Marine and lacustrine ice fracture detection

Abstract

Ice fracturing has been extensively studied and modeled. With increased interest in ice mechanics and fracturing in recent years in climate science, fisheries, and for cultural impacts, detecting and classifying fracturing events has become an important problem to consider. Fractures primarily occur due to stress relief within an ice sheet during temperature shifts and ice movement. These events create mechanical waves within the sheet that couple into the water column which can then be detected as pressure and particle velocity fluctuations. Machine learning algorithms will be used to detect and classify ice cracking events though their acoustic signature. Different models will be compared to one another for effectiveness and accuracy. Data will be shown from several different locations, including Northern Alaska and the Great Lakes.