This chapter introduces the technical details that underlie the committee’s conclusions and recommendations. The role of the equilibrium climate sensitivity (ECS) in determining temperature changes is described. Several additional relevant climate metrics that reflect the state of the literature are discussed.
The first question in the committee’s charge is to consider the merits and challenges associated with a near-term revision of the distribution of the ECS. A broad perspective on the relationship between emissions (a key input to the physical climate/carbon cycle model in the social cost of carbon integrated assessment models [SCC-IAMs]) and global mean temperature (the output) is considered in this chapter. Four metrics are of particular importance to the discussion: ECS, transient climate response (TCR), transient climate response to emissions (TCRE), and the initial pulse-adjustment time (IPT); see Box 3-1. In comparison with other metrics used to summarize the relationship between emissions and temperature change, researchers have noted that the ECS is not necessarily the most relevant physical parameter over the nearer-term timeframes particularly important to determining the SCC (e.g., Otto et al., 2013b).
Modeling the effect of CO2 emissions on global mean surface temperature entails estimating the effect of emissions on atmospheric CO2 concentrations, the effect of CO2 concentrations on radiative forcing, and the effect of forcing on temperature. Although this path is complex, the result appears to be a simpler relationship between temperature and cumulative emissions than between temperature and forcing. As described below (“The Carbon Cycle and TCRE”), the relationship between cumulative CO2 emissions and global mean temperature change is approximately linear and can be summarized by a single parameter: the transient climate response to cumulative carbon emissions in the industrial era (TCRE). TCRE measures, on a time scale of decades, the ratio between CO2-induced warming and cumulative emissions, expressed in units of °C/Tt C, where 1 Tt C is 1 trillion tons of fossil carbon or 3.67 trillion tons of CO2. TCRE is, in turn, determined primarily by TCR (see Matthews et al., 2009; Gillett et al., 2013).
Calculating the SCC entails estimating a baseline temperature trajectory and the temperature response to a pulse of CO2. The multidecade-to-century warming caused by a pulse of CO2 can be approximated as the product of the TCRE and the total cumulative amount of carbon injected. The speed of this response, determined by the IPT, is also important for estimating the SCC; see discussion below (“Implications for Estimation of the SCC”).
In Chapter 4, the committee details the implications of the discussion in this chapter for calculation of the SCC. The importance of ECS, relative to TCR, depends on the time pattern of damages associated with a time pattern of global temperature change. The higher the fraction of the present discounted value of damages that occur in the first century after emissions, the more important is the TCR relative to the ECS, since the TCR is a much better predictor of climate
response on time scales of less than a century. In Chapter 4, the committee outlines tests that could be applied to the simple climate models used to generate the SCC to determine whether the central projections of these models agree with those of the class of Earth system models used by the Intergovernmental Panel on Climate Change (IPCC).14
The concepts of ECS and TCR arise, in their simplest form, from the conservation of energy. In equilibrium, the incoming solar radiation absorbed by Earth balances the outgoing longwave infrared radiation emitted by the planet to space. If either the absorbed solar radiation or the outgoing longwave radiation is perturbed from an equilibrium state, the heat content of the climate system will change at a rate set by the magnitude of the imbalance. The absorbed solar radiation is controlled by the amount of incoming solar radiation and by the Earth’s albedo, which is the fraction of the incoming solar radiation reflected away by the atmosphere or the surface. The amount of outgoing longwave radiation is set primarily by the planet’s radiative temperature—the temperature of the atmospheric level from which, on average, infrared radiation can be emitted through the “haze” of infrared-absorbing greenhouse gases and clouds to space. Because the radiative temperature increases as the climate system absorbs heat (thereby increasing outgoing longwave radiation) and declines as the climate system loses heat (thereby decreasing outgoing longwave radiation), the imbalance, and thus the rate of temperature change in response to a perturbation, declines over time until a new equilibrium is reached.
A climate forcing (measured in W/m2 [watts per square meter]) refers to a decrease in net outgoing energy, relative to some initial state in which the planet was in equilibrium, driven by an exogenous factor, such as a change in greenhouse gas or aerosol concentrations. The change in temperature caused by a forcing triggers climate feedbacks: additional changes in the planet’s albedo or emissivity that amplify or dampen the energy imbalance and thus cause additional changes in temperatures. Feedbacks involving greenhouse gases and clouds affect emissivity; those involving aerosols, clouds, and land surface characteristics affect albedo. For example, water vapor, which increases in concentration with temperature and thereby decreases emissivity, gives rise to one important amplifying feedback—sea ice—which decreases in surface area with temperature and thereby increases albedo, giving rise to another (amplifying) feedback.
To a good approximation, the equilibrium change in global mean temperature is proportional to the forcing applied. This magnitude is captured by ECS. However, the equilibrium response to a forcing may take centuries to be realized. Within the context of SCC estimates, it is therefore necessary to understand the transient response both to the range of human-caused forcings and to a pulse of CO2, the marginal impact of which the SCC estimates. One common metric of the transient response is the TCR, which is defined as the global mean surface temperature change at the time of CO2 doubling for a benchmark forcing scenario, specifically, an increase in CO2 concentrations at a rate of 1 percent per year: see Figure 3-2. Under such a scenario, the time of CO2 doubling occurs at year 70, and the TCR estimate is generally made by averaging global mean surface temperature over years 61-80. Just as ECS is a general measure of the equilibrium response to any indefinitely sustained radiative forcing, TCR is a general measure of the transient response to a gradually increasing radiative forcing. Because
the climate system does not instantaneously re-equilibrate in response to a forcing, TCR is always less than ECS.
One source of the difference between ECS and TCR can be observed in a simple “one-box” energy balance model, in which all the heat taken up by the climate system as a result of a forcing is distributed evenly through the climate system as a whole. In such a simple model, the rate of increase in global mean temperature is directly proportional to the rate of heat uptake, with the proportionality set by the heat capacity of the climate system. In response to an instantaneous change in forcing, the temperature of a one-box climate will evolve toward its equilibrium response following an exponential decay with a single timescale. The timescale is directly proportional to both the heat capacity and ECS (see, e.g., Hansen et al., 1985). TCR therefore increases with ECS at a substantially less-than-linear rate: if the TCR is 2°C with an ECS of 3°C, then with the same heat capacity and an ECS of 6°C, TCR will be just 2.8°C.
In contrast to this simple one-box model, full-complexity climate models exhibit two dominant timescales of temperature change in response to a forcing: a fast timescale, associated with the response of the atmosphere and ocean mixed layer (the surface ~100 meters of the ocean), and a slow timescale, associated with the response of the deep ocean. The atmosphere and the mixed layer respond on a timescale of years to a change in forcing, while the deep ocean takes decades to centuries to warm, which slows down the overall response (see, e.g., Hansen et al., 1984; Held et al., 2010). In the scenario used to measure TCR, the mixed layer is nearly fully equilibrated with the applied forcing at the time TCR is assessed, but the deep ocean can be far from equilibrium. These two timescales can be adequately represented in a “two-box” simple
The magnitude of ECS is uncertain due to a number of factors. First, the historical forcing, particularly the historical aerosol forcing, is uncertain (Myhre et al., 2013). Second, as noted, warming lags any radiative forcing, with the strong response implied by a high ECS that takes longer to realize than a weaker response associated with a low ECS. This lag makes it more challenging to distinguish values of ECS observationally. Third, the rate and magnitude of the heat flux from the mixed layer into the deep ocean are uncertain; accordingly, the same transient response can be produced either with a low ECS and faster ocean mixing, or a higher ECS and slower ocean mixing.
A fourth challenge has been identified in recent years: state-dependent feedbacks. Earth’s outgoing longwave radiation depends not only on the average radiative temperature, but also on the spatial pattern of temperature, which changes as the planet warms. Accordingly, the rate of energy loss to space also depends on how far the system is from equilibrium (Held et al., 2010). As one example, cloud feedbacks can exhibit state dependence that is represented in atmosphere-ocean global circulation models and Earth system models but not in the simple climate models that specify a fixed ECS value.15 State-dependent feedbacks can also be related to long-term changes in ocean circulations, land-surface conditions, ocean carbon uptake, and the cryosphere.
This state dependence gives rise to an effective climate sensitivity—not ECS, equilibrium climate sensitivity—that is constrained by observations of the recent energy budget constraint. Winton et al. (2010) found that, in 17 of the 22 global climate models participating in the Coupled Model Intercomparison Project Phase 3 (CMIP3),16 the effective climate sensitivity at the time of CO2 doubling was less than ECS. Estimates of ECS based on recent climate observations are actually estimates of effective climate sensitivity and may therefore significantly underestimate the true equilibrium response. Unfortunately, there are no clear observational constraints on the relationship between effective and equilibrium climate sensitivity, but this distinction does explain why different approaches to estimating ECS can provide very different ranges (depending on whether or not they assume, implicitly, a specific relationship between the two sensitivity parameters). Although paleoclimatic observations can provide additional constraints on ECS, they are hampered by uncertainties in past forcing and climate data.
Because of these four challenges and the associated uncertainties, the uncertainty in ECS is quite large, with a positively skewed tail of possible high values. A major source of this uncertainty can be seen from the simple treatment of Roe and Baker (2007), whose analysis gave rise to the form of the probability distribution for ECS currently used in the U.S. government’s SCC analysis; see Figure 3-3. In the absence of any climate feedbacks other than the “Planck feedback” (by which changes in surface temperature stabilize radiative temperature), ECS would be about 1.2°C (e.g., Hansen et al., 1981). However, other feedbacks come into play. Using f to indicate the total magnitude of these feedbacks on temperature change and ECS0 the value of ECS including only the Planck feedback gives . The different processes contributing to f add linearly. The positive skewness of the ECS distribution arises from those values of f that approach 1. The Roe and Baker (2007) distributional form for ECS arises simply by assuming
16CMIP provides a standard experimental protocol for IPCC-class global circulation models, and provides community-based support for climate model diagnosis, validation, intercomparison, documentation, and data access.
that f has a truncated normal distribution; the associated long right tail would also arise from many other symmetric distributions of f.
Because larger equilibrium responses caused by higher ECS take longer to realize, TCR is less skewed than ECS (Baker and Roe, 2009). Moreover, whereas ECS may take centuries of data to constrain, the transient response by definition plays out over timescales of less than a century; it can therefore be better constrained by observations (e.g., Gregory and Forster, 2008; Otto et al., 2013a). Box 3-2 describes how IPCC statements regarding the ECS and TCR have evolved from the Fourth to the Fifth Assessment Report (AR4 and AR5), as well as research since the AR5.
ECS and TCR by definition exclude feedbacks—such as those involving dust, vegetation, or land ice—that have not traditionally been represented in coupled climate models. If these other feedbacks are predominantly positive, for the timescales on which they are operative, measures such as ECS and TCR may understate the expected warming. Indeed, geological data suggest that omitted feedbacks may significantly amplify warming relative to that expected based on ECS alone (e.g., PALAEOSENS Project, 2012). Attempts to include relevant processes in Earth system models are a major area of active research. Some simple climate models also attempt to incorporate feedbacks traditionally absent from coupled climate models.
Because the experiments to assess ECS and TCR prescribe only CO2 concentrations, these metrics also exclude carbon cycle feedbacks. The next section highlights the important role of land and ocean carbon cycle feedbacks in giving rise to CO2 warming processes operating over millennia.
The discussion so far has focused on the response of global mean surface temperature to a particular level or time path of greenhouse gas concentrations in the atmosphere. To fully
understand the response of temperature to CO2 emissions, one must also understand how CO2 emissions translate into atmospheric concentrations, how atmospheric CO2 concentrations translate into forcing, and how forcing translates into temperature change.
For CO2, the relationship between concentrations and forcing is fairly straightforward. To a good approximation, the radiative forcing of CO2 is logarithmic in concentration (Arrhenius, 1896). This logarithmic relationship means that, for higher CO2 concentrations, further incremental increases in the CO2 concentration yields a diminishing increase in the CO2 forcing.
The relationship between emissions and concentrations is more complicated, as it involves the full carbon cycle, including some crucial feedbacks between concentrations, temperatures, and fluxes. When a ton of CO2 is emitted into the atmosphere, a small fraction, about 20 percent, is removed within the first 5 years by the land biosphere and by the ocean, so that about 80 percent is still airborne; see Figure 3-4. After 20 years, about 40 percent of the emitted ton has been thus taken up, and about 60 percent is still airborne; after 100 years, about 60 percent has been removed from the atmosphere and about 40 percent is still airborne. Over
the course of the following centuries, the oceans become the major repository of the added carbon.
There are two major bottlenecks in the ocean uptake of CO2. The first is across the air-sea interface: the CO2 partial pressure in the surface oceans, i.e., the pressure pushing CO2 back into the atmosphere, increases with carbon uptake and the accompanying decrease in pH. The second is below the mixed layer, where carbon is mixed into the deeper ocean on multicentennial timescales. Yet even on multicentennial timescales, the carbonate chemistry and the ocean volume dictate that oceans cannot absorb 100 percent of the added carbon, and about 20 percent will remain in the atmosphere after a millennium (Broecker et al., 1979). The ultimate carbon sink occurs through weathering reactions and sedimentation on the ocean floor, which takes place on time scales of hundreds of thousands of years (Archer et al., 2009; Ciais et al., 2013).
The effect of climate change on the carbon cycle gives rise to an amplifying feedback between atmospheric CO2 and temperature. Warming accelerates decomposition on land faster than CO2 fertilization increases the rate of photosynthesis, weakening the land-carbon sink (Friedlingstein et al., 2006). Warming also further stratifies the oceans, slowing the penetration of heat and carbon to the deep ocean. The decreasing pH and the warmer temperatures (decreasing solubility) also shift the equilibrium of the carbonic acid/bicarbonate buffer and reduce the ocean absorption of CO2 from the atmosphere (Archer and Brokin, 2008).
The weakening of the land and ocean carbon sinks as a result of warming increases the atmospheric residence time of CO2 (Jones et al., 2013), giving rise to a convex relationship between cumulative carbon emissions and atmospheric CO2 concentrations. When the convex relationship between emissions and concentrations is combined with the concave relationship between concentrations and forcing, the result is a coincidental cancellation that results in a nearly linear relationship between cumulative CO2 emissions and radiative forcing.
The global mean surface temperature also responds approximately linearly to a continually increasing effective radiative forcing (Flato et al., 2013). Hence, provided the forcing is increasing slowly relative to the response time of the ocean mixed layer (Held et al., 2010), there is a linear relationship between the forcing at any given time and the resulting warming at that time. (Note that this warming is generally not in equilibrium with the forcing.) When the nearly linear relationship between cumulative CO2 emissions and forcing is combined with the linear relationship between forcing and temperature, the result is a simple, nearly linear relationship between cumulative carbon emissions and the resulting warming (Goodwin et al., 2015).
Another cancellation, between the gradual decline of atmospheric CO2 and the slow approach of the ocean to thermal equilibrium, causes temperatures to remain nearly constant for centuries following a complete cessation of CO2 emissions (Matthews and Caldeira, 2008; Solomon et al., 2009). This cancellation arises because both of these processes operate on similar timescales set by the mixing of carbon and heat into the deep ocean; see Figure 3-5.
A series of papers published in the late 2000s (e.g., Allen et al., 2009; Matthews et al., 2009; Meinshausen et al., 2009; Zickfeld et al., 2009) pointed out that, as a consequence of the longevity of the warming associated with CO2 emissions, temperatures in any given year were largely determined by cumulative CO2 emissions up to that time. Gillett et al. (2013) updated the carbon budget estimates in these early papers to take into account updated estimates of the strength of anthropogenic forcing and the evolution of global temperatures since 2000; see Figure 3-6. In the AR5, Collins et al. (2013, p. 1033) concluded that “the principal driver of long-term warming is total emissions of CO2 and the two quantities are approximately linearly related.”
The approximately linear relationship between cumulative CO2 emissions and the warming it causes simplifies the estimation of the climate response to CO2 emissions. It means
that the global mean temperature change induced by CO2 can be largely summarized by a single parameter, the TCRE.
Herrington and Zickfeld (2014) show that the TCRE is a robust measure of the climate response, regardless of the timescale of injection for realistic emission scenarios, although the response takes up to a century to stabilize for very large instantaneous CO2 pulses. The AR5 (Collins et al., 2013) suggested the TCRE was approximately constant for cumulative injections up to 2 teratons of carbon (Tt C), although Herrington and Zickfeld (2014), consistent with Allen et al. (2009), suggest it declines slightly for cumulative emissions in excess of 1.5 Tt C. Gillett et al. (2013), with a broader range of models, find evidence for a smaller decline, with any departure from linearity being small relative to uncertainty in the response (Figure 3-6).
TCRE is determined by two quantities. First, TCRE depends on the cumulative airborne fraction at the time of CO2 doubling (or, approximately equivalently, after a cumulative release of 1 Tt C in the form of CO2) following a gradual increase in concentrations over a multidecade period. This fraction is between 0.47 and 0.67 in current climate models (Gillett et al., 2013). Second, TCRE depends on the warming at the time of CO2 doubling following a gradual increase in concentrations, or the TCR. Most of the uncertainty in TCRE arises from the uncertainty in TCR, leading to an AR5 consensus “likely” range, accounting for both model and observational lines of evidence, of 0.8°C-2.5°C/Tt C, which is equivalent to 0.2°C-0.7°C/1000 Gt CO2 (Collins et al., 2013).
While uncertainty in ECS gives rise to uncertainty in the additional warming that would occur over centuries if atmospheric CO2 concentrations were stabilized, current comprehensive Earth system models indicate that uncertainty in ECS is largely irrelevant to TCRE and hence to the temperature response to a pulse injection of CO2. This irrelevancy occurs because, after a
cessation of emissions, atmospheric CO2 concentrations do not stabilize, but rather fall just fast enough that the “recalcitrant” warming reflected by ECS (Held et al., 2010) never materializes (Matthews and Caldeira, 2008; Solomon et al., 2009).
The constancy of the TCRE indicates that the multidecade-to-century-timescale climate response to any CO2 injection can be accurately approximated by a constant temperature increase set by the total cumulative amount of carbon injected and the TCRE. A key remaining aspect of the response that is relevant to the SCC is the form and speed of the adjustment immediately following a pulse injection of carbon. The most comprehensive study to date to address this question was the multimodel comparison of Joos et al. (2013). They examined the impact of a 100 Gt C pulse injection of CO2, relative to a baseline scenario in which CO2 concentrations were held constant at 389 ppm following a historical transition to that point in a range of simple climate models and Earth system models of both intermediate and full complexity.
Results are shown in Figure 3-7, with solid lines corresponding to full-complexity models, dashed lines to intermediate-complexity models, and dotted lines to simple models. The full-complexity models display large fluctuations that can be understood entirely as random internal variability, given the small size of the temperature response even to a pulse of this magnitude (comparable to about a decade of CO2 emissions at 2015 levels). Strikingly, all models, including the most complex, adjust relatively rapidly, with temperatures rising to about 0.2°C within 10 to 20 years of the pulse and then remaining constant for the remainder of a century. A slight decline is observed over the millennium (right panel).
In modeling the carbon cycle response to this pulse injection, Joos et al. (2013) find a very rapid IPT of only a few years and very slow subsequent adjustments on multidecade and multicentury timescales. The short IPT in Figure 3-7 is primarily set by the ocean mixed-layer thermal response time, which is known, on physical grounds, to be of the order of a decade or less (Held et al., 2010). The adjustment to a pulse injection of CO2 can thus be adequately characterized by an initial adjustment within a timeframe of 4 years to a decade, followed by stable temperatures for a century and slow decline thereafter.
Ricke and Caldeira (2014) use a simple climate model that combines a carbon cycle model fit to the results of Joos et al. (2013) with a simple model of the thermal response (similar to that of Held et al. ) to obtain the temperature response to a pulse CO2 emission. They find that temperature peaks between 7 and 31 years after the emissions (90% probability range, median of 10 years). While their simple climate model suggests a modest decline in temperatures after the peak, this finding arises from the lack of explicit climate carbon cycle feedback in their composite model, and it is not evident in the more complex models on which their composite model is based (Joos et al., 2013). In order for a simple climate model to generate the near linearity of warming in cumulative carbon emissions, as well as the longevity of the associated warming, it is necessary to use either a carbon cycle model that includes the effects of pH and warming on CO2 solubility (e.g., Glotter et al., 2014), the impact of warming on land carbon sinks (e.g., Friedlingstein et al., 2006; Allen et al., 2009), or a direct approximation of the linearity in cumulative carbon emissions (e.g., Kopp and Mignone, 2013).
On the basis of the current evidence, the committee concludes that the likely approximation for characterizing the response over annual-to-centennial timescales of the climate system to a pulse of emissions of CO2 is a simple rapid adjustment (4 years to one decade) to the level of warming indicated by the TCRE, with modest decline for at least a
To estimate the SCC, it is necessary to project both the physical climate changes associated with a baseline emissions trajectory and the effect of a small, additional pulse of CO2 emitted on top of that baseline trajectory.18
While the TCRE and IPT are relevant for capturing the response to cumulative or pulse emissions of CO2, other measures are relevant for computing a baseline climate, which may be influenced by CO2 emissions high enough (greater than approximately 1.5 Tt C) that the TCRE is not constant and is also affected by non-CO2 forcers. The relative importance of TCR and ECS in characterizing the SCC depends on the relative proportion of net present value damages that occur in roughly the first century of emissions. By construction, TCR is a much better predictor than ECS of the climate response on timescales of less than a century.19 As a result, Otto et al. (2013b) found that in their simple model for estimating the SCC, for a moderate emissions trajectory20 and a quadratic damage function, reducing uncertainty in TCR leads to a greater reduction in SCC uncertainty than reducing uncertainty in ECS, provided that the discount rate is at least about 1 percent higher than the growth rate of consumption; see Figure 3-8. For highly convex damage functions and discount rates sufficiently close to the consumption growth rate, Otto et al. (2013b) found that learning about ECS leads to a greater reduction in SCC uncertainty than learning about TCR.
Factors that increase the fraction of the SCC due to damages after the first century, and thus increase the importance of ECS in comparison with TCR, include an increase in baseline temperatures as well as economic factors. In climate damage functions, such as those used in the SCC-IAMs, faster economic growth for a given discount rate or a lower discount rate for given economic growth will both tend to increase the importance of the more distant future and thus the ECS. In this context, it is worth noting that the IWG analysis holds the discount rate constant but assumes a decrease in growth rates after 2100, thereby increasing the importance of TCR over ECS relative to a constant growth-rate scenario or one in which the discount rate declines when the growth rate declines. In the 21st century, the average economic growth rate in the IWG scenarios ranges between 2.0 and 2.4 percent per year, while over 2100-2300 it ranges between
17Joos et al. (2013) found that the magnitude of the temperature response to a pulse injection (0.20 ± 0.12°C/100 Gt C) is comparable to—though slightly higher than—the AR5 range for TCRE, although their analysis was based on a subset of the models used by the AR5 for its statement on TCRE. In single-model studies, Herrington and Zickfeld (2014) and Zickfeld and Herrington (2015) found that TCRE falls with both the speed and magnitude of a pulse injection, while Krasting et al. (2014) found that TCRE is larger for both small (~2 Gt C/yr) and large (~20 Gt C/yr) rates of emissions than for current rates of emission (~10 Gt C/yr).
18This requirement can be seen in a simple, typical model: If damages are equal to economic output times a power function of temperature, D(T) = aTb, then the change in damages associated with an emission pulse that shifts temperature from T to T + ΔT at time t is proportional to T(t)b−1 ΔT(t). Thus, the physical climate model underlying the SCC calculation must provide reasonable projections for both T(t) and ΔT(t); that is, both the baseline temperature response and the long-term temperature changes due to an emissions pulse. The economic valuation also depends on the relative sizes of the growth rate of consumption and the rate at which damages are discounted.
19This finding can be seen from the 1 percent/year CO2 concentration growth scenario used to define TCR, in which ECS provides no additional information about the temperature response until after year 70.
0.5 and 0.9 percent per year.21 In the context of the Otto et al. (2013b) results, the low growth rates after 2100 suggest that TCR will be a more important determinant than ECS of the SCC calculated using the IWG methodology, even at the lowest discount rate used (2.5% per year).
21The IWG used aggregate output (GDP) as a socioeconomic input, not macroeconomic aggregate consumption, which is a component of aggregate output. For this purpose, it is reasonable to think of consumption growth as proportional to output growth.
TCRE is the crucial parameter determining the contribution of the physical climate system response to the SCC, since it determines the magnitude of multidecade-to-century timescale warming resulting from a pulse injection of CO2. TCRE is primarily determined by TCR, not ECS. Revisions to ECS are therefore relevant to SCC estimation, principally through their possible implications for baseline warming after a century or more. TCR and IPT determine temperature changes over shorter time periods, including the response to a small pulse emission of CO2. Hence, the revision of the “likely” range of ECS from 2.0°C to 4.5°C in the AR4 to 1.5°C to 4.5°C in the AR5 should have a minimal impact on estimates of the SCC.