High-Resolution Global Simulation Of The Climatic Effects Of Increased Greenhouse Gases

HIGH-RESOLUTION GLOBAL SIMULATION OF THE CLIMATIC EFFECTS OF INCREASED GREENHOUSE GASES
B. Govindasamy* and P. B. Duffy Climate and Carbon Cycle Modeling Group Atmospheric Science Division Lawrence Livermore National Laboratory CA 94550, USA

1. INTRODUCTION Typical state of the art Atmospheric General Circulation Models (AGCMs) used in climate change studies have approximately 300 kmresolution in the horizontal. As computing power increases, many climate modeling groups are exploring the possibility of enhancing the resolution of AGCMs. In principle, high-resolution models have several advantages. Dynamics, topography and the land sea mask are better resolved compared to their coarse resolution counterparts. More physical processes can be explicitly represented reducing dependence onsemiempirical parameterizations. With a highresolution model, regional spatial scale details are simulated and hence prediction of regional climate change becomes more credible. Because high-resolution models simulate a wider spectrum of spatial scales and their nonlinear interactions, in principle, even the larger scale features should be better simulated by them. This was found to be the caseby Williamson (1999) and Duffy et al. (2002). Higher resolution global models have the added advantage over regional models that they avoid the numerical problems associated with lateral boundary conditions. They also avoid the scale separation issues that are faced by regional models driven by coarse resolution global models. In this paper, we report on the simulation of global climate changeusing the highest resolution (T170) yet performed for the global domain with an AGCM. Here, our primary motivation is to investigate if climate sensitivity on both global and regional scales depends on the resolution of the model. This
_____________________________________ *Corresponding author address: Bala Govindasamy, L-103, Atmospheric Science Division, Lawrence Livermore National Laboratory,Livermore, CA 94550; e-mail: bala@LLNL.GOV

is an important issue in global climate models whenever their resolution is changed. Because of the high cost of computing, we compare only two resolutions in this study, T42 and T170. Caution should be exercised in interpreting our results because we have performed climate simulations using a single atmospheric general circulation model driven byprescribed sea surface temperatures and sea ice. It lacks the feedbacks associated with a fully dynamically coupled ocean, and those associated with ocean and land carbon cycles. Traditional estimates of climate model sensitivity use AGCMs coupled to a mixed layer ocean. Instead, we have used a simplified model formulation for our estimate of climate sensitivity. Therefore, it is possible that otheratmospheric GCMs coupled to ocean and carbon cycle models would yield qualitatively and quantitatively different results (Hansen et al., 1999). 2. THE MODEL We use National Center for Atmospheric Research (NCAR) CCM3 (Kiehl et al., 1996). CCM3 is a spectral model with a specified number of spherical harmonics to represent the horizontal structure of prognostic variables. In the vertical, a hybridsigmapressure coordinate system is used. For our experiments we used 42 and 170 waves in the horizontal: the horizontal resolution is ~2.8° (grid spacing of 300 km) in latitude and longitude for T42 and 0.70 (~ 75 km) for T170. The model has 18 levels in the vertical. The model has been extensively "tuned" to optimize results at T42. In part as a result of this tuning, the standard configuration (T42)has now very little systematic bias in the topof-atmosphere and surface energy budgets. We adopted a version of CCM3 that uses

prescribed sea surface temperature and sea ice as lower boundary condition. 3. EXPERIMENTS Two simulations are performed at each resolution (T42 and T170): 1) a control simulation with present day climatological SSTs and sea ice extent, 2) a "2100 AD" simulation...