Arctic sea ice extent has drawn increasing interest and alarm from geoscientists, owing to its rapid decline. In this article, we propose a Bayesian spatio-temporal hierarchical statistical model for binary Arctic sea ice data over two decades, where a latent dynamic spatio-temporal Gaussian process is used to model the data-dependence through a logit link function. Our ultimate goal is to perform inference on the dynamic spatial behavior of Arctic sea ice over a period of two decades. Physically motivated covariates are assessed using autologistic diagnostics. Our Bayesian spatio-temporal model shows how parameter uncertainty in such a complex hierarchical model can influence spatio-temporal prediction. The posterior distributions of new summary statistics are proposed to detect the changing patterns of Arctic sea ice over two decades since 1997.
BA Webinar: https://www.youtube.com/watch?v=yg4df9QBQL4&t=88s.
The acknowledgements section was corrected on 26 May 2020 by adding more details to Bohai Zhang’s research support.
"Bayesian Inference of Spatio-Temporal Changes of Arctic Sea Ice." Bayesian Anal. 15 (2) 605 - 631, June 2020. https://doi.org/10.1214/20-BA1209