2007 Annual Meeting and Arctic Forum | Abstracts

Arctic Atmospheric Modeling: Should We Believe the Models?

Michael Tjernström^1
1Department of Meteorology, Stockholm University, Arrhenius Laboratory, Stockholm, SE-106 91, Sweden, Phone +46 8 163110, Fax +46 8 157185, michaelt@misu.su.se

Climate models form a backbone for climate research. Not only are they the only way by which we can say something about the future, but they are also fundamental tools for understanding the climate system. The climate system is so complex and non-linear that it is basically impossible for the human mind to comprehend without the aid of models describing as much as possible of all the processes and their interactions within the system.

But what is a climate model and why should we believe in them? From a philosophical point, a climate model is the aggregate of all our understanding of how the system, and the parts thereof, work. It is the sum total of all hypotheses about the processes we understand, refined to the point possible given the uncertainty of our current knowledge and the extent to which current computing power allows an implementation of this understanding into computer code. Climate models thus contain all the things we are pretty sure about, quite a few things we have a fair but not complete understanding about but also processes were we know we could do a lot better but are hindered by the practicalities of current modeling techniques. And there are also the things we sense might be important but do not know how to formalize into equations the way a model most be designed, and - off course - all the processes that we don't know that we don't know about; these could be important or not and since we don't know, how can we make an informed judgment?

In practice, we thus tend to believe in climate models for two reasons: 1) They are capable of reconstructing present climate, given the climate forcing of today; 2) They (are supposed to) take into account all the processes we think are important, including their interactions. At the heart of this problem lies the scale separation between resolved and unresolved scales. As computing power increases, the resolved scales becomes finer and finer and today's models properly resolve the "synoptic scale", or what we commonly refer to as the "weather systems". But a paradox in climate modeling is that the processes that drives the climate - and climate change - will likely "forever" remain unresolved in these models. In the atmosphere this is the case for radiation, cloud and aerosols, precipitation, boundary-layer turbulence, the energy balance and the exchange processes at the surface. These are all processes that have to be parameterized - described as functions of the resolved scale variables, applying some understanding that ultimately always relies on empirical evidence. This implies a fundamental uncertainty, and the IPCC AR4 report also points to the cloud processes as a fundamental reason for different models having different climate sensitivity. For the Arctic this is a particular problem since the body of empirical evidence for any process is small, certainly smaller than for other regions - or climate regimes - on Earth.

In this presentation I aim to show that both the fundamental cornerstones for believing in climate models may in fact be rather shaky at best. Current global climate models do not necessarily do a very good job of reconstructing the current arctic climate. And while the relevant processes are there, they are parameterized based on empirical evidence that does not come from the Arctic and thus, due to aspects of some climate processes that are unique to the Arctic, they don't always work very well. On top of these issues are off course the large inherent climate variability of the entire arctic climate system. Thus the paradox that although climate change is the fastest in the Arctic, the tools we use to understand this change are at the same time the poorest. The challenge therefore becomes to improve the representation of arctic climate in global climate models while at the same time preserving or improving their overall performance. This can only happen if we learn more about climate processes in the Arctic, which imply in-situ process studies in the Arctic. This is what IPY is - or should - be all about.