The improvement of wind-wave forecasts in the Great Lakes

The GLERL-Donelan parametric wind-wave model is presently used by the National Weather Service for Great Lakes forecasts. The suitability of this model is questionable, however. Wind sea and swell cannot be independently resolved, and shallow water effects are neglected. Following a review, the more advanced SHALWV wave model was selected for comparison to the GLERL-Donelan model. The SHALWV model has a spectral structure, distinguishes between wind sea and swell, includes shallow water effects, and has manageable computational requirements. Comparisons were made using idealized tests and Lake Erie lakewide and nearshore hindcasts. The standardized SWAMP and SWIM tests were first used to establish basic model characteristics. Idealized tests were then carried out using a Lake Erie bathymetry grid. The results of the idealized tests simplified analysis of the hindcasts. Lake Erie lakewide hindcasts were selected to coincide with buoy data. Predictions were evaluated both graphically and statistically. In turning wind cases the GLERL-Donelan model's lack of swell resolution led to a temporary underprediction of wave height, followed by excessive re-growth. The SHALWV model responded realistically to turning winds, but retained too much swell and consistently overpredicted wave periods. Nearshore hindcasts were made using data from the Lake Erie Coastal Boundary Layer Experiment. Nested-grid simulations showed that SHALWV refraction was generally accurate, but dissipation was underpredicted. GLERL-Donelan hindcasts confirmed major wave height overpredictions in shallow water. It was concluded that use of the SHALWV model would upgrade Great Lakes wave forecasts, but that retuning was advisable. Tuning was performed using the same hindcasts as in the model evaluation stage. The principle target was to minimize wave height overprediction, while eliminating the underprediction of wave height maxima. The success of each trial adjustment was determined statistically. A variable wind-input algorithm was found to mitigate excessive growth rates close to full sea development. The final, tuned model was named SHALWV-GL. An independent, verification hindcast confirmed that the use of SHALWV-GL would significantly improve the accuracy of Great Lakes wind-wave forecasts.