Predicting Breakup Ice Jams Using Logistic Regression

Abstract

Breakup ice jams form suddenly and with little warning. The lack of forewarning hinders emergency response and ice jam mitigation efforts. Present knowledge of breakup jam processes does not allow for the development of a deterministic ice jam prediction model. Probabilistically based prediction models include linear regression, discriminant function analysis, and empirical cluster-type analyses. In this paper, the use of logistic regression to predict breakup ice jam occurrence is presented, with an example application for the Platte River at North Bend, Nebraska.