Ensemble Simulation of Twenty-First Century Climate Changes: Business- as-Usual versus CO2 Stabilization

As the world economy and population grow in the twenty-first century, atmospheric concentrations of CO 2 and other greenhouse gases are likely to continue to rise (Naki enovi and Swart 2000). Credible projections of future climate changes induced by these increases have many practical applications (e.g., for climate impact assessments and to guide future emissions controls). Most previous simulations of such changes using coupled ocean–atmosphere general circulation models (GCMs; Manabe et al. 1991; Cubasch et al. 1992; Mitchell et al. 1995; Roeckner et al. 1999; Russell and Rind 1999; Boer et al. 2000; Meehl et al. 2000a,b; Dai et al. 2001a) have been carried out using a single realization for each emissions scenario. However, it is well-known that individual realizations show considerable differences from run to run (e.g., Delworth and Knutson 2000), since each run includes its own specific realization of the model’s internal variability. These differences cause uncertainties in modelsimulated climate changes that are in addition to uncertainties associated with a model’s climate sensitivity and other intermodel differences (Allen et al. 2000) and with future emissions (Kattenberg et al. 1996). While areal and temporal averaging can partly reduce uncertainties related to internal variability, intraensemble differences can still make climate changes on regional (L ~ 10 km) and smaller scales from single realizations noisy and irreproducible (Dai et al. 2001a). The preferred solution to this problem is to carry out multiple realizations (an ensemble) for each emissions case (starting from different initial conditions), and to average these. Ensembles have been used in climate studies with GCMs (e.g., Cubasch et al. 1994; Zwiers 1996; Hansen et al. 1997; Rowell 1998; Delworth and Knutson 2000; Wehner 2000; Stott et al. 2000, 2001). Most previous work, however, has employed only relatively short ensemble simulations or Ensemble Simulation of Twenty-First Century Climate Changes: Businessas-Usual versus CO2 Stabilization