A Bayesian Approach to Significant Wave Height Modeling Using Measurement Data

Abstract Safe design and assessment of offshore structures rely heavily on statistical modeling of environmental variables such as significant wave height. While the models are typically constructed using hindcast data from numerical simulation or recorded observations, they face two key challenges: uncertainty in how well hindcast data represents environmental physics, and limited availability of site-specific observations. These factors contribute to epistemic uncertainty in the modeling process, which is due to lack of knowledge. This study investigates a Bayesian approach to update the statistical model by integrating the hindcast data and measurement data from offshore buoy. The method leverages hindcast data for the target site to construct the prior distribution for the model parameters, which are then sequentially updated as new measurement data becomes available. The probability model considered in this study is the generalized lognormal distribution as it is found to provide the best fit to the environmental variable at the target site. Model uncertainties are quantified through Bayesian posterior distributions, providing a comprehensive understanding of the prediction confidence. This Bayesian framework provides a systematic approach for reducing epistemic uncertainty by incorporating new measurement data to update probabilistic models. The results demonstrate the effectiveness of Bayesian updating in improving the accuracy of environmental variable modeling for offshore applications.