(non) Emergent Constraints

Thursday, 18 December 2014: 4:40 PM
Charles S Jackson1, Mohammad Wasef Hattab2 and Gabriel Huerta2, (1)University of Texas at Austin, Austin, TX, United States, (2)University of New Mexico Main Campus, Department of Mathematics and Statistics, Albuquerque, NM, United States
Emergent constraints are observable quantities that provide some physical basis for testing or predicting how a climate model will respond to greenhouse gas forcing. Very few such constraints have been identified for the multi-model CMIP archive. Here we explore the question of whether constraints that apply to a single model, a perturbed parameter ensemble (PPE) of the Community Atmosphere Model (CAM3.1), can be applied to predicting the climate sensitivities of models within the CMIP archive. In particular we construct our predictive patterns from multivariate EOFs of the CAM3.1 ensemble control climate. Multiple regressive statistical models were created that do an excellent job of predicting CAM3.1 sensitivity to greenhouse gas forcing. However, these same patterns fail spectacularly to predict sensitivities of models within the CMIP archive. We attribute this failure to several factors. First, and perhaps the most important, is that the structures affecting climate sensitivity in CAM3.1 have a unique signature in the space of our multivariate EOF patterns that are unlike any other climate model. That is to say, we should not expect CAM3.1 to represent the way another models within CMIP archive respond to greenhouse gas forcing. The second, perhaps related, reason is that the CAM3.1 PPE does a poor job of spanning the range of climates and responses found within the CMIP archive. We shall discuss the implications of these results for the prospect of finding emergent constraints within the CMIP archive. We will also discuss what this may mean for establishing uncertainties in climate projections.