Author(s): Regan P. Mcdonald; Kathleen D. White; Steven F. Daly; Darrell D. Massie

Linked Author(s):

Keywords: No Keywords

Abstract: Break-up jams occur suddenly and often cause rapid increases in upstream water levels. Break-up ice jam prediction methods are desirable to provide early warning and allow effective mitigation. The lack of an analytical description of the complex physical processes involved in break-up ice jam formation has limited development of effective prediction models. Current methods include empirical, threshold-type models and statistical methods such as logistic regression and discriminant function analysis. A neural network method developed to predict break-up ice jams at Oil City, Pennsylvania proved more accurate than other methods previously attempted at this site. This paper will discuss the neural network input vector determination, including a watershed model of Oil Creek, and the methods used to appropriately account for the relatively low occurrence of jams. Discussion of how those vectors are estimated and how users of the predictor interface the software with a web-based package are presented.

DOI:

Year: 2002