High-mountain landscapes are becoming increasingly dynamic under climate change, with environmental changes impacting ecosystems and societies in and near mountains. The mountain cryosphere responds to rising atmospheric temperatures with declining snow and ice cover and thawing permafrost. The snow and ice cover decline modifies the surface energy balance, allowing increased heat transfer from the atmosphere into the ground and affecting mountain permafrost. Coupled with reduced mechanical ice support by retreating glaciers, high mountain landscapes destabilise and respond with increased bedrock erosion and sediment production. Mountain glaciers become increasingly covered by supraglacial debris of variable thickness that modifies their response to climate change by altered melt rates. Observations of increased rockfall activity, catastrophic slope failures, and cascading hazards related to climate change highlight the need for large-scale monitoring and detection techniques to analyse recent landscape changes and identify regions at risk of warming-related natural hazards. In this thesis, I evaluate the potential and limitations of thermal infrared remote sensing to detect and analyse glacial landscape dynamics based on land surface temperature (LST) observations acquired from an unpiloted aerial vehicle (UAV) and multidecadal time series (1984-2022) of combined Landsat sensors. In the first study, I measured a significant portion of the diurnal temperature cycle of a debris-covered glacier section in Switzerland using UAV-acquired very high-resolution LST data and estimated spatially distributed debris thickness using two distinct approaches. The results showed that the nonlinearity in the relationship between LST and debris thickness varies throughout the day, which is relevant for interpreting satellite-derived debris thickness estimates at fixed acquisition times. Sub-daily LST measurements further allow quantifying surface energy balance components that satellite-based approaches cannot capture. Consequently, UAV-derived debris thickness maps can be used to calibrate and correct satellite-derived estimates, bridging the gap between spatial and temporal scales. Despite the relatively large RMSE in debris thickness estimates, this study is the first to discuss the limits and opportunities of UAV-derived LST to study the evolution of debris-covered glaciers. While UAV-derived LST can provide insights into glacial landscape changes over short time scales and relatively small spatial extent, detecting and monitoring longer-term changes at mountain range scale require long LST records and high spatial resolution to account for the steep altitudinal temperature gradients. In Chapter 3, I access patterns and trends in Landsat-derived LST data across the Swiss Alps using a harmonic model including a linear trend component. Comparison with LST time series from 119 weather stations revealed good accuracy of Landsat LST and LST trends. However, LST trends are biased due to orbital changes that cause variations in acquisition times, affecting the temporal coherence of LST measurements. Analysis of high temporal resolution LST from alpine weather stations shows that the linear change in Landsat acquisition time can explain a trend of 0.045 K y-1. However, the LST trend bias varies with topography, and I used modelled incoming shortwave radiation as a proxy to estimate the spatial variability of the LST trend bias. The corrected LST trends suggest that the highest LST warming rates are found in regions where mean annual land surface temperature is between -5 and 0°C and snow cover loss and permafrost thaw are expected to be large. In Chapter 4, I used several land cover datasets and created binary masks to compare the distributions of LST trends and their elevational differences for selected land cover and land cover change categories across the European Alps. Despite significant uncertainties in some of the binary masks, mean LST trends significantly vary with land cover and land cover changes, with a spatially averaged mean LST trend of 0.09 Kyr-1. The analysis further shows that warming rates of regions with significant snow cover loss increase with elevation, while other land cover types do not show such patterns. The studies in this thesis are the first to access LST and LST trends at unprecedented high spatial resolution, contributing to a better understanding of the climate sensitivity of glacial landscape dynamics and identifying regions prone to warming-related natural hazards.
