Back to list of UTIG abstract submissions, Fall Agu 2003

Charles Jackson and Qiaozhen Mu
Institute for Geophysics
John A. and Katherine G. Jackson School of Geosciences
The University of Texas at Austin

One potential source of the large spread in sensitivities that exist among climate models in their response to projected increases in atmospheric CO2 concentrations is the way different models treat convection and clouds and the subjective choices that are made to specify bulk (non-observable) parameters. Quantifying these uncertainties is made much more difficult by the fact that the optimal choice of any single parameter value depends on the values of other key parameters. The solution requires that a multi-dimensional probability distribution be generated that describes regions of parameter space that enable the model to be most consistent with observational data. Toward this end we have devised a new EOF-based multivariate measure of model performance that compares model predictions of 12 quantities with observational or reanalysis data. Using this EOF-based measure of model performance, we have considered 5 key parameters within the NCAR atmospheric climate model (CCM3.10) that are important to convection and clouds and coarsely mapped their linear and non-linear sensitivities to changes in these parameters. We have found that there are multiple regions of parameter space that provide a close match to observational data, which suggests one reason why different modeling efforts may ‘settle’ on different values for the same bulk-parameters.