Applying the structural-stratigraphic analysis method, the shallow seismic profile across the fault in southwestern Haizhou depression is interpreted through borehole constraints. On this basis, five important structural interfaces of Yongji subsection of Haizhou section of northern Zhongtiaoshan fault are traced, the activity patterns of fault and sedimentation rates of strata are quantitatively analyzed, and the neotectonic activity episodes in Yongji subsection of Haizhou section of northern Zhongtiaoshan fault are established.

In 2019, the Beiliu M5.2 and Jingxi M5.2 earthquakes occurred successively in different tectonic regions of Guangxi. To obtain the parameter characteristics and probability prediction results of aftershock sequences based on statistical seismology models, we use the ETAS model to calculate the model parameters of two earthquake sequences, and analyze the probability of occurrence and occurrence rate of aftershocks and evaluate the prediction efficiency. The results show that both Beiliu and Jingxi earthquake sequences exhibit high α-value in early stage, indicating weaker ability to trigger secondary aftershocks. The p-value shows that two earthquake sequences decay rapidly in early stage. The prediction results of probability of occurrence and occurrence rate are consistent with the actual low occurrence rate of aftershocks. The ETAS model performs better in predicting aftershocks of Jingxi earthquake sequence than Beiliu earthquake sequence. The strike-slip Beiliu earthquake sequence has lower α-value than the ultra-shallow thrust-type Jingxi earthquake sequence in early stage, which may be greatly affected by the high geothermal heat flow, but the p-value is relatively high. The μ-value of Jingxi earthquake sequence is significantly greater than that of Beiliu earthquake sequence in early stage, suggesting more significantly affected by fluid triggering.

In order to verify the feasibility of CPS software in the strain wave simulation, and examine whether CPS is useful for seismic source parameter inversion constrained by borehole strain waves. We use CPS and QSSP software to calculate the strain waves of virtual stations, respectively. The body wave and surface wave obtained from both methods exhibit a high degree of correlation. Taking the M_W7.0 earthquake in Kyushu Island, Japan on August 8, 2024 as an example, the strain waves from six stations with epicenter distance greater than 30° are simulated using CPS software. The results indicate that the simulated waves show good agreement with the observed S-waves, and the average correlation among the five stations is significant. Theoretically, S-wave combined with CPS software can be used to invert the seismic source rupture process.

We use the joint tomography of first-arrival and reflected waves based on the adjoint techniques and conventional optimization methods to compare the combination methods of three commonly used optimization methods (CG, CGDESCENT, L-BFGS) and line search methods (LNS and CWI), and test the six combination algorithms using anomaly body model and chessboard model. The results show that: 1) For high-speed mutation models, CGDESCENT+LNS has slightly higher accuracy than L-BFGS+CWI, but CGDESCENT did not converge in low-speed models, while both L-BFGS+LNS and L-BFGS+CWI perform well. CGDESCENT is more suitable for large-scale models. 2) Under the same conditions, the accuracy of CWI is slightly higher than that of LNS, but CWI is prone to local minimum and is very unstable.

A two-dimensional strain tensor model is established to analyze the spatio-temporal deformation characteristics of a landslide. The results indicate that the landslide deformation exhibits spatio-temporal heterogeneity. During the initial monitoring stage, the front edge of landslide was dominated by compressive deformation, while the middle and rear edges showed a combination of expansion and shear deformation, with intensity rapidly decreasing over time. After the reinforcement of retaining wall, a brief period of strong compressive deformation occurred at the front edge of landslide, with a maximum compressive strain rate of -6 μstrain. In April 2021, a localized landslide occurred at the rear edge of landslide, characterized by strong localized tensile and shear deformation, with a maximum shear strain rate of 150 μstrain and a surface expansion strain peak of 30.9 μstrain. After the sliding event, only the surface expansion intensity diminished rapidly. Moreover, strain extremes along the crack between PB03 and PB04 persisted over time, showing a polarity reversal six months before the landslide, revealing potential risks even during the macroscopically stable deformation stage.

A dam deformation prediction model is proposed that integrates secondary mode decomposition with butterfly optimization algorithm(BOA)-optimized light gradient boosting machine(LightGBM). First, the training set data is decomposed using complete ensemble empirical mode decomposition with adaptive noise(CEEMDAN), and the composite entropy of decomposed subsequences is calculated. Second, the K-means clustering algorithm is used to categorize the decomposed subsequences into high and low frequency components, and variational mode decomposition(VMD) is applied to high frequency signals. Finally, the BOA-optimized LightGBM model is employed for prediction. The case studies demonstrate that this method effectively processes deformation data and enhances data stationarity, achieving significant accuracy improvements compared with conventional approaches. The nMAPE, MSE, and MAE indexes improve 16.2%-22.5%, 16.8%-28.1%, and 16.2%-22.5%, respectively.

To address the issue of reduced accuracy in antenna rotation inversion due to the noise in ground-based phase wind-up(GPWU), we propose a real-time yaw angle inversion model that combines low-pass filtering, an improved reset Kalman filtering, and elevation angle weighting. After epoch differencing and low-pass denoising of the geometry-free measurements, the model identifies the time points of yaw angle change based on the variations of epoch-differenced geometry-free observations. We then use the reset Kalman filtering to predict yaw rate in real-time. Finally, the model optimizes the station antenna rotation rate calculated from all satellites, weighted by elevation angles, to obtain the final antenna rotation rate. The results show that the proposed method achieves a yaw rate prediction accuracy of 0.587° before Kalman filtering convergence, which improves to 0.245° after convergence, with a time error of less than 1 second.

High-precision polar motion prediction is crucial for applications such as real-time satellite orbit determination and deep space probes navigation. We combine the singular spectrum analysis(SSA) and weighted least squares(WLS) models, which takes into account the temporal variability of periodic components of polar motion. Combining with the auto-regressive(AR) model, we prove the effectiveness of SSA+WLS+AR model in improving the accuracy of polar motion prediction. We analyze the impact of detrending, the length of basic sequence, and the weighting method on the final prediction accuracy of the SSA+WLS+AR model. The results show that in the X direction, the model combination of "no detrending+50-year basic sequence+inverse weighting" has the highest prediction accuracy, while in the Y direction, the model combination of "detrending +50-year basic sequence+inverse weighting" has the highest prediction accuracy. When the prediction span is from 150 to 360 days, the prediction accuracy of SSA+WLS+AR model is better than that of bulletin A provided by international Earth rotation service(IERS), with maximum improvements of 33% and 26% in the X and Y directions, respectively. The prediction results show that the first-day prediction accuracy of polar motion is significantly better using the C04 20 sequence compared to the C04 14 sequence.

To address the low accuracy of zenith tropospheric delay(ZTD) models and the scarcity of water vapor data in Antarctica, we use ZTD data from six global navigation satellite system(GNSS) stations in Antarctica, ECMWF reanalysis v5(ERA5) data, and radiosonde(RS) data from 2018 to 2021. First, the ZTD accuracy of UNB3m, EGNOS, and GPT3 models in Antarctica is analyzed. Then, the ZTD correction and PWV conversion models are constructed separately based on extreme gradient boosting(XGBoost). The results show that the average root mean square errors(RMSEs) of UNB3m-ZTD, EGNOS-ZTD, and GPT3-ZTD of six GNSS stations in Antarctica are 98.12 mm, 116.13 mm, and 24.31 mm, respectively. After correction using XGBoost, the average RMSEs of UNB3m-ZTD, EGNOS-ZTD, and GPT3-ZTD from 2020 to 2021 are 10.46 mm, 10.50 mm, and 10.60 mm, respectively, demonstrating higher accuracy and closeness between the models. The average RMSEs of PWV converted from the corrected UNB3m-ZTD, EGNOS-ZTD, and GPT3-ZTD using XGBoost are 1.71 mm, indicating high accuracy.

The convolution regression calculation of pressure effect on well water level is carried out using MATLAB, and the pressure step response function of well water level is obtained. By calculating the determination coefficient fitted by pressure step response function and comparing the AIC value, the time interval and maximum time delay parameters in calculation process can be determined. Combining with other existing research results, preliminary judgments and even quantitative calculations of aquifer parameters of observation well can be made.