Author(s): Liming Zhao; Faye Hicks; Aminah Robinson Fayek; Nadia Kovachis

Linked Author(s):

Keywords: No Keywords

Abstract: River ice breakup is a dramatic event for many northern rivers. Floods associated with ice jam formation and release events are particularly dangerous because of the speed with which water levels can rise (e. g. ~0.8 m/min increases have been documented by the second author). It would be especially helpful for many riverside communities to have the ability to predict the anticipated timing of such events; however, for most cases this is not yet possible. This study aims to improve this situation by exploring the applicability of Artificial Neural Networks (ANN) for forecasting breakup timing. This paper describes the training and testing of an ANN for forecasting the onset of breakup, using the Hay River at Hay River, NWT, Canada as a demonstration case. Several key indicators of dynamic breakup are employed in this prediction, including: freeze-up stage, accumulated degree-days of freezing over the winter, ice thickness, accumulated degree-days of thaw during the pre-breakup period, accumulated degreedays of freezing during the pre-breakup period and the total spring rainfall before breakup. The resulting ANN model is expected to provide the local community with short lead time (1 to 2 days) warning of ice jam flood timing, based on weather forecasts and observed water level data.

DOI:

Year: 2010