From Concept to Application: Machine Learning for Near-Real-Time River Ice Breakup Prediction Using SAR and Meteorological Data

Using the reanalysis datasets and the Canadian Ice Island Drift, Deterioration and Detection database, a probabilistic model was developed to quantify ice island fracture probability under various atmospheric and oceanic conditions. The model identified water temperature as the most dominant variable behind ice island fracture events, while ocean currents played a minor role. The developed model offers a predictive capability and could be of particular interest to offshore and marine activities.