Schemes for estimating the standard deviation of wind direction

Abstract

Wind direction is a circular variable. The algorithms used to find the standard deviation for this type of variables need the storage of all the data point in the sample. This causes problems in remote data acquisition systems since storage capacity would be reduced. For this reason, various estimators of the standard deviation of wind direction were developed. In this study, existing predictors were analyzed and attempts were made to improve them. Subsequently a set of estimators was developed using new approximations to the algorithm for the standard deviation of wind direction. All the predictors were tested using real data and statistical tests were carried out to assess their performance. Results show that the new predictors have a smaller bias and correlate better than the existing ones. It was also foreseen that the new approximations developed could provide new theoretical insight on circular variables. Some possible improvements on existing statistics have been proposed.