The PI will investigate statistical methods, based on first moments or climate averages that can contribute to the goal of quantitative evaluation of the credibility of climate change models. Space data on resolved thermal radiances and radio refractivities obtained from Global Positioning System (GPS) occultations are particularly suitable for this task. The statistical approach is Bayesian, with three levels of inference. The first involves detecting a specific forcing in the climate record. The methodology is that of optimal fingerprinting and this level of inference corresponds to existing "detection and attribution" studies. The second level of inference permits an objective evaluation of the quality of the model, and the third allows conclusions to be drawn about model improvement. The Bayesian paradigm permits the direct use of observed data (radiances or refractivities) without inversion to physical climate variables. Synthetic data will be constructed by means of forward calculations from the output of fully coupled climate models. By the end of this grant the methodology should have advanced to a point at which it can be combined with other work on statistical second moments to give the basis for a systematic approach to testing, improving and evaluating climate models. The intellectual merit of this project lies in the use of Bayesian methods to understand the relationship of observed data from space to the development of credible predictions from numerical climate models. The broader impacts of this project lie in (1) the potential to improve the quality of climate projections and to increase their credibility, and (2) the provision of objective criteria for climate observing systems. Both are central to the societal goals described in the recently issued Strategic Plan for the U.S. Climate Change Science Program. Two graduate students will be involved.
