Experimental Wave Predictions

The Great Lakes Experimental Wave Model was developed based off of the operational Great Lakes Wave Unstructured v2.1 (GLWUv2.1) model, built using WAVEWATCH III, and available at https://www.weather.gov/greatlakes (Abdolali et al. 2024).

This modified experimental model configuration incorporates an ice-wave damping scheme intended to provide more accurate and comprehensive estimates of wave conditions in the presence of ice cover, and makes minor changes to the model mesh (Hu et al. 2025). Wind forcing is from the HRRR for the first 48 hours and GFS for the remainder of the forecast window. Ice forcing is based off of the US NIC Great Lakes Ice Charts and held constant through the forecast window.

The Great Lakes Experimental Wave Model runs two times per day, each producing 3-day forecasts of wave heights throughout the Great Lakes. Recent nowcast data is available for a 2-week rolling window, and recent forecast data is available for a 3-day rolling window. Because this is not an operational system, we are unable to provide model archives outside of these rolling real-time windows.

Abdolali, A., Banihashemi, S., Alves, J.H., Roland, A., Hesser, T.J., Anderson Bryant, M., McKee Smith, J. 2024. Great Lakes wave forecast system on high-resolution unstructured meshes. Geoscientific Model Development, 17(3), pp.1023-1039. https://doi.org/10.5194/gmd-17-1023-2024

Hu, H., Titze, D., Fujisaki-Manome, A., Mroczka, B., Wang, J., Hawley, N., Orendorf, S., Frank, K., Ruberg, S. 2025. Winter ice‐wave modeling with WAVEWATCH III in Lake Erie. Journal of Geophysical Research: Oceans, 130(1). https://doi.org/10.1029/2024JC021146

This Great Lakes Experimental Point-Based Ensemble Forecast System supplements existing Great Lakes wave forecast products by generating a range of possible wave conditions to understand uncertainty in forecast conditions.

The point-based ensemble forecast system was developed using a machine learning approach. Models were trained on data from the National Data Buoy Center (NDBC) using the eXtreme Gradient Boost (XGBoost) library, based upon the methods of Hu et al. 2021. Site-specific models were developed for buoy locations throughout the Great Lakes with sufficient historical records for model training.

The ensemble forecast is generated by forcing these models with winds from the 31-member NOAA Global Ensemble Forecast System (GEFS) to produce a corresponding 31-member ensemble forecast of significant wave height. Results from the GLERL Experimental WAVEWATCH III model are plotted for comparison, along with results from an XGBoost model forced with HRRR (-12-48hr) and GFS (48-120hr) winds. The effects of ice cover are not considered in the machine learning forecasts at this time.

The Great Lakes Experimental Point-Based Ensemble Forecast System runs two times per day, each producing 10-day forecasts of wave heights at stations throughout the Great Lakes. Recent forecast data is available for a 3-day rolling window. Because this is not an operational system, we are unable to provide model archives outside of this rolling real-time window.

Hu, H., van der Westhuysen, A.J., Chu, P., Fujisaki-Manome, A. 2021. Predicting Lake Erie wave heights and periods using XGBoost and LSTM. Ocean Modelling, 164, p.101832. https://doi.org/10.1016/j.ocemod.2021.101832