Evaluation of Bias Correction Methods for Nearshore Wave Modeling

Models that seek to predict environmental variables invariably demonstrate bias when compared to observations. Bias correction (BC) techniques are common, even requisite, in the climate and hydrological modeling community but have seen few applications in the field of nearshore wave modeling. Overall, there remains little investigation or guidance as to which BC methodology is most applicable in this new context. This paper introduces and compares a subset of BC methods with the goal of clarifying a “best practice” methodology. Specific focus is paid to comparing parametric vs. empirical methods as well as univariate vs. bivariate methods. The methods are tested on a dataset produced by WAVEWATCH III and compared to available buoy data. Both wave height and wave period are considered in order to investigate effects on inter-variable correlation. Results show that all methods perform relatively uniformly in terms of properly correcting statistical moments of individual variables. When comparing parametric and empirical methods, the parametric method is found to perform poorly when directly comparing the bivariate cumulative distribution functions (CDFs). This result is not shown when looking at bulk statistics such as correlation coefficients or statistical moments. Between bivariate methods and univariate methods, it is found that bivariate methods greatly improve inter-variable correlations. In summary, this study demonstrates that bias correction methods are effective when applied to wave model data and that it is essential to employ methods that consider the relationships between variables.