A REGRESSION MODEL FOR STORM SURGE PREDICTION

Abstract

The practical difficulties arising in the solution of the hydrodynamic equations for storm surges by numerical methods are reviewed. It is concluded that some of these can be avoided by means of a statistical approach if sufficient records of past surges are available. A statistical approach, based on dynamic principles, is presented and the recent literature of similar studies is reviewed. A test is made of this statistical method by applying it to recorded storm surges on Lake Erie. The agreement between the surge values computed by the statistical method and the observed values is considered good. The test verified the statistical approach but did not lead to an operational prediction system, because of recent changes in the observational practices at some of the weather stations bordering Lake Erie. One somewhat unexpected result was a finding that a prediction scheme based on the assumption of wind stress proportional to wind speed is not significantly inferior to one-based on the assumption of a quadratic wind stress law.