Time Series Modeling and Prediction of Circular Data

This project, supported by CSIRO and Data61, will explore time series modeling and prediction of circular data. Wind direction data in particular has observations that naturally fall around a (360 degree) circle. While circular methods have been developed to characterise the centre and spread of data around the circle, these methods have only been extended into supervised learning methods such as regression in limited ways. This project aims to extend circular methods into the time-series domain, allowing better understanding and inference related to circular time series data. Further, these methods will be extended in such a way as to create synthetic circular time-series data that can recapture key features of such data in observed data sets, in particular for fine-scale wind direction data. Project researchers are Professor Alan Welsh (RSFAS) and Dr Carolyn Huston (Data 61).