A machine learning approach to nearshore wave modeling in large lakes using land-based wind observations

Abstract

This study presents a novel machine learning-based (ML) framework that utilizes the ConvLSTM-1D model to simulate wave heights at a nearshore location in Lake Michigan using a nonuniform land-based array of wind observations from a broad geographic region as an input. This approach was applied to Lake Michigan to perform a 70-year ice-free hindcast of waves near Chicago, IL (USA). Because of annual winter gaps in Great Lakes buoy observations, output from the Wave Information System model (WIS) was used for the training, validation, and testing of the ML model. Models with different numbers of wind stations were tested, showing that a single wind station as an input feature produced a reasonably accurate wave height. However, the wave height model accuracy increased as more wind input data was included from around the lake, largely plateauing beyond the inclusion of four stations that spanned Lake Michigan’s southern basin. The optimized model lookback period was found to be 10 h for all models, suggestive of a fetch-limited temporal coupling between the wind observations and nearshore waves. The extended time series showed a statistically significant increase in the interannual standard deviation of the storm wave height and the storm height over the mean wave height while no correlation was found with the annual water level time series. The ML framework offers a promising avenue for utilizing historical wind records to extend wave height time series for nearshore locations, particularly in enclosed and semi-enclosed basins where waves are strongly linked to local winds.