A Bayesian hierarchical approach for spatial analysis of climate model bias in multi-model ensembles

Abstract

Coupled atmosphere–ocean general circulation models are key tools to investigate climate dynamics and the climatic response to external forcings, to predict climate evolution and to generate future climate projections. Current general circulation models are, however, undisputedly affected by substantial systematic errors in their outputs compared to observations. The assessment of these so-called biases, both individually and collectively, is crucial for the models’ evaluation prior to their predictive use. We present a Bayesian hierarchical model for a unified assessment of spatially referenced climate model biases in a multi-model framework. A key feature of our approach is that the model quantifies an overall common bias that is obtained by synthesizing bias across the different climate models in the ensemble, further determining the contribution of each model to the overall bias. Moreover, we determine model-specific individual bias components by characterizing them as non-stationary spatial fields. The approach is illustrated based on the case of near-surface air temperature bias in the tropical Atlantic and bordering regions from a multi-model ensemble of historical simulations from the fifth phase of the Coupled Model Intercomparison Project. The results demonstrate the improved quantification of the bias and interpretative advantages allowed by the posterior distributions derived from the proposed Bayesian hierarchical framework, whose generality favors its broader application within climate model assessment.