Since the Third Assessment Report (TAR), the scientific community has undertaken the largest coordinated global coupled climate model experiment ever attempted in order to provide the most comprehensive multi-model perspective on climate change of any IPCC assessment, the World Climate Research Programme (WCRP) Coupled Model Intercomparison Project phase three (CMIP3), also referred to generically throughout this report as the ‘multi-model data set’ (MMD) archived at the Program for Climate Model Diagnosis and Intercomparison (PCMDI). This open process involves experiments with idealised climate change scenarios (i.e., 1% yr–1 carbon dioxide (CO2) increase, also included in the earlier WCRP model intercomparison projects CMIP2 and CMIP2+ (e.g., Covey et al., 2003; Meehl et al., 2005b), equi- librium 2 × CO2 experiments with atmospheric models coupled to non-dynamic slab oceans, and idealised stabilised climate change experiments at 2 × CO2 and 4 × atmospheric CO2 levels in the 1% yr–1 CO2 increase simulations).
In the idealised 1% yr–1 CO2 increase experiments, there is no actual real year time line. Thus, the rate of climate change is not the issue in these experiments, but what is studied are the types of climate changes that occur at the time of doubling or quadrupling of atmospheric CO2 and the range of, and difference in, model responses. Simulations of 20th-century climate have been completed that include temporally evolving natural and anthropogenic forcings. For projected climate change in the 21st century, a subset of three IPCC Special Report on Emission Scenarios (SRES; Nakićenović and Swart, 2000) scenario simulations have been selected from the six commonly used marker scenarios. With respect to emissions, this subset (B1, A1B and A2) consists of a ‘low’, ‘medium’ and ‘high’ scenario among the marker scenarios, and this choice is solely made by the constraints of available computer resources that did not allow for the calculation of all six scenarios. This choice, therefore, does not imply a qualification of, or preference over, the six marker scenarios. In addition, it is not within the scope of the Working Group I contribution to the Fourth Assessment Report (AR4) to assess the plausibility or likelihood of emission scenarios.
In addition to these non-mitigation scenarios, a series of idealised model projections is presented, each of which implies some form and level of intervention: (i) stabilisation scenarios in which greenhouse gas concentrations are stabilised at various levels, (ii) constant composition commitment scenarios in which greenhouse gas concentrations are fixed at year 2000 levels, (iii) zero emission commitment scenarios in which emissions are set to zero in the year 2100 and (iv) overshoot scenarios in which greenhouse gas concentrations are reduced after year 2150.
The simulations with the subset A1B, B1 and A2 were performed to the year 2100. Three different stabilisation scenarios were run, the first with all atmospheric constituents fixed at year 2000 values and the models run for an additional 100 years, and the second and third with constituents fixed at year 2100 values for A1B and B1, respectively, for another 100 to 200 years. Consequently, the concept of climate change commitment (for details and definitions see Section 10.7) is addressed in much wider scope and greater detail than in any previous IPCC assessment. Results based on this Atmosphere-Ocean General Circulation Model (AOGCM) multi-model data set are featured in Section 10.3.
Uncertainty in climate change projections has always been a subject of previous IPCC assessments, and a substantial amount of new work is assessed in this chapter. Uncertainty arises in various steps towards a climate projection (Figure 10.1). For a given emissions scenario, various biogeochemical models are used to calculate concentrations of constituents in the atmosphere. Various radiation schemes and parametrizations are required to convert these concentrations to radiative forcing. Finally, the response of the different climate system components (atmosphere, ocean, sea ice, land surface, chemical status of atmosphere and ocean, etc.) is calculated in a comprehensive climate model. In addition, the formulation of, and interaction with, the carbon cycle in climate models introduces important feedbacks which produce additional uncertainties. In a comprehensive climate model, physical and chemical representations of processes permit a consistent quantification of uncertainty. Note that the uncertainties associated with the future emission path are of an entirely different nature and not considered in this chapter.
Figure 10.1. Several steps from emissions to climate response contribute to the overall uncertainty of a climate model projection. These uncertainties can be quantified through a combined effort of observation, process understanding, a hierarchy of climate models, and ensemble simulations. In a comprehensive climate model, physical and chemical representations of processes permit a consistent quantification of uncertainty. Note that the uncertainty associated with the future emission path is of an entirely different nature and not addressed in Chapter 10. Bottom row adapted from Figure 10.26, A1B scenario, for illustration only.
Many of the figures in Chapter 10 are based on the mean and spread of the multi-model ensemble of comprehensive AOGCMs. The reason to focus on the multi-model mean is that averages across structurally different models empirically show better large-scale agreement with observations, because individual model biases tend to cancel (see Chapter 8). The expanded use of multi-model ensembles of projections of future climate change therefore provides higher quality and more quantitative climate change information compared to the TAR. Even though the ability to simulate present-day mean climate and variability, as well as observed trends, differs across models, no weighting of individual models is applied in calculating the mean. Since the ensemble is strictly an ‘ensemble of opportunity’, without sampling protocol, the spread of models does not necessarily span the full possible range of uncertainty, and a statistical interpretation of the model spread is therefore problematic. However, attempts are made to quantify uncertainty throughout the chapter based on various other lines of evidence, including perturbed physics ensembles specifically designed to study uncertainty within one model framework, and Bayesian methods using observational constraints.
In addition to this coordinated international multi-model experiment, a number of entirely new types of experiments have been performed since the TAR to quantify uncertainty regarding climate model response to external forcings. The extent to which uncertainties in parametrizations translate into the uncertainty in climate change projections is addressed in much greater detail. New calculations of future climate change from the larger suite of SRES scenarios with simple models and Earth System Models of Intermediate Complexity (EMICs) provide additional information regarding uncertainty related to the choice of scenario. Such models also provide estimates of long-term evolution of global mean temperature, ocean heat uptake and sea level rise due to thermal expansion beyond the 21st century, and thus allow climate change commitments to be better constrained.
Climate sensitivity has always been a focus in the IPCC assessments, and this chapter assesses more quantitative estimates of equilibrium climate sensitivity and transient climate response (TCR) in terms of not only ranges but also probabilities within these ranges. Some of these probabilities are now derived from ensemble simulations subject to various observational constraints, and no longer rely solely on expert judgement. This permits a much more complete assessment of model response uncertainties from these sources than ever before. These are now standard benchmark calculations with the global coupled climate models, and are useful to assess model response in the subsequent time-evolving climate change scenario experiments.
With regard to these time-evolving experiments simulating 21st-century climate, since the TAR increased computing capabilities now allow routine performance of multi-member ensembles in climate change scenario experiments with global coupled climate models. This provides the capability to analyse more multi-model results and multi-member ensembles, and yields more probabilistic estimates of time-evolving climate change in the 21st century.
Finally, while future changes in some weather and climate extremes (e.g., heat waves) were addressed in the TAR, there were relatively few studies on this topic available for assessment at that time. Since then, more analyses have been performed regarding possible future changes in a variety of extremes. It is now possible to assess, for the first time, multi-model ensemble results for certain types of extreme events (e.g., heat waves, frost days, etc.). These new studies provide a more complete range of results for assessment regarding possible future changes in these important phenomena with their notable impacts on human societies and ecosystems. A synthesis of results from studies of extremes from observations and model is provided in Chapter 11.
The use of multi-model ensembles has been shown in other modelling applications to produce simulated climate features that are improved over single models alone (see discussion in Chapters 8 and 9). In addition, a hierarchy of models ranging from simple to intermediate to complex allows better quantification of the consequences of various parametrizations and formulations. Very large ensembles (order hundreds) with single models provide the means to quantify parametrization uncertainty. Finally, observed climate characteristics are now being used to better constrain future climate model projections.