Diagnosing Decadal-Scale Climate Variability in Current Generation Coupled Models for Informing Near-term Climate Change Impacts

As the relevance of climate change information grows, demand for that information, in particular covering the next 1ΓÇÉ2 decades increases. On the decadal timescale, both natural and anthropogenic factors will influence the evolution of the climate. The scientific community, particularly the international modeling community, has been working towards predictions/projections that consider both the changes in atmospheric composition, relevant to climate change projections, and initial oceanic conditions, relevant to decadalΓÇÉscale climate variability predictions. Initialization of dynamical models, while a very new effort, is considered crucial to reducing uncertainty in the nearΓÇÉterm climate projections. Even with initialized models, questions still exist on the degree to which they exhibit realistic variability on decadal time scales. It is imperative that we examine and document the characteristics of decadalΓÇÉscale variability in CGCMs, particularly in the context of initialized predictions, in order to prepare for the experimental decadal predictions that are starting to emerge from modeling centers. The three objectives that this proposal will address are: 

1) Determine the fidelity of the surface expression of oceanic decadal variability, and the associated climate teleconnections, in several stateΓÇÉofΓÇÉthe-art CGCMs, with particular emphasis on the impact of initialization. 

2) Develop metrics and baselines for estimating the quality of decadal predictions 

3) Design climate information products for climate risk management and planning. 

We have configured a team of researchers containing scientists from the top international modeling centers involved in the generation of climate projections and experimental decadal predictions, and scientists who focus on assessing and designing information that can benefit regional climate risk management. We will assess decadal predictability through the use of baselines, and by examining the impact of initialization. In any prediction system, one must know the quality of the models, particularly in a prediction context. One must also know how to handle biases that may be masking predictability, how to appropriately quantify uncertainty, and ultimately how to communicate that information in a meaningful, yet compact and flexible manner. The results of our project will contribute to progress in all these areas through work with existing and anticipated model simulations and hindcasts from state-ofΓÇÉthe-art CGCMs.