The method section is structured in four parts: 1) A description of the system boundaries and the coverage of global NCGG emissions, 2) An approach to construct the MACs (provided in more detail in supplement S2), 3) The development of the “optimistic, default and pessimistic” MACs and 4) A description of the scenario analysis.
System boundaries
The MAC curves and scenario assessment in this study are based on the emission source categories of the IMAGE 3.2 model [2, 3], representing all anthropogenic NCGGs. The MAC curves in this study cover 92% of the present-day NCGG emissions and 96% of the projected emissions in 2100 (see supplement S1). The MAC curves represent potential emission reductions under CO2 equivalent prices up to 4000 $(2005)/tCeq (or 1446 $(2020)/tCO2eq.), the maximum price that is applied in the IMAGE IAM framework. Emissions and emission reductions are calculated for the 26 global IMAGE regions (see supplement S3). Regional differences in present-day emission intensities and activities are fully represented in the scenario assessment. Regional emissions in the base year (2015 to 2020, depending on the source) are calibrated with data from several detailed databases covering different emissions sources; CEDS [24], GAINS [23], EDGAR 4.2.3 [25], [26].
Construction of the MAC curves
The MACs are built up from individual source-specific measures and assumptions on long-term developments (See supplement S2 for a more detailed description). The relative reduction potential (RP) (in %) of each mitigation measure in year t and region r is determined by Eq. 1. The maximum reduction potential (MRP) (in %) is the maximum relative abatement compared to baseline source emissions when all source-specific measures are implemented (Eq. 2).
With (all in %): TA: Technical applicability, this is the part of the baseline that can technically be covered by the measure. This is often 100% but can be lower, e.g., if only a sub-process is targeted or if regional climatic circumstances are unsuitable. RE: Reduction efficiency, i.e., the relative reduction in case a measure can be applied, generally based on multiple case studies. IP: Implementation potential represents (the lack of) non-technical barriers. This is assumed to increase in time due to improved technology diffusion and policy acceptance. OVcorr: Correction for overlap between measures that target the same emissions. If a subsequent measure is applied, it has a diminished benefit due to lower remaining emissions. Note that this correction increases with time as IP increases (based on [27], see S2). TP: Technological progress, increase of the reduction potential with time as a result of new or improved technologies. This is the only factor that is larger than 100% (see S2). Bcorr: Correction for regional emission reductions that already occur in the baseline scenario, e.g., due to zero or negative cost measures, such as the use of fugitive CH4 emissions as an energy source, or non-climate policy reductions, such as from air quality measures.
Marginal costs
The combination of measures with the highest estimated maximum reduction potential is used to construct MAC curves. It is assumed that the least costly measures are implemented first. When multiple measures are used, mitigation costs increase due to diminishing returns when measures overlap, with for any measure x:
Cost newx = Cost oldx * 1/OVcorrx (3)
Regional differences
Regional differences in mitigation potential are included if these are known. These differences are reflected in the parameters: technical applicability, reduction efficiency, and costs. Partly, these are due to socio-economic circumstances (e.g., different present-day emission intensities and different levels of advancements in farming techniques) that can have short-term implications on mitigation potentials. However, in the case of similar biophysical circumstances across regions, we assume convergence in mitigation potentials (i.e., in minimum emission intensities) in the long term and at maximum carbon prices. Where differences in mitigation potentials are known to be caused by biophysical differences, such as regional temperature, precipitation, geography, etc., this has been taken into account in the form of quantitative constraints of the components underlying the MACs. In this study, we differentiated between regions with high, medium and low technical applicability for enteric fermentation and CH4 manure (see supplement S5), based on the GAINS model global CH4 mitigation potentials for livestock in 2030 and 2050 [22]. Regional differences in reduction efficiency are incorporated in the measure ‘anaerobic digestion’, which has different known impacts in warm and cold environments. Regional differences in costs are incorporated based on region-specific cost assessments (see tables S5.2 and S5.3).
Emission categories
The MACs for the agricultural emission sources (CH4 from rice production, CH4 from enteric fermentation in ruminants, CH4 and N2O from manure, and N2O from fertilizer) have been constructed fully bottom-up, using the described methodology, as was also used in [1]. Here, we have updated the agricultural MAC curves by including (mostly) reduction efficiency data from ±120 recent studies. For the Monte Carlo analysis, ranges have been defined for all underlying MAC components (see section 2.3).
All non-agricultural sources are directly based on [1], with only a few, minor modifications to the default values for the maximum reduction potentials (MRPs). For the development of the pessimistic and optimistic MACs, MRP ranges have been varied, based on literature (see supplement S6). Waste and industry MACs (CH4 from landfills/solid waste, CH4 from sewage and wastewater, N2O from adipic and nitric acid production, N2O from transport, and N2O from domestic sewage), are based on data up to 2030 [28-30] but have added assumptions on the technological progress up to 2100, largely based on current best practices [1]. Fossil energy MACs (CH4 from coal, oil and gas production) are based on a dataset from the GAINS model [23, 31] with added long-term (MRP) assumptions on including promising technologies that are currently not in use on a large scale. The default F-gas MACs (HFCs, PFCs and SF6) are directly used from [1], including recent calibrations by [26] and [32].
MAC uncertainty range
Agriculture: Monte Carlo analysis
The uncertainty analysis for agricultural sources is based on a Monte Carlo (MC) analysis where the underlying parameters have been randomly varied and subsequently run 1000 times. The outcome of the MC analysis is a range in relative reductions at all carbon eq. prices between zero and 4000$/tC. The pessimistic, default and optimistic MACs are based on the 5th, 50th and 95th percentile in reductions for each carbon price, respectively.
Each MAC component value within a range is given equal weight (i.e., uniform distribution) (see supplement S5 for the input values, assumptions and motivation). The minimum and maximum for the reduction efficiency (RE) component are based on case studies found in the literature. For each measure, the highest and lowest outliers were excluded to prevent the distribution from being skewed. The minimum and maximum of the distributions of the other MAC components are based on a delta value (all in ±%points, since uncertainty is expected to be equally large at high and low values, except for costs, which is given in US$ and where absolute uncertainty is expected to be proportional to values) around the default component value (unless new information was available, this was based on ref. [1]. The default delta values are (in ±%points): TA(40), OVcorr(30), IP(30), TP(10) (note, this applies to the “diff” term, explained in S1) and (in ±%): Cost(80). The cost delta value is large because of particularly large uncertainty. The values of all components can never be lower than 0 and higher than 100%. Where found relevant, based on existing literature, the sampling was constrained by technical limits (e.g., a TA value is never allowed to be higher than 70% if it is known that 30% of the baseline emissions cannot be reduced by a certain measure).
Non-agricultural sources: range in maximum reduction potentials
The optimistic, default and pessimistic MACs for the non-agricultural sources have been developed by varying the maximum reduction potentials (MRPs) in 2050 and 2100 and scaling them in intermediate years. A full MC analysis is not possible for these sources, since most values of the underlying parameters are unknown, as the short-term MAC data is based on external databases. However, reduction potentials are generally higher, implying lower uncertainty and lower residual emissions in stringent climate scenarios [21]. The default MACs are largely equal to those developed by [1], with some small modifications (see supplement S6 for the quantitative assumptions by source). Where known, estimates of current technical reduction potentials (based on projections by GAINS and US-EPA [12, 22, 33]) were used as a minimum value for the pessimistic MACs.
Scenario analysis
The MAC curves have been used as an input to IMAGE 3.2 [2, 3] in conjunction with Shared Socio-economic Pathway (SSP) based scenario assumptions [34]. The scenarios are described in Table 1. The core set to assess the implications of the MAC uncertainty is based on SSP2, a scenario with middle-of-the-road socio-economic and technological development assumptions. In these scenarios, a 1.5- and 2-degrees Celsius target should be reached in 2100 (represented by 2.0 W/m2 and 2.6 W/m2 radiative forcing targets), under optimistic, default and pessimistic NCGG MAC assumptions (i.e., with low (L), medium (M) and high (H) reduction potentials, respectively). The mitigation scenario implications are compared to a no climate policy baseline (Base). Pre-2100 temperature overshoots are allowed.
In addition, the analysis includes two additional SSP narratives (in a 2-degree case) to assess the additional uncertainty due to human activities: SSP1 and SSP3, with low and high GHG-emitting activities, respectively. The underlying scenario assumptions for SSP1 and SSP3 are described in [35] with included updates [3]. SSP1 is combined with optimistic MAC assumptions (H) and SSP3 with pessimistic assumptions (L) to represent the extremes in NCGG emissions. The goal of the scenario analysis is to analyze the effect of MAC uncertainty and uncertainty in human NCGG emitting activities on:
- Feasibility of scenarios
- NCGG emission reductions (total and source-specific)
- Climate policy costs
- Remaining global carbon budgets, i.e., the need for CO2 mitigation
The scenarios used to assess uncertainty in GHG-emitting activities (2H_SSP1 and 2L_SSP3) have been used for the feasibility and carbon budget calculations only. Policy costs and NCGG reduction are not directly comparable due to different cost and baseline emission assumptions.
Table 1: Scenario setup
Scenario
|
NCGG MAC reduction potential
|
Human GHG-emitting activities
|
Radiative forcing target 2100 (W/m2)
|
Base
|
n.a.
|
Medium (SSP2)
|
n.a. *
|
2H
|
High / Optimistic
|
Medium (SSP2)
|
2.6
|
2M
|
Medium
|
Medium (SSP2)
|
2.6
|
2L
|
Low / Pessimistic
|
Medium (SSP2)
|
2.6
|
1.5H
|
High / Optimistic
|
Medium (SSP2)
|
2.0
|
1.5M
|
Medium
|
Medium (SSP2)
|
2.0
|
1.5L **
|
Low / Pessimistic
|
Medium (SSP2)
|
2.0
|
|
|
|
|
2H_SSP1
|
High / Optimistic
|
Low (SSP1)
|
2.6
|
2L_SSP3 **
|
Low / Pessimistic
|
High (SSP3)
|
2.6
|
* No target set. Default SSP2 baseline settings lead to a forcing level of 6.0 – 6.2 W/m2
** Infeasible scenarios (see Results)