Simulating Great Lakes Water Levels For Erosion Prediction

Abstract

An investigation of monthly average lake levels is provided to enable the predictive portion of the lake level signal to be separated from random noise residual. Various empirical linear models are proposed and compared in a predictive mode to a portion of the historical record not used in the model parameter estimation. Models utilized for prediction and hence simulations assume a deterministic seasonal cycle as well as a longer term lake level pseudocycle consisting of correlated random noise. Residual analysis is performed to enable the models to be utilized in a simulation mode. The models developed are shown to provide satisfactory results for use in simulating future scenarios of the longer term lake level cycles and pseudocycles to drive shoreline erosion models.