The Qinghai–Tibet Plateau is one of the most geologically active regions in the world, characterized by significant geomorphic variation and a wide range of geological hazards. The multifactorial coupling of tectonic movements, geomorphological evolution, climate variability, and lithological characteristics contributes to the pronounced spatial heterogeneity of the disaster-inducing environment. Identifying key controlling factors and their driving mechanisms is crucial for effective regional disaster prevention and mitigation. This study adopts a systematic framework based on regional disaster systems theory, integrating tectonic activity, engineering geology, topography, and precipitation to construct a multi-factor zoning system. Using the Random Forest model, we quantify factor contributions and delineate eight distinct disaster-inducing environment zones. Zones I–III (Himalayas–Hengduan Mountains–Qilian Mountains) are characterized by a dominant coupling mechanism of “tectonic fragmentation—topographic relief—precipitation erosion” and account for the majority of large-scale disasters. In contrast, Zones IV–VIII, primarily located in the central–western Plateau basins, are constrained by limited material sources, resulting in lower disaster densities. The findings indicate that geological structures and lithological fragmentation provide the material foundation for hazard occurrence, while topographic potential and hydrodynamic forces serve as critical triggering conditions. This nonlinear coupling of factors shapes a disaster geographic pattern characterized by “dense in the east and sparse in the west”. Based on these results, the targeted recommendations proposed offer valuable theoretical insights and methodological guidance for disaster mitigation and region-specific management across the Qinghai–Tibet Plateau.

Simulating potential glacier collapses can provide crucial support for local disaster prevention and mitigation efforts. The Xiaomagou Glacier in the Amnye Machen Mountains, Qinghai–Tibetan Plateau, has experienced five collapses in the past two decades. Field investigation and remote sensing observations indicate that the topography and bedrock characteristics of the Qushi’an No. 22 Glacier, which is 3.5 km south of the Xiaomagou Glacier, are similar to those of the Xiaomagou Glacier. More importantly, the mass movement of the Qushi’an No. 22 Glacier since 2018 closely resembles that of the Xiaomagou Glacier exhibited before its previous collapses. Therefore, in the context of rising temperatures, it is possible that the Qushi’an No. 22 Glacier will collapse in the near future. Based on remote sensing imagery and the glacier’s surface elevation changes, we reconstructed the 2004 collapse process of the Xiaomagou Glacier via numerical simulation. The key parameters of the mass flow model were optimized based on the actual deposition area of the 2004 collapse. The model with optimized parameters was then used to simulate the potential Qushi’an No. 22 Glacier collapse. Two collapse scenarios were set for the Qushi’an No. 22 Glacier. In Scenario 1, the lower half of the tongue detaches; in Scenario 2, the whole tongue detaches. Simulation results show that, in Scenario 1, the maximum mass flow depth is 72 m, the maximum mass flow speed is 51.6 m/s, and the deposition area is 5.40 × 10⁶ km²; in Scenario 2, the maximum mass flow depth is 75 m, the maximum mass flow speed is 59.7 m/s, and the deposition area is 6.32 × 10⁶ km². In both scenarios, the deposition area is much larger than that of the Xiaomagou Glacier 2004 collapse, which had a deposition area of 2.21 × 10⁶ km². The simulation results suggest that the Qushi’an No. 22 Glacier collapse could devastate the pastures and township roads lying in front of the glacier, seriously affecting local transportation and livestock farming; furthermore, it may deposit in the Qinglong River, forming a large, dammed lake. At present, the Qushi’an No. 22 Glacier remains in an unstable state. It is crucial to strengthen monitoring of its surface morphology, flow speed, and elevation.

The Alpine Periglacial Weathering Zone (APWZ) is a critical transitional belt between alpine vegetation and glaciers, and a highly sensitive region to climate change. Its dynamic variations profoundly reflect the surface environment’s response to climatic shifts. Taking Gongga Mountain as the study area, this study utilizes summer Landsat imagery from 1986 to 2024 and constructs a remote sensing method based on NDVI and NDSI indices using the Otsu thresholding algorithm on the Google Earth Engine platform to automatically extract the positions of the upper limit of vegetation and the snowline. Results show that over the past four decades, the APWZ in Gongga Mountain has exhibited a continuous upward shift, with the mean elevation rising from 4101 m to 4575 m. The upper limit of vegetation advanced at an average rate of 17.43 m/a, significantly faster than the snowline shift (3.9 m/a). The APWZ also experienced substantial areal shrinkage, with an average annual reduction of approximately 13.84 km², highlighting the differential responses of various surface cover types to warming. Spatially, the most pronounced changes occurred in high-elevation zones (4200–4700 m), moderate slopes (25–33°), and sun-facing aspects (east, southeast, and south slopes), reflecting a typical climate–topography coupled driving mechanism. In the upper APWZ, glacier retreat has intensified weathering and increased debris accumulation, while the newly formed vegetation zone in the lower APWZ remains structurally fragile and unstable. Under extreme climatic disturbances, this setting is prone to triggering chain-type hazards such as landslides and debris flows. These findings enhance our capacity to monitor alpine ecological boundary changes and identify associated disaster risks, providing scientific support for managing climate-sensitive mountainous regions.

Grassland fire risk perception constitutes a fundamental element of fire risk assessment and underpins the evaluation of response capacities in grassland regions. This study examines Qinghai Province, the fourth-largest pastoral region in China, as a case study to develop an evaluation index system for assessing residents’ perceptions of grassland fire risk. Using micro-level survey data, the study quantifies these perceptions and applies a quantile regression model to investigate influencing factors. The results indicate that: (1) the average grassland fire risk perception index among residents in Qinghai Province’s grassland areas is 0.509, with response behaviors contributing the most and response attitudes contributing the least; (2) Residents in agricultural areas perceive higher risks than those in semi-agricultural/semi-pastoral or purely pastoral areas, and individuals in regions with moderate dependency ratios and moderate fire-susceptibility conditions demonstrate the highest performance, whereas those in pastoral and high-susceptibility zones exhibit signs of “risk desensitization”; (3) risk communication and information dissemination are the primary drivers of enhanced perception, followed by climate variables, whereas individual characteristics of residents attributes exert no significant effect. It is recommended to monitor the impacts of climate change on fire risk patterns, update risk information dynamically, address deficits in residents’ cognition and capabilities, strengthen behavioral guidance and capacity-building initiatives, and foster a transition from passive acceptance to active engagement, thereby enhancing both cognitive and behavioral responses to grassland fires.

Frequent grassland fires have severely affected regional ecosystems as well as the production and living conditions of local residents. Grassland fire prevention capabilities constitute an integral part of the disaster prevention and mitigation system and play an important role in improving grassroots governance. To gain a deeper understanding of the practical foundation and influencing mechanisms of grassland fire prevention capabilities, establish an evaluation index system for prevention capabilities covering the four dimensions of disaster prevention, disaster resistance, disaster relief, and recovery. Combining micro-level survey data, a quantile regression model is used to analyze the influencing factors. The research findings indicate that (1) disaster resistance (0.49) plays a prominent role in grassland fire prevention capabilities, with economic foundations and individual disaster relief capabilities being particularly critical for overall improvement. Although residents have strong fire prevention awareness, their organizational collaboration capabilities are relatively weak, and there are significant differences in prevention capabilities across regions, necessitating tailored, precise enhancements. (2) There are significant differences in prevention capabilities among residents of different agricultural and pastoral production types, with semi-agricultural and semi-pastoral areas having the strongest comprehensive capabilities and pastoral areas relatively weaker. (3) A significant analysis of factors influencing grassland fire prevention capabilities: effective and diverse risk communication is a prerequisite for enhancing residents’ prevention capabilities; the level of panic regarding grassland fires and road infrastructure are important influencing factors, but residents’ understanding of climate change and grassroots organizations’ capacity for mechanism construction have insignificant impacts. Therefore, in future grassland fire disaster prevention and mitigation efforts, it is essential to strengthen risk communication, improve infrastructure, monitor environmental changes and the spatiotemporal patterns of grassland fires, enhance residents’ understanding of climate change, reinforce the emergency response capabilities of grassroots organizations, and stimulate public participation awareness to collectively build a multi-tiered grassland fire prevention system.

Ice thickness is a key parameter for glacier mass estimations and glacier dynamics simulations. Multiple physical models have been developed by glaciologists to estimate glacier ice thickness. However, obtaining internal and basal glacier parameters required by physical models is challenging, often leading to simplified models that struggle to capture the nonlinear characteristics of ice flow and resulting in significant uncertainties. To address this, this study proposes a convolutional neural network (CNN)-based deep learning model for glacier ice thickness estimation, named the Coordinate-Attentive Dense Glacier Ice Thickness Estimate Model (CADGITE). Based on in situ ice thickness measurements in the Swiss Alps, a CNN is designed to estimate glacier ice thickness by incorporating a new architecture that includes a Residual Coordinate Attention Block together with a Dense Connected Block, using the distance to glacier boundaries as a complement to inputs that include surface velocity, slope, and hypsometry. Taking ground-penetrating radar (GPR) measurements as a reference, the proposed model achieves a mean absolute deviation (MAD) of 24.28 m and a root mean square error (RMSE) of 37.95 m in Switzerland, outperforming mainstream physical models. When applied to 14 glaciers in High Mountain Asia, the model achieves an MAD of 20.91 m and an RMSE of 27.26 m compared to reference measurements, also exhibiting better performance than mainstream physical models. These comparisons demonstrate the good accuracy and cross-regional transferability of our approach, highlighting the potential of using deep learning-based methods for larger-scale glacier ice thickness estimation.