Wind Wave Modeling in the Lake Erie

Session: 47. - Physical Processes in Lakes

Haoguo Hu, CIGLR, University of Michigan, Ann Arbor, [email protected]
Philip Chu, NOAA/GLERL, [email protected]

Abstract

An unstructured WaveWatch III (WW3) model was implemented in the Lake Erie for simulating wind waves. The preliminary results were compared with wave height from NOAA/GLERL Great Lakes Coastal Forecasting System. WW3 results are also compared with Buoy data in the Lake Erie. Errors are analyzed and it shows that the errors are mainly caused by errors of wind data. The Mean Absolute error of wave height is 0.16, variance is 0.06, skewness =1.72, and kurtosis = 9.29.

A novel machine learning method is introduced for data analyzing and for model training.  We used 2010-2015 data for training a model, and use the trained model to predict 2016 wave features. The results are inspiring. The Mean Absolute error of wave height is 0.19, variance is 0.09, skewness =1.57, and kurtosis = 6.36.

1. Keyword
waves

2. Keyword
modeling

3. Keyword
Lake Erie

4. Additional Keyword
Machine Learning