Energy and Climate Change CommitteeSupplementary written evidence submitted by the Met Office

Supplementary Evidence on Carbon Uptake in Future Climate Scenarios

During the evidence session of 12 June 2013, Dr Jason Lowe offered to provide a more detailed explanation of what may happen to the future atmospheric CO2 burden following a peak in anthropogenic emissions and subsequent rapid decline.


Climate models coupled to carbon cycle models are used to study the uptake of carbon by the atmosphere, ocean and land surface for potential response to future emissions scenarios. They include both climate and CO2 concentration feedbacks on to the carbon cycle.

A range of different models show that when carbon emissions are reduced rapidly following a peak in the near future, it is plausible that atmospheric CO2 concentrations will peak before the anthropogenic emissions have reached zero.

The latest complex earth system model results from CMIP5 reinforce the findings from a simpler (traceable) climate model set up and used in 2008, which also showed this behaviour.

However, the Met Office also considers it prudent to repeat the assessment of budgets as new understanding of how the earth system works becomes available and can be quantified.

Idealised Case of Zeroing Emissions

1. Climate model experiments using idealised emission or concentration pathways are useful to better understand complex changes in the earth system, and to bound a range of future responses before more credible and policy relevant pathways are considered. Several studies have performed an experiment in which anthropogenic emissions of carbon dioxide are instantaneously set to zero—the so called “zero emission commitment” experiment. This provides useful understanding of the manner in which carbon sinks evolve, and global temperatures respond, following a rapid reduction in anthropogenic emissions. The results from three different models are reported in Matthews and Weaver (2010), who conclude that “in response to an abrupt elimination of carbon dioxide emissions, global temperatures either remain approximately constant, or cool slightly as natural carbon sinks gradually draw anthropogenic carbon out of the atmosphere”.

2. In the study of Lowe et al. (2009) the evolution of simulated carbon sinks were examined in detail. Following cessation of emission, the atmospheric CO2 concentration declined at a rate of between around 20 and 75ppm%ury. The relative contributions to the reduction of atmospheric CO2 from land and ocean sinks, and from different parts of the land sinks, varied depending on the timing of the emission reduction and the level of atmospheric CO2 when emissions were reduced. These studies highlight that carbon sinks can continue to take up carbon at significant rates after the magnitude of the emissions has been reduced, and that the uptake is affected by the state of the carbon cycle components prior to the cessation of emissions. In turn, this depends on the pathway of earlier emissions.

3. This is not surprising and is entirely consistent with our understanding of fundamental carbon cycle processes which can continue to respond to past changes or emissions for many years and are not purely determined by emissions from a single year. Rapid reduction in emissions can lead to periods of time where significant natural sinks exceed emissions and lead to CO2 reductions even while anthropogenic emissions are still positive.

The Policy Relevant Scenarios Used in the Recent IPCC Assessments

4. Numerous studies have produced scenarios or emissions pathways that peak carbon dioxide emissions early in the 21st century and have rapid declines following the peak. These studies then use a variety of climate models and carbon cycle models to understand the evolution of atmospheric CO2 and global temperatures. We present these to demonstrate the credibility of the situation where atmospheric CO2 concentration is no longer rising, or is even declining slowly, despite non-zero anthropogenic emissions.

5. Plattner et al. (2008) examined the climate and carbon cycle response for eight models and a number of stabilisation scenarios. One of those models is the BERN2.5CC model, which clearly simulates a number of cases where CO2 concentrations in the atmosphere level off or even peak but anthropogenic emissions remain above zero (see their Figure 11). Furthermore the authors state that “allowable emissions for stabilization at 450 ppm (SP450 and OSP450) remain above zero throughout the simulation in all of the EMICs”, where EMICs refers to the models considered.

6. The most recent climate model inter-comparison exercises have tended to use the, so called, Representative Concentration Profiles (RCPs; Moss et al., 2010) as input. These provide another opportunity to examine the response of the climate-system when atmospheric CO2 is declining slowly following a peak. We focus on RCP2.6, which is the most similar case to the global pathway considered by the Committee on Climate Change that gives approximately a 50% chance of limiting global average near-surface warming to 2°C. RCP2.6 was produced using the IMAGE model (eg van Vuuren et al. 2011) and shows a peak CO2 concentration occurring in 2050 followed by a decline. The emissions of CO2 in 2050 and 2060 (a decade after CO2 concentrations started to decline) were 3.4 GtC/yr and 2GtC/yr respectively, compared to a 1990 value of 7.5GtC/yr.

7. Using the CMIP5 ensemble of more complex earth system models, a recent study by Jones et al. (2013) examined the relationship between atmospheric CO2 concentrations, carbon sinks and implied anthropogenic emissions in more detail. The study found that when models were constrained to follow the concentrations of the RCP2.6 pathway the implied anthropogenic CO2 emissions in 2050 (the time of atmospheric CO2 concentration peak) were still positive and had reduced from 1990 levels by between 14 to 96% depending on the model. A decade later the emissions were also clearly still above zero in the majority of the models. The requirement for negative emissions to achieve this CO2 decline is mixed—for some models emissions stayed positive for the whole of the 21st century.



8. In terms of carbon storage, the land carbon store increased throughout the 21st century in RCP2.6 in the majority of models, with the amount of increase varying with model. Uptake by the ocean also continued during the 21st century simulations in all of the models, but with a decreasing rate during the 21st century. The cumulative airborne fractions evaluated for the model ensemble between 2006 and 2100 were less than the values estimated for the 1990s for the same models for RCP2.6.

The Simple Model Calculations used for the First Carbon Budget Calculations

9. The modelling philosophy applied for the global scenarios used to inform the first carbon budget calculations is well documented1 and the model outputs are available on the website of the Committee on Climate Change. The ability of the model to credibly emulate the C4MIP simulations is included in the documentation. When this modelling approach is applied to the 2016R4L pathway, which peaks CO2 emissions in 2016 and has a long-term decline rate of 4% per year, a range of atmospheric CO2 concentrations can be simulated depending on the choice of key model parameters within the chosen uncertainty ranges. The median simulated atmospheric CO2 concentration shows a peak in the mid-21st century (peaking at around 445ppm in 2050) followed by a small decline (of around 15ppm over 50 years) later in the 21st century during a period when the anthropogenic emissions have declined significantly, but are still above zero. This is consistent with the behaviour of more recent and complex models reported above. It is important to note that during this (2050 to 2100) period the median temperature response does not show a decline, with the reduction in forcing being buffered by the top of the atmosphere energy imbalance built up during the earlier part of the experiment when atmospheric CO2 concentrations and forcing were increasing. It is also important to note that the precise behaviour of the climate system and the carbon cycle remain uncertain. Narrowing this uncertainty is a key challenge of climate science.

Supplementary Evidence on Climate Sensitivity


Equilibrium climate sensitivity (ECS) and Transient Climate Response (TCR) are two measures of the sensitivity of the climate to changes in the atmospheric concentration of greenhouse gases, such as CO2.

Precise values of ECS and TCR are not known but ranges of certainty can be estimated.

There are many different ways to estimate ranges of ECS and TCR, each with different advantages and disadvantages.

The HadGEM2-ES model is at the higher end of complex climate model estimates of ECS and TCR but within the range estimates from a number of different approaches.

Climate policy advice from the Met Office draws on a range of models and on observations. It is tailored to address relevant issues.


10. The Met Office Hadley Centre has used a number of climate models over the period 2008 to present, in order to provide evidence to inform policy and decision making. These include the HadCM3, HadGEM1 and HadGEM2-ES models. Additionally, traceable simple climate models have been used for some purposes. Equilibrium climate sensitivity (ECS) and Transient Climate Response (TCR) are two measures of the sensitivity of the climate to changes in the atmospheric concentration of greenhouse gases, such as CO2. The ECS and TCR of several complex Met Office GCMs are:













11. The simple climate model used to provide information for the original carbon budgets in 2008 used a distribution of ECS based on Murphy et al. (2004) which had a 5th to 95th percentile range of 2.4 to 5.4°C. This distribution was considered in the IPCC 4th assessment. The overall “likely”2 range presented by the IPCC 4th assessment for ECS was 2 to 4.5°C, with a best estimate of 3°C and a note that values substantially higher than 4.5°C cannot be excluded. TCR was concluded to be “very likely” to be greater than 1°C and “very unlikely” to be greater than 3°C.

12. Recent research shows that we still can not yet estimate a single precise value of ECS, but improved uncertainty ranges can be estimated using a variety of techniques. These techniques include: using the recent measurements of temperature, ocean heat uptake and estimated radiative forcing, using palaeo climate measurements, using complex models, and using a combination of models and observations. No single method can be identified as being the best approach and there are pros and cons for each approach. Within a particular class of approach, such as using the recent measurements of temperature, ocean heat uptake and estimated radiative forcing, there are still methodological issues of debate, such as the type of assumed priors3.

13. There have been a number of recently published estimates of ECS and TCR based on measurements of temperature, ocean heat uptake and estimated radiative forcing. These typically focus on the energy balance of the earth. The study by Otto et al. (2013) presented a number of estimated ranges based on considering various recent periods. When only the 2000 to 2009 period was used to represent near present day conditions the 5th to 95th percentile ranges of ECS and TCR were 1.2 to 3.9°C and 0.9 to 2.0°C, respectively. However, the article highlights that whilst there are advantages of using the most recent period because it is somewhat better observed than earlier periods, there are limitations to using a single decade too. The authors state that “caution is required in interpreting any short period, especially a recent one for which details of forcing and energy storage inventories are still relatively unsettled: both could make significant changes to the energy budget. The estimates of the effective radiative forcing by aerosols in particular vary strongly between model-based studies and satellite data. The satellite data are still subject to biases and provide only relatively weak constraints”. The numerical values in the article also clearly demonstrate the variability in ECS and TCR estimates when using a single decade. When the longer 1970 to 2009 was used (which includes data from the most recent decade alongside earlier data) in the calculation, their 5th to 95th percentile estimates for ECS and TCR were 0.9 to 5.0°C and 0.7 to 2.5°C respectively.

14. Forster et al., 2013 calculates that for the CMIP5 set of models (the recent models used in the forthcoming IPCC 5th assessment) there is a range of ECS from 2.1 to 4.7°C. TCR varies in the range 1.1 to 2.5°C. A recent comprehensive estimate using the palaeo climate approach (Palaeosense, 2013) estimates a likely range for ECS to be 2.2 to 4.8°C (noting that this covers a probability range of 68%). A recent estimate by the MOHC (Harris et al., 2012) using a combination of climate model results and observational constraints together gives a 5th to 95th percentile range of 2.4 to 4.3°C.

15. Thus, we conclude the HadGEM2-ES model is at the upper end of the range from current models for both ECS and TCR. However, we also conclude that it falls inside the ECS and TCR 5th to 95th percentile ranges estimated by a number of different approaches—including from a method that includes the recent observations. The climate sensitivity uncertainty distribution used by the Met Office in earlier carbon budget analyses within a simple climate model framework also has a significant overlap with many newer estimates.

16. Finally, we highlight that when considering the utility and skill of climate models it is prudent to consider more than their global average temperature response to atmospheric CO2 concentration changes. The spatial behaviour of complex climate models, such as HadGEM2-ES, are rigorously evaluated against observations for numerous variables. Natural variability and long-term response to external forcings of different types are considered. The HadGEM2-ES model provides one of the most comprehensive platforms for studying a wide range of earth system processes. We have also combined HadGEM2-ES with observational constraints to estimate ranges of future climate change for particular pathways of future greenhouse gas concentration increase.


Jones et al. J Climate, 2013.

Lowe et al, Env Res Lett, 2009.

Matthews and Weaver, Nature Geoscience 3, doi:10.1038/ngeo813, 2010.

Moss et al., Nature, 2010.

Plattner et al, J Climate, 21, 2008 .

Van Vuuren et al. Climatic Change, 109, 2011.

Elfron, Science, 380, 2013.

Forster et al., JGR- atmos, 2013.

Harris et al., J Clim, 2013.

Murphy et al, Nature, 2004.

Otto et al., Nature GeoSci, 2013.

Palaeosens project members, Nature, 2013.

26 June 2013


2 IPCC 4th assessment uses “likely” to mean >66% probability, “very likely” to be >90% probability and “very unlikely” to be <10% probability.

3 Bayesian approaches can be used to refine prior estimates of climate sensitivity but typically contain a subjective element. There is continued debate about the appropriate choice of prior information, such as whether it should be informative or uninformative. See for instance Efron, 2013.

Prepared 3rd October 2013