A Summit on Climate Prediction

Climate Summit: What Direction Now?

I’ve been going to a lot of meetings lately where we are talking about where the climate modeling community is talking about what to do next? What comes after the IPCC Fourth Assessment Report (AR4) as the results of this assessment permeates society?

The World Modeling Summit for Climate Predictions: The first meeting I attended was hosted by the European Center for Medium-range Weather Forecasts and sponsored by the World Climate Research Program, the World Weather Research Program, and the International Geosphere-Biosphere Program. This meeting of more than 100 scientists had people from both the weather and climate community. Much of the discussion was, in fact, about the efficacy of starting climate prediction efforts that are in the spirit of weather forecasts. At the core of the discussion was a recent paper by Tim Palmer and co-authors on “Seamless Prediction,” the idea of looking at scales from regional to global weather forecasting, to monthly, to El Nino, to decadal, to century long forecasts in an integrated, consistent matter. This is a subject of significant controversy because the method, practice, and evaluation of model performance is quite different in the weather and climate communities.

While there is a still evolving message coming from this summit, it is safe to say that all agreed that a much more robust ability to predict climate on decadal scales is needed. This will be required to provide information that is good enough for resource managers to make decisions about, especially, water resources. Much of the discussion was whether or not we have to wait until we have global models capable of modeling, explicitly, cloud systems. These cloud-resolving models are many years and several generations of computers away. It is my opinion that we cannot “wait” for these models; we have the potential to provide much more robust information with current and next-generation models.

I am of the camp that we should focus on developing a generation of climate prediction models that are at comparable resolution to the current generation of weather forecast models; that is, approximately a 20 km grid size on the Earth’s surface. At this resolution we should be able to represent the mechanisms, low level jet streams, which supply moisture to the continents. We should also do a reasonable representation of topography, land-ocean contrast, as well as the representation of desert boundaries. With this capability, then we should be able to provide much more robust analysis of regional impacts of global warming.

Both of these paths are necessary, but the question of priority is a question of resources. Again, thinking about time scales is important. We will need information for policy, infrastructure expenditures, and adaptation in the next 2 years, 5 years, and 10 years. Scientific investigation can contribute to answering these questions. Therefore, a set of requirements for the scientific community arises from outside the scientific community. This can, and often does, stand in contrast to the requirements generated by scientists. Again, opinion, I believe our community must organize and allocate resources to address these questions as best as we can at any given time.

One of the truths of climate and weather prediction is the need for high-performance computing. Climate and weather modeling has, from its beginning, been one of the drivers of high-performance computing. The development of computational platforms is not straightforward and the underlying hardware has undergone great changes. Twenty years ago custom hardware was made for scientific computing. This business model was too costly, so today high-performance computers are based on commodity hardware, and it is much harder to program large codes, like climate models, on these computers. A new challenge that we are facing today is the fact that computer chips are coming up against physical limits of size and their ability to cool.

The links above give you access to all of the presentations from the Summit.

The other meetings I have attended include an excellent workshop on the numerical techniques for future climate models (see also, this discussion and on facebook?) and, right now, the annual meeting of the Community Climate System Model, where a model for the next IPCC assessment is being configured.

Are you interested in reports from these meetings?

r