The thermal stability of permafrost under complex environment (climate scenarios, permafrost types and regional air temperatures) directly affects the long-term service performance of highway or railway. This study uses a large amount of valuable soil temperature monitoring and simulation data to examine the stability of typical crushed-rock embankments (CREs) along Qinghai-Tibet Railway, which is located in the permafrost re-gion on the Roof of the World. Firstly, a novel numerical model for CREs considering a complex heat transfer environment is established and verified. Then, alteration characteristics of recent-term thermal regime of permafrost across time and space under three CREs are revealed based on a decade of field monitoring data. Finally, long-term thermal regime of permafrost under three CREs under complex environment is analyzed by numerical simulation. Results include: 1) Soil temperature near the ground surface under the three CREs shows a decreasing trend, whereas the overall temperature around the deep permafrost increases over time under a recent climate warming. 2) The warming rate of permafrost under CREs rises with the acceleration of climate scenarios and regional air temperature and the decrease of regional ground temperatures. 3) U-shaped crushed-rock embankment is the most suitable CRE for managing complex environment, especially when the mean annual ground temperature is 0 similar to -1 degrees C or climate scenarios are RCP 2.6 and RCP 4.5. 4) Transition from low-temperature permafrost to warm permafrost is a warning signal of permafrost degradation under climate warming. 5) With an increase of permafrost degradation rate (the thawing and warming rates of permafrost), embankment stability becomes worse. These findings will not only serve as a scientific basis for the embankment damage prevention of the Qinghai-Tibet Railway, but also provide important technical supports for the successful building of infrastructure in permafrost regions under complex environment.

Northern high-latitude permafrost holds the largest soil carbon pool in the world. Understanding the responses of permafrost to wildfire is crucial for improving our ability to predict permafrost degradation and further carbon emissions. Recently, studies have demonstrated that wildfires in the pan-Arctic region induced the thickening of the active layer based on site or fire event observations. However, how this induced thickening is influenced by vegetation and permafrost types remains not fully understood due to the lack of wall-to-wall analysis. Therefore, this study employed remotely sensed fire data and modelled active layer thickness (ALT) to identify the fireinduced ALT change (& UDelta;ALT) for the pan-Arctic region, and the contributions of vegetation and permafrost were quantified using the random forest (RF) model. Our results showed that the average & UDelta;ALT and the sensitivity of & UDelta;ALT to burn severity both increased with decreasing ground ice content in permafrost. The largest values were detected in thick permafrost with low ground ice content. Regarding vegetation, the average and sensitivity of & UDelta;ALT in tundra were highest, followed by those in forest and shrub. When the individual environmental factors were all taken into account, the results showed that the contribution of vegetation types was much higher than that of permafrost types (20.2 % vs. 3.5 %). Our findings highlighted the importance of environmental factors in regulating the responses of permafrost to fire.