The scientific, statistical, and computational strategies that are used for uncertainty quantification are key to the future of climate model development. The objective of this proposal is to develop Bayesian statistical methodologies that can leverage high-performance computing (HPC) resources to reduce biases in future versions of National Center for Atmospheric Research's Community Atmosphere Model (CAM). In particular we shall use ideas of Gaussian Markov random fields to create a multi-variate metric that takes into account spatial and field dependencies that avoids many of the mathematical and scientific limitations imposed by more traditional strategies based on singular value decomposition and empirical orthogonal functions. The project will also identify the biases that most affect scatter in model projections of climate through use of multi-thousand member "perturbed physics" ensembles. This calculation will use ensembles of the slab ocean version of CAM3.1 and CAM5 to correlate biases in the regions and fields of the control (1xCO₂) climate with feedbacks affecting quantities of interest in projections of future (2xCO₂) climate. Finally, we will propose strategies for incorporating bias information within statistical measures of model acceptability. These efforts will support the development of a new paradigm in which statistical inference and HPC resources can accelerate the scientific activities associated with model development and uncertainty quantification. The software, experiment results, and documentation will be made freely available to the academic community via a web ftp data server.