Drift and deterioration of Petermann ice islands

Abstract

The eastern Canadian waters is an important part of Canadian shipping lanes and subjected to occasional ice island presence, which could pose serious hazards to offshore and shipping activities in this region. It is, therefore, important to better characterize the dynamics of glacial ice features for safe and cost-effective activities in the region. This thesis presents advanced predictive models to provide a better understanding of how atmospheric and oceanic variables influence ice island drift and deterioration. To understand how ice islands drift under the influence of atmospheric and oceanic forces, a deterministic model was presented, where the relative contribution of various forces governing the drift of four tracked ice islands was quantified. The results showed that in low sea ice concentrations and ice island speeds, ocean current and sea surface tilt forces dominated ice island force balance (63% on average). Wind, however, played a minor role (< 5%), and Coriolis and sea ice forces were significant only at higher ice island speeds and sea ice concentrations, respectively. Atmospheric and oceanic variables were further studied using a probabilistic Bayesian approach to investigate their relative influences on the fracture events and drift velocities of hundreds of Petermann ice islands tracked in the Canadian Ice Island Drift, Deterioration and Detection database. The presented models identified water temperature and ocean currents as the most important contributor to ice island fracture events and drift velocities, respectively. It was revealed that under severe conditions of wind, current, waves, and air/water temperatures, ice islands are more likely to fracture, with fracture probability reaching as high as 75% in extreme conditions. It was also revealed that under stronger currents, ice islands are most likely to drift at higher speeds and in close proximity to ocean current direction. Models were validated using the 5-fold cross-validation approach and errors up to 39% and 29% were reported in the fracture and drift probability estimations, respectively. The presented models have predictive capabilities for future drift and deterioration of Petermann ice islands. However, further training and testing of the developed models is necessary before they can be used as operational forecasting tools.