Thermokarst lakes play a crucial role in regulating hydrological, ecological, and biogeochemical processes in permafrost regions. However, due to the limited spatial resolution of earlier satellite imagery, small thermokarst lakes—highly sensitive to climate change and permafrost degradation—have often been overlooked, hindering accurate spatiotemporal analyses. To address this limitation, five water indices—Modified Normalized Difference Water Index (MNDWI), Multi-Band Water Index (MBWI), Automated Water Extraction Index (AWEI_sh and AWEI_nsh), and Red Edge Water Index (RWI)—were employed based on Sentinel-2 imagery from 2021 to extract thermokarst lakes in the Qinghai–Tibet Highway (QTH) region, China. Visual validation indicated that the Red Edge Water Index (RWI) yielded the best performance, with an error of only 10.21%, significantly lower than other indices (e.g., MNDWI: 41.36%; MBWI: 38.80%). Seasonal comparisons revealed that the applicability of different water indices varies, with autumn months (September to October) being the optimal period for lake extraction due to stable and unfrozen surface conditions. Using the RWI, 56 thermokarst lake drainage events were identified in the study area from 2016 to 2025 (as of September 2025), most occurring after 2019—likely associated with climatic factors—and small lakes were found to be more prone to drainage, accompanied by notable surface subsidence in drained regions. These findings are applicable across the Qinghai–Tibet Plateau (QTP) and provide a scientific basis for monitoring thermokarst lakes, delineating accurate lake boundaries, and exploring drainage mechanisms.

Accurate mapping of landslide-affected land surfaces from very-high-resolution optical imagery is essential for mountainous land monitoring and hazard-related land management, yet it remains difficult in complex terrain because landslide bodies are fragmented, elongated, shadowed, and easily confused with bare soil, terraces, roads, and erosion features. This study presents TM-Net as a task-oriented decoder-centric architecture for landslide-affected land surface mapping. The Inception Token Mixer is adopted from InceptionNeXt as an encoder-side component, whereas TM-Block is newly designed as a decoder-side twin-path refinement module for boundary recovery, slender target preservation, and complex-background suppression. On the fused public benchmark, TM-Net achieved the highest IoU (73.56%) and F1 (84.76%) under the identical evaluation protocol. On XBLD, TM-Net achieved the highest IoU (44.94%) and F1 (62.01%). These results indicate that TM-Net provides a favorable balance between missed detections and false-positive predictions for mapping visually identifiable landslide-affected land surfaces in heterogeneous mountainous landscapes.

Landslide susceptibility mapping (LSM) in mountain–basin transition zones remains challenging because conventional approaches rely mainly on historical inventories and static conditioning factors, whereas independent deformation evidence is seldom incorporated to refine susceptibility zonation. This study proposes an integrated LSM framework for the Xining Basin by coupling a Mamba-based model (Mamba-LSM) with SBAS-InSAR-based deformation-informed bidirectional reclassification, with the key innovation lying in the use of independent deformation evidence to refine susceptibility zonation after model prediction. Specifically, Mamba-LSM integrates six-channel neighborhood patches, CNN-based local spatial encoding, and Mamba-based latent feature transformation to improve the representation of local terrain context for landslide susceptibility assessment. Results show that Mamba-LSM achieved the highest AUC among the evaluated models, reaching 0.9011 with an F1-score of 0.7431. After deformation-informed bidirectional reclassification, the high- and very-high-susceptibility classes occupied only 25.31% of the study area but contained 69.84% of the mapped landslides, and were concentrated mainly in valley–mountain transition belts, river-incised slopes, and engineering-disturbed sectors where SBAS-InSAR deformation hotspots were also preferentially distributed. These findings demonstrate that integrating independent SBAS-InSAR deformation evidence can improve both the spatial concentration of landslides in high-susceptibility zones and the physical interpretability of susceptibility zonation.

Permafrost warming is reshaping cold region surface water systems, where thermokarst lake drainage can abruptly alter lake abundance, hydrological connectivity and exposed thaw terrain. Yet future drainage risk trajectories remain poorly constrained for thermokarst lakes on the Mongolian Plateau (MP), a mid-latitude permafrost region undergoing strong climatic and cryospheric change. Here, we developed a future-compatible drainage risk assessment framework that combines Landsat-derived drainage mapping, environmental predictors, eXtreme Gradient Boosting (XGBoost) modelling, and fixed 2020 baseline risk thresholds with climate projections. The model was trained with 2003–2020 annual lake observations and applied to 31,786 undrained candidate lakes. Independent validation using 2021–2025 drainage events showed that 987 of 1188 events (83.1%) exceeded the fixed 2020 top 10% risk threshold, corresponding to an enrichment ratio of 8.31. Future projections showed a substantial upward shift in drainage risk levels relative to the 2020 baseline. These outputs represent relative risk levels referenced to the fixed 2020 distribution rather than calibrated probabilities of drainage within a specified future period. Under the Shared Socioeconomic Pathway 5-8.5 (SSP5-8.5) scenario, 12,448 lakes (39.16%) exceeded the fixed top 10% threshold by 2081–2100, while 17,980 lakes (56.57%) exceeded the fixed top 25% threshold. Among the top 10% high-risk lakes, 9597 were consistently identified by all six global climate models (GCMs), indicating strong cross-model agreement. These GCM-supported high-risk lakes formed spatially coherent clusters in the northern, western and northeastern MP. The Stefan-type active layer sensitivity test retained high spatial overlap with the main projection (Jaccard = 0.917). Environmental association analysis showed that mean annual ground temperature (MAGT) was the most closely related factor, with consistently negative Spearman correlations across scenarios and projection periods (−0.45 to −0.43), followed by thawing degree days (TDD), freezing degree days (FDD), June–September precipitation and river network distance. These findings identify areas with elevated relative drainage risk levels as warming continues.

Regional landslide susceptibility screening can support sustainable land management by improving the allocation of limited geological-survey and field-verification resources toward locations requiring further investigation. An uncertainty-aware framework integrating ensemble learning, spatial validation, inter-model agreement, and probability uncertainty was proposed for regional landslide investigation prioritization in the Hualong–Xunhua region of the Upper Yellow River Basin, China. Using 281 manually interpreted landslide locations, 281 pseudo-absence samples, and twelve conditioning factors, four tree-based models (Random Forest, XGBoost, LightGBM, and CatBoost) were developed, and their probability outputs were combined through arithmetic averaging. Model performance was evaluated using stratified random validation and 10 km spatial block validation, while agreement and probability divergence were incorporated to define screening priorities. Random-validation AUC values ranged from 0.885 to 0.898, and spatial-validation AUC values ranged from 0.852 to 0.873; the Mean Ensemble achieved AUC values of 0.8961 and 0.8637, respectively. NDVI showed the highest mean normalized permutation importance (0.754), although sensitivity analysis demonstrated that other factors retained useful discrimination after removing NDVI and land-cover information. The final priority screening zone covered 862.72 km² (19.04% of the valid mapped area). The framework provides a transparent decision-support approach for sustainable land management in mountainous regions by identifying investigation priorities while accounting for model consistency and uncertainty.

Conditioning factor reduction can decrease redundancy and modelling complexity in landslide susceptibility mapping. These factors describe the environmental conditions associated with slope instability. Statistical screening, redundancy diagnostics, importance rankings and internal validation do not establish that the remaining conditioning factors can be safely removed. Whether aggressive reduction preserves the environmental information required for reliable regional landslide susceptibility mapping therefore remains uncertain. This study developed an integrated framework for evaluating the reliability of conditioning-factor reduction in the Hualong–Xunhua region of the Upper Yellow River, China, combining statistical diagnostics, model interpretation, GeoDetector analysis, and random and spatial validation. The compact system selected during development was further compared with the complete system using a reserved evaluation subset, followed by regional susceptibility mapping and spatial comparison. The results showed that, first, precipitation showed weak univariate but strong model and interaction evidence, illustrating their complementarity. Second, Top5, comprising normalised difference vegetation index (NDVI), distance to rivers, precipitation, lithology and terrain ruggedness index (TRI), met development criteria: mean area under the receiver operating characteristic curve (AUC) values were 0.866 under random validation and 0.846 under spatial validation, compared with 0.875 and 0.853 for the complete 17-factor system (Full17), respectively. Third, in the post-selection reserved comparison, mean AUC was 0.865 for Top5 and 0.895 for Full17, with paired uncertainty supporting this decline; favourable development performance did not ensure reliable substitution. Finally, Full17 was retained for final mapping; absolute score differences greater than 0.10 covered 32.02% of mapped area, indicating potential spatial information loss. Similar predictive performance did not imply spatial equivalence. These findings show that factor reduction requires evaluation of predictive performance, generalisation, model dependence, and spatial information preservation.