In order to investigate the effect of different loading models on the nonlinear variations in Global Positioning System (GPS) height time series, the characteristics of annual signals (amplitude and phase) of GPS time series, loading series from Deutsche GeoForschungsZentrum, Germany (GFZ) and School and Observatory of Earth Sciences, France (EOST) at 633 global GPS stations are processed and analyzed. The change characteristics of the root mean square (RMS) reduction rate, annual amplitude and phase of GPS time series after environmental loading corrections (ELCs) are then detected. Results show that ELCs have a positive effect on the reduction in the nonlinear deformation contained in most GPS stations around the world. RMS reduction rates are positive at 82.6% stations after GFZ correction and 87.4% after EOST correction, and the average reduction rates of all stations are 10.6% and 15.4%, respectively. As for the environmental loading series from GFZ and EOST, their average annual amplitudes are 2.7 and 3.1 mm, which explains ~40% annual amplitude of GPS height time series (7.2 mm). Further analysis of some specific stations indicates that the annual phase difference between GPS height time series and the environmental loading series is an important reason that affects the reduction rates of the RMS and annual amplitude. The linear relationship between the annual phase difference and the annual amplitude reduction rate is significant. The linear fitting results show that when there is no annual phase difference between GPS and loading series, the reduction rates of the RMS and annual amplitude will increase to the maximum of 15.6% and 41.6% for GFZ, and 22.0% and 46.6% for EOST.

During an earthquake sequence, there are often multiple recurring landslides. Understanding the spatial distribution of the landslides triggered by the first earthquake can help us predict the landslide susceptibility for subsequent shakes over a short term. This study used two landslide inventories from the Lombok earthquake sequence in Indonesia in 2018 to construct a short-term secondary disaster prediction model and an overall spatial prediction model using four machine learning algorithms. The average accuracy of the positive samples predicted by the prediction model was 7.1% lower than that of the short-term model. The highest accuracy of the overall prediction model was 14.9% higher, on average, and the area under the ROC curve (AUC) score was 8.1% higher, on average, but the corresponding probability thresholds were lower. The reason for this difference is that, in the short-term prediction model, since most of the landslides in the first landslide inventory were prone to fail two or more times due to the effect of multiple earthquakes, the prediction results have a high positive rate. This feature of the short-term prediction model makes it suitable for landslide rescue guidance in a sequence of earthquakes. In contrast, the overall prediction model can better represent the spatial distribution of the earthquake-triggered landslides in the area.

Seasonal variations in the vertical Global Positioning System (GPS) time series are mainly caused by environmental loading, e.g., hydrological loading (HYDL), atmospheric loading (ATML), and nontidal oceanic loading (NTOL), which can be synthesized based on models developed by various institutions. A comprehensive comparison among these models is essential to extract reliable vertical deformation data, especially on a regional scale. In this study, we selected 4 HYDL, 5 ATML, 2 NTOL, and their 40 combined products to investigate their effects on seasonal variations in vertical GPS time series at 27 GPS stations in Yunnan, southwest China. These products were provided by the German Research Center for Geosciences (GFZ), School and Observatory of Earth Sciences (EOST), and International Mass Loading Service (IMLS). Furthermore, we used the Cross Wavelet Transform (XWT) method to analyze the relative phase relationship between the GPS and the environmental loading time series. Our result showed that the largest average Root-Mean-Square (RMS) reduction value was 1.32 mm after removing the deformation associated with 4 HYDL from the vertical GPS time series, whereas the RMS reductions after 5 ATML and 2 NTOL model corrections were negative at most stations in Yunnan. The average RMS reduction value of the optimal combination of environmental loading products was 1.24 mm, which was worse than the HYDL (IMLS_GEOSFPIT)-only correction, indicating that HYDL was the main factor responding for seasonal variations at most stations in Yunnan. The XWT result showed that HYDL also explained the annual variations reasonably. Our finding implies that HYDL (IMLS_GEOSFPIT) contributes the most to the environmental loading in Yunnan, and that the ATML and NTOL models used in this paper cannot be effective to correct seasonal variations.

Most machine learning (ML) studies on predicting co-seismic landslides have relied on Peak Ground Acceleration (PGA). The PGA of the ground strongly correlates with the relative position and azimuth of the seismogenic faults. Using the co-seismic landslide records of the 2008 Wenchuan earthquake, we show that the ML model using the distances and azimuths from the epicenter to sites performs better than the PGA model regarding accuracy and actual prediction results. The distances and azimuths are more accessible than the PGA because obtaining accurate and realistic large-scale PGAs is difficult. Considering their computational efficiency and cost-effectiveness, the ML models utilizing distances and azimuths from the epicenter to the sites as inputs could be an alternative to PGA-based models for evaluating the impact of co-seismic landslides. Notably, these models prove advantageous in near-real-time scenarios and settings requiring high spatial resolution, and provide favorable assistance in achieving the goal of sustainable development of society.

Most machine learning (ML) studies in predicting co-seismic landslides relied on the Peak Ground Acceleration (PGA). Yet, the accuracy of PGA-based models can be compromised due to the limited availability of near-source observations and spatial information regarding seismic faults. This limitation can result in significant discrepancies between the accumulated PGA and the actual ground shaking, which cannot be overlooked. Here we show the ML model using the distance and azimuth to the epicenter performs better than the model using PGA in terms of accuracy as well as actual prediction results. Considering their computational efficiency and cost-effectiveness, the ML models utilizing distance and azimuth to the epicenter as inputs could be used as an alternative to PGA-based models for evaluating the impact of co-seismic landslides. Particularly, these models prove advantageous in near real-time scenarios and settings that require high spatial resolution.