Measure-correlate-predict (MCP) methods infer the long-term wind resource climatology at prospective site from the wind climate of a predictor site using a relational model based on short-term concurrent wind data. Published relational models continue to rely on various forms of regression or on parametric joint wind speed distributions with marginal probability density functions (PDFs) of Weibull form. However, as shown in recent literature, the adequacy of Weibull model for describing marginal wind speed PDFs is less than general, a situation that cannot improve for the multivariate case. Furthermore, kinetic energy flux density and turbine output are non-linear functions, the distributions of which are highly sensitive to errors in the wind speed PDF.

An MCP method based on a direct orthogonal expansion of the joint turbine output density function is demonstrated, which avoids non-linear error amplification and the fitting limitations of simpler parametric techniques. The method is compared to a recently proposed method based on a parametric joint wind speed distribution model in light of the ability of each method to duplicate the statistics of wind turbine output – as inferred from observed wind speed data – over a long-term validation data set.