A comparison of Measure-Correlate-Predict Methodologies using LiDAR as a candidate site measurement device for the Mediterranean Island of Malta

Abstract

This study compares various MCP methodologies in predicting wind speed and direction at various heights. The candidate site measurements were obtained by means of a Light Detection and Ranging System (LiDAR) deployed on a building on the coast in the northern part of the Mediterranean Island of Malta. MCP methodologies tested Artificial Neural Networks, Support Vector Regression and Decision Trees, apart from the traditional regression techniques. The performance of the MCP techniques was analysed by means of coefficients of determination, together with the Mean Squared Error and the Mean Absolute Error of the residuals. Conclusions reached are that the results depend on the LiDAR measurement height and on the Measure-Correlate-Predict methodology used. Another conclusion drawn from the analysis is that although some regression methodologies show a better behaviour in correlating the candidate and reference site, they might show a different behaviour when used for prediction. Hence, there is no methodology which can be classified as being the best overall, but it is best to analyse various methodologies when applying the Measure-Correlate-Predict technique.