Simulation of Current and Future Climate in Antarctic Peninsula Region

B.D. Santer¹ and P.B. Duffy²

¹Program for Climate Model Diagnosis and Intercomparison, Lawrence Livermore National Laboratory, Livermore, California

²Atmospheric Science Division, Lawrence Livermore National Laboratory, Livermore, California

We examine numerical model simulations of present-day and future climate in the Antarctic Peninsula region. Simulations were performed with the 16 coupled atmospheric/ocean General Circulation Models (A/OGCMs) participating in Phase II of the Coupled Model Intercomparison Project (CMIP). Two 80-year integrations were conducted with each model: a control run with atmospheric CO2 fixed at roughly present-day levels, and a climate-change experiment with an idealized 1% per year (compound interest) CO2 increase, starting from a present-day climate state. The CMIP results help to identify systematic model errors in the simulation of present-day Antarctic Peninsula climate, and also provide valuable information on the model dependence of predicted climate changes.

The CMIP II A/OGCMs differ markedly in their horizontal and vertical resolution, and hence in their representation of bathymetry and orography in the Antarctic Peninsula study area considered here (55°-85°S, 120°-50°W). There are also pronounced differences in model treatment of the dynamics and rheology of sea-ice. In the CMIP II control runs, there is no clear link between differences in model resolution and sea-ice treatment and the strength of the Antarctic Circumpolar Current (ACC) or the extent of Antarctic sea-ice. Simulation of the strength of the ACC appears to be more closely tied to the mean strength of the westerlies and the surface wind stress. The climatological annual mean near-surface temperature pattern is reasonably well simulated by roughly half of the CMIP II models.

One simple measure of the model dependence of predicted climate changes is the signal-to-noise ratio, S. Here the "signal" is the projected change in a given variable, averaged over all 16 models, and the "noise" is the between-model variability of the signal. The projected surface temperature changes in the CO2 perturbation experiment have low S values (= 1) over large areas of the Southern Ocean, primarily due to large model differences in the magnitude and location of sea-ice changes. For precipitation, values of S are much higher in this region (1 to 2.5) due to a common increase in the poleward transport of moisture. This suggests that in the Southern Ocean, the precipitation signal in response to increasing CO2 may be less model-dependent than the temperature signal. Most mid- and low-latitude regions yield the opposite result. Over the Antarctic Peninsula itself, S is relatively low for both temperature and precipitation changes.