Statistical Emulator for NARCCAP Surface Precipitation and Temperature Using Spatially Varying Bayesian Model Averaging

• Activity Number: 446
• Type: Contributed
• Date/Time: Tuesday, August 2, 2016 : 2:00 PM to 3:50 PM
• Sponsor: Section on Statistics and the Environment
• Author(s): Marcela Alfaro Cordoba* and Montse Fuentes and Joseph Guinness and Dorit Hammerling and Doug Nychka
• Companies: North Carolina State University and North Carolina State University and North Carolina State University and National Center for Atmospheric Research and National Center for Atmospheric Research
• Keywords: BMA ; Spatial ; Statistical Emulator ; Climate Models ; RCM

Abstract

Regional Climate Models provide high-resolution climate change simulations that are important to investigate uncertainties in regional scale projections of future climate, and generate climate change scenarios for use in impacts research. We present a statistical framework to emulate Regional Climate Model (RCM) output variables using output variables from Global Circulation Models (GCM) as covariates, as an alternative to the RCM runs which are computationally intensive. To this end, we use spatial-temporal measures of temperature and surface precipitation, produced by three different regional models and 50 variables from a GCM, grouped in different sets as covariates. With this, we develop a spatially varying version of Bayesian Model Averaging (BMA), capable of selecting the locations where each of the sets of covariates explains the RCMs output variables best, weighting the estimated response accordingly. We apply our methods to regional model output from the North American Regional Climate Change Assessment Program (NARCCAP) experiment introducing efficient computational algorithms.