A Hierarchical Bayesian Model for Regression-Based Climate-Change Detection and Attribution

• Activity Number: 116
• Type: Topic Contributed
• Date/Time: Monday, August 4, 2014 : 8:30 AM to 10:20 AM
• Sponsor: Section on Bayesian Statistical Science
• Author(s): Matthias Katzfuss*+ and Dorit Hammerling and Richard L. Smith
• Companies: Texas A&M and NCAR and University of North Carolina
• Keywords: Empirical orthogonal functions ; Gaussian Markov random field ; Low-rank models ; Temperature ensembles ; Uncertainty

Abstract

Climate change detection and attribution methodology has evolved substantially within the climate community over the last decades and has started to attract the attention of statisticians. We propose to add to this development by presenting a Bayesian hierarchical statistical model for regression-based climate-change detection and attribution. The proposed model addresses some of the challenges of current approaches, such as high-dimensional covariance estimation and how to optimally choose the truncation of basis function expansions. Another feature of the proposed model is the comprehensive treatment and propagation of uncertainties from all components of the model. We apply the methodology to global temperature-change data.