A new study developed a new AI tool that helps predict the safety of northern ice roads

ReSEC collaborated with C-CORE to the new study published in the Cold Regions Science and Technology journal, which developed an AI model to help predict when ice roads in northern Canada become unsafe, offering a promising tool for communities that rely on winter transportation. Ice roads, critical for delivering fuel, food, and supplies to remote Indigenous communities, are becoming less reliable due to climate change. The research focused on the Délı̨nę ice road on Great Bear Lake in the Northwest Territories, where warming temperatures are shortening the safe operating season.

The team used a machine learning approach called Long Short-Term Memory (LSTM) to predict ice surface temperatures up to seven days in advance. Unlike traditional methods that rely on air temperature, this approach directly estimates surface temperature, an important factor controlling ice strength and melt. The model showed strong performance, predicting temperatures within about 4-5°C of observed values. Importantly, the study found that rising surface temperatures closely track real-world ice melt and breakup, captured through satellite imagery. This means the model can provide early warning of dangerous conditions, helping authorities make better decisions about road closures and load limits.

The research highlights how combining satellite data, climate reanalysis, and AI can improve safety and planning for northern infrastructure.

Read the full article here