models, and key technical and scientific issues remain. For example, regional modeling efforts have been limited by constraints on computing resources, uncertainties and complexities associated with data assimilation and parameterization, the lack of a well-developed framework for downscaling, and the limitations of the large-scale simulations on which the downscaling is performed (Held and Soden, 2006; NRC, 2009k). An additional challenge for regional projections is representing regional modes of variability, such as ENSO and the Pacific Decadal Oscillation (described earlier in this chapter). Not only do these regional modes have a strong influence on local and regional climate change, but many also have global signatures, and they could potentially change themselves as the climate system warms. Finally, climate forcing scenarios that project human influences on local and regional climate, such as regional aerosol loading and land use change, are needed because these forcings may have a large influence on local and regional climate change (CCSP, 2008c).
The most comprehensive suite of climate modeling experiments performed to date were completed in 2005 as part of the World Climate Research Programme’s Coupled Model Intercomparison Project phase 3 (CMIP3; Meehl et al., 2007b) in support of the IPCC’s Fourth Assessment Report. CMIP3 included 23 different state-of-the-art models from groups around the world, all of which were run with a specific set of emissions scenarios (based on the SRES report described above) to facilitate comparison and synthesis of results. As described in detail by the IPCC (Meehl et al., 2007a), the CMIP3 climate models project increases in mean surface temperatures over the 21st century ranging from 2.0°F to 11.5°F (1.1°C to 6.4°C), relative to the 1980-1999 average, by the end of the century.
Figure 6.19 shows projected global temperature changes associated with three representative scenarios of high, medium-high, and low future GHG emissions. The separation between the three curves illustrates the uncertainty associated with the choice of scenario, while the uncertainties associated with differences among different models in simulating the climate system can be inferred from the shading surrounding each curve. The “commitment warming” associated with emissions through the year 2000 and, for two of the future forcing scenarios, through 2100 are also shown. These “commitment” runs, which are performed by instantaneously stabilizing atmospheric GHG concentrations, show that the climate system will continue to warm for several centuries after GHG emissions are stabilized—illustrating the inherent time lag between GHG emissions and the long-term climate response.