Analyzing environmental data within their natural sample space is essential for optimizing energy, economics, and transport systems. Wind direction, for example, significantly impacts aircraft landing, power transmission, and wind energy generation. However, its circular nature presents unique challenges. Previous research on wind direction distribution has employed methods such as Newton’s optimization of von Mises distribution mixtures, the EM algorithm, and evolutionary algorithms, focusing primarily on the approximation of individual functional observations. In this study, we advanced this field by introducing a compositional periodic spline representation of wind direction data within the Bayes spaces framework. The presented approach is efficient for processing directional data using functional data analysis techniques. Our theoretical framework will be validated on empirical wind direction datasets.