Coupling physical understanding and statistical modeling to estimate ice jam flood frequency in the northern Peace-Athabasca Delta under climate change

Abstract

The Peace-Athabasca Delta (PAD) of northeastern Alberta is one of the largest inland freshwater deltas in the world, laying at the confluence of the Peace and Athabasca Rivers. The PAD is internationally recognized as having a unique ecological and cultural significance, and periodic ice jam flooding from both rivers is an important feature of its current ecology. Past studies have debated whether a change in ice jam flood frequency on the Peace River has recently occurred, and what factors might be driving any perceived changes. This study contributes to this debate by addressing two questions: (1) what factors are most predictive of large Peace River ice jam floods, and (2) how might climate change impact large ice jam flood frequency? This work starts with a physically-based conceptual model of the necessary conditions for a large Peace River ice jam flood, and the factors that indicate whether those conditions are met. Logistic regression is applied to the historical flood record to determine which combination of hydroclimatic and riverine factors best predict large ice jam floods and the uncertainty in those relationships given the available data. Climatic factors including winter precipitation and temperature are most predictive of large Peace River ice jam floods. Other factors including winter river flows and freeze-up elevation provide little additional predictive power. The best logistic regression model is then forced with downscaled climate change scenarios from multiple climate models to project large ice jam flood frequency for a variety of plausible futures. Parametric uncertainty in the best logistic regression model is propagated into the projections using a parametric bootstrap to sample many plausible statistical models. Although there is variability across emissions scenarios and climate models, all projections indicate that the frequency of large Peace River ice jam floods is likely to decrease substantially in the coming decades, and that average waiting times between future large ice jam floods will likely surpass recent experience.