Evaluating the Forecast Skill of Two Coastal Wave Simulation Systems

Abstract

A wave transformation model, SWAN, is used to predict wave conditions at a high spatial resolution in Prince William Sound (Alaska) and in Massachusetts Bay (Gulf of Maine). The motivation is to provide forecasts to assist mariners, oil spill modelers, recreational boaters, and others. The US National Oceanic and Atmospheric Administration (NOAA) provides both wind and wave fields for these domains, but they are on a much coarser grid which does not account for local topographic variations. However, their data can be used as forcing functions and boundary conditions for the regional wave prediction systems developed here. Two PC and internet-based systems are established to automatically provide 12-hour hindcasts and 48-hour forecasts of the wave conditions at a resolution of 0.5 km or less. Simulated results are compared with measurements from four buoys for different time periods to evaluate the reliability of the wave forecasting system. The 48-hour forecasts show reasonably good agreement with data, and in some cases the comparison is much better as the forecast lead time decreases. The reliability of the results demonstrated here suggests that they may be used for many practical applications.