The Contribution of Non-CO 2 Greenhouse Gas Mitigation to Achieving Long-Term Temperature Goals

: This paper analyses the emissions and cost impacts of mitigation of non-CO 2 greenhouse gases (GHGs) at a global level, in scenarios aimed at meeting a range of long-term temperature goals (LTTGs). The study combines an integrated assessment model (TIAM-Grantham) representing CO 2 emissions (and their mitigation) from the fossil fuel combustion and industrial sectors, coupled with a model covering non-CO 2 emissions (GAINS), using the latest global warming potentials from the Intergovernmental Panel on Climate Change’s Fifth Assessment Report. We illustrate that in general non-CO 2 mitigation measures are less costly than CO 2 mitigation measures, with the majority of their abatement potential achievable at US2005$100/tCO 2 e or less throughout the 21st century (compared to a marginal CO 2 mitigation cost which is already greater than this by 2030 in the most stringent mitigation scenario). As a result, the total cumulative discounted cost over the period 2010–2100 (at a 5% discount rate) of limiting global average temperature change to 2.5 ◦ C by 2100 is $48 trillion (about 1.6% of cumulative discounted GDP over the period 2010–2100) if only CO 2 from the fossil fuel and industrial sectors is targeted, whereas the cost falls to $17 trillion (0.6% of GDP) by including non-CO 2 GHG mitigation in the portfolio of options—a cost reduction of about 65%. The criticality of non-CO 2 mitigation recommends further research, given its relatively less well-explored nature when compared to CO 2 mitigation.


Introduction
Achieving stringent mitigation of greenhouse gases (GHGs) is likely to require a multi-gas approach.As such, it is important to understand the contribution of non-CO 2 mitigation to achieving different long-term temperature goals.This requires simulations of future energy, industrial and agricultural systems to account for all GHGs together, in as consistent a manner as possible.This paper presents a multi-model approach to such a challenge, to analyse the emissions and cost impacts of mitigation of both CO 2 and non-CO 2 GHGs at a global level, in scenarios which are focused on meeting a range of long-term temperature goals (LTTGs).The objectives are threefold:

•
First, to demonstrate how an integrated assessment model (TIAM-Grantham) representing CO 2 emissions (and their mitigation) from the energy and industrial sectors is coupled with a model covering non-CO 2 emissions (GAINS) in order to provide a complete picture of GHG emissions in a reference scenario in which there is no mitigation of either CO 2 or non-CO 2 gases, as well as in scenarios in which both CO 2 and non-CO 2 gases are mitigated in order to achieve different LTTGs.

•
Second, to demonstrate the degree of indirect mitigation of non-CO 2 gases that results from mitigation of CO 2 sources.This principally applies to methane (CH 4 ) emission reductions which result from reduced extraction and distribution of fossil fuels in CO 2 mitigation scenarios which see a shift from fossil fuel energy sources to renewables and nuclear.

•
Third, to analyse the costs associated with mitigating non-CO 2 GHGs to varying degrees, by considering different levels of CO 2 e prices applied to the non-CO 2 GHG-emitting sectors, relative to the CO 2 prices that result from the CO 2 mitigation scenarios.This provides a picture of the marginal impact (in terms of temperature change in 2100) of varying the relative degree of effort in mitigating non-CO 2 gases when compared to CO 2 mitigation effort.
Non-CO 2 GHG emissions, at about 14 GtCO 2 e in 2010 (compared to 37 for CO 2 emissions) constituted about 28% of total GHG emissions in that year, measured on a CO 2 -equivalent (CO 2 e) basis using the Intergovernmental Panel on Climate Change (IPCC) fifth assessment report 100-year global warming potentials (GWP100) with climate-carbon feedback effects for each gas [1,2].Agricultural CH 4 and nitrous oxide (N 2 O) emissions, at between 5.2 and 5.8 GtCO 2 e in 2010, are the largest contributor to non-CO 2 GHG emissions.Over the last three decades (comparing 1980-1989, 1990-1999 and 2000-2009) CH 4 and N 2 O emissions from agriculture increased from about 4 to over 5 GtCO 2 e per year, with CH 4 emissions from livestock (enteric fermentation, mainly from cattle) accounting for just under half of this level throughout this period.Emissions growth from most agricultural sources (enteric fermentation, manure and fertiliser) in Africa, Asia and the Americas has been offset to some extent by emissions reductions in Europe [3], but future demand for food from these regions could be a major driver of emissions growth over the coming decades.Waste, fossil fuel extraction, transmission and distribution, and industrial production are other significant sources of non-CO 2 GHGs, principally CH 4 and N 2 O.
As well as making a significant contribution to warming of the climate, some non-CO 2 species also lead to relatively large amounts of warming per tonne emitted.CH 4 for example, by mass, has a global warming potential over 100 years (GWP100) which is 34 times larger than that of CO 2 when taking account of carbon-climate feedback effects in the atmosphere [1].It is important to note that this value is higher than the value (25) used in the previous (fourth) IPCC assessment report [4], when such feedbacks had not been adequately considered.This comparative measure of warming-that of an equivalent mass of CO 2 -is the basis for emissions accounting and allows one method of comparing the cost effectiveness of mitigation measures across different gas species for a given timeframe.The major sources and mitigation options for non-CO 2 GHGs are shown in Appendix A.
In addition to the technical supply side measures shown in Appendix A, mitigation could also come through changes in consumer preferences for meat and dairy products and reduced losses and waste of food [3,5,6] although there is in general less evidence on these demand-side emissions mitigation options [3].
There have been relatively fewer studies on the mitigation potential of non-CO 2 GHGs compared to CO 2 from the energy and industrial sectors.A number of sector specific studies were carried out in the late 1990s and early 2000s [7][8][9], many of which formed the basis of more comprehensive assessments [10][11][12].These studies were undertaken in order to construct marginal abatement cost curves for 2010, which were then extrapolated for use in integrated assessment studies [13,14].Further work [15] extended the marginal abatement costs (MACs) more systematically to 2100.This analysis, as well as some more recent analyses [16], has formed the basis of relatively recent estimates of long-term mitigation in for example the agricultural sector [17].
A consistent message from the multi-gas modelling studies is that the cost of mitigation to achieve a given temperature goal is less when mitigation of non-CO 2 GHGs is included amongst the mitigation options available.For example, Rao and Riahi [13] found that carbon prices associated with achieving a radiative forcing level of 4.5 W/m 2 by 2100 when using a multi-gas mitigation approach are about half those when using a CO 2 -only set of mitigation measures.Kurosawa [18] found that a multi-gas approach (again, to achieve a 4.5 W/m 2 forcing level by 2100) leads to a global mitigation cost of 3.8% of GDP by 2100, compared to 8.6% of GDP with a CO 2 -only approach.Lucas et al. [15] found that a multi-gas approach lowers mitigation costs between 3%-21% (by 2050) and 4%-26% (by 2100) compared to a CO 2 -only approach, to achieve a 550 ppm CO 2 e stabilisation concentration of GHGs.Recent analysis [19] shows that non-CO 2 mitigation to higher CO 2 e prices can relax the cumulative CO 2 budget required to meet specific temperature targets, which could be a useful facet of flexibility if there are specific technology constraints which prevent the achievement of stringent CO 2 budgets.
This analysis is the first of which we are aware that examines the impact of non-CO 2 gases on overall mitigation costs when using the latest GWP100 figures produced by the IPCC [1].In addition, this study, unlike the more recent studies on non-CO 2 gases [20,21], explicitly sets out the relative contribution of non-CO 2 gases to 2100 warming when taking into account differences between the marginal mitigation cost of CO 2 and non-CO 2 .This is of particular policy relevance when considering the acceptable level of marginal mitigation effort that might be allocated to different gases.Although the "what, when and where" flexibility of mitigation advocated by Stern [22] suggests that the most cost-efficient mitigation strategies should see CO 2 e prices equilibrate between sectors, gases and regions, it is clearly not going to be possible in all practical cases to achieve this.A further policy-relevant aspect of this analysis is that-in light of the Paris Agreement's [23] goal to achieve global warming levels of "well below 2 • C"-it is instructive to consider the degree to which this might be achieved without a contraction of an already-challenging CO 2 budget, but rather with additional effort in the non-CO 2 emitting sectors.In examining scenarios in which there is greater mitigation effort in the non-CO 2 emitting sectors than the CO 2 emitting sectors, this analysis allows a consideration of the potential viability of such strategies.

Materials and Methods
The model used in this assessment, the Greenhouse Gas Air Pollution Interactions and Synergies (GAINS) model, has a comprehensive, multi-country and region representation of non-CO 2 GHG emissions sources, as well as the measures and costs for their mitigation [24,25].The cost data used here is from the 2013 update of the GAINS model.It has been used in recent studies of the mitigation potential of CH 4 [26], as well as other climate forcing species such as black carbon, with a view to assessing not just climate but also air quality, health and agricultural crop yield benefits of mitigating these short-lived species [27].As such, it has been chosen because of its relatively recent development, its state-of-the-art level of detail of mitigation options for the non-CO 2 GHGs, as well as its geographical detail which allows aggregation of countries into regions which closely match the 15 regions represented in Imperial College London's global TIMES (The Integrated MARKAL-EFOM System) Integrated Assessment Model (TIAM-Grantham) [28,29].This model represents the global energy and industrial system in these regions, including low-carbon technologies and their costs, and associated CO 2 emissions.It is an inter-temporal optimisation model which finds the welfare maximising solution to the objective of meeting future energy service and industrial product demands across all economic sectors within a given climate or CO 2 emissions constraint.It has been used in a model inter-comparison study as part of the AVOID 2 research programme to analyse the technologies and costs of a range of long-term temperature targets [30].It should be noted that this analysis covers the well-mixed GHGs and does not explicitly model emissions of aerosols and precursors, for example black carbon-for each scenario these have been estimated using the methods described below.
There are in many cases interactions between measures that mitigate different GHGs.For example, mitigation of CO 2 frequently consists of substituting non-fossil energy sources for fossil fuels, which results in reduced fugitive CH 4 emissions from the extraction and distribution of these fuels [21].In addition to accounting for such interactions, it is important to ensure a high level of consistency between the drivers of energy and industrial CO 2 emissions and those for non-CO 2 emissions sources, principally agricultural activity responsible for CH 4 and N 2 O emissions.
In order to maximise consistency between the energy and industrial CO 2 mitigation modelling in the TIAM-Grantham model, and the non-CO 2 mitigation modelling in the GAINS model, a number of steps have been undertaken, as described in detail in Appendix B. In summary:

•
For each LTTG (in this study 2100 temperature change levels of 2, 2.5 and 4  The TIAM-Grantham model has been used to produce an unmitigated reference scenario, as well as mitigation scenarios based on these estimated CO 2 budgets, using a standard set of socio-economic drivers, specifically the OECD variant of the Shared Socio-Economic Pathways 2 (SSP2), which has been used in order to represent a future world in which recent socio-economic trends continue [33]; The GAINS model, also using SSP2 socio-economic inputs, as well as energy price and fossil fuel supply and demand outputs from the TIAM-Grantham model scenarios, has been used to produce a "baseline" level of non-CO 2 emissions for each TIAM-Grantham scenario, as well as marginal abatement cost (MAC) curves for each ten-year time point (2020, 2030, 2040, etc.) for each non-CO 2 GHG species (Methane (CH 4 ), Nitrous Oxide (N 2 O), Fluorinated gases (F-gases, which are perfluorocarbons, PFCs, Sulphur hexafluoride, SF 6 , and hydrofluorocarbons, HFCs));

•
For each scenario, the 2100 temperature when mitigating non-CO 2 GHGs to different prices (on a GWP100 basis, with prices relative to the CO 2 price for each TIAM-Grantham scenario) has been calculated, using the same version of the MAGICC used to estimate the initial CO 2 budgets;

•
Where the non-CO 2 and CO 2 prices are equal, if there is a major (in this case, greater than 0.1 • C) difference in the calculated 2100 temperature change relative to the initially-intended LTTG, a revision to the initial CO 2 budget has been made and the process repeated.
As indicated above, the MAC curves derived from GAINS allow analysis of non-CO 2 mitigation up to a CO 2 e price equal to the CO 2 price which was output from the TIAM-Grantham model (thereby equating marginal mitigation "effort" for CO 2 and the non-CO 2 GHGs) as well as at CO 2 e prices at different fractions of the TIAM-Grantham CO 2 price (thereby considering different marginal effort levels for non-CO 2 GHGs when compared to CO 2 mitigation effort).This approach allows analysis of the 2100 median temperature change and overall mitigation cost (i.e., considering both CO 2 and non-CO 2 mitigation options) when considering lower and higher levels of "effort" of non-CO 2 GHG mitigation measures compared to CO 2 mitigation measures.For each mitigation scenario, as well as the 2100 temperature change, the cumulative discounted cost (using a discount rate of 5% per year) of both CO 2 and non-CO 2 GHG mitigation is calculated, relative to the reference (unmitigated) scenario.

Mitigation of Non-CO 2 Emissions
Figure 1 shows the emissions level for each GHG in the unmitigated reference scenario where there is no price or constraint on any of the GHGs, using the GWP100 equivalence measure (as taken from the IPCC's fifth assessment report [1]).This unmitigated scenario follows from running the TIAM-Grantham model to produce a scenario for a least-cost energy system that meets future energy needs under the SSP2 shared socio-economic pathways assumptions [33], but with no climate constraints.Emissions rise from 50 GtCO 2 e/year in 2010 to 132 GtCO 2 e/year in 2100.The resulting median warming in 2100 is 4.6 • C. For both 2100 emissions and temperature change, these figures are closer to the upper end of the range for the high emissions scenarios presented in the IPCC's 5th Assessment Report, WGIII [2], reflecting the relatively strong socio-economic growth throughout the century represented by the SSP2 input scenarios.It can be seen that CO 2 is the largest contributor to GHG emissions throughout the period (reaching 93 GtCO 2 e/year by 2100), with CH 4 and N 2 O continuing to remain significant.By comparison, the RCP8.5 pathway, which has the highest emissions of the RCPs, sees global GHG emissions reaching 120 GtCO 2 e/year in 2100, albeit with much lower global GDP by 2100 (a seven-fold increase over the 21st century [34], compared to an 11-fold increase in this study).Of this 120 GtCO 2 e/year, approximately 80 GtCO 2 e/year is from CO 2 and the remainder from non-CO 2 gases (compared to 93 and 39 GtCO 2 e/year respectively in this study) [35].
Energies 2017, 10, 602 5 of 24 median warming in 2100 is 4.6 °C.For both 2100 emissions and temperature change, these figures are closer to the upper end of the range for the high emissions scenarios presented in the IPCC's 5th Assessment Report, WGIII [2], reflecting the relatively strong socio-economic growth throughout the century represented by the SSP2 input scenarios.It can be seen that CO2 is the largest contributor to GHG emissions throughout the period (reaching 93 GtCO2e/year by 2100), with CH4 and N2O continuing to remain significant.By comparison, the RCP8.5 pathway, which has the highest emissions of the RCPs, sees global GHG emissions reaching 120 GtCO2e/year in 2100, albeit with much lower global GDP by 2100 (a seven-fold increase over the 21st century [34], compared to an 11-fold increase in this study).Of this 120 GtCO2e/year, approximately 80 GtCO2e/year is from CO2 and the remainder from non-CO2 gases (compared to 93 and 39 GtCO2e/year respectively in this study) [35].The approximately linear nature of these GHG emissions levels is primarily because of the approximately linear nature of the CO2 emissions level, which is coincidental only.Emissions values follow from detailed modelling of the energy and non-CO2 GHG emitting sectors in 10 year time-steps from the TIAM-Grantham and GAINS models used in this analysis.As shown in Appendix C which decomposes the CO2 emissions levels, the CO2 emissions pathway results from non-linearly increasing GDP, offset by non-linearly decreasing primary energy intensity of GDP, combined with approximately constant CO2 intensity of primary energy.
Hence, both RCP 8.5 and this study see non-CO2 emissions accounting for about a third of the total GHG emissions by 2100, slightly higher than the upper end of the range (16-27%) in recent multi-gas mitigation scenarios [20].In fact in these recent scenarios the maximum 2100 non-CO2 emissions level (across the six models compared) is 30 GtCO2e/year, with CH4 emissions at 15 GtCO2e/year.This is about 10 GtCO2e/year below the emissions in this study, mainly because in this study CH4 makes up 25 GtCO2e/year in 2100.This results from three main factors: first, the relatively high-growth socio-economic assumptions driving future emissions growth in this study; second, the considerably higher GWP100 value for CH4 (34) taken from the IPCC's latest (i.e., fifth) assessment report compared to the lower value (25) used in the recent multi-gas mitigation scenarios [20] and also the IPCC's earlier fourth assessment report [1], and third, the considerably higher CH4 emissions from global oil production in GAINS' historical emission inventories compared with others (e.g., [20,35,36]), which is a result of a more in-depth estimation method [37] and which also has implications for future emissions from this source.
Table 1 shows the estimated CO2 budgets as well as the median temperature change that results from mitigation of non-CO2 GHGs to a CO2e price (using GWP100) equal to the CO2 price from the TIAM-Grantham model for each budget (taking the scenarios with delayed action to 2020).Also Figure 1.Global greenhouse gas emissions in the (unmitigated) reference scenario.Notes: The approximately linear nature of these GHG emissions levels is primarily because of the approximately linear nature of the CO 2 emissions level, which is coincidental only.Emissions values follow from detailed modelling of the energy and non-CO 2 GHG emitting sectors in 10 year time-steps from the TIAM-Grantham and GAINS models used in this analysis.As shown in Appendix C which decomposes the CO 2 emissions levels, the CO 2 emissions pathway results from non-linearly increasing GDP, offset by non-linearly decreasing primary energy intensity of GDP, combined with approximately constant CO 2 intensity of primary energy.
Hence, both RCP 8.5 and this study see non-CO 2 emissions accounting for about a third of the total GHG emissions by 2100, slightly higher than the upper end of the range (16-27%) in recent multi-gas mitigation scenarios [20].In fact in these recent scenarios the maximum 2100 non-CO 2 emissions level (across the six models compared) is 30 GtCO 2 e/year, with CH 4 emissions at 15 GtCO 2 e/year.This is about 10 GtCO 2 e/year below the emissions in this study, mainly because in this study CH 4 makes up 25 GtCO 2 e/year in 2100.This results from three main factors: first, the relatively high-growth socio-economic assumptions driving future emissions growth in this study; second, the considerably higher GWP100 value for CH 4 (34) taken from the IPCC's latest (i.e., fifth) assessment report compared to the lower value (25) used in the recent multi-gas mitigation scenarios [20] and also the IPCC's earlier fourth assessment report [1], and third, the considerably higher CH 4 emissions from global oil production in GAINS' historical emission inventories compared with others (e.g., [20,35,36]), which is a result of a more in-depth estimation method [37] and which also has implications for future emissions from this source.
Table 1 shows the estimated CO 2 budgets as well as the median temperature change that results from mitigation of non-CO 2 GHGs to a CO 2 e price (using GWP100) equal to the CO 2 price from the TIAM-Grantham model for each budget (taking the scenarios with delayed action to 2020).Also shown is the median 2100 temperature change resulting from the unmitigated TIAM-Grantham and GAINS scenarios (i.e., resulting from the emissions levels shown in Figure 1).Figure 2 shows the non-CO 2 GHG emissions for a 2 • C mitigation scenario with global mitigation action starting in 2020 (and weak country/regional policy actions to 2020), after CO 2 mitigation has occurred to meet the cumulative CO 2 budget, but before any specific mitigation has occurred in the non-CO 2 sectors.Also shown is the completely unmitigated level of non-CO 2 GHG emissions that derive from the reference scenario with no mitigation action for any GHGs (which in the case of non-CO 2 means action beyond that prescribed in existing legislation as of September 2015).In other words, Figure 2 shows the indirect mitigation of the non-CO 2 GHGs that occurs as a result of changes in the energy system when transitioning to low-carbon (and in particular lower fossil fuel reliance) over the century.There is significant mitigation of CH 4 (about 9 GtCO 2 e/year by 2100) resulting from reduced fossil fuel extraction and distribution, and therefore lower fugitive CH 4 emissions.The importance of accounting for this indirect mitigation effect has been highlighted in recent studies [21].
Energies 2017, 10, 602 6 of 24 shown is the median 2100 temperature change resulting from the unmitigated TIAM-Grantham and GAINS scenarios (i.e., resulting from the emissions levels shown in Figure 1).Figure 2 shows the non-CO2 GHG emissions for a 2 °C mitigation scenario with global mitigation action starting in 2020 (and weak country/regional policy actions to 2020), after CO2 mitigation has occurred to meet the cumulative CO2 budget, but before any specific mitigation has occurred in the non-CO2 sectors.Also shown is the completely unmitigated level of non-CO2 GHG emissions that derive from the reference scenario with no mitigation action for any GHGs (which in the case of non-CO2 means action beyond that prescribed in existing legislation as of September 2015).In other words, Figure 2 shows the indirect mitigation of the non-CO2 GHGs that occurs as a result of changes in the energy system when transitioning to low-carbon (and in particular lower fossil fuel reliance) over the century.There is significant mitigation of CH4 (about 9 GtCO2e/year by 2100) resulting from reduced fossil fuel extraction and distribution, and therefore lower fugitive CH4 emissions.The importance of accounting for this indirect mitigation effect has been highlighted in recent studies [21].Figure 3 shows the effect of indirect mitigation for a range of long-term temperature goals.As expected, the degree of mitigation increases as the temperature goal decreases, resulting from an Figure 3 shows the effect of indirect mitigation for a range of long-term temperature goals.As expected, the degree of mitigation increases as the temperature goal decreases, resulting from an increasingly marked shift from a fossil fuel-based energy system to a low-carbon system in which non-fossil sources such as nuclear and renewables dominate.
Energies 2017, 10, 602 7 of 24 increasingly marked shift from a fossil fuel-based energy system to a low-carbon system in which non-fossil sources such as nuclear and renewables dominate.Figure 4 shows the further mitigation of non-CO2 GHGs resulting from mitigation measures targeted specifically towards these gases, for the 2 °C scenario with delayed action to 2020.Also shown are the levels of each non-CO2 GHG for the indirectly mitigated case.The figure shows for each time step the mitigation of non-CO2 GHGs up to the CO2e price that is equal to the CO2 price in the TIAM-Grantham model (i.e., the shadow price of CO2 associated with achieving the least cost mitigation pathway to meet the specified 21st century cumulative CO2 budget).As such, this equates a level of mitigation effort for CO2 and non-CO2 GHGs according to the marginal cost of abatement at any given time.In the case of Figure 4, this marginal abatement cost is calculated on a GWP100 basis.
It can be seen that there is significant abatement of all non-CO2 GHGs up to this CO2e price, such that by 2100 the fully mitigated level of non-CO2 GHGs is just under 13 GtCO2e/year, compared to 39 GtCO2e/year in the unmitigated reference scenario.Of the 27 GtCO2e/year reduction, 69% occurs through the direct mitigation of the non-CO2 GHGs and 31% through the indirect mitigation (mostly of CH4) that follows from CO2 mitigation.Of the unmitigated reference 2100 level of each non-CO2 GHG, 67% of CH4, 37% of N2O and 99% of F-gases are mitigated, leaving 7.8, 4.5 and 0.1 GtCO2e/year of CH4, N2O and F-gases respectively.These reductions compare to recent modelled scenarios (focusing specifically on the issue of non-CO2 GHG mitigation) in which by 2100 up to 71% of CH4, 42% of N2O and 90% of F-gases are mitigated [20], as well as the broader IPCC fifth assessment report database [38] in which, across all of the most stringent mitigation scenarios, 44-74% of CH4, 9-42% of N2O and 45-90% of F-gases are mitigated by 2100, compared to the relevant unmitigated baseline scenario for each model used.In this database, the range of 2100 CH4 emissions is 12-25 GtCO2e/year (using the most current CH4 GWP100 value of 34) in the reference scenario and 4-11 GtCO2e/year in the mitigation scenarios, compared to 25 GtCO2e/year and 7.8 GtCO2e/year in the 2 °C scenario of this study.The database's range of 2100 N2O emissions is 3.0-8.8GtCO2e/year in the reference and 2.1-8.1 GtCO2e/year in the mitigation scenarios, compared to 7.0 and 4.5 GtCO2e/year in this study.The database's range of 2100 F-gases emissions is 1.2-10 GtCO2e/year in the reference and 0.06-1.7 GtCO2e/year in the mitigation scenarios, compared to 7.2 and 0.08 GtCO2e/year in this study.Figure 4 shows the further mitigation of non-CO 2 GHGs resulting from mitigation measures targeted specifically towards these gases, for the 2 • C scenario with delayed action to 2020.Also shown are the levels of each non-CO 2 GHG for the indirectly mitigated case.The figure shows for each time step the mitigation of non-CO 2 GHGs up to the CO 2 e price that is equal to the CO 2 price in the TIAM-Grantham model (i.e., the shadow price of CO 2 associated with achieving the least cost mitigation pathway to meet the specified 21st century cumulative CO 2 budget).As such, this equates a level of mitigation effort for CO 2 and non-CO 2 GHGs according to the marginal cost of abatement at any given time.In the case of Figure 4, this marginal abatement cost is calculated on a GWP100 basis.
It can be seen that there is significant abatement of all non-CO 2 GHGs up to this CO 2 e price, such that by 2100 the fully mitigated level of non-CO 2 GHGs is just under 13 GtCO 2 e/year, compared to 39 GtCO 2 e/year in the unmitigated reference scenario.Of the 27 GtCO 2 e/year reduction, 69% occurs through the direct mitigation of the non-CO 2 GHGs and 31% through the indirect mitigation (mostly of CH 4 ) that follows from CO 2 mitigation.Of the unmitigated reference 2100 level of each non-CO 2 GHG, 67% of CH 4 , 37% of N 2 O and 99% of F-gases are mitigated, leaving 7.8, 4.5 and 0.1 GtCO 2 e/year of CH 4 , N 2 O and F-gases respectively.These reductions compare to recent modelled scenarios (focusing specifically on the issue of non-CO 2 GHG mitigation) in which by 2100 up to 71% of CH 4 , 42% of N 2 O and 90% of F-gases are mitigated [20], as well as the broader IPCC fifth assessment report database [38]  Hence, the reference and mitigation emissions levels in this study are within the AR5 database range, although the CH 4 reference emissions are at the higher end of the range, reflecting the relatively high socio-economic growth path and industrial output growth over the 21st century, as previously mentioned.
Energies 2017, 10, 602 8 of 24 Hence, the reference and mitigation emissions levels in this study are within the AR5 database range, although the CH4 reference emissions are at the higher end of the range, reflecting the relatively high socio-economic growth path and industrial output growth over the 21st century, as previously mentioned.

Costs of Mitigation Considering Non-CO2 Gases
Figure 5 shows the time-dependent global marginal abatement cost curves for the total non-CO2 GHGs starting from the point at which any indirect mitigation occurring as a result of CO2 mitigation has already occurred, for the 2 °C scenario in which global mitigation action starts in 2020.Of note is that, even in 2100, there is expected to be significant abatement potential at marginal costs of $50/tCO2e or less, with the majority of abatement in all years available at below $100/tCO2e.The increase in mitigation potential between 2050 and 2100 is entirely driven by changes in activity levels, e.g., population, economic growth and changes in the energy-system.No effects of learning or technological development are taken into account in the assessments of future mitigation potentials.A reason is that there is a lack of empirical basis for adopting general assumptions about the rate at which non-CO2 regulations would drive long-term technological development.Most likely, this drive is not as strong for non-CO2 as for CO2, where regulations reinforce already existing incentives to improve energy efficiency in order to save on energy costs.Hence, in the absence of a firm basis for assumptions on technological development of non-CO2 mitigation measures, the estimated future potentials for non-CO2 mitigation should be considered conservative rather than optimistic.
Also of note is the presence of some significantly negative cost mitigation measures in all years.These measures are not profitable with today's energy prices, but expected to become profitable in the future, conditional on a rise in future energy prices.This effect is not accounted for in the reference scenario as it is defined as a scenario without further mitigation actions.Whether measures that become profitable in the future as a result of rising energy prices will be taken up automatically or not depends on more factors than pure short-run cost-effectiveness [39].Without additional regulations in place, the presence of x-inefficiency, institutional inertia and uncertainty regarding future regulations and energy prices, are likely to discourage investments in mitigation in the reference scenario.To avoid speculation, such investment opportunities appear here as negative cost mitigation measures in the cost curves and are likely to be among the first measures to be taken up once regulations have been introduced.

Costs of Mitigation Considering Non-CO 2 Gases
Figure 5 shows the time-dependent global marginal abatement cost curves for the total non-CO 2 GHGs starting from the point at which any indirect mitigation occurring as a result of CO 2 mitigation has already occurred, for the 2 • C scenario in which global mitigation action starts in 2020.Of note is that, even in 2100, there is expected to be significant abatement potential at marginal costs of $50/tCO 2 e or less, with the majority of abatement in all years available at below $100/tCO 2 e.The increase in mitigation potential between 2050 and 2100 is entirely driven by changes in activity levels, e.g., population, economic growth and changes in the energy-system.No effects of learning or technological development are taken into account in the assessments of future mitigation potentials.A reason is that there is a lack of empirical basis for adopting general assumptions about the rate at which non-CO 2 regulations would drive long-term technological development.Most likely, this drive is not as strong for non-CO 2 as for CO 2 , where regulations reinforce already existing incentives to improve energy efficiency in order to save on energy costs.Hence, in the absence of a firm basis for assumptions on technological development of non-CO 2 mitigation measures, the estimated future potentials for non-CO 2 mitigation should be considered conservative rather than optimistic.
Also of note is the presence of some significantly negative cost mitigation measures in all years.These measures are not profitable with today's energy prices, but expected to become profitable in the future, conditional on a rise in future energy prices.This effect is not accounted for in the reference scenario as it is defined as a scenario without further mitigation actions.Whether measures that become profitable in the future as a result of rising energy prices will be taken up automatically or not depends on more factors than pure short-run cost-effectiveness [39].Without additional regulations in place, the presence of x-inefficiency, institutional inertia and uncertainty regarding future regulations and energy prices, are likely to discourage investments in mitigation in the reference scenario.To avoid speculation, such investment opportunities appear here as negative cost mitigation measures in the cost curves and are likely to be among the first measures to be taken up once regulations have been introduced.Measures with a negative marginal abatement cost are assumed to be cost-saving if adopted.They are not assumed to be taken up in the reference scenario, as a result of uncertainty, inertia and inefficiency in current practices.
Figure 6 shows the marginal abatement cost curves for 2050, for three different LTTGs (2, 2.5 and 4 °C median global warming by 2100), in scenarios with global mitigation action starting in 2020.At higher LTTGs, there is less indirect mitigation, which means that the total direct mitigation potential at a given CO2 price is greater.
Table 2 sets out some significant mitigation options for each non-CO2 GHG within different cost ranges.
Figure 6.Global marginal abatement cost curves in 2050 for the total non-CO2 GHGs (GWP100 basis) for different LTTGs, relative to the case where indirect non-CO2 GHG mitigation resulting from CO2 mitigation has already occurred.Notes: Measures with a negative marginal abatement cost are assumed to be cost-saving if adopted.They are not assumed to be taken up in the reference scenario, as a result of uncertainty, inertia and inefficiency in current practices.Measures with a negative marginal abatement cost are assumed to be cost-saving if adopted.They are not assumed to be taken up in the reference scenario, as a result of uncertainty, inertia and inefficiency in current practices.
Figure 6 shows the marginal abatement cost curves for 2050, for three different LTTGs (2, 2.5 and 4 • C median global warming by 2100), in scenarios with global mitigation action starting in 2020.At higher LTTGs, there is less indirect mitigation, which means that the total direct mitigation potential at a given CO 2 price is greater.
Table 2 sets out some significant mitigation options for each non-CO 2 GHG within different cost ranges.Measures with a negative marginal abatement cost are assumed to be cost-saving if adopted.They are not assumed to be taken up in the reference scenario, as a result of uncertainty, inertia and inefficiency in current practices.
Figure 6 shows the marginal abatement cost curves for 2050, for three different LTTGs (2, 2.5 and 4 °C median global warming by 2100), in scenarios with global mitigation action starting in 2020.At higher LTTGs, there is less indirect mitigation, which means that the total direct mitigation potential at a given CO2 price is greater.
Table 2 sets out some significant mitigation options for each non-CO2 GHG within different cost ranges.
Figure 6.Global marginal abatement cost curves in 2050 for the total non-CO2 GHGs (GWP100 basis) for different LTTGs, relative to the case where indirect non-CO2 GHG mitigation resulting from CO2 mitigation has already occurred.Notes: Measures with a negative marginal abatement cost are assumed to be cost-saving if adopted.They are not assumed to be taken up in the reference scenario, as a result of uncertainty, inertia and inefficiency in current practices.for different LTTGs, relative to the case where indirect non-CO 2 GHG mitigation resulting from CO 2 mitigation has already occurred.Notes: Measures with a negative marginal abatement cost are assumed to be cost-saving if adopted.They are not assumed to be taken up in the reference scenario, as a result of uncertainty, inertia and inefficiency in current practices.Notes: All CO 2 e prices calculated using GWP100 basis; many mitigation options span a range of costs, depending on region, practices and local costs-hence figures are illustrative and do not reflect all details of estimated cost curves.
Figure 7 shows, for the different scenarios explored, the total cumulative discounted cost over the period 2010-2100 (at a discount rate of 5%) associated with mitigation of CO 2 to 2100, as well as mitigation of non-CO 2 GHGs to 2100 at a range of CO 2 e prices, the latter as a percentage of the CO 2 price from the TIAM-Grantham model for each time point.This cost is calculated by combining two costs: the first is the present value (using a discount rate of 5%) of the additional cost of the energy system in the TIAM-Grantham model when comparing the 2 • C scenario with the unmitigated reference scenario; the second is the present value (again at a discount rate of 5%) of the sum of annual non-CO 2 mitigation costs as calculated from the area under the marginal abatement cost curve for each year in the GAINS model.Mitigation at a zero price on non-CO 2 (thereby allowing only negative cost measures) results in a 2100 median temperature change of just under 2.5 • C.This is because the cumulative CO 2 budget for the fossil fuel and industrial sectors in order to produce a 2100 median warming level of 2 • C is appropriate only if there is also significant abatement of non-CO 2 GHGs [40] (broadly in line with the level of mitigation achieved in the RCP 2.6 scenario [41]).
Energies 2017, 10, 602 11 of 24 Figure 7 shows, for the different scenarios explored, the total cumulative discounted cost over the period 2010-2100 (at a discount rate of 5%) associated with mitigation of CO2 to 2100, as well as mitigation of non-CO2 GHGs to 2100 at a range of CO2e prices, the latter as a percentage of the CO2 price from the TIAM-Grantham model for each time point.This cost is calculated by combining two costs: the first is the present value (using a discount rate of 5%) of the additional cost of the energy system in the TIAM-Grantham model when comparing the 2 °C scenario with the unmitigated reference scenario; the second is the present value (again at a discount rate of 5%) of the sum of annual non-CO2 mitigation costs as calculated from the area under the marginal abatement cost curve for each year in the GAINS model.Mitigation at a zero price on non-CO2 (thereby allowing only negative cost measures) results in a 2100 median temperature change of just under 2.5 °C.This is because the cumulative CO2 budget for the fossil fuel and industrial sectors in order to produce a 2100 median warming level of 2 °C is appropriate only if there is also significant abatement of non-CO2 GHGs [40] (broadly in line with the level of mitigation achieved in the RCP 2.6 scenario [41]).Mitigation of non-CO2 GHGs even to a small fraction (20%) of the price of CO2 from fossil fuels and industry leads to significant abatement of non-CO2 GHGs, and a 2100 median temperature change of much closer to 2 °C (about 2.04 °C), at an additional cumulative discounted cost of around 0.08% of 2010-2100 GDP.Even at this 20% fraction of the fossil fuel and industry CO2 price, the non-CO2 GHG price rises to $1170/tCO2e by 2100.For this reason Figure 7 also shows the median warming (as well as total mitigation cost) at sustained prices of (2005) $50/tCO2e and $100/tCO2e throughout the century, reflecting the significant degree of mitigation potential available up to these prices, as shown in Figure 5.As expected, the scenarios with these CO2e prices lead to median warming levels which are lower than the 2.5 °C median warming that results when a zero CO2e price is applied to non-CO2 GHGs.
However, the scenarios with a uniformly-applied CO2e price are not as cost-efficient as the scenarios in which the CO2e price is applied as a fixed fraction of the (rising) CO2 price, which is to say that they do not achieve as low a level of 2100 median warming at the same cumulative cost as the fractional price scenarios.For example, Figure 7 shows that applying a CO2e price of 20% of the CO2 price throughout the mitigation period (during which the CO2e price rises from $0/tCO2e in 2020 to $38/tCO2e in 2030, $62/tCO2e in 2040 and then to $1170/tCO2e in 2100) is actually less costly, and achieves a lower 2100 temperature change, than applying a $50/tCO2e price uniformly from 2020 to 2100.This is because, with the uniform non-CO2e prices, some of the mitigation effort in the early  Mitigation of non-CO 2 GHGs even to a small fraction (20%) of the price of CO 2 from fossil fuels and industry leads to significant abatement of non-CO 2 GHGs, and a 2100 median temperature change of much closer to 2 • C (about 2.04 • C), at an additional cumulative discounted cost of around 0.08% of 2010-2100 GDP.Even at this 20% fraction of the fossil fuel and industry CO 2 price, the non-CO 2 GHG price rises to $1170/tCO 2 e by 2100.For this reason Figure 7 also shows the median warming (as well as total mitigation cost) at sustained prices of (2005) $50/tCO 2 e and $100/tCO 2 e throughout the century, reflecting the significant degree of mitigation potential available up to these prices, as shown in Figure 5.As expected, the scenarios with these CO 2 e prices lead to median warming levels which are lower than the 2.5 • C median warming that results when a zero CO 2 e price is applied to non-CO 2 GHGs.
However, the scenarios with a uniformly-applied CO 2 e price are not as cost-efficient as the scenarios in which the CO 2 e price is applied as a fixed fraction of the (rising) CO 2 price, which is to say that they do not achieve as low a level of 2100 median warming at the same cumulative cost as the fractional price scenarios.For example, Figure 7 shows that applying a CO 2 e price of 20% of the CO 2 price throughout the mitigation period (during which the CO 2 e price rises from $0/tCO 2 e in 2020 to $38/tCO 2 e in 2030, $62/tCO 2 e in 2040 and then to $1170/tCO 2 e in 2100) is actually less costly, and achieves a lower 2100 temperature change, than applying a $50/tCO 2 e price uniformly from 2020 to 2100.This is because, with the uniform non-CO 2 e prices, some of the mitigation effort in the early part of the century which targets short-lived CH 4 and F-gases (particularly over the decades 2020-2040, in which the uniformly applied CO 2 e price is on average higher than the steadily-rising fractional CO 2 e price) has no impact on the 2100 median warming level, and is in some ways therefore "wasted" effort (and cost) with regard to the 2100 median temperature change.This additional mitigation cost of the uniform non-CO 2 e price, which does not achieve a benefit in terms of 2100 warming, outweighs the (discounted) cost saving of the uniform price being lower than the fractional price in later decades.
This result is, however, highly dependent on the discount rate used (with lower discount rates de-emphasising the cost of applying a uniform price in the short term, compared to the higher fractional cost in the long term).Perhaps more importantly, it is feasible that early action on mitigation of non-CO 2 gases would reap benefits in terms of learning and associated cost reductions in future mitigation measures.In addition, early mitigation of non-CO 2 GHGs may be seen as highly advantageous in order to lower near-term warming, regardless of its impact on warming in 2100.Hence, further analysis is required before any policy conclusions can be drawn on the timing and degree of effort in mitigating short-lived gases such as CH 4 .
Also of note from Figure 7 is that where the non-CO 2 e price is higher than the CO 2 price (the "120%" point) there is relatively little impact on 2100 median temperature change, since the vast majority of non-CO 2 abatement is already taken up at lower non-CO 2 e prices.
A similar analysis is shown for the 2.5 and 4 • C scenarios, in Figures 8 and 9 respectively.Of note is that the overall mitigation cost is significantly lower than the 2 • C pathway, which gives a sense of the relative degree of challenge involved in meeting the 2 • C long-term goal.In fact, in the case of the 4 • C scenario, the very low mitigation costs for CO 2 are slightly outweighed by negative cost measures for non-CO 2 gases, leading to an overall marginal negative cost of meeting the 4 • C goal.Whether this is realisable in practice depends on the realism of achieving these measures.These stem principally from recycling in developing countries, with the assumption that recycled products would be sold at international market prices-in practice the recycled products may have less economic value than this if they cannot reach these markets.
In both the 2.5 and 4 • C scenarios, there is actually over-achievement of the long-term goal (i.e., temperature change is less than 2.5 and 4 • C respectively) when the CO 2 and non-CO 2 prices are equal, indicating that the target may be achieved in a less costly way with a little less CO 2 mitigation effort.Nevertheless, the final estimated median temperature changes are sufficiently close to the desired goals to prove useful as an indicative scenario of the costs and measures associated with meeting these goals.
Energies 2017, 10, 602 12 of 24 part of the century which targets short-lived CH4 and F-gases (particularly over the decades 2020-2040, in which the uniformly applied CO2e price is on average higher than the steadily-rising fractional CO2e price) has no impact on the 2100 median warming level, and is in some ways therefore "wasted" effort (and cost) with regard to the 2100 median temperature change.This additional mitigation cost of the uniform non-CO2e price, which does not achieve a benefit in terms of 2100 warming, outweighs the (discounted) cost saving of the uniform price being lower than the fractional price in later decades.This result is, however, highly dependent on the discount rate used (with lower discount rates de-emphasising the cost of applying a uniform price in the short term, compared to the higher fractional cost in the long term).Perhaps more importantly, it is feasible that early action on mitigation of non-CO2 gases would reap benefits in terms of learning and associated cost reductions in future mitigation measures.In addition, early mitigation of non-CO2 GHGs may be seen as highly advantageous in order to lower near-term warming, regardless of its impact on warming in 2100.Hence, further analysis is required before any policy conclusions can be drawn on the timing and degree of effort in mitigating short-lived gases such as CH4.
Also of note from Figure 7 is that where the non-CO2e price is higher than the CO2 price (the "120%" point) there is relatively little impact on 2100 median temperature change, since the vast majority of non-CO2 abatement is already taken up at lower non-CO2e prices.
A similar analysis is shown for the 2.5 and 4 °C scenarios, in Figures 8 and 9 respectively.Of note is that the overall mitigation cost is significantly lower than the 2 °C pathway, which gives a sense of the relative degree of challenge involved in meeting the 2 °C long-term goal.In fact, in the case of the 4 °C scenario, the very low mitigation costs for CO2 are slightly outweighed by negative cost measures for non-CO2 gases, leading to an overall marginal negative cost of meeting the 4 °C goal.Whether this is realisable in practice depends on the realism of achieving these measures.These stem principally from recycling in developing countries, with the assumption that recycled products would be sold at international market prices-in practice the recycled products may have less economic value than this if they cannot reach these markets.
In both the 2.5 and 4 °C scenarios, there is actually over-achievement of the long-term goal (i.e., temperature change is less than 2.5 and 4 °C respectively) when the CO2 and non-CO2 prices are equal, indicating that the target may be achieved in a less costly way with a little less CO2 mitigation effort.Nevertheless, the final estimated median temperature changes are sufficiently close to the desired goals to prove useful as an indicative scenario of the costs and measures associated with meeting these goals.Figure 10 shows a subsection of the data from Figures 7-9, so as to demonstrate the change in total mitigation cost and median 2100 temperature change as the non-CO2 GHG price (in $/tCO2e using GWP100) changes as a fraction of the CO2 price (in $/tCO2).This demonstrates first the significant additional cost of meeting the 2 °C target compared to the 2.5 and 4 °C targets, as well as the significant reduction in 2100 temperature change achievable by including non-CO2 GHG options in the overall mitigation portfolio.This is most clearly illustrated with reference to the vertical dashed line around the 2.5 °C mark in Figure 10: this LTTG is achievable either at a cost of $48 trillion by focusing only on CO2 mitigation, or alternatively at $17 trillion by including non-CO2 GHG mitigation in the portfolio of options, a cost reduction of about 65%.This compares to the figures discussed in Section 2, in which Rao and Riahi [13] found an approximate halving of carbon price, Kurosawa [18] found an approximate 55% cost saving by 2100, and Lucas et al [15] found a 4-26% mitigation cost reduction by 2100, when achieving a 550 ppm CO2e stabilisation concentration using a multi-gas approach compared to a CO2-only approach.The greater percentage cost savings in this study are most likely to stem from the fact that the scenarios shown in Figure 10 are for global mitigation action starting from 2020, whereas the above-quoted cases are immediate action scenarios.As such, this makes it more challenging and costly to meet any given long-term target with CO2 alone as a result of lock-in to CO2-intensive infrastructure and technologies with delayed action, thereby increasing the benefit of including non-CO2 GHG mitigation.Figure 10 shows a subsection of the data from Figures 7-9, so as to demonstrate the change in total mitigation cost and median 2100 temperature change as the non-CO 2 GHG price (in $/tCO 2 e using GWP100) changes as a fraction of the CO 2 price (in $/tCO 2 ).This demonstrates first the significant additional cost of meeting the 2 • C target compared to the 2.5 and 4 • C targets, as well as the significant reduction in 2100 temperature change achievable by including non-CO 2 GHG options in the overall mitigation portfolio.This is most clearly illustrated with reference to the vertical dashed line around the 2.5 • C mark in Figure 10: this LTTG is achievable either at a cost of $48 trillion by focusing only on CO 2 mitigation, or alternatively at $17 trillion by including non-CO 2 GHG mitigation in the portfolio of options, a cost reduction of about 65%.This compares to the figures discussed in Section 2, in which Rao and Riahi [13] found an approximate halving of carbon price, Kurosawa [18] found an approximate 55% cost saving by 2100, and Lucas et al [15] found a 4-26% mitigation cost reduction by 2100, when achieving a 550 ppm CO 2 e stabilisation concentration using a multi-gas approach compared to a CO 2 -only approach.The greater percentage cost savings in this study are most likely to stem from the fact that the scenarios shown in Figure 10 are for global mitigation action starting from 2020, whereas the above-quoted cases are immediate action scenarios.As such, this makes it more challenging and costly to meet any given long-term target with CO 2 alone as a result of lock-in to CO 2 -intensive infrastructure and technologies with delayed action, thereby increasing the benefit of including non-CO 2 GHG mitigation.Figure 10 shows a subsection of the data from Figures 7-9, so as to demonstrate the change in total mitigation cost and median 2100 temperature change as the non-CO2 GHG price (in $/tCO2e using GWP100) changes as a fraction of the CO2 price (in $/tCO2).This demonstrates first the significant additional cost of meeting the 2 °C target compared to the 2.5 and 4 °C targets, as well as the significant reduction in 2100 temperature change achievable by including non-CO2 GHG options in the overall mitigation portfolio.This is most clearly illustrated with reference to the vertical dashed line around the 2.5 °C mark in Figure 10: this LTTG is achievable either at a cost of $48 trillion by focusing only on CO2 mitigation, or alternatively at $17 trillion by including non-CO2 GHG mitigation in the portfolio of options, a cost reduction of about 65%.This compares to the figures discussed in Section 2, in which Rao and Riahi [13] found an approximate halving of carbon price, Kurosawa [18] found an approximate 55% cost saving by 2100, and Lucas et al [15] found a 4-26% mitigation cost reduction by 2100, when achieving a 550 ppm CO2e stabilisation concentration using a multi-gas approach compared to a CO2-only approach.The greater percentage cost savings in this study are most likely to stem from the fact that the scenarios shown in Figure 10 are for global mitigation action starting from 2020, whereas the above-quoted cases are immediate action scenarios.As such, this makes it more challenging and costly to meet any given long-term target with CO2 alone as a result of lock-in to CO2-intensive infrastructure and technologies with delayed action, thereby increasing the benefit of including non-CO2 GHG mitigation.

Discussion
The 21st century cumulative CO 2 budgets estimated for the 2 • C, 2.5 • C and 4 • C long-term temperature goals achieve 2100 median temperature changes of 2.00, 2.45 and 3.88 • C, once mitigation of non-CO 2 GHGs is taken into account up to a CO 2 e price equal to the CO 2 price in the fossil fuel combustion and industrial process sectors.
Significant mitigation of non-CO 2 GHGs, particularly CH 4 , results from the transition from a fossil fuel intensive to low-carbon energy and industrial system.This is primarily because fugitive CH 4 emissions from oil, coal and gas extraction, transmission and distribution activities decline with total primary fossil fuel demand.Further, significant mitigation of the majority of non-CO 2 GHGs is available at relatively low CO 2 e prices, with the majority of options below $100/t CO 2 e (when calculated on a GWP100 basis).This means that the mitigation of non-CO 2 GHGs even to CO 2 e prices at a fraction of the fossil fuel and industrial CO 2 price yields significant reductions in 2100 median temperature change.The most cost-effective non-CO 2 mitigation options include:

•
For CH 4 , increased recycling and energy recovery of biodegradable solid waste instead of landfill, reduced leakage from gas pipelines in Russia and Eastern Europe, extended recovery of associated waste gas from gas and oil production, and farm-scale anaerobic digestion of manure on large pig farms;

•
For N 2 O, reduced emissions from nitric acid production through improved technologies and catalytic reduction, as well as optimised wastewater treatment practices and improved fertiliser application regimes in agriculture;

•
For F-gases, reduction of leakage of HFCs from refrigeration, as well as replacement of HFCs with low-GWP alternatives in refrigeration and air conditioning.
The total cumulative discounted cost over the period 2010-2100 (at a 5% discount rate) of limiting global average temperature change to 2.5 • C by 2100 is $48 trillion (about 1.6% of cumulative discounted GDP over the period 2010-2100) if only CO 2 from the fossil fuel and industrial sectors is targeted, whereas the cost falls to $17 trillion (0.6% of GDP) by including non-CO 2 GHG mitigation in the portfolio of options-a cost reduction of about 65%.
If non-CO 2 GHGs are mitigated to the same CO 2 e price level as for CO 2 from the fossil fuel and industrial emissions sectors, then there is significant abatement of all non-CO 2 GHGs up to this CO 2 e price, such that in the 2 • C scenario, by 2100 the fully mitigated level of non-CO 2 GHGs is just under 13 GtCO 2 e, compared to more than 39 GtCO 2 e in the unmitigated reference scenario.Of this approximate 27 GtCO 2 e reduction, 69% occurs through the direct mitigation of the non-CO 2 GHGs and 31% through the indirect mitigation (mostly of CH 4 ) that follows from CO 2 mitigation.For each non-CO 2 GHG (CH 4 , N 2 O, and aggregated F-gases) the absolute emissions in the reference and mitigation scenarios are within the ranges of the database of scenarios presented in the IPCC's fifth assessment report, although the CH 4 reference emissions are at the higher end of the range, reflecting the relatively high socio-economic growth path and industrial output growth over the 21st century that underlies the scenario projections in this study.Furthermore, the percentage emissions reductions of each non-CO 2 GHG resulting from both direct and indirect mitigation are also comparable to those in the fifth assessment report database.
For the most stringent mitigation scenario explored in this study (that aimed at achieving a 2 • C median temperature change in 2100), mitigation of non-CO 2 GHGs beyond the CO 2 carbon price does not yield significant reductions in 2100 temperature.This is because the marginal abatement cost curve for non-CO 2 GHGs is relatively steep at higher CO 2 e prices, so that little marginal mitigation happens as CO 2 e prices increase.This indicates that mitigation to achieve the Paris Agreement's "well below 2 • C" goal is likely to have to come through a contraction of the 21st century CO 2 cumulative budget (relative to the budget which achieves the 2 • C goal), rather than through increased action on non-CO 2 GHGs alone.
The analysis also indicates that a rising CO 2 e price could result in a more cost-efficient mitigation strategy than a constant CO 2 e price over time, where earlier action on short-lived gases such as CH 4 does not contribute to a significant impact on 2100 temperature change.Three factors must be accounted for in further considering this point: (i) the discount rate used and the value of early mitigation costs compared to later mitigation costs; (ii) the fact that early action on CH 4 mitigation could foster learning and cost reductions, making later mitigation cheaper, and (iii) the fact that short-term mitigation of CH 4 may be an important facet of avoiding any dangerous overshoots of temperature change.
Although the mitigation potentials and costs in the GAINS model take account of purely technical barriers to adoption on a regional basis, there are other barriers which are more difficult to account for e.g., behavioural or institutional.Such barriers may add to costs at the local level.On the other hand, the purely technical nature of the cost estimates also means not accounting for potential co-benefits of mitigation in terms of improved health and reduced agricultural damages from methane as an ozone precursor [27].In addition, a number of mitigation options associated with demand-side measures, notably human dietary changes, are not included.These could yield significant additional non-CO 2 emissions reductions [17,42].Finally, the analysis does not assume technological development and associated cost reductions in the non-CO 2 mitigation measures over time.Implementation of climate policies, which incentivise the wide-spread adoption of non-CO 2 abatement technology, are likely to drive the development of cheaper and more effective abatement technology as time and learning progress.
In summary, a number of major policy implications derive from this analysis.First, indirect mitigation of non-CO 2 gases from CO 2 mitigation strategies is significant, which reaffirms the policy imperative to focus on shifting from fossil-fuel intensive to low-carbon technologies.Second, direct mitigation of non-CO 2 gases is potentially very cost-effective as part of an overall mitigation strategy, with cost savings in the 2.5 • C scenario shown in this study even greater than those for similar scenarios in previous studies, owing to the increased difficulty of achieving this mitigation goal through CO 2 mitigation alone, as a result of significant global mitigation action starting from 2020 only.Each of the major cost-effective non-CO 2 GHG mitigation measures highlighted should therefore be fully incentivized through appropriate policy measures.Third, the steepness of the marginal abatement cost curve for non-CO 2 GHGs after relatively low-cost abatement (at typically below $100/tCO 2 e) has been taken up implies that increasingly stringent long-term temperature goals, such as the Paris Agreement's 1.5 • C aspiration, are likely to rely in a large part from further mitigation of CO 2 .However, an important caveat is that this study does not include demand side changes such as dietary changes, which could yield further cost-effective mitigation of non-CO 2 GHGs.A fourth implication for policy makers is that it will be critical to weigh up the relative costs of early mitigation of short-lived non-CO 2 GHGs, which may not yield benefits in terms of longer-term warming (assuming they are eventually mitigated) but which could drive cost reductions in non-CO 2 mitigation technologies and measures through learning, whilst achieving important near-term reductions in warming.
This analysis, combined with the fact that non-CO 2 GHG mitigation options, particularly on the demand side, remain relatively less well explored compared to CO 2 options, highlights the importance in undertaking further research into the drivers, barriers and costs of mitigating these gases, so that policy makers can understand the trade-offs between early, gradual and delayed adoption of non-CO 2 mitigation measures.

Appendix B. Deriving Temperature Goal-Consistent 21st Century CO 2 Budgets and Emissions Profiles
The TIAM-Grantham [28][29][30] and IIASA GAINS [24,25] models are used to derive time profiles of emissions of CO 2 , CH 4 , N 2 O and total F-Gas emissions from a given cumulative CO 2 budget for fossil fuels and industry (FFI) in order to meet a given long-term temperature goal (LTTG)-the temperature change in 2100.In order to make climate projections (verifying the CO 2 budgets) the total F-Gas emissions must be broken down into constituent species and emissions of other gases must also be estimated.The process of constructing the full set of emissions required and the iterative process used to determine the 21st century (i.e., 2000-2100) CO 2 FFI budget is detailed here.A schematic of the information flow through the RCPs, TIAM-Grantham, GAINS and Met Office Hadley Centre (MOHC) calculations is illustrated in Figure A1.

1.
Projections of global temperature change for the four RCPs is made using emissions relating to the RCPs [48].Emissions are used rather than concentrations as this takes fuller account of uncertainty carbon cycle feedbacks.Following Bernie and Lowe [49], probabilistic projections are made using values of equilibrium climate sensitivity from models in the fifth Coupled Model Inter-comparison Project (CMIP5) [50] along with uncertainty distributions of ocean mixing and carbon cycle feedbacks.

2.
In each year land use emissions of CO 2 are linearly interpolated from the RCPs on the basis of each RCP's median 2100 projected temperature and the intended LTTG of the scenario.

3.
Initial estimates of 21st century cumulative CO 2 emissions from the FFI sectors are also linearly interpolated from the RCPs on the basis of future temperature projections and the new scenario's LTTG. 4.
The cumulative CO 2 FFI budget thus derived from the RCP projections is then used to calculate emissions of CO 2 from FFI, CH 4 , N 2 O and F-gases: a. time profile of CO 2 emissions from FFI is then calculated from the cumulative CO 2 FFI along with a carbon price profile; b.
The CO 2 FFI emissions profile and aspects of the underlying energy system structure (in particular the fossil fuel energy mix) are then passed to GAINS to calculate non-CO 2 GHG no-mitigation scenarios and corresponding MAC curves; c.
The CO 2 FFI profile from TIAM-Grantham and the non-CO 2 GHG no-mitigation scenarios and MAC curves from GAINS are then used to calculate the emissions of CH 4 , N 2 O and total F-Gas emissions, at different levels of CO 2 e price applied to the non-CO 2 GHGs (using GWP100 values).

5.
Individual F-gas emissions are then needed, but the constituent F-gases in the categories used by GAINS do not exactly match those used by MAGICC.Whilst this has a very small influence on the overall CO 2 e emissions, the individual gas species are needed by MAGICC.To estimate emissions of individual F-gases it is assumed that the relative emissions rate of each F-gas to the total F-gas emissions will change with time in line with the "unmitigated" RCP 8.5 scenario.
Based on this assumption the emissions of each F-gas in RCP 8.5 are scaled by a ratio of the total F-gas emissions from GAINS to the total F-gas emissions in the unmitigated reference scenario.So for example if the F-gas emissions from GAINS are 20% of the unmitigated F-gas emissions for that scenario, then this factor is applied to emissions of each individual F-gas from RCP 8.5.This approach circumvents the issue of different gases being included in the calculation by GAINS and those needed by MAGICC.While other assumptions are possible, given the relatively small effect of differences in F-gas emissions between the RCPs, this is an appropriate level of detail for the scope of the current study.

6.
The emissions of non-Kyoto GHG and other gases needed by MAGICC (principally NOx, CO, NMVOC, SO 2 ) are all based on the ratio of the emissions of each gas to the emissions of CO 2 from the FFI sector in the RCPs being applied to the CO 2 FFI emissions from TIAM-Grantham.For example if the CO 2 FFI emissions from GAINS in a given year where 80% of the way between RCP 4.5 and RCP 6.0, the SO 2 emissions would be the product of the CO 2 FFI from TIAM-Grantham multiplied by a weighted mean of the ratio of SO 2 to CO 2 FFI in those two RCPs, with four times more weight given to the ratio from RCP 6.0.7.
Projected median 2100 temperature change is then calculated and if within 0.1 • C of the original LTTG, the CO 2 FFI budget is accepted, or else the CO 2 budget for the scenario is re-estimated, before repeating the above procedure to re-calculate 2100 median temperature change.
the FFI sector in the RCPs being applied to the CO2 FFI emissions from TIAM-Grantham.For example if the CO2 FFI emissions from GAINS in a given year where 80% of the way between RCP 4.5 and RCP 6.0, the SO2 emissions would be the product of the CO2 FFI from TIAM-Grantham multiplied by a weighted mean of the ratio of SO2 to CO2 FFI in those two RCPs, with four times more weight given to the ratio from RCP 6.0.7. Projected median 2100 temperature change is then calculated and if within 0.1 °C of the original LTTG, the CO2 FFI budget is accepted, or else the CO2 budget for the scenario is re-estimated, before repeating the above procedure to re-calculate 2100 median temperature change.It should be noted again that the temperatures resulting from the emissions derived from a given budget are verified as meeting the target.With the cumulative CO2 FFI being the only variable here the process used in iterating its value for each target warming level is unimportant.However, the use of a simple interpolation of cumulative CO2 emissions to determine eventual warming is a notion that has become widely accepted in recent years [51][52][53].Its use here to initially estimate the CO2 budget for specific target warming levels implicitly assumes that the contribution of non-CO2 gases to warming is linearly related to the emissions of CO2.While this may appear to be broadly the case across the wide range of scenarios from the IPCC's AR5 WGIII report [54], the wide spread in IAM construction and the experimental design across the scenarios available is likely to obscure more subtle relations from IAM scenarios constructed under specific sets of assumptions on constraints.For example two scenarios with similar CO2 emissions profiles but which focus on either energy demand reduction or the heavy use of bio-energy with carbon capture and storage (BECCS) would likely have different non-CO2 contributions to warming.Similarly, emissions scenarios with different climate targets derived from a common approach, such as here, would not necessarily produce a robustly linear relation of warming to CO2 when the nuances of the underlying technological, economic and social assumptions and constraints are considered.
While the breakdown of the relation of cumulative emissions to temperature demonstrated by the need for iteration in developing these scenarios in small, it illustrates the inherent uncertainty in this relation and warrants careful verification of projections developed on this basis.It should be noted again that the temperatures resulting from the emissions derived from a given budget are verified as meeting the target.With the cumulative CO 2 FFI being the only variable here the process used in iterating its value for each target warming level is unimportant.However, the use of a simple interpolation of cumulative CO 2 emissions to determine eventual warming is a notion that has become widely accepted in recent years [51][52][53].Its use here to initially estimate the CO 2 budget for specific target warming levels implicitly assumes that the contribution of non-CO 2 gases to warming is linearly related to the emissions of CO 2 .While this may appear to be broadly the case across the wide range of scenarios from the IPCC's AR5 WGIII report [54], the wide spread in IAM construction and the experimental design across the scenarios available is likely to obscure more subtle relations from IAM scenarios constructed under specific sets of assumptions on constraints.For example two scenarios with similar CO 2 emissions profiles but which focus on either energy demand reduction or the heavy use of bio-energy with carbon capture and storage (BECCS) would likely have different non-CO 2 contributions to warming.Similarly, emissions scenarios with different climate targets derived from a common approach, such as here, would not necessarily produce a robustly linear relation of warming to CO 2 when the nuances of the underlying technological, economic and social assumptions and constraints are considered.The CO2 emissions level changes over time in the reference scenario, as a result of changes in global GDP, the primary energy (PE) intensity of global GDP, and the CO2 intensity of primary energy.This follows from a basic decomposition identity of CO2 emissions as shown in Equation (A1): Figure A3 shows the changes in each of the three factors on the right-hand side of Equation (A1).The Figure shows that whereas GDP increases non-linearly to 2100, primary energy intensity of GDP decreases (again non-linearly) over this period.CO2 intensity of primary energy remains approximately constant over the period to 2100, as a result of the fossil fuel share of primary energy remaining at just over 80% throughout the century.The approximately linearly increasing nature of CO2 emissions is therefore a purely coincidental result of non-linear changes in two of the three underlying components.The CO 2 emissions level changes over time in the reference scenario, as a result of changes in global GDP, the primary energy (PE) intensity of global GDP, and the CO 2 intensity of primary energy.This follows from a basic decomposition identity of CO 2 emissions as shown in Equation (A1): Figure A3 shows the changes in each of the three factors on the right-hand side of Equation (A1).The Figure shows that whereas GDP increases non-linearly to 2100, primary energy intensity of GDP decreases (again non-linearly) over this period.CO 2 intensity of primary energy remains approximately constant over the period to 2100, as a result of the fossil fuel share of primary energy remaining at just over 80% throughout the century.The approximately linearly increasing nature of CO 2 emissions is therefore a purely coincidental result of non-linear changes in two of the three underlying components.

Figure 1 .
Figure 1.Global greenhouse gas emissions in the (unmitigated) reference scenario.Notes:The approximately linear nature of these GHG emissions levels is primarily because of the approximately linear nature of the CO2 emissions level, which is coincidental only.Emissions values follow from detailed modelling of the energy and non-CO2 GHG emitting sectors in 10 year time-steps from the TIAM-Grantham and GAINS models used in this analysis.As shown in Appendix C which decomposes the CO2 emissions levels, the CO2 emissions pathway results from non-linearly increasing GDP, offset by non-linearly decreasing primary energy intensity of GDP, combined with approximately constant CO2 intensity of primary energy.

Figure 2 .
Figure 2. Non-CO2 GHG emissions in unmitigated reference scenario, with indirect savings resulting from fossil fuel and industry CO2 mitigation measures in 2 °C scenario with global mitigation action delayed to after 2020.Notes: Fluorinated gas (F-gas, which are Perfluorocarbons, PFCs, Sulphur hexafluoride, SF6, and Hydrofluorocarbons, HFCs) emissions presented here do not consider the effect of recent HFC phase-down under the Kigali Amendment of the Montreal Protocol from October 2016.

Figure 2 .
Figure 2. Non-CO 2 GHG emissions in unmitigated reference scenario, with indirect savings resulting from fossil fuel and industry CO 2 mitigation measures in 2 • C scenario with global mitigation action delayed to after 2020.Notes: Fluorinated gas (F-gas, which are Perfluorocarbons, PFCs, Sulphur hexafluoride, SF 6 , and Hydrofluorocarbons, HFCs) emissions presented here do not consider the effect of recent HFC phase-down under the Kigali Amendment of the Montreal Protocol from October 2016.

Figure 3 .
Figure 3. Non-CO2 GHG emissions savings (relative to unmitigated reference scenario) from CO2 mitigation measures, in a range of scenarios targeting different long-term temperatures with global mitigation action delayed to 2020.Notes: Percentage figures show the share of fossil fuels in total primary energy supply in 2010 and 2100 for each scenario.

Figure 3 .
Figure 3. Non-CO 2 GHG emissions savings (relative to unmitigated reference scenario) from CO 2 mitigation measures, in a range of scenarios targeting different long-term temperatures with global mitigation action delayed to 2020.Notes: Percentage figures show the share of fossil fuels in total primary energy supply in 2010 and 2100 for each scenario.
Figure4shows the further mitigation of non-CO 2 GHGs resulting from mitigation measures targeted specifically towards these gases, for the 2 • C scenario with delayed action to 2020.Also shown are the levels of each non-CO 2 GHG for the indirectly mitigated case.The figure shows for each time step the mitigation of non-CO 2 GHGs up to the CO 2 e price that is equal to the CO 2 price in the TIAM-Grantham model (i.e., the shadow price of CO 2 associated with achieving the least cost mitigation pathway to meet the specified 21st century cumulative CO 2 budget).As such, this equates a level of mitigation effort for CO 2 and non-CO 2 GHGs according to the marginal cost of abatement at any given time.In the case of Figure4, this marginal abatement cost is calculated on a GWP100 basis.It can be seen that there is significant abatement of all non-CO 2 GHGs up to this CO 2 e price, such that by 2100 the fully mitigated level of non-CO 2 GHGs is just under 13 GtCO 2 e/year, compared to 39 GtCO 2 e/year in the unmitigated reference scenario.Of the 27 GtCO 2 e/year reduction, 69% occurs through the direct mitigation of the non-CO 2 GHGs and 31% through the indirect mitigation (mostly of CH 4 ) that follows from CO 2 mitigation.Of the unmitigated reference 2100 level of each non-CO 2 GHG, 67% of CH 4 , 37% of N 2 O and 99% of F-gases are mitigated, leaving 7.8, 4.5 and 0.1 GtCO 2 e/year of CH 4 , N 2 O and F-gases respectively.These reductions compare to recent modelled scenarios (focusing specifically on the issue of non-CO 2 GHG mitigation) in which by 2100 up to 71% of CH 4 , 42% of N 2 O and 90% of F-gases are mitigated[20], as well as the broader IPCC fifth assessment report database[38] in which, across all of the most stringent mitigation scenarios, 44-74% of CH 4 , 9-42% of N 2 O and 45-90% of F-gases are mitigated by 2100, compared to the relevant unmitigated baseline scenario for each model used.In this database, the range of 2100 CH 4 emissions is 12-25 GtCO 2 e/year (using the most current CH 4 GWP100 value of 34) in the reference scenario and 4-11 GtCO 2 e/year in the mitigation scenarios, compared to 25 GtCO 2 e/year and 7.8 GtCO 2 e/year in the 2 • C scenario of this study.The database's range of 2100 N 2 O emissions is 3.0-8.8GtCO 2 e/year in the reference and 2.1-8.1 GtCO 2 e/year in the mitigation scenarios, compared to 7.0 and 4.5 GtCO 2 e/year in this

Figure 4 .
Figure 4. Non-CO2 GHG direct emissions savings (relative to baseline) as a result of applying a CO2e price equal to the fossil fuel and industry CO2 price, 2 °C scenario with action delayed to 2020.Notes: From 2030 onwards, F-gas emissions are around 0.1 GtCO2e per year.

Figure 4 .
Figure 4. Non-CO 2 GHG direct emissions savings (relative to baseline) as a result of applying a CO 2 e price equal to the fossil fuel and industry CO 2 price, 2 • C scenario with action delayed to 2020.Notes: From 2030 onwards, F-gas emissions are around 0.1 GtCO 2 e per year.

Energies 2017, 10 , 602 9 of 24 Figure 5 .
Figure5.Time-dependent global marginal abatement cost curves for the total non-CO2 GHGs (GWP100 basis) for 2 °C scenario with global mitigation action starting in 2020, relative to the case where indirect non-CO2 GHG mitigation resulting from CO2 mitigation has already occurred.Notes: Measures with a negative marginal abatement cost are assumed to be cost-saving if adopted.They are not assumed to be taken up in the reference scenario, as a result of uncertainty, inertia and inefficiency in current practices.

Figure 5 .
Figure5.Time-dependent global marginal abatement cost curves for the total non-CO 2 GHGs (GWP100 basis) for 2 • C scenario with global mitigation action starting in 2020, relative to the case where indirect non-CO 2 GHG mitigation resulting from CO 2 mitigation has already occurred.Notes: Measures with a negative marginal abatement cost are assumed to be cost-saving if adopted.They are not assumed to be taken up in the reference scenario, as a result of uncertainty, inertia and inefficiency in current practices.

Energies 2017, 10 , 602 9 of 24 Figure 5 .
Figure5.Time-dependent global marginal abatement cost curves for the total non-CO2 GHGs (GWP100 basis) for 2 °C scenario with global mitigation action starting in 2020, relative to the case where indirect non-CO2 GHG mitigation resulting from CO2 mitigation has already occurred.Notes: Measures with a negative marginal abatement cost are assumed to be cost-saving if adopted.They are not assumed to be taken up in the reference scenario, as a result of uncertainty, inertia and inefficiency in current practices.

Figure 6 .
Figure6.Global marginal abatement cost curves in 2050 for the total non-CO 2 GHGs (GWP100 basis) for different LTTGs, relative to the case where indirect non-CO 2 GHG mitigation resulting from CO 2 mitigation has already occurred.Notes: Measures with a negative marginal abatement cost are assumed to be cost-saving if adopted.They are not assumed to be taken up in the reference scenario, as a result of uncertainty, inertia and inefficiency in current practices.

Figure 7 .
Figure 7. Cost of meeting 2100 temperature change levels with non-CO2 GHG mitigation at a range of CO2e prices relative to CO2 mitigation, for the 2 °C scenario with delayed action to 2020.Notes: Figures in parenthesis on Y-axis show costs as a share of cumulative 2000-2100 discounted GDP (at 5% per year discount rate); blue points on chart are for non-CO2 GHG prices which vary over time (as a fixed fraction of CO2 prices) whereas red points show time-invariant non-CO2 GHG prices.

Figure 7 .
Figure 7. Cost of meeting 2100 temperature change levels with non-CO 2 GHG mitigation at a range of CO 2 e prices relative to CO 2 mitigation, for the 2 • C scenario with delayed action to 2020.Notes: Figures in parenthesis on Y-axis show costs as a share of cumulative 2000-2100 discounted GDP (at 5% per year discount rate); blue points on chart are for non-CO 2 GHG prices which vary over time (as a fixed fraction of CO 2 prices) whereas red points show time-invariant non-CO 2 GHG prices.

Figure 8 .
Figure 8. Cost of meeting 2100 temperature change levels with non-CO2 GHG mitigation at a range of CO2e prices relative to CO2 mitigation, 2.5 °C scenario with delayed action to 2020.Notes: Figures in parenthesis on Y-axis show costs as a share of cumulative 2000-2100 discounted GDP (at 5% per year discount rate).

2 120%Figure 8 .
Figure 8. Cost of meeting 2100 temperature change levels with non-CO 2 GHG mitigation at a range of CO 2 e prices relative to CO 2 mitigation, 2.5 • C scenario with delayed action to 2020.Notes: Figures in parenthesis on Y-axis show costs as a share of cumulative 2000-2100 discounted GDP (at 5% per year discount rate).

Figure 9 .
Figure 9. Cost of meeting 2100 temperature change levels with non-CO2 GHG mitigation at a range of CO2e prices relative to CO2 mitigation, 4 °C scenario with delayed action to 2020.Notes: Figures in parenthesis on Y-axis show costs as a share of total cumulative 2000-2100 discounted GDP (at 5% per year discount rate).

Figure 10 . 9 .
Figure 10.Costs of achieving different long-term temperature goals with varying degrees of non-CO2 mitigation (in terms of CO2e prices as a fraction of CO2 prices).

Energies 2017, 10 , 602 13 of 24 Figure 9 .
Figure 9. Cost of meeting 2100 temperature change levels with non-CO2 GHG mitigation at a range of CO2e prices relative to CO2 mitigation, 4 °C scenario with delayed action to 2020.Notes: Figures in parenthesis on Y-axis show costs as a share of total cumulative 2000-2100 discounted GDP (at 5% per year discount rate).

Figure 10 .
Figure 10.Costs of achieving different long-term temperature goals with varying degrees of non-CO2 mitigation (in terms of CO2e prices as a fraction of CO2 prices).

Figure 10 .
Figure 10.Costs of achieving different long-term temperature goals with varying degrees of non-CO 2 mitigation (in terms of CO 2 e prices as a fraction of CO 2 prices).
FigureA2shows the emissions levels of each GHG in the reference scenario as shown in Figure1, indexed to its 2010 value.This figure demonstrates that only CO 2 emissions increase approximately linearly over the period 2010-2100.

Figure A2 .
Figure A2.Increase in each GHG in the (unmitigated) reference scenario, indexed to its starting value in 2010 (which is given a value of 1).

Figure A2 .
Figure A2.Increase in each GHG in the (unmitigated) reference scenario, indexed to its starting value in 2010 (which is given a value of 1).

Figure A3 .
Figure A3.CO2 emissions in the reference scenario, and decomposed to show the change in each major component over the period 2010-2100.

Figure A3 .
Figure A3.CO 2 emissions in the reference scenario, and decomposed to show the change in each major component over the period 2010-2100.
• C are assessed) a cumulative 2000-2100 global CO 2 budget for the fossil fuel and industrial (FFI) sectors has been estimated from a simple interpolation of the budget from the Representative Concentration Pathways (RCPs) and projections of their corresponding global temperature change when simulated with a probabilistic version of the Model for Greenhouse gas Induced Climate Change [32]ICC) (as detailed in[31]) using a distribution of equilibrium climate sensitivity from the Fifth Coupled Model inter-comparison Project (CMIP5), as detailed in[32];

Table 1 .
Estimates of 2000-2100 cumulative CO 2 from fossil fuel combustion and industry sectors, with associated calculated 2100 median temperature change.

Table 1 .
Estimates of 2000-2100 cumulative CO2 from fossil fuel combustion and industry sectors, with associated calculated 2100 median temperature change.

Table 2 .
Major non-CO 2 GHG mitigation measures in different cost ranges.

Cumulative discounted mitigation cost, 2010-2100, (2005)$trillion 2100 median temperature change, O C
Schematic illustrating the process used to derive emissions scenarios from CO 2 budgets and iterate for target temperature levels where appropriate.RCP: Representative Concentration Pathway; GHG: greenhouse gas; FFI: fossil fuels and industry; MAC: marginal abatement cost; MOHC: Met Office Hadley Centre; NMVOC: non-methane volatile organic compounds; and MAGICC: Model for Greenhouse gas Induced Climate Change.