Statistical, remote sensing-based approach for estimating the probability of catastrophic drainage from moraine-dammed lakes in southwestern British Columbia

Safe development of glacierized regions requires a systematic and objective method for assessing the hazard posed by moraine-dammed lakes. Empirical relations exist for estimating outburst flood magnitude, but, until now, no standardized procedures have been developed for estimating outburst flood probability. To make quick and inexpensive preliminary assessments that are reproducible, we propose using a statistical, remote sensing-based approach to estimate the probability of catastrophic drainage of moraine-dammed lakes. We completed a comprehensive inventory of 175 moraine-dammed lakes in the southern Coast Mountains of British Columbia, Canada. By applying logistic regression analysis to the data set, we identified and weighted the following four independent predictor variables that best discriminate drained lakes from undrained lakes: moraine height-to-width ratio, presence/absence of an ice-core in the moraine, lake area, and main rock type forming the moraine. With an appropriate classification cutoff value, the predictive model correctly classifies 70% of drained lakes and 90% of undrained lakes, for an overall accuracy of 88%. Our model provides engineers and geoscientists with a tool for making first-order estimates of the probability of catastrophic drainage from moraine-dammed lakes in southwestern British Columbia.