CLIMATE RESEARCH GROUP

Optimizing multi-model climate predictions

Global climate models are central for the study of climate. Estimates of future climate change, including the latest IPCC report, depend largely on the insight gained from such models. However, climate models are far from perfect. Our still incomplete theoretical understanding of climate and simplifications needed to make the calculations feasible introduce well-known shortcomings into their simulations. Those deficits have attracted criticism, arguing that model based projections of climate are too unreliable to serve as a basis for public policy.

Performance index I2 for individual models (circles) and model generations (rows). Best performing models have low I2 values and are located towards the left. Circles sizes indicate the length of the 95% confidence intervals. Letters and numbers identify individual models (see supplemental online material); flux corrected models are labeled in red. Grey circles show the average I2 of all models within one model group. Black circles indicate the I2 of the multi-model mean taken over one model group. The green circle (REA) corresponds to the I2 of the NCEP/NCAR reanalyses (Kalnay et al. 1996). Last row (PICTRL) shows I2 for the preindustrial control experiment of the CMIP 3 project. Adapted from Reichler and Kim, 2008.

The overall goal of this project is to improve the certainty and detail of future climate predictions by investigating how well models simulate current climate. In other words, we develop a comprehensive metric of model performance that objectively describes the ability of models to simulate observed climate. The metric is then being used to devise an optimal strategy for the combination of climate predictions from different models. We use various innovative statistical approaches, which are then applied to the data archive provided by the IPCC-AR4 project (CMIP-3).

Further reading

Pennell, C. and T. Reichler (2008): On the effective number of climate models, Geophys. Res. L., , (sub judice). [PDF]

Reichler, T., and J. Kim (2008): How Well do Coupled Models Simulate Today's Climate?, Bull. Amer. Meteor. Soc, 89, 303-311. [PDF]


This page has been visited [an error occurred while processing this directive] times since Friday, 31-Oct-2008 19:08:01 MDT