Analysis and Prediction of Mid Winter Breakups on Northern Canadian Rivers Using Time Series Data Mining

Abstract:

River ice breakup is a prominent yearly event in most Canadian rivers. If the breakup mechanism is dynamic, resulting from a sudden and rapid increase in discharge while the ice is still intact, these events carry the risk of triggering ice jamming in rivers. An increase in unseasonal winter temperatures in recent years has also led to a significant number of Mid Winter Breakup (MWB) events, carrying the risk of more severe ice jam flooding in affected communities. These jams have the potential to freeze in place if temperatures return to seasonal levels, creating a semi permanent dam in the river. This study focuses on the spatial, temporal, and attributional analysis of MWB events contained in the Canadian River Ice Database (CRID), a newly assembled dataset from Environment Canada containing detailed ice related observations from 196 rivers throughout the country. The analysis of MWBs along with the preceding and following ice seasons identify specific patterns of variability and changes in the timing and severity of MWBs allowing classification analysis of the MWBs to be performed. The study also investigates peak seasonal water levels and their return periods and identifies proportions of specific return periods caused by MWBs at each measuring station. Finally, an attribution analysis is performed to identify the links between the primary drivers of ice breakup and the known MWB events. In combination with MWBs classification analysis, an index for the occurrence of severe MWBs that can providing early indication of potential flooding that may occur during any MWBs is proposed.