Supplementary Information to “Copenhagen Accord Pledges imply higher costs for staying below 2°C warming.“

  1. Model description

The FAIR–SiMCaP model is a combination of the abatement costs model of the FAIR model and the SiMCaP model (den Elzen et al., 2007). The FAIR cost model distributes the difference between baseline and a global emission pathway following a least-cost approach using regional Marginal Abatement Costs (MAC) curves for the different emissions sources (den Elzen et al., 2007). The model itself is part of the IMAGE[1] 2.4 integrated assessment framework (van Vuuren et al, 2007a). The IMAGE framework consists of several consistently integrated models that describe land use and the energy system including the associated emissions. The IMAGE framework provides information on baseline emissions and mitigation costs to the FAIR–SiMCaP model. The latter is done by creating time-, baseline- and regional-specific marginal abatement cost (MAC) curves for CO2 and fluorinated gasses using the TIMER[2] energy model (van Vuuren, 2007b) and MAC curves for the other non-CO2 greenhouse gases following the method of Lucas et al. (2007).

To generate the MAC curves, we interpolate here between different MAC sets in order to approximate the likely dynamic response.A MAC curve for energy and industry-related CO2 emissions is determined with TIMER by imposing a carbon tax and recording the induced reduction of CO2 emissions.Among the set of pricepaths used in TIMER to derive the MACs there are pricepaths included that start with close to zero prices, after which prices quickly increase. The full set consists of a linear, cubic, or cubic root shape, starting in 2010 and ending in 2020,2030,2040,..,2100, and with price levels of 0, 10, 20, …1000 $/tC at the end year. These tax profiles cover a large range of options of future climate policy timing, including a path that assumes very little mitigation action until 2020 or later and then attempting strong reductions.

On the basis of the MACs, FAIR–SiMCaP generates global emission pathways[3] by minimizing cumulative discounted mitigation costs[4] (here a discount rate of 5% is used) over the model simulation period (2010-2100 starting with Kyoto targets[5] or 2020-2100 starting from Copenhagen Accord pledges), subject to a radiative forcing target in 2100. At each moment in time, the model allows for full global substitution among different gases and sources. In the calculations, there are no intermediate restrictions with respect to emissions, radiative forcing or the probability of exceeding a temperature target. This implies that the model allows for an overshoot of the radiative forcing target (in the period before 2100).

In a subsequent step, we use the MAGICC 6 model 600 times using probabilistic parameter settings (Meinshausen et al. 2009) for each emission scenario, based on observational constraints on the model’s parameters and past climate simulations.Theassessment of each pathway’s probability to stay below 2°C is based on a wide range of net radiative forcing levels by 2100.

  1. Important model assumptions

The TIMER model explicitly models capital stock and its replacement. It does not include the option ofearly retirement of capital. The lifetime of installations and speed of technology dispersion therefore limits the maximum speed of emission reductions of the energy system and the inertia to a change in reduction speed. The maximum rate of reduction and rates of change over the entire 2010-2100 time period can be estimated by implementing a series of extreme price paths and recording reductions of emissions against a (zero price) baseline scenario.This leads to a maximum reduction rate of 3.5% without and 4.5% with BECCS.In the model these rates function as a maximum and are not prescribed for any of the scenarios. Den Elzen et al (2010) further discuss limits to reduction rates in our and other frameworks.Combining the MAC curves (that inherently do no contain information on limits to reduction speed and its inertia) and the deduced limits to reduction speed and inertia, the dynamics of the energy system are well covered by the model that calculates costs and emissions pathways.

For non-CO2 gasses, the following feasible annual reduction rates compared to baseline were used (based on expert opinion, estimating average lifetime, per source per gas):

CH4from coal production / 4.5%
CH4from oil production / 5.0%
CH4from gas production / 2.5%
CH4from landfills / 4.5%
CH4from sewage/waste water / 4.0%
CH4from rice cultivation / 2.7%
CH4from animals/enteric fermentation / 5.0%
CH4from animal waste / 2.5%
N2O from transport / 4.3%
N2O from adipic acid production / 9.8%
N2O from nitric acid production / 9.0%
N2O from fertilizers / 3.5%
N2O from animal waste / 1.8%
N2O from domestic sewage / 1.0%
HFC / 4.5%
PFC / 4.0%
SF6 / 4.0%

If the above reduction rates are reached, the MAC curve is cut off at that point so that no additional reductions are possible for the specific gas in that sector.

International aviation emissions are explicitly modeled in the TIMER energy model as part of the transport model. Marine bunker emissions are included as a separate sector. The emissions for these sectors respond to a carbon price (implicitly in the case of marine transport), assuming improvements in efficiency and design.

  1. Scenario Design
  2. Baseline design

The baseline applied in this analysis is based on the IEA World Energy Outlook (IEA, 2007) and described more extensively in van Vuuren et al. (2010). The economic growth and emissions in this scenario were updated to account for the economic crisis of 2008/2009 (den Elzen et al., 2011) on the basis of IMF economic data of July 2009 IMF (2009). On average, this led to a negative adjustment for the 2009 GDP growth rate for each world region of 3 to 5 percentage-points, a somewhat smaller impact for 2010, and a return to the original growth path after this period. The economic crisis resulted in a decrease in baseline greenhouse gas emissions without climate policy of about 10% by 2010, and 8% by 2020 (some 4 GtCO2eq), compared to baseline emissions without the crisis. Finally, the baseline does not include the impact of current climate policies such as Kyoto targets or other domestic policies. Not including such policies in the baseline means that implementation of Kyoto targets induces a carbon price.

The model framework is calibrated to 1995 and 2000 historical data (predominantly from UNFCCC, but from many others as well). The 2010 emission levels are therefore projections based on modeling and scenario assumptions. The current estimates of emissions can therefore easily deviate from very recent global emission estimates (such as Olivier et al (2011) or Montzka(2011)). Global emission level in 2010 in our model are 44 GtCO2eq, while estimated emissions by Montzka are as high as 49 GtCO2eq[6] in 2009. Van Vuuren et al (2010) discuss causes and implications of incongruence of emissions from long term scenarios and short term emission inventories. Concluding, while there is an obvious preference for close accordance, long term models do not capture all processes that determine short term drivers of emission levels. But that this does not make them irrelevant for long term scenario exercises.

  1. Scenario on land use emissions

Land use emission presented in the paper are net emissions including re-growth of natural forests and timber pools. Net CO2 emissions from land use and land use change are projected to fall, to nearly zero by 2075 and become negative (due to re-growth of natural forest on abandoned agricultural lands) afterwards. This is caused by continued technology improvement and declining population growth. In addition we assume for all scenarios, that CO2 emissions from land use and land-use change decline linearly, as a fraction of baseline,as a result of mitigation actions proposed by major deforesting countries by 2020 to near zero by 2050.

For 2010, we assume for all mitigation scenarios that land use emissions are equal to baseline levels, by 2020 around 75% of baseline levels and by 2030 50% of baseline levels.The reductionsby 2020 result from the implementation of national plans proposed at the 16th Conference of the Parties inCopenhagen by Brazil, Mexico and Indonesia(three countries with high deforestation rates and emissions) combined with other best available data as described inden Elzen etal.(2010). Other goals, like the EU goal of halting forest loss by 2030[7], were taken into consideration as well for the assumptions on land use emission reductions for the scenarios. We assume zero land use CO2 emissions by 2050, in line with the above reduction targets, recent political willingness in creating a framework to include avoided deforestation in the mitigation portfolio part of the Cancun agreements,and, finally, the declining trends in the land use CO2baseline emissions.

These reduction percentages would be attainable according to the MAC curves for the G4M model, as presented in Kindermann et al. (2008). However, it is important to emphasize the uncertainty in CO2 emissions from the land use sector and the potential for reductions, also in relation to current policies. Finally, while costs for mitigation of land-use CO2 emissions can be a relevant fraction of annual costs around 2020 (e.g. Kindermann et al., 2008), the fraction of cumulative (from 2010 to 2100) discounted costs will be minimal. Costs for reductions in land use CO2 emissions are therefore not included in the presented cost calculations.

Estimates of current emissions from land use and landuse change are already very uncertain (Friedlingstein et al, 2011), with uncertainty levels in the order of several GtCO2. Future estimates are therefore also very likely to be uncertain. This can have a significant impact on timing of emission reductions in 2°C emission pathways.

Figure 1. CO2 emission from land use and land use changesas used in this study, and the original baseline projection calculated by IMAGE model.

Finally, in our modeling framework the production of biofuels does not lead to deforestation for bio-energy production, as bio-energy production is only allowed on natural savannas and abandoned agricultural lands. There is a considerable contribution of agro residues suitable for energy production as well. These assumptions result in similar land use related emissions with and without BECCS, and therefore, we assume the same land use emissions for all scenarios.

The potential for bio-energy production is the same for all scenarios, based on the availability of bio-based residuals and available land for crop production after satisfying demand for food and excluding natural forests. Crops and residuals have a fifty-fifty contribution to total potential. The methodology for determining potentials of bio-energyis more extensively described in van Vuuren et al (2009). The primary difference among scenarios including and excluding BECCS is the end use of these energy sources. Without BECCS,bio-energy is transformed into biofuels and consumed in the transport sector. With BECCS, energy crops and bio residuals will be consumed in the power sector. The decarbonization of the transport sector would then be realized with an earlier and more extensive introduction of hydrogen based transport.

  1. Evaluating pledges

In this paper we use a country by country interpretation of pledges as presented by den Elzen et al. (2011). However, total global emissions differ from those in this paper. This is due to the applied methodology of harmonizing emissions and the assumed land use CO2emissions. The paper by den Elzen focuses specifically on 2020 emissions and harmonizes emission levels based on historic data.Rogelj et al. (2011) discuss a range of issues related to harmonization including different methods and reasons for harmonization, but also limitation, including for instance the differences in historical databases.The method that is usually applied is the one developed by Van Vuuren et al (2008) - and was also used for the Representative Concentration Pathways: emissions get harmonized to reach a historical emission level in the base year and next the offset factor slowly converges to 1 (Meinshausen et al , 2011) .

The relevant point is whether harmonization in this case would have influenced the results of the study. As Rogelj et al (2011) show, if the purpose of the paper was to estimate as accurately as possible the emission level resulting from the Copenhagen pledges, harmonization would be a relevant approach. However, the central question of the paper is the impact on costs and climate of postponed 2°C emission reduction pathways. The non-harmonized 2020 emission levels are relevant proxies for the shape of these postponed emission reduction pathways and the results are not sensitive to whether or not harmonized data is used. Harmonization would increase 2020 emission levels by around 1GtCO2eq, but this would not change the conclusions that i) Copenhagen pledges are a move away from least cost pathways, ii) delay of reductions seems possible to some extent without increasing the risks of exceeding 2°C - possibly even past the proposed levels of 40-46 GtCO2eq, as proposed by UNEP(2010), and iii) current policy progress scenarios clearly show there is a limit to this delay.

The authors recognize the value of harmonization (naturally, if this impacts conclusions), particularly for comparison of model results. Given that for the results first of all the relative positions of the scenarios are relevant and secondly, harmonization would only to a limited degree influence that long-term concentration levels (see Meinshausen et al., 2011) here we prefer to use the simplest method and stay to use the original model data.

The relative reductions against base year (1990 or 2005 for Annex I countries, and 2020 for non-Annex I countries), as shown in table 1 of den Elzen et al. (2011), have been applied to the non-harmonized data, resulting in a slightly different global emission estimate for 2020(around 1 GtCO2eq). Next to that the land use CO2 emissions in the current study are based on the IMAGE baseline, whereas in den Elzen et al. (2011) these are based on the reference emissions provided by the countries, or if not available, based on IIASA model projections of Kindermann et al. (2008), which are both higher in general.

  1. Feasibility of other related scenario designs

Not al permutations of the assumptions described in the scenario setup (2020 emission level, availability of BECCS, long term climate target, and assuming for all scenarios a reduction of land use emissions) lead to scenarios that are feasible[8] within this model framework. The least ambitious interpretation of the Copenhagen pledges in combination with a 2.9 W/m2 climate target is an example of this. We have looked into other combinations that deserve mentioning:

-A climate target such as 2.6 W/m2 without BECCS is not feasible for the Optimal or Copenhagen scenarios.

-None of the Copenhagen scenarios would be feasible without any reduction of land use CO2 emissions.

-A 2.9 W/m2 target without reduction of land use CO2 emissions, but with BECCS would be a feasible scenario.

  1. Scenario Results

Table1. Probabilities to exceed various temperature targets of the presented scenarios.

Scenario name / Probability to exceed 1.5°C by 2100 / Probability to exceed 2°C by 2100 / Probability to exceed 2.5°C by 2100
% / % / %
Optimal 2.9 W/m2 / 90 / 41 / 11
Optimal BECCS 2.6 W/m2 / 79 / 27 / 5
Copenhagen Potential 2.9 W/m2 / 90 / 42 / 11
Copenhagen Potential BECCS 2.6 W/m2 / 79 / 27 / 5
Current Copenhagen 3.2 W/m2 / 96 / 63 / 21
Current Copenhagen BECCS 2.8 W/m2 / 88 / 40 / 10

Emission scenarios are designed until 2100. Beyond 2100,surface air temperatures are still rising in the scenarios without BECCS. The probability to exceed a temperature threshold in the longerterm (eg until 2200) could therefore be higher and depends on, inter alia, emissions in 2100-2200 period.

  1. Uncertainty related to discount rates

The authors have analyzed discount rates that decrease over time, such as advocated by theUK Green Book(Treasury, 2003) for long-term cost-benefit analyses. Such decreasing discount rates did not influence the conclusions on the effect of discount rates as presented in the paper. This is due to the fact that the distinction between flat and decreasing discount rates are most pronounced on the long term. However, the long-term contribution of costs to total discounted cumulative costs is limited. Therefore, the effect of decreasing discount rates on the timing of reductionsis limited, although it does influence total cumulative costs.

The analysis shows that pathways with lower discount rates show considerably more early action than medium discount rates, with the pathways with higher discount rates showing delayed action. There are three effects that could account for this dissimilarity, with the first probably having the most significant effect.

  1. The value of future costs with increasing discount rate is exponentially asymptotic to zero. While after 75 year discounting at 1% valuation of costs is still 47%, at 2% this falls to 22%. The impact on valuation of future costs of an increasing discount rate quickly diminishes.
  2. Due to the general exponential shape of MAC curves, postponing mitigation causes a more than linear increase of the carbon price. This will offset part or all of the lower valuation of future costs due to discounting.
  3. Finally, postponing action reduces future benefits of learning increasing technology costs. Again offsetting the decreased valuation of future costs due to higher discount rates.

The effect of discounting is analyzed in the paper in combination with ambitious climate targets. An ambitious target limits to some extent the flexibility in timing of emission reductions under varying discount rates. Flexibility in timing requires both sufficient mitigation potential and the possibility of reducing emissions at faster rates, both of which can be limited under ambitious climate targets. It can be expected that under less ambitious targets there is a larger spread among pathways.

References

den Elzen, M. G. J., D. P. van Vuuren and J. van Vliet (2010). "Postponing emission reductions from 2020 to 2030 increases climate risks and long-term costs." Climatic Change99(1): 313-320.

den Elzen, M. G. J., A. F. Hof, A. Mendoza Beltran, et al. (2010). Evaluation of the Copenhagen Accord: Chances and risks for the 2C climate goal., PBL, Netherlands Environmental Assessment Agency,

den Elzen, M. G. J., A. F. Hof and M. Roelfsema (2011). "The emissions gap between the Copenhagen pledges andthe 2C climate goal: Options for closing and risks that could wident he gap." Global Environmental Change(21): 733-743.

den Elzen, M. G. J., A. F. Hof, A. Mendoza Beltran, G. Grassi, M. Roelfsema, B. van Ruijven, J. van Vliet and D. P. van Vuuren (2011). "The Copenhagen Accord: Abatement costs and carbon prices resulting from the submissions." Environmental Science and Policy14(1): 28-39.