10.5.3 Global Mean Responses from Different Scenarios
The TAR projections with an SCM presented a range of warming over the 21st century for 35 SRES scenarios. The SRES emission scenarios assume that no climate policies are implemented (Nakićenović and Swart, 2000). The construction of Figure 9.14 of the TAR was pragmatic. It used a simple model tuned to AOGCMs that had a climate sensitivity within the long-standing range of 1.5°C to 4.5°C (e.g., Charney, 1979; and stated in earlier IPCC Assessment Reports). Models with climate sensitivity outside that range were discussed in the text and allowed the statement that the presented range was not the extreme range indicated by AOGCMs. The figure was based on a single anthropogenic-forcing estimate for 1750 to 2000, which is well within the range of values recommended by TAR Chapter 6, and is also consistent with that deduced from model simulations and the observed temperature record (TAR Chapter 12.). To be consistent with TAR Chapter 3, climate feedbacks on the carbon cycle were included. The resulting range of global mean temperature change from 1990 to 2100 given by the full set of SRES scenarios was 1.4°C to 5.8°C.
Since the TAR, several studies have examined the TAR projections and attempted probabilistic assessments. Allen et al. (2000) show that the forcing and simple climate model tunings used in the TAR give projections that are in agreement with the observationally constrained probabilistic forecast, reported in TAR Chapter 12.
As noted by Moss and Schneider (2000), giving only a range of warming results is potentially misleading unless some guidance is given as to what the range means in probabilistic terms. Wigley and Raper (2001) interpret the warming range in probabilistic terms, accounting for uncertainties in emissions, the climate sensitivity, the carbon cycle, ocean mixing and aerosol forcing. They give a 90% probability interval for 1990 to 2100 warming of 1.7°C to 4°C. As pointed out by Wigley and Raper (2001), such results are only as realistic as the assumptions upon which they are based. Key assumptions in this study were that each SRES scenario was equally likely, that 1.5°C to 4.5°C corresponds to the 90% confidence interval for the climate sensitivity, and that carbon cycle feedback uncertainties can be characterised by the full uncertainty range of abundance in 2100 of 490 to 1,260 ppm given in the TAR. The aerosol probability density function (PDF) was based on the uncertainty estimates given in the TAR together with constraints based on fitting the SCM to observed global and hemispheric mean temperatures.
The most controversial assumption in the Wigley and Raper (2001) probabilistic assessment was the assumption that each SRES scenario was equally likely. The Special Report on Emissions Scenarios (Nakićenović and Swart, 2000) states that ‘No judgment is offered in this report as to the preference for any of the scenarios and they are not assigned probabilities of occurrence, neither must they be interpreted as policy recommendations.’
Webster et al. (2003) use the probabilistic emissions projections of Webster et al. (2002), which consider present uncertainty in SO2 emissions, and allow the possibility of continuing increases in SO2 emissions over the 21st century, as well as the declining emissions consistent with SRES scenarios. Since their climate model parameter PDFs were constrained by observations and are mutually dependent, the effect of the lower present-day aerosol forcing on the projections is not easy to separate, but there is no doubt that their projections tend to be lower where they admit higher and increasing SO2 emissions.
Irrespective of the question of whether it is possible to assign probabilities to specific emissions scenarios, it is important to distinguish different sources of uncertainties in temperature projections up to 2100. Different emission scenarios arise because future greenhouse gas emissions are largely dependent on key socioeconomic drivers, technological development and political decisions. Clearly, one factor leading to different temperature projections is the choice of scenario. On the other hand, the ‘response uncertainty’ is defined as the range in projections for a particular emission scenario and arises from the limited knowledge of how the climate system will react to the anthropogenic perturbations. In the following, all given uncertainty ranges reflect the response uncertainty of the climate system and should therefore be seen as conditional on a specific emission scenario.
The following paragraphs describe the construction of the AR4 temperature projections for the six illustrative SRES scenarios, using the SCM tuned to 19 models from the MMD (see Section 8.8). These 19 tuned simple model versions have effective climate sensitivities in the range 1.9°C to 5.9°C. The simple model sensitivities are derived from the fully coupled 2 × and 4 × CO2 1% yr–1 CO2 increase AOGCM simulations and in some cases differ from the equilibrium slab ocean model sensitivities given in Table 8.2.
The SRES emission scenarios used here were designed to represent plausible futures assuming that no climate policies will be implemented. This chapter does not analyse any scenarios with explicit climate change mitigation policies. Still, there is a wide variation across these SRES scenarios in terms of anthropogenic emissions, such as those of fossil CO2, CH4 and SO2 (Nakićenović and Swart, 2000) as shown in the top three panels of Figure 10.26. As a direct consequence of the different emissions, the projected concentrations vary widely for the six illustrative SRES scenarios (see panel rows four to six in Figure 10.26 for the concentrations of the main greenhouse gases, CO2, CH4 and N2O). These results incorporate the effect of carbon cycle uncertainties (see Section 10.4.1), which were not explored with the SCM in the TAR. Projected CH4 concentrations are influenced by the temperature-dependent water vapour feedback on the lifetime of CH4.
In Figure 10.26, the plumes of CO2 concentration reflect high and low carbon cycle feedback settings of the applied SCM. Their derivation is described as follows. The carbon cycle model in the SCM used here (Model for the Assessment of Greenhouse-gas Induced Climate Change: MAGICC) includes a number of climate-related carbon cycle feedbacks driven by global mean temperature. The parametrization of the overall effect of carbon cycle feedbacks is tuned to the more complex and physically realistic carbon cycle models of the C4MIP (Friedlingstein et al, 2006; see also Section 10.4) and the results are comparable to the BERN-CC model results across the six illustrative scenarios. This allows the SCM to produce projections of future CO2 concentration change that are consistent with state-of-the-art carbon cycle model results. Specifically, the C4MIP range of CO2 concentrations for the A2 emission scenario in 2100 is 730 to 1,020 ppm, while the SCM results presented here show an uncertainty range of 806 ppm to 1,008 ppm. The lower bound of this SCM uncertainty range is the mean minus one standard deviation for low carbon cycle feedback settings and the 19 AOGCM tunings, while the upper bound represents the mean plus one standard deviation for high carbon cycle settings. For comparison, the 90% confidence interval from Wigley and Raper (2001) is 770 to 1,090 ppm. The simple model CO2 concentration projections can be slightly higher than under the C4MIP because the SCM’s carbon cycle is driven by the full temperature changes in the A2 scenario, while the C4MIP values are driven by the component of A2 climate change due to CO2 alone.
Figure 10.26. Fossil CO2, CH4 and SO2 emissions for six illustrative SRES non-mitigation emission scenarios, their corresponding CO2, CH4 and N2O concentrations, radiative forcing and global mean temperature projections based on an SCM tuned to 19 AOGCMs. The dark shaded areas in the bottom temperature panel represent the mean ±1 standard deviation for the 19 model tunings. The lighter shaded areas depict the change in this uncertainty range, if carbon cycle feedbacks are assumed to be lower or higher than in the medium setting. Mean projections for mid-range carbon cycle assumptions for the six illustrative SRES scenarios are shown as thick coloured lines. Historical emissions (black lines) are shown for fossil and industrial CO2 (Marland et al., 2005), for SO2 (van Aardenne et al., 2001) and for CH4 (van Aardenne et al., 2001, adjusted to Olivier and Berdowski, 2001). Observed CO2, CH4 and N2O concentrations (black lines) are as presented in Chapter 6. Global mean temperature results from the SCM for anthropogenic and natural forcing compare favourably with 20th-century observations (black line) as shown in the lower left panel (Folland et al., 2001; Jones et al., 2001; Jones and Moberg, 2003).
The radiative forcing projections in Figure 10.26 combine anthropogenic and natural (solar and volcanic) forcing. The forcing plumes reflect primarily the sensitivity of the forcing to carbon cycle uncertainties. Results are based on a forcing of 3.71 W m–2 for a doubling of the atmospheric CO2 concentration. The anthropogenic forcing is based on Table 2.12 but uses a value of –0.8 W m–2 for the present-day indirect aerosol forcing. Solar forcing for the historical period is prescribed according to Lean et al. (1995) and volcanic forcing according to Ammann et al. (2003). The historical solar forcing series is extended into the future using its average over the most recent 22 years. The volcanic forcing is adjusted to have a zero mean over the past 100 years and the anomaly is assumed to be zero for the future. In the TAR, the anthropogenic forcing was used alone even though the projections started in 1765. There are several advantages of using both natural and anthropogenic forcing for the past. First, this was done by most of the AOGCMs the simple models are emulating. Second, it allows the simulations to be compared with observations. Third, the warming commitments accrued over the instrumental period are reflected in the projections. The disadvantage of including natural forcing is that the warming projections in 2100 are dependent to a few tenths of a degree on the necessary assumptions made about the natural forcing (Bertrand et al., 2002). These assumptions include how the natural forcing is projected into the future and whether to reference the volcanic forcing to a past reference period mean value. In addition, the choice of data set for both solar and volcanic forcing affects the results (see Section 2.7 for discussion about uncertainty in natural forcings).
The temperature projections for the six illustrative scenarios are shown in the bottom panel of Figure 10.26. Model results are shown as anomalies from the mean of observations (Folland et al., 2001; Jones et al., 2001; Jones and Moberg, 2003) over the 1980 to 2000 period and the corresponding observed temperature anomalies are shown for comparison. The inner (darker) plumes show the ±1 standard deviation uncertainty due to the 19 model tunings and the outer (lighter) plumes show results for the corresponding high and low carbon cycle settings. Note that the asymmetry in the carbon cycle uncertainty causes global mean temperature projections to be skewed towards higher warming.
Considering only the mean of the SCM results with mid-range carbon cycle settings, the projected global mean temperature rise above 1980 to 2000 levels for the lower-emission SRES scenario B1 is 2.0°C in 2100. For a higher-emission scenario, for example, the SRES A2 scenario, the global mean temperature is projected to rise by 3.9°C above 1980 to 2000 levels in 2100. This clear difference in projected mean warming highlights the importance of assessing different emission scenarios separately. As mentioned above, the ‘response uncertainty’ is defined as the range in projections for a particular emission scenario. For the A2 emission scenario, the temperature change projections with the SCM span a ±1 standard deviation range of about 1.8°C, from 3.0°C to 4.8°C above 1980 to 2000 levels in 2100. If carbon cycle feedbacks are considered to be low, the lower end of this range decreases only slightly and is unchanged to one decimal place. For the higher carbon cycle feedback settings, the upper bound of the ±1 standard deviation range increases to 5.2°C. For lower-emission scenarios, this uncertainty range is smaller. For example, the B1 scenario projections span a range of about 1.4°C, from 1.5°C to 2.9°C, including carbon cycle uncertainties. The corresponding results for the medium-emission scenario A1B are 2.3°C to 4.3°C, and for the higher-emission scenario A1FI, they are 3.4°C to 6.1°C. Note that these uncertainty ranges are not the minimum to maximum bounds of the projected warming across all SCM runs, which are higher, namely 2.7°C to 7.1°C for the A2 scenario and 1.3°C to 4.2°C for the B1 scenario (not shown).
The SCM results presented here are a sensitivity study with different model tunings and carbon cycle feedback parameters. Note that forcing uncertainties have not been assessed and that the AOGCM model results available for SCM tuning may not span the full range of possible climate response. For example, studies that constrain forecasts based on model fits to historic or present-day observations generally allow for a somewhat wider ‘response uncertainty’ (see Section 10.5.4). The concatenation of all such uncertainties would require a probabilistic approach because the extreme ranges have low probability. A synthesis of the uncertainty in global temperature increase by the year 2100 is provided in Section 10.5.4.6.