The noise components contained in the deformation monitoring data are difficult to effectively filter out, which leads to the unsatisfactory accuracy of the prediction results, so we propose a new noise reduction method applied in the field of deformation monitoring. First, we use VMD to decompose the original monitoring sequence into k band-limited natural mode functions(BIMF) with different center frequencies, and then directly remove the high-frequency BIMF with sample entropy greater than a set threshold as the noise component, and finally reconstruct the remaining BIMF to obtain the noise reduction sequence. The effectiveness and feasibility of the new method are verified by simulation and engineering examples; the results show that compared with EEMD and CEEMD, the new method has the best evaluation index, and the noise reduction results obtained by the VMD-SE method can provide a reliable basis for further deformation analysis and prediction.

The cross-sea bridge system is disturbed by external influences, and its deformation is accompanied by chaos. Chaotic identification in bridge deformation monitoring data is realized, delay time of time series is calculated in C-C method, the G-P method is used to obtain the best embedding dimension. Bridge deformation monitoring data is compared with the obtained time delay. Spatial reconstruction lays the foundation for the establishment of the chaotic time series prediction model; the chaotic time series prediction model is established based on the RBF neural network, the horizontal displacement of the bridge deformation is predicted from the measured data, and we conduct comparative analysis of the prediction results of the chaotic time series based on the largest Lyapunov exponent and the measured data. The results shows that the prediction results of chaotic time series based on RBF neural network are better than those of chaotic time series based on maximum Lyapunov index; short-term prediction has good effect.

Aiming at issues such as slow learning speed and ease of falling into local extremum existing in the BP neural network, using the mixed algorithm of whale optimization algorithm(WOA) and wolf pack algorithm(WPA) to optimize the weights and the thresholds of BP neural network, we construct the WPA-WOA-BP neural network model and apply it to predict PM2.5 concentration. The experimental results prove the stability and feasibility of applying the WPA-WOA-BP neural network model to predict PM2.5 concentration. The precision of this model is higher than BP neural network, WPA-BP neural network and WOA-BP neural network.

The atmospheric water vapor conversion coefficient (K) is one of the key parameters affecting the precision of GNSS precipitable water vapor (GNSS-PWV) inversed from ground-based GNSS meteorology. Based on the data from 7 radiosonde stations in the Yangtze river delta, we calculate the K values from 2016 by numerical integration and use a multivariate linear fitting method to establish two new models. One model is elevation independent (Emardson-I), and the other one is elevation dependent (Emardson-H). Finally, we validate these two new models using K values from 2017. The experimental results show that the MAE and RMS of the Emardson-H forecasting model are 0.001 297 and 0.001 616, which are slightly better than the Emardson-I forecasting model (0.001 303 and 0.001 620 respectively). GNSS-PWV inversion precisions of the two models are equivalent, and their MAE and RMS are both better than 0.6 mm. Therefore, as for the GNSS-PWV inversion in Yangtze river delta, Emardson-I model has better efficiency advantages in real-time application of ground-based GNSS meteorology as there is no need to measure meteorological parameters and elevation.

We use the GNSS occultation data of Fengyun 3C meteorological satellite independently developed by China to retrieve the atmospheric temperature profile from March 1, 2017 to February 28, 2018. We estimate tropopause height change and verify it using the sounding balloon data and ERA5 model data. With a spatial resolution of 2°×2°, China is divided into 608 cells according to latitude and longitude. For each cell, we calculate the tropopause parameters, and a grid model is established to analyze the temporal and spatial distribution characteristics of tropopause parameters in China. The results show that the tropopause parameters have obvious zonation characteristics with latitude, but not obvious change in longitude, which shows that the north-south region is asymmetric, and the east-west region is more symmetrical. The tropopause parameters have obvious seasonal characteristics.

In this paper, we analyze the difference between GLONASS broadcast ephemeris and GLONASS precise ephemeris. The results show that uncalibrated equipment delay parameter is the predominant factor of the GLONASS broadcast ephemeris clock error. Based on pseudorange residuals analysis, the IFBs of stations and satellites are separated. A modified model of IFB and pseudorange positioning of broadcast ephemeris is proposed. The modified model is verified by dynamic positioning. The results show that the precision has been improved by 51.1%, 41.7%, 48.3% in N,E,U components, respectively.

We process BDS/GPS data from 5 IGS tracking stations and vehicle experiment by adopting precise satellite orbits and clocks. The paper comparatively analyzes the effectiveness of three velocity determination methods: Doppler measurements, the carrier phase derived Doppler measurements, and the kinematic PPP velocimetry. The results show that in the simulated kinematic experiment, the precision of three methods in horizontal direction could be up to 10 mm/s, 4 mm/s, 5 mm/s, respectively. In the dynamic environment, the precision of the Doppler measurements from carrier phase and the kinematic PPP velocimetry in horizontal direction could be up to 5 cm/s and 8 cm/s, respectively. The accuracy of the two methods is limited by the velocity of the carrier. Velocity determination accuracy of BDS is comparable to GPS.

We calculate the scale parameters of the daily solutions of the IGS second reprocessing campaign with respect to the IGS14 from 7 analysis centers(ACs), analyze the scale parameters using the maximum likelihood method, and investigate the impact of 2 different GNSS antenna phase center correction models on noise characteristics, the long-term rate and seasonal signals of scale parameters. The results show that all the scale parameters exhibit correlated noise that is better represented by a white plus power law noise model. The rates estimated from the scale parameters fluctuate between -0.22 and -0.15 mm/a depending on different ACs. The seasonal signals show the significant annual signals and the weak semi-annual signals. The change of GNSS antenna phase center correction models has a significant effect on the mean scale offset, but there are no obvious variations of noise characteristics, the rate and seasonal signals.

We use principal component analysis(PCA) to analyze the coordinate time series of 224 GNSS reference stations of CMONOC. First, the original coordinate sequence of the reference station is preprocessed by mutation fitting, gross error elimination and missing data interpolation. Then, we performed PCA separately on the continuous residual GNSS coordinate time series matrix to calculate principal components(PCs) and corresponding spatial eigenvectors in three directions: N,E and U. According to the PCs of each direction and their corresponding spatial eigenvectors, we analyze the common mode error(CME), regional distribution characteristics of sites spatial response, and abnormal site impact on PCA results. The results indicate that a single PC is no longer able to reflect the whole spatial and temporal patterns of the CME in China; the first three PCs are required to be considered to analyze the CME. In addition, there are relatively uniform spatial responses in the northwest region, north China and Yunnan province, which imply water reserves vary significantly. After removing the abnormal sites, the first two PCs, especially the vertical direction, exhibit obvious variations in contribution and spatial response. The contribution rate of the first PCs increased by 2.0% (N), 3.9% (E) and 5.7% (U) respectively, while the second PCs decreased by 1.1% (N), 1.9% (E) and 6.7% (U) respectively. The spatial response of the station is significantly improved after the removal of abnormal sites.

Aiming at the problem of pedestrian trajectory tracking in an unknown environment, to achieve accurate tracking in the magnetic field, we propose an improved trajectory tracking algorithm based on pedestrian dead reckoning. We use a double threshold binarization algorithm to analyze the acceleration data in step estimation, and an adaptive fusion algorithm is proposed in course estimation. This fusion algorithm fuses the angular velocity and acceleration to obtain a heading angle based on the updating differential equation, and then determines the weight of the heading angle and the magnetometer to measure the heading fusion through the magnetic induction intensity threshold. The effectiveness and superiority of the proposed algorithm are validated by an experiment. The experimental results show that the trajectory error obtained by the proposed tracking algorithm is less than 1.1% when the total walk distance is 32 m, and the probability that the positioning error is less than 0.75 m and 1 m is 59% and 80%, which effectively verifies its excellence.

We select four waveform retrackers including SAR multi-threshold, ICE1, IceSheet and SAMOSA, to retrack the Sentinel-3A SAR waveforms in the coastal areas. We then analyze the accuracy of the four retrackers within 20 km offshore by comparison with observations from 27 tide gauges. The results show that the SAR multi-threshold retracker can retain the most valid data, and obtain the highest correlation as well as the smallest root mean square error of difference between the derived sea surface heights and the tide gauge sea levels within 6 km offshore. The SAMOSA retracker has the highest stability related to the geoid beyond 5 km from the shore, which can provide high-precision sea surface height in open ocean.

To study the activity characteristics of the Maqin-Maqu section of the east Kunlun fault, through the double-difference location algorithm, we relocate the small and medium seismic events in the Maqin-Maqu section of the east Kunlun fault from January 2009 to September 2018. Based on the CAP(cut and paste) method, we invert the focal mechanism solutions of earthquakes over magnitude 3.5 in the study area. The results show that the segmentation characteristics of seismic activity along the fault zone are obvious: the earthquake distribution around 101° E unfolds along the Awancang fault and the earthquake airspace phenomenon from Maqin to Zoige in the past 10 years has continued. The focal depth of the Maqin-Maqu segment of the east Kunlun fault is mainly distributed in the range of 2 to 30 km. The main compressive stress in the studied area is NEE, while the tensile stress is towards NNW. The results of the study provide seismological evidence for the conclusion that the slip rate of the Maqin-Maqu section weakened due to the absorption conversion of the Awancang fault in the east Kunlun fault.

We obtain the Receiver functions of 609 broadband seismic stations in south China block and adjacent areas from about 900 teleseismic events(M>5.7) during 2007 to 2012. By analyzing the waveform changes of the receiver functions, we can understand the crustal structure characteristics under each station. Using the H-κ stacking and searching method, we obtain the crustal thickness and velocity ratio beneath the stations. Results show that the crustal structure of the study area presents a gradual and complex change from coastal to inland and the crustal thickness also presents a gradual increase trend from coastal to inland and from south to north. Guangdong, Fujian, Zhejiang, Shanghai, Jiangxi, Guangxi, southern Anhui, central and eastern Hubei, central and eastern Hunan and southeastern Yunnan in the study area are mainly represented by standard receiver functions. The crust is relatively stable, and the thickness of the crust is generally below 35 km, which is close to the Earth average model. The western Hubei, the western Hunan, Guizhou and the eastern, central and northern part of Yunnan mainly show transitional receiver functions. The thickness of the crust in this area increases gradually from about 35 km to about 45 km, corresponding to the area of crustal thickening and deformation. Chongqing, Sichuan, southern Shaanxi, southern Gansu and northern Yunnan in the studied area mainly show complex receiver functions. The crustal thickness increases from 45 to 60 km, or even 70 km, corresponding to the region with intense crustal thickening.

This paper uses the 2017 M6.6 Jinghe earthquake in Xinjiang as a case. By collecting continuous waveform data from geomagnetic stations in different azimuths and different distances around the epicenter, based on the Parkinson vector method and the improved method of Chen, we calculate the Parkinson vector azimuth anomalies of each station in different frequency bands. All the vector azimuth anomalies are superimposed by multiple stations in the same frequency band to extract and analyze the short-period variation anomaly characteristics of geomagnetic fields around the epicenter before the earthquake. The temporal and spatial characteristics of geomagnetic anomalies at each station are obvious, and they all show short-term and impending anomalies. The epicenter position can be better estimated through the comprehensive superposition analysis of anomalies. The research results are significant for future short-term and imminent earthquake prediction and earthquake prevention and disaster reduction.

In this article, we briefly review the gCAP (general cut and paste) and the full-waveform inversion focal mechanism solution inversion theories using the regional seismic waveform. These methods are applied to the focal mechanism solution inversion of the Ludian MS6.5 event. The main idea of this work is to evaluate the reliability of the inversion results and to compare the two methods in real data application. According to the comparison, we find that the bad azimuthal coverage decreases the reliability of inversion results. In conclusion, the source mechanism is strike 73°-77°, dip 72°-85°, and rake 157°-180°, inversed by gCAP and full-waveform method.

We analyze the relationship between fault H2 concentration and air temperature and air pressure to obtain a regression equation using correlation and regression analysis methods. Through the R-value grading method, we test the earthquake prediction capability of fault H2 concentration regression residuals. The results show that: 1) Air temperature is the main influence factor of fault H2 concentration change in Aksu with high positive correlation. Air pressure has little effect on fault H2 concentration variation in Aksu with low-medium degree negative correlation. 2) The regression residual R value of fault H2 concentration is 0.690 7, and the earthquake prediction capability is good. The mean corresponding period of M5 earthquake prediction is 23 d. The dominant seismogenic zone is the western section of southern Tianshan.

 Based on mathematical or MATLAB language, this paper improves the geomagnetic low-point displacement method and software. This paper replaces the manual drawing with the automatic drawing technology of geomagnetic abrupt boundary to eliminate the random error. The idea is to realize automatic mapping by establishing the geometric analytical relationship between the abrupt boundary and the geomagnetic low-point time of the station network. Voronoi subdivision is done for the geomagnetic network, and the trend of the abrupt boundary is limited to the Dirichlet tessellation line of the geomagnetic network. Then the low-point time isoline is calculated by nearest neighbor interpolation. Finally, the low-point time contour map is transformed into the second order contour map, so as to extract the abrupt boundary. We replace the original GMT with the local time of geomagnetic station to eliminate the low point time deviation caused by longitude effect. The idea is to establish the functional relationship between station GMT, station longitude and local time, then input station GMT and station longitude to output the station local time. This paper takes a geomagnetic low-point displacement phenomenon on July 20, 2017 related to the Jiuzhaigou MS7.0 earthquake as an example to evaluate the improved method.

This paper improves the methods of downward continuation of gravity anomaly based on the Lagrange mean-value theorem. We apply the difference between upward continuation and linear extrapolation to obtain the precise vertical gradients of gravity anomaly used in the downward continuation. We first introduce the principles and the procedures of the proposed methods. Based on EGM2008 global gravity model, we validate the strong linear distribution pattern of gravity vertical gradients between 0 m and 10 000 m in Taiwan area and justify the linear extrapolation of these gradients. We conduct the simulated downward continuation experiment in Taiwan area and the accuracy reaches 2.34 mGal. Finally, we downward continue  the airborne gravity anomaly at 5 155 m in Taiwan to the 3 360 land gravity points and the accuracy reaches 10.13 mGal.

Applying Beidou technology in seismic observation is the developing trend of future seismic observation instrument. We elaborate on the present situation of timing technology of seismic observation instrument, provide statistics, and analyze defects of seismic observation instrument’s timing terminal. We target the proposed the split design of the “antenna & circuit” for the timing terminal based on Beidou/GPS dual mode technology. We then develop the terminal’s software and hardware to achieve the function of collecting to the seismic data collector. We verify the design and timing accuracy of the timing terminal by the test result of running in the seismic station, achieving the expected design effect.

In recent years, as well as being used in deep seismic sounding, PDS seismographs are also used in ultra-dense seismic array observations and ambient noise imaging. Unfortunately, while PDS seismographs have no real-time GNSS time synchronization service, it is critical to get seismic data with high-accuracy sampling time. This paper introduces the hardware and firmware design of a new GNSS-synchronized automatic device based on microcontroller Apollo2. Through laboratory testing and field experiments, this time synchronization and automatic control device for PDS seismographs can well meet the needs of ultra-dense seismic array observations.