Accurate automated extraction of coseismic deformation from Synthetic Aperture Radar (SAR) data can be challenging owing to interference from inherent atmospheric noise. Particularly, the limited displacement of small-to-moderate earthquakes (Mw<6.5) can easily be obscured by phase errors and/or noise. To address this issue, we developed an autoencoder model based on a deep learning framework (i.e., Pytorch) to automate the accurate extraction of coseismic displacement from Interferometric SAR (InSAR) interferograms. We constructed a training dataset using simulated interferograms. Our trained model performed well for interferograms with real noise. When applied to worldwide real earthquakes of various rupture styles, the model produced clear coseismic displacement with less noise and a better fit to coseismic fault models compared to the differential InSAR method without noise correction. Additionally, it achieved co-seismic deformation similar to popular InSAR time series and GNSS methods. The approach will enhance the proceduralization and popularization of InSAR applications in earthquake monitoring, providing improved constraints on the kinematic characteristics of earthquakes.

The rupture patterns of large earthquakes in transpressional systems can provide important information for understanding oblique motion and strain partitioning between tectonic blocks. The 1927M8.0 Gulang earthquake occurred on the transpressional boundary between the Tibetan and Gobi‐Alashan blocks. Our results, combined with those of previous studies show that the Lenglongling fault (LLLF) and Southern Wuwei Basin fault (SWBF) have both ruptured during the Gulang earthquake, but they exhibited different motions. An ~120‐km‐long surface rupture zone formed along the LLLF, with a left‐lateral strike‐slip motion and a coseismic horizontal offset of ~2.4–7.5 m. Bending, bifurcation, and change of the slip sense occurs at both ends of the fault. An ~42‐km‐long rupture zone formed along the SWBF, with a coseismic vertical offset of ~0.6–2.8 m. Thus, the Gulang earthquake is a complex rupture event where strike‐slip and thrust faults ruptured simultaneously. Analysis of deep and shallow structures and three‐dimensional finite‐element modeling reveal that the north dipping LLLF and the SWBF may converge downward to a low‐angle décollement, accommodating the strike‐slip and thrust motions during the earthquake, respectively. This pattern of deformation partitioning is similar to some other earthquakes where oblique block convergence is generally partitioned into strike‐slip motion on steeply dipping faults and vertical motion on gently dipping faults. Our study also suggests that the strain partitioning pattern in the NE Tibetan Plateau may be controlled by changes of the regional principal compressive stress directions and fault geometry at depth.

Peak ground displacement (PGD) and peak ground velocity (PGV) are critical parameters during earthquake early warning, as they can provide rapid magnitude estimation before rupture end. In this study, we used the high-rate Global Navigation Satellite System (GNSS) data from 55 continuous stations to estimate the magnitude of the 2021 Maduo earthquake in western China. We used the relative positioning method and variometric approach to acquire real-time GNSS displacement and velocity waveforms, respectively. The results showed the amplitude of displacement and velocity waveforms gradually decreased with increasing hypocentral distance. Our results showed that the fluctuation of PGD magnitudes over time is smaller than that of PGV magnitudes. Nonetheless, the earthquake magnitudes estimated from both methods were consistent with their counterparts (Mw 7.3) reported by the United States Geological Survey (USGS). The final magnitude estimated from the PGD and PGV methods were Mw 7.25 and Mw 7.31, respectively. In addition, our results highlighted how the number of high-rate GNSS stations could influence the stability and convergence time of magnitude estimation.

Investigating the interaction between slow‐slip events (SSEs) and large earthquakes provides insights into earthquake‐triggering mechanisms and is crucial for earthquake hazard assessment. In this study, we used Sentinel‐1 Interferometric Synthetic Aperture Radar and Global Positioning System (GPS) data to estimate the source parameters of the 2018 $Mw$ 7.2 Pinotepa, Mexico, earthquake. The results show that the earthquake ruptured both the seismogenic zone and a long‐term SSE area, with two asperities ruptured during the event. GPS time‐series data indicate that an SSE was initiated approximately during March 2017 below Oaxaca; the SSE ruptured an area below the source region of the Pinotepa earthquake and intruded into the seismogenic zone. The spatiotemporal proximity may suggest that the SSE triggered the Pinotepa earthquake. We propose that the triggering mechanism may either have been Coulomb stress loading or weakening of the source area by fluid migration. Furthermore, we calculated the seismic moment deficit and Coulomb failure stress changes and conclude that the Oaxaca area still has the potential for large earthquakes.

Accurate and reliable land cover data are essential for environmental monitoring, climate research, and sustainable land management. However, the proliferation of multi-source global land cover datasets with long time series poses challenges for selecting the best products for specific applications. Existing assessments often lack systematic comparisons of classification accuracy and time consistency across geographic areas. This study addresses the critical gap in cross-product comparability by systematically comparing five recent global time-series land cover products (GLC_FCS30D, Esri Land Cover, MCD12Q1, ESA CCI, and Dynamic World) against a reference dataset (CGLS-LC100). Through a unified classification system, resolution resampling, and random sampling validation, we assessed their classification accuracy and time-series change accuracy across three transitional regions representing diverse environmental contexts: rapidly urbanizing regions, agriculturally intensive zones, and high-latitude forested areas. The results indicate that while datasets exhibit spatial consistency, significant discrepancies exist in land cover classification, with each dataset demonstrating varying levels of accuracy depending on the environmental context and land cover type. High-resolution products (e.g., GLC_FCS30D, Dynamic World) are optimal for monitoring fragmented landscapes and urban expansion, whereas long-term datasets (e.g., ESA CCI, MCD12Q1) suit climate trend analysis in stable ecosystems. Based on the evaluation, we provide generalized guidance for dataset selection aligned with land cover types and monitoring objectives, emphasizing the need for region-specific and application-oriented choices. This study highlights challenges in dynamic datasets, including classification system discrepancies, resolution effects, and reference data limitations, and suggests that future advancements should focus on improving classification algorithms, refining sampling methods, and developing assessment systems that incorporate high-precision, real-time validation data.