Clustering for Climate Science Insights

Clustering is a popular unsupervised methodology used to divide individuals of a population or data points into groups based on their similarity. Accurately measuring similarities between data points is crucial to understanding the innate groupings in a dataset. In this talk, we will discuss some challenges in clustering spatiotemporal data, mainly when the data are mixtures of cross-sectional and time series data. A review of clustering in climate data will be given, along with applications that may inspire new clustering methods for weather and climatology data. We will discuss various strategies for estimating similarity and clustering through smoothing methodologies for mixed-type data. The Q\&A period will be open to discussing problems of climate science that may be addressed through clustering.

Bio:
Dr. John R.J. Thompson is an Assistant Professor at the University of British Columbia (Okanagan campus) whose areas of expertise are nonparametric and applied statistics and machine learning. His research interests lie in smoothing, distance metric learning, clustering, and change-point analysis. This research is motivated by applications to behavioural finance and environmental science, including estimating forest fire spread using anisotropic smoothing techniques and estimating fuel types in regional satellite imagery, clustering the trading behaviours of Canadian investors under the guidance of financial advisors, and designing effective financial measures and Robo-tools that aid financial advisors in supporting their clients’ investment portfolios.