Special environmental condition with strong wind and large diurnal temperature range is not conducive for the control of crack development during the curing of mass concrete in the cold environment of Northwest China. In this study, the temperature at four measured points along the vertical profile of a mass concrete was measured over 14 days. Then, a thermal-mechanical model with the optimized initial condition, thermal parameters, boundary conditions, and the effects of rebar on temperature stress of concrete was built. Based on the observations and numerical model, the effects of environmental factor on the peak temperature, temperature fluctuation amplitude, and maximum temperature stress of the mass concrete are analyzed. The result showed that increasing wind speed can reduce the temperature of the concrete, but an opposite result is observed with increase in diurnal temperature range. Moreover, the more obvious diurnal variation of concrete temperature with the increase in wind speed and diurnal temperature range is not conducive to the thermal stability of the structure, especially for horizontal and side surfaces. Accordingly, increasing wind speed and diurnal temperature range will increase the peak temperature stress of the concrete and the overall stress level. Strong wind and large diurnal temperature range significantly increase the temperature stress and its duration for the side surface. The high temperature stress and intensive daily variation of temperature stress at the horizontal surface affect the durability of the concrete material. Therefore, the temperature stresses of the horizontal surface under wind speed levels 3, 4, 5, and 1.25 times large diurnal temperature range easily exceed the tensile strength; then, the temperature crack will appear. Combined with the research results, the temperature control and crack prevention measures are proposed under the condition of ensuring concrete strength. The study will further guide the improvement and optimization of mass concrete construction in an extreme environment.

The Qinghai-Tibetan Plateau (QTP) has undergone an increase in rainfall and a drastic alteration in the moisture-heat regime in active layers and engineering. To investigate the water and heat responses of natural ground and engineering to rainfall, the differences in energy on the ground surface and the thermal-moisture dynamics of different permafrost underlying surfaces were discussed. Based on the meteorological data in 2013 observed at the Beiluhe observation station, three types of underlying surfaces (i.e., natural ground, asphalt pavement, and gravel pavement) were selected to compare the differences in energy balance at the ground surface, water-energy transport process, and coupling mechanism in active layers under rainfall conditions by a coupled water vapor-heat model of the unsaturated frozen soil. The results show that the asphalt pavement greatly increases the surface net radiation and soil surface heat flux, decreases the surface evaporation latent heat, and cuts off the moisture migration between the atmosphere and the active layer. The gravel pavement significantly increases the surface evaporation latent heat to lower the soil surface heat flux, and the amount of moisture in shallow soil is strongly influenced by rainfall and evaporation. Therefore, the moisture migration and accumulation under the asphalt pavement are dominated by the water vapor flux under thermal gradients, whereas the liquid water under the water potential gradients is the major source of moisture migration under the gravel pavement. The heat transfer in the shallow active layer is dominated by heat conduction. The effect of heat conduction, water vapor migration, and phase transition on the soil temperature is evident for the asphalt pavement, while the impact of liquid water migration on the shallow soil temperature for the gravel pavement is significant in the thawing period. As a result, the soil temperature relationship between different underlying surfaces is asphalt pavement>gravel pavement>natural ground. The thickness of the active layer gradually decreases. Although rainfall infiltration promotes the liquid water convection of the gravel pavement, the decrease in heat flux is less than the increase in thermal conductivity. In general, the construction of asphalt pavement and gravel pavement accelerates the degradation of permafrost. The results can provide theoretical and simulated guidance for the stability prediction and analysis of various underlying surfaces in the central QTP where rainfall is increasing.

The choice of spatial mapping units substantially shapes the accuracy, spatial detail, and interpretability of landslide susceptibility assessment (LSA). To investigate how mapping-unit selection affects LSA performance, this study used Hejing County, Xinjiang, China, as a case study. Based on a historical landslide inventory and a multi-source environmental-factor database, two mapping-unit schemes were constructed: 30 m raster cells and slope units extracted through a hydrology-based delineation method. Based on twelve selected conditioning factors, landslide susceptibility was modeled using four machine-learning algorithms, including logistic regression (LR), multilayer perceptron (MLP), random forest (RF), and XGBoost. In addition, the SHAP method was applied to examine differences in model decision mechanisms and factor contributions between the two mapping-unit schemes. The findings indicate that both mapping-unit schemes are capable of representing the spatial variability of LSA across the study area. The raster-cell-based scheme generally showed better predictive performance, with the mean AUC of the four models increasing from 0.798 under the slope-unit scheme to 0.859 under the raster-cell scheme. This indicates that raster cells are more effective in capturing localized environmental heterogeneity and small-scale high-susceptibility patches. In contrast, the slope-unit-based scheme reduced inter-model performance differences and generated more spatially continuous and stable susceptibility zonation consistent with natural slope boundaries. SHAP analysis further indicates that the dominant conditioning factors were generally consistent between the two schemes; however, raster-cell-based models were more sensitive to localized disturbances and micro-topographic variability, whereas slope-unit-based models emphasized the integrated constraints of geomorphological structure and hydrological processes. Collectively, the predictive performance, spatial representation, and interpretability of LSA models are largely governed by how mapping units are delineated. Raster cells are more suitable for site-specific landslide susceptibility assessment and infrastructure siting, whereas slope units are more appropriate for regional-scale susceptibility mapping and disaster-mitigation planning.