| Summary: | Abstract Prediction of permafrost stability is associated with challenges, such as data scarcity and climate uncertainties. Here we present a data-driven framework that predicts permafrost thaw threat based on present ground ice distributions and ground surface temperatures predicted via machine learning. The framework uses long short-term memory models, which account for the sequential nature of climate data, and predicts ground surface temperature based on several climate variables from reanalysis products and regional climate models. Permafrost thaw threat is then assessed for three cases in northern Canada: Hudson Bay Railway, Mackenzie Northern Railway, and Inuvik–Tuktoyaktuk Highway. The models predict ground surface warming in all studied areas under both moderate and extreme climate change scenarios. The results also suggest that all studied cases are already under threat, with the northern sections of the Hudson Bay Railway and Inuvik–Tuktoyaktuk Highway facing an increasing threat by the end of the century.
|
|---|