Regional-scale climate change detection using a Bayesian decision method

Abstract

We use Bayesian statistics for a regional climate change detection problem and show an application for the East Asian surface air temperature ( SAT) field. Detection variables are constructed from a data-independent advection-diffusion model for SAT. Two scenario cases, namely a control scenario ( CTL) and a CO2-induced climate change scenario ( G), are derived from model integrations. The Bayesian decision process starts from prior probabilities, goes through the likelihood function where the observations enter, and finally produces posterior probabilities. We select the scenario of larger posterior probability given the observations, by which the theoretical decision error becomes a minimum. The application results for the East Asian SAT reveal strong G signals since 1990s insensitive to prior probabilities. The signal is carried on temporal scales longer than 1 year and spatial scales larger than 6000 km.