The Arctic is warming at least two to four times as fast as the global average. This leads to an increase in extreme events and changes in the spatial patterns of snow cover, directly affecting transportation, health, and the economic situation of northern communities. Due to their increased vulnerability, accurate information on snow conditions is needed to adequately mitigate the impacts of climate change on these communities. However, there is a lack of high spatial and temporal resolution information on snow cover in the Arctic. 

This seminar will present a methodology using machine learning algorithms trained with on-site airborne LIDAR for 1 meter resolution snow depth predictions, to create snow depth maps for use by Inuit communities and with their cooperation. Preliminary results show a R^2 of 0.95 for a Random Forest model using topographical features, while being orders of magnitude faster than physics-based models.