As is evident from the preceding discussion, models are prone to error from a variety of sources. A major thrust of the various MIPs is to examine such errors with the aim of developing better models. Of special relevance to regional climate models, GCMs and LSMs (with or without coupling to an atmospheric model) is sub-optimal parameterization. For simulations of the Arctic region, the parameterization of snow and sea ice albedo stands out. From Chapter 5, we know that the factors controlling snow and sea ice albedo are quite complex. The level of detail at which albedo is treated can impact strongly on simulated snow and ice mass, the surface energy balance and hence the strength of ice-albedo feedbacks. For atmospheric models, another problem with first-order impacts on the surface energy budget is the simulation of Arctic cloud cover. Especially for global climate models, the relatively coarse representations of topography, sea ice margins and coasts can have strong expressions on the simulated atmospheric circulation, such as in the placement of storm tracks. Global and regional models, including coupled ice-ocean models, also vary widely in their treatment of the ocean, which, among other things, will impact on sea ice growth and melt. In turn, different assumptions regarding ice interaction may lead to different realizations of ice circulation and ice thickness distributions. Turning back to the land surface, we saw that projected future changes in the Arctic's terrestrial carbon budget are critically dependent on assumptions regarding the function and structure of Arctic ecosystems and the appropriateness of "upscaling" results from a small area to a full Arctic view.
Many of the model types reviewed in this chapter require specified driving fields (e.g., "stand alone" LSMs and sea ice models, active layer thickness models and regional climate models). The quality of the model output will depend in part on the quality of these driving fields. Some simulations rely on sparse station data or remotely sensed fields (e.g., snow water equivalent) of uncertain quality. However, for many simulations, the driving fields represent output from another model, in particular, from NWP. Outputs from NWP models have their own sources of error, variously from difficulties in specifying the initial atmospheric state to biases in surface fields related to sub-optimal parameterization of albedo, evaporation and convective precipitation. In conclusion, while modeling is one of the most important tools for understanding the Arctic climate system, model output must always be viewed with appropriate caveats.
Was this article helpful?