Modeling regional impacts of climate teleconnections using functional data analysis

Subscribe to email list

Please select the email list(s) to which you wish to subscribe.

User menu

You are here

Modeling regional impacts of climate teleconnections using functional data analysis

TitleModeling regional impacts of climate teleconnections using functional data analysis
Publication TypeJournal Article
Year of Publication2014
AuthorsBonner, SJ, Newlands, NK, Heckman, NE
Date PublishedMAR
Type of ArticleArticle
KeywordsAgriculture, Climate variability, Functional data analysis, Functional principal components analysis, Teleconnections
AbstractTeleconnections are quasi-periodic changes in atmospheric circulation that oscillate over long periods of time and impact climate over large regions. These patterns are often linked to long-term variations in climate and extreme weather events and may explain regional differences in climate vulnerability. We apply methods of functional data analysis to examine regional impacts of teleconnections on climate in British Columbia, Canada, between 1951 and 2000. We focus on monthly mean temperature as an overall determinant of crop growth and apply functional principal components analysis (FPCA) to study variations in the impacts of four major teleconnection indices affecting the Northern Hemisphere (the Southern Oscillation Index, the Pacific North American (PNA), Pacific Decadal Oscillation, and the North American Oscillation indices). Two challenges we consider are that the impacts of teleconnections cannot be observed directly and that fine scale data required to study regional variations may come from different sources with highly varied records. We first fit thin-plate regression splines to the raw data to construct complete series of pseudo-data at fixed grid points. Regression models incorporating Bayesian P-splines were then fit to the pseudo-data to estimate the impacts of the four teleconnections over time. Finally, FPCA was then applied to study regional variations in these effects. Our analysis identified strong variations in mean temperature associated with the PNA. The resulting spatial patterns also reveal areas of increased/decreased temperature variability that may have higher climate risk or be suitable for expansion of agricultural activity.