this area remains high after the earthquakes.

In order to explore the influence of ENSO events on GNSS ZTD(PWV) and their relationship, this paper takes Hebei province as an example to conduct a study on the influence of ENSO events on GNSS ZTD and its periodic variation. Firstly, using the method of fast Fourier transform to select the common cycle of southern oscillation index(SOI) and GNSS water vapor, and we use the wavelet transform to extract the common cycle of SOI and GNSS water vapor in high frequency. We analyze the correlation between the reconstructed high frequency term and SOI, finding that the negative correlation between SOI and GNSS ZTD in Hebei province. We conclude that the ENSO event is related to the periodic variation of GNSS ZTD. Then, using the method of fast Fourier transform to extract the GNSS ZTD change cycles under ENSO events and normal climate and analyze the influence of ENSO events on the change of GNSS ZTD cycle. The result shows that ENSO warm event(El Niño event) has a significant influence on the longest significant cycle of GNSS ZTD. The longest significant cycle of GNSS ZTD is slightly affected by ENSO cold event(La Niña event).

Considering that ionospheric total electron content(TEC) presents as non-stationary, nonlinear and  high noise characteristics, we use ionospheric TEC data provided by international GPS service(IGS) to predict 3 days TEC using the back propagation(BP) neural network model and the ARMA model. We compare the prediction performance and accuracy of each model in different seasons, different ionospheric active intensity and different sample lengths. The results show that the two models can well reflect the change characteristics of TEC in different seasons, among which ARMA model is slightly better than BP neural network in spring and winter. In the quiet period, the average relative accuracy of the two models is 87.3 % (BP) and 87.5% (ARMA), which means the predictive effect is similar. In the active period, the average relative accuracy of the two models is 78.5% (BP) and 75.5% (ARMA). The accuracy of BP neural network is 3% higher than that of ARMA model. With the increase of sample length, the accuracy of BP neural network model reaches the maximum on the 21st day, the prediction accuracy of ARMA model decreases with the increase of sample length.

GPS cycle slip is one of the important factors affecting the accuracy of monitoring results. Its multi-value and the method of judging the cycle slip of each frequency are particularly important. In this paper, for the detection of dual-frequency cycle slip, we use ΔΦ=ΔN₁－(λ₂/λ₁)ΔN₂ as the master equation and propose a method to detect the cycle slip using the Grubbs criterion, restricting the ΔN₁、ΔN₂ search range of dual-frequency cycle slip to calculate the integer solution satisfying the master equation. Through the verification of GPS data with a sampling rate of 1 s, 5 s and 15 s, the method in this paper not only solves the problem of multiple values, but also accurately detects the cycle slip at all frequencies with a success rate of 100%.

The Gaussian sum filter refines the non-Gaussian noise stochastic model by Gaussian mixture model(GMM) to improve estimation accuracy. However, the dynamic and complex of the navigation measurement environment brings time-varying characteristics to non-Gaussian noise, resulting in distortion in the filter solution. To solve this problem, we propose a Gaussian sum extended Kalman filter(GSEKF) algorithm, which is improved by adaptive estimation of displacement parameters. We discuss the influence of GMM displacement parameters on the fitting accuracy of non-Gaussian noise, and GMM is refined by displacement parameter adaptive method. These improvements make the GSEKF algorithm more stable. The experimental results show that the proposed algorithm has less fluctuation and stronger anti-interference ability compared to the traditional Gaussian sum filtering algorithm when there is time-varying non-Gaussian noise in GNSS/SINS measurement model, which can further improve the estimation accuracy and stability in practical application.

From the perspective of CNES BDS real-time PPP users, this paper studies the orbit/clock difference between the BDS real-time and archival products for 10 consecutive days, and analyzes their impacts on BDS PPP. The results show that the orbit difference between these two products is below 1 mm, but the clock difference reaches ±0.1 ns. These two products are applied to BDS static PPP of four IGS MGEX stations. It is found that the difference between the two products can impact the PPP convergence speed, and the positioning accuracy difference after convergence can reach 3 cm. Therefore, we cannot ignore the difference between CNES real-time and archival orbit/clock products.

Aiming at the problem that satellite clock bias cannot be accurately modeled, this paper applies the Elman neural network with strong memory function and powerful computing ability to satellite clock bias prediction and proposes an Elman model suitable for satellite clock bias prediction. First, we perform one-difference process on the original clock bias data. Then, we select the appropriate neural network structure to establish the Elman clock bias prediction model with the best forecast effect. Finally, precise clock bias data provided by the international GNSS service(IGS) is used to predict GPS satellite clock bias, and comparison analysis is performed with the quadratic polynomial model, the polynomial model with additional periodic terms, and the gray system model. The experimental results show that the accuracy of  1-day, 7-days, and 30-days clock bias predictions of the Elman model have been significantly improved, reaching sub-nanosecond, nanosecond, and microsecond levels, respectively, proving that clock bias prediction performance of the model is superior to the three commonly used models and establishing the feasibility of the new model in satellite clock bias prediction.

High precision vertical deformation velocity field can be obtained by long-term GPS continuous observation, but it is not clear whether GPS campaign observation data can be used for vertical deformation monitoring. In this paper, we process the observation data of 260 continuous stations and 2 000 GPS campaign stations in Chinese mainland from 1999 to 2019. The time series and velocity field of each station are obtained. We compare and analyze the vertical deformation velocity field of 226 continuous stations and 226 campaign stations. The results show that the error of vertical deformation rate of 95% of continuous GPS stations are less than 0.5 mm/a and 50% of campaign GPS stations reach more than 1 mm/a. The residual value of the vertical deformation rate of about 40% of campaign stations and continuous stations are greater than 2 mm/a, and the vertical deformation trend of about 50% of campaign GPS stations are inconsistent with that of continuous stations. The vertical deformation monitoring accuracy of campaign GPS stations in the western region is higher than that in the eastern region, which may be related to the differences in observation environment and station stability between the eastern and western regions. Based on the analysis of the accuracy and residuals of vertical deformation velocity field of GPS continuous stations and campaign stations, we consider that the campaign GPS observations from mainland China are not suitable for monitoring the area with vertical deformation less than 2 mm/a.

In this paper, we use the GPS velocity fields of 1999-2007 and 2009-2013, and the negative dislocation-block model of TDEFNODE to invert fault locking and slip deficit velocity of the Red river fault zone. The results show that the Red river fault zone is dominated by right-lateral strike-slip, with some extension and compression. Before 2008, the average strike-slip rate and dip-slip rate of the Red river fault zone is 3.8±2.5 mm/a and 1.2±2.5 mm/a, the northern and middle segments exhibit tension, and the southern segment exhibits compression. The locking fraction of the northern and middle segments between 0 to 15 km depth are from 0.8 to 0.99, and the slip deficit rate is 4.5 mm/a, for which it is easy to accumulate strain energy quickly. The locking fraction of the southern segment between 0 to 10 km depth is from 0.8 to 0.95, and the slip deficit rate is 4.5 mm/a. After 2008, the fault locking and slip deficit rate in the middle segment decrease, and the fault locking and slip deficit rate in the southern segment increase.

Taking the Wuhan Greenland Center, which is under construction, as the research object, a set of technical methods for monitoring and analyzing the dynamic characteristics of super high-rise buildings are established. Based on the proposed methods, we extract the dynamic characteristics of the building, such as dynamic displacement time series, amplitude and significant frequency during the monitoring period. We analyze the correlation between the displacement and temperature change at the roof of the building based on grey relational analysis method. The results show that the maximum horizontal amplitudes at the top, 400 m height and 300 m height of the building in monitoring direction, are 5.12 mm, 4.22 mm and 3.11 mm respectively, and the standard deviations reach the submillimeter level. The natural frequency of the building is detected as 0.200 2 Hz. The building is also significantly affected by other frequencies because of construction vibration and temperature. For example, the displacement time series of monitoring direction at the roof of the building is still significantly affected by the frequencies of 0.308 2/8.033 9 Hz. The grey relational grade between displacement and temperature time series at the roof of the building is close to 1, which indicates the displacement may be closely related to temperature change. 

Based on induced ordered weighted averaging (IOWA) operator, we combine the difference autoregressive integrated moving average (ARIMA) model and Holt-Winters exponential smoothing model. We use the SBAS-InSAR monitoring value to predict mining area surface subsidence, and compare this prediction with the results of each single model. The results show that the prediction accuracy of the combined model based on IOWA operator is significantly improved compared with that of the single model. For the combined model, the mean square error (MSE) and the mean absolute error (MAE) of each point reaches 1.458 mm and 2.175 mm respectively, which can be used for the monitoring and prediction of mining area surface subsidence.

In view of the disadvantages of the SVM model, such as difficulty in parameter selection and single-point data modeling in the field of settlement prediction of foundation pit, we establish a neighbor-point PSO-SVM model. Selecting the PSO-SVM model for the optimal training sample quantity research, the results show that short-term samples have the best prediction effect. We introduce the settlement deformation value of neighbor points as a factor affecting the settlement of foundation pit into the improved PSO-SVM model. The example shows that the fitting accuracy of the PSO-SVM model considering the neighbor points under short-term sample data is better than that of the PSO-SVM model. The prediction accuracy is poor under medium and long-term sample conditions. Aiming at this shortcoming, we propose a combination of multi-scale one-dimensional wavelet decomposition function and Cauchy distribution function to improve the PSO-SVM model that takes into account the neighbor points. The experimental results show that the improved PSO-SVM model effectively solves the difficulty in parameter selection and single-point data modeling. The model is suitable for the prediction of settlement deformation under different sample sizes, and has high prediction accuracy.

In the process of micro-seismic signal acquisition, a large number of interference signals with different frequencies make it difficult to pick up the primary arrival of the signal. We propose a method based on empirical wavelet transform(EWT) combined with EWT component threshold reconstruction rule and singular value decomposition(SVD) technology for signal denoising. The component signals are obtained by the EWT method, which decomposes the micro-seismic signals with the characteristics of adaptive decomposition and conquering modal mixing. For the high signal-to-noise ratio signal, the reconstruction of inherent modal components with a correlation coefficient greater than 0.3 and a variance contribution rate greater than 15% has a better denoising effect. For the low signal-to-noise ratio signal, based on the high signal-to-noise ratio denoising method, we propose a new denoising method using SVD to denoise the high-frequency components and reconstructing with low-frequency effective components decomposed by EWT. Through experimental analysis, the signal-to-noise ratio of the construction signals with different signal-to-noise ratio and the actual micro-seismic signals are obviously improved. The modified energy ratio method and fractal dimension method are used to pick up the primary arrival of actual micro-seismic denoising signals for verification. The relative error of picking up is less than 1%. The test results show the effectiveness of the denoising method.

To effectively denoise the seismic data collected by the JCZ series ultra-wideband seismograph, we discuss the actual effect of wavelet threshold method in different seismic data. Firstly, this paper introduces the basic theory of wavelet method and the standard of denoising. We analyze the effect of different wavelet base selection, different decomposition, reconstruction scale and different threshold selection and processing methods on the final denoising effect of wavelet threshold method. Through the control variable method, and selecting recorded data from the seismic station in Wuhan University, we explore how to obtain better denoising effect. In order to compare the effect of the same method in two cases of small earthquakes and medium-strong earthquakes, we use the selected wavelet base and geometric scale to denoise the seismic data. In actual seismic denoising, the threshold selected based on sample number estimation has certain limitations, and the same threshold has different results in denoising in different magnitudes. Finally, the wavelet threshold is used to denoise the seismic data collected by the JCZ series ultra-wideband seismographs and we find that it is also helpful for the identification and judgment of the P wave first motion.