The AGTP is an emission-metric-based emulator of the climate response and a well-established method that enables us to quantify and compare global temperature impacts of a large number of sources and scenarios in a transparent and, in terms of computer resources, cost-effective manner. The approach is described in detail in the literature (Fuglestvedt et al., 2010; Shine et al., 2005; Aamaas et al., 2013); here we give a brief outline.

The AGTP gives the global-mean surface temperature response per kilogram of species emitted as a function of time after an emission pulse, i.e., an instantaneous one-off emission. At time H after the emission, the AGTP for species i is given (for each sector and region) by 1AGTPiH=∫t=0HFitIRFTH-tdt, where Fi is the radiative efficiency. Emissions of SLCFs can have both direct and indirect radiative effects. For BC, OC, and SO2 we account for the direct, semi-direct, and indirect radiative forcing (RF) as described below. AGTPs for NOx, CO, and VOCs include the forcing due to tropospheric ozone production and (for NOx) nitrate aerosol formation, as well as the longer-term effect on methane lifetime and methane-induced ozone loss. The AGTP for methane includes the direct forcing, as well as the effect of OH-induced changes in its lifetime and effects on tropospheric ozone and stratospheric water vapor. See Aamaas et al. (2013) for details and analytical expressions for the AGTP of individual species.

For CO2 and methane, we calculate the global-mean F for year 2014 global concentrations (i.e., the year that is considered present-day in our emissions data; see below) using the equations from Etminan et al. (2016). Compared to the approach used by the IPCC Fifth Assessment report (AR5) (Myhre et al., 2013), this increases the radiative efficiency of methane by 14 %. For NH3, we use the IPCC AR5 best estimate for global mean radiative efficiency for all regions. For the remaining short-lived species, we derive values of Fi that depend on the location of the emission and calculate region-specific AGTPs for BC, OC, SO2, and the ozone precursors. The regional radiative efficiencies (i.e., the global radiative forcing per unit of regional emissions) for BC, OC, sulfate, nitrate, and ozone (in response to NOx, CO and VOC) are derived using radiative kernels (Samset and Myhre, 2011) and atmospheric concentrations from simulations performed with the global chemistry transport model OsloCTM3 (Søvde et al., 2012) for the second phase of the Hemispheric Transport of Air Pollution (HTAP2) (Janssens-Maenhout et al., 2015). Details about the chemistry and aerosol parameterizations and properties can be found in Lund et al. (2018). In addition to their direct radiative effects, aerosols also affect the energy balance through modifications of clouds and atmospheric heating rates (indirect and semi-direct effects). To account for the additional negative RF resulting from aerosol–cloud interactions, we scale the AGTP of SO2 by a factor of 2.1 based on the ratio of total global RF of sulfate to that due to direct effects alone from the IPCC AR5 (Myhre et al., 2013). Due to lack of available information about geographical dependence of the radiative efficiency, the same scaling factor is applied for all regions, recognizing that this is a simplification as the indirect effect also likely varies with location of emission. We also account for the semi-direct effect of BC (i.e., the rapid adjustments of the atmosphere to the local heating (Smith et al., 2018). Here we use the multi-model ratio between semi-direct and direct BC RF from Stjern et al. (2017) and calculate an average adjustment factor for the rapid adjustments of -15 %. This is then applied to the AGTP of BC for all regions except South Africa, where Stjern et al. (2017) found a small positive forcing from rapid adjustments. Radiative forcing of BC deposition on snow and ice is not included in our estimates.

IRFT in (Eq. 1) is the impulse response function used to estimate the temperature response to a given radiative forcing: 2IRFTt=λ∑j=1Jcjdjexp-tdj, where cj and dj are constants and timescales of the fast and slow model of the climate system response, respectively, and λ is the equilibrium climate sensitivity (ECS). An IRF is also used to represent the atmospheric decay of CO2. Several different IRFs exist in the literature. Here we use the IRFT from Geoffroy et al. (2013) (G13) and the IRFCO2 from Joos et al. (2013). Values of cj, dj, and λ derived from the analytical solution of the two-layer energy balance model used by G13 are given in Table 1. Compared to the IRFT from Boucher and Reddy (2008; B&R08) used in the bulk of previous metrics studies including IPCC AR5, G13 has shorter timescales and yields a lower ECS (0.885 K (W m-2)-1) compared to 1.06 K (W m-2)-1) from B&R08. To place our values in the context of previous literature and explore sensitivities to the choice of IRFs, we perform additional calculations using different combinations of IRFT and IRFCO2; see Sect. 1 of the Supplement.

Finally, we consistently account for the climate–carbon feedback (CCf) in the AGTPs. The IRFCO2, derived from complex models, implicitly includes the CCf. However, this is not the case for other components. This inconsistency was first highlighted in Myhre et al. (2013), where a first attempt to include the CCf was made for halocarbons based on an earlier study by Collins et al. (2013). This method has since been refined. Here we use the framework developed by Gasser et al. (2017) where a separate IRF for the CCf was derived using the simple Earth system model OSCARv2.2. This IRF is used to calculate a Δ AGTPi(H), which is then added to the AGTPi(H) without CCf. The difference between this method and the approach taken by Myhre et al. (2013) is discussed in Gasser et al. (2017). We also perform a sensitivity test to quantify the impact on our estimated temperature responses of excluding the CCf; see Sect. 4.1. Furthermore, as different methods to account for the CCf exist in the literature, we provide both sets of AGTPs for further use (see “Data Availability”).

As described above, we investigate the role and global temperature impacts of SLCF and CO2 from two different perspectives. First, the AGTPs at two given time horizons H (here 10 and 100 years) are multiplied by year 2014 emissions from the Community Emission Data System (CEDS) (Hoesly et al., 2018) for each species, sector, and region. The result is the near- and long-term global temperature response, ΔTi(H), to present-day regional and sectoral emissions.

Next, we quantify the temperature response to temporally evolving emissions from 1900 to 2100. The AGTP framework can readily be extended from pulse-based calculations since any scenario can be viewed as a series of pulse emissions and analyzed through convolution (Aamaas et al., 2013). The temperature response ΔT at time t for species i is (for each region and sector) given by 3ΔTit=∫0tEit′AGTPit-t′dt′. Importantly, the AGTPs are linear in that they do not account for the potential changes in radiative efficiency with changing background pollution levels; see Sect. 4 for further discussion.

Historical emissions are from the CEDS database, while future emissions follow the SSP–RCP scenarios. Gridded and harmonized emissions are available via ESFG from the Integrated Assessment Modeling Community (IAMC) for the nine SSP–RCP combinations that form the core of the Coupled Model Intercomparison Project Phase 6 (CMIP6) experiments (Gidden et al., 2019): SSP1-1.9, SSP1-2.6, SSP2-4.5, SSP3-7.0, SSP3-lowNTCF, SSP4-3.4, SSP4-6.0, SSP5-3.4, and SSP5-8.5. The gridded SSP–RCP data product, including the methodology for country- and sector-level emission mapping, is documented by Feng et al. (2020). We extract regional emission scenarios using the geographical definitions and spatial mask from HTAP2 (Janssens-Maenhout et al., 2015). Furthermore, we consider the energy (ENE), agriculture (AGR), waste (WST), residential (RES), industry plus solvents (IND), transport (TRA), and shipping (SHP) sectors, as they are defined in the CEDS-SSP inventory (Feng et al., 2020; Hoesly et al., 2018). Due to the large spread in historical estimates and lack of emissions consistent with CEDS, we do not include CO2 emissions due to land use and land cover change. Additionally, agricultural waste burning is excluded as these are more difficult to mitigate and estimates of future CO2 emissions are not available.

We first discuss the global mean surface temperature response to 1 year of present-day (i.e., year 2014) emissions for global total emissions and broken down by key contributing sectors and geographical source regions as shown in Fig. 2. While we here select 10- and 100-year time horizons to represent near- and long-term impacts, we recognize that other choices may affect the relative importance and even sign of the temperature response from some of the SLCFs, or that they may be more relevant for certain applications. For this reason, we also provide the full time series of our AGTPs (see Data Availability).

Globally, current emissions result in an approximate balance between cooling and warming SLCFs in the near-term, with main warming contributions from BC and CH4and cooling from SO2 and NOx (Fig. 2a). The total net effect after 10 years is therefore only slightly larger than that due to CO2 alone. As the impact of the SLCFs decays over years to decades upon emission, the total net temperature impact after 100 years is predominantly determined by CO2. As clearly seen in Fig. 2a, CO2 emissions also cause a notable contribution to near-term warming. While both of these features are well known in the scientific community, the role of CO2 as driver also of near-term warming is not always fully acknowledged in the discussions of LLGHGs versus SLCFs.

Differences in the mix of emissions result in net impacts on global temperature that vary significantly in both magnitude and sign between sectors and regions. Of the economic sectors, energy (ENE), agriculture (AGR), and waste management (WST) give the largest net near-term warming (i.e., after 10 years) (Fig. 2b). For AGR and WST, this is a result of strong methane-induced warming. The energy sector (ENE) is also characterized by a significant warming due to methane (originating from fossil fuel mining and distribution), as well as CO2, but also by a considerable cooling from high emissions of SO2. Our results hence reinforce the importance of methane as a driver of near-term warming but show that the net effect on global temperature of SLCF mitigation may be small in the case of the energy sector if simultaneous reductions in SO2 take place. A particular feature of the energy sector, however, is that a significant portion of methane mitigation from oil and gas (production and distribution) can be done independently from other energy-related (combustion) emissions. An explicit distinction between production and combustion emissions was not available in the gridded CEDS inventory, but, as illustrated in Sect. 3.2, mitigation strategies targeting one category or the other can result in distinctly different temperature outcomes. Global emissions from industry (IND) and shipping (SHP) cause a net cooling impact despite a considerable warming from CO2 emissions. In the long term, the net impact of AGR and WST is small, while energy is the largest individual contributor to warming due to its high CO2 emissions (note that N2O is not included in the present analysis as emissions are not included in the gridded CEDS and SSP database but would add a small contribution to the long-term impact of AGR). The second largest driver of long-term temperature change is IND, demonstrating the importance of non-CO2 emissions for shaping relative weight over different time frames. Aviation is not included here but was recently evaluated by Lund et al. (2017).

The largest regional contribution to net near-term warming is caused by emissions in East Asia (EAS) and North America (NAM), followed by Southeast Asia (SEA) and South Africa (SAF) (Fig. 2c). However, the relative contributions from individual species vary. In EAS and NAM, as well as Europe (EUR), the impact of current emissions of cooling and warming SLCFs approximately balance in the near-term, and these regions cause comparable net warming impacts on 10- and 100-year timescales, as seen by comparing the white and grey circles in Fig. 2c. These balancing characteristics do not imply that SLCF emissions should not be reduced but that the net benefits on global temperature may be low if mitigation measures that simultaneously affect both cooling and warming SLCFs are implemented, in turn placing added focus on the need to reduce CO2 in order to mitigate warming in both the near and long term. In SEA, SAF, and South and Central America (SAM and MCA) methane and BC emissions are presently high while emissions of CO2 and cooling aerosols are low compared to other regions, resulting in a net warming impact after 10 years that is substantially higher than that of CO2 alone. This, in turn, suggests that using SLCF emission reduction to limit near-term warming would be more effective here than in many other regions. Such detailed characteristics at the emission source level are needed for the design of effective mitigation strategies.

Breaking down the temperature impacts further into economic sectors within each region (not shown), we find that the results largely mirror the relative role of species and sectors on the global level in Fig. 2b. The warming contributions in South America and Africa, and hence higher potential for net temperature reductions, stem primarily from the agriculture, waste management, and energy production sectors. In SAF, mitigation of BC emissions from the residential and transport sectors also play an important role. In most regions, emissions from IND cause a net negative impact on global temperature change, while in the ENE sector impacts of cooling and warming SLCFs compete and warming from CO2 is a key driver of both near- and long-term warming.

Overall, the potential for global temperature reductions inherent in the present SLCF emissions is highly inhomogeneous, and co-emitted species – including CO2 – must be taken into account in any targeted climate policy for reduction of near-term warming. We emphasize that mitigation of SLCFs, while important, need to be sustained and complementary to strong cuts in CO2 for long-term reduction in global warming.

The results above suggest that strategies for emission reductions clearly can play out very differently in terms of net impact on global temperature across source region and sector. To further illustrate the importance of considering co-emissions and demonstrate the applicability of our dataset, we calculate the effect on global temperature in the near- and long-term following simplified examples of emission reduction packages in three of the global sectors (ENE, AGR, and SHP). The measures are broadly assumed to be motivated by either (i) air quality improvements (package 1, P1), (ii) methane reductions (as part of the SDG agenda or climate mitigation) (P2), or (iii) CO2 reductions and climate targets (P3). Table 2 shows the set of species reduced in each case, with the percentage reduction given in parentheses. We note that these reductions are based on expert judgment given underlying assumptions, e.g., for the reduction in shipping speed, and are associated with uncertainties. Furthermore, they are assumed to occur instantaneously. However, as the primary purpose here is illustrative, the examples are kept idealized and should be interpreted as such.

The global temperature effect resulting from elimination of emissions in each package on 10- and 100-year time horizons is shown in Fig. 3. The energy sector can be subdivided into fossil fuel production and distribution and combustion categories. An air-quality-driven set of measures (P1), e.g., end-of-pipe measures such as scrubbers, filters, and catalysts, could therefore be implemented that would strongly reduce SO2 and NOx emissions but not noticeably affect the key methane or CO2 contribution. Such measures are well understood, i.e., their efficiencies, costs, and technical implementation have been well documented and their real-life application is already widespread, but there is still large potential, especially in fast-growing economies. As shown by the top bar on the left in Fig. 3, the subsequent near-term temperature impact would be a warming contribution due to removal of cooling aerosols, adding to the already large net warming impact of the sector (Fig. 2b). As seen from the right-hand side of Fig. 3, the long-term effect would also be minor, leaving the dominating CO2 warming. A significant fraction of methane emissions, originating from the production and distribution of fossil fuels, could be mitigated separately from several other SLCFs, for instance by addressing venting and leaks from oil, gas and coal exploration, and upstream and downstream gas flaring. Respective measures would include capture, recovery, and use of gas, as well as reduced and improved flaring, with added benefits in terms of reduced CO2 and BC (P2). This results in a notable reduction in the near-term impact of the sector. Finally, P3 shows the impact of a dedicated climate strategy, here illustrated by the difference between a middle-of-the-road and a below-two-degrees scenario (in 2050, obtained from the GAINS model (Klimont et al., 2017), where more substantial CO2 mitigation also results in larger reduction of the sector's long-term temperature impact than in P2.

Due to the dominating contribution from methane to the temperature impact of the agriculture sector, measures that primarily target other emissions, such as improving nitrogen use efficiency (P1), unsurprisingly bring low net climate benefits unless accompanied by simultaneous measures for methane reductions (P2). An example of the latter is promoting dietary changes, leading to lower meat consumption and consequently lower livestock numbers. Reducing NH3 and NOx (P1) could, however, bring important local air quality benefits, and our results suggest that these would come with relatively small trade-offs from unmasking of aerosol cooling, at least in terms of global mean temperature on this timescale. Only small additional benefits (at a global scale) were estimated for the increased use of biogas (P3) based in utilization of livestock manures. The net impact of the shipping sector (SHP) is a cooling in the near-term, as shown in several previous studies (e.g., Berntsen and Fuglestvedt, 2008; Fuglestvedt et al., 2009). Measures that eliminate shipping emissions of SO2 (low sulfur fuels, scrubbers) and NOx (selective catalytic reduction) hence result in an added near-term warming also when simultaneous elimination of the sector's CO2 emissions occur (P2, P3).

This example is simplified and illustrative, and we calculate pulse-based temperature impacts following instantaneous emission reductions. However, since our pulse-based emission metrics can easily be used to study changes over time to any emission or policy scenario through convolution (Aamaas et al., 2013), our dataset has broad applicability. In the next section, we use precisely this method to quantify the impact of temporally evolving emissions according to the most recent scenarios.

While knowledge of the present-day emission composition and net temperature impact over time is essential to support mitigation design and implementation, real-world emissions will evolve following a combination of socioeconomic developments, technological advancement and policy adoption. Next, we investigate plausible pathways for the future impact of SLCFs and CO2 by quantifying the global temperature change over the period 1900–2100 to regional and sectoral emissions following the SSP–RCP scenarios. In the following paragraphs, we show results from four of the nine SSP–RCP scenarios used in the present analysis (SSP1-1.9, SSP2-4.5, SSP3-7.0 and SSP5-8.5). These span the range of future emission evolutions, but we recognize that the realism of SSP5-8.5 is debated in the literature due to its very high emissions (e.g., Ritchie and Dowlatabadi, 2017).

Figure 4 shows the evolution of temperature response under the SSP–RCPs for our source regions, with corresponding results for the global economic sectors given in Fig. S3. Our emissions regions have large differences not only in terms of present-day emissions but also of past evolution. This historical contribution, which was not captured in the analysis of the first half of the paper, brings NAM and EUR as the two largest contributors to the present-day warming (Fig. 4a) due to their much higher past CO2 emissions, in line with previous literature (Höhne et al., 2011; Skeie et al., 2017). While presently being the largest emission source, EAS only surpasses EUR and NAM in terms of contribution to temperature change between 2020 and 2030 when the cumulative effect of CO2 is accounted for. In SSP1-1.9, where emissions of CO2 decline strongly during the first half of the century in all regions, the net temperature response levels off or starts to decline in the second half of the century. In the remaining scenarios, the net temperature impact increases over the century for all regions. EAS remains the largest contributor, whereas in SSP5-8.5 South Asia (SAS) overtakes NAM as the second most important region by 2100 and SAF reaches the same order of magnitude as EUR. This shows a projected shift in emissions and increasing importance of the developing world. We note that since our primary focus here is on quantifying the contributions to, and potential for further reduction of, near- and long-term temperature impacts, we do not include negative CO2 emissions, which is already a mitigation measure. Furthermore, the gridded SSP–RCP inventory only provide negative CO2 as a separate category without information for mapping these emissions to economic sectors. We do, however, include the negative CO2 category in our inventory of regional scenarios for further analyses beyond our study (see “Data Availability”).

In our calculations, the net temperature response to emissions from the global energy (ENE) sector becomes larger than that due to AGR and RES in the early 2000s (Fig. S1a), after which ENE remains the largest individual sector until 2100 in all scenarios. The relative importance of AGR and ENE historically is yet another example of how including SLCFs can change relevance over different time frames, as also demonstrated by Reisinger and Clark (2018) for non-CO2 livestock emissions. In our results, both the warming due to CH4 from AGR and the contributions from cooling emissions from ENE act to shape the relative role of the two sectors over time. The global mean temperature impact of IND switches from a net cooling to a net warming in the late 20th century as the warming due to CO2 accumulates and overwhelms the cooling from SO2.

While the contribution from CO2 to the net warming becomes dominant by 2100 for most regions and sectors in all scenarios, the relative importance of SLCFs and CO2 continue to be highly variable across emission source over time, in particular under SSP3-7.0 and SSP5-8.5. This can be seen in Fig. 4b, where we break down the future net temperature response in 2030, 2050, and 2100 into individual contributions from methane, CO2, BC, and the sum of SO2 and NOx. Here we show a selection of the source regions that differ notably in composition and temporal trend. See Fig. S4 for remaining regions.

The SSP–RCPs differ in both climate forcing targets and stringency of air pollution control, as well as underlying socioeconomic development. SSP1-1.9 is characterized by low societal challenges to mitigation and adaptation, and strong climate and air quality policies, resulting in rapidly declining emissions of both SLCFs and CO2. However, even for strong air pollution there is a differentiation between high-, medium- and low-income countries, with a substantial time lag in the latter two (Rao et al., 2017). For example, emissions of SO2 in SAS and SAF decline less than in other regions, subsequently maintaining a significant cooling contribution to the temperature impact. In the intermediate scenario, SSP2-4.5, there is a reduction in emissions, but this is delayed and slower compared to SSP1-1.9. In SSP3-7.0, the world follows a path with more inequality and conflict, where only weak air pollution control is implemented and the end-of-century climate forcing, and hence CO2 emissions, is higher. Subsequently, emission trends and SLCF contributions display more regional heterogeneity. There is a particularly strong projected increase in methane emission in South Asia, Africa, and South America in this scenario. While previous decades have seen a southeastward shift in air pollution emissions, from high-income regions at northern latitudes to East Asia and South Asia, these findings suggest that a second shift may be underway, towards low- and middle-income countries in the developing world. Further studies are needed to improve the knowledge about the resulting climate and environmental consequences, as well as how to strengthen the mitigation options in these regions. While EAS remains the region with the largest warming impact by 2100 in all scenarios, the contributions to warming from methane and BC in SAF and SAS surpasses those of EAS in 2100 in both SSP3-7.0 and SSP5-8.5. The net temperature response to emissions in SAS increases from close to zero to a significant warming as CO2 emissions increase. SSP5-8.5 is characterized by high challenges to mitigation and high climate forcing in 2100 but still assumes strong air pollution control since the high use of fossil fuels would otherwise result in unbearable air pollution levels. Combined, this leads to increasing temperature impact due to increasing CO2 emissions with lower SLCF impacts than in SSP3-7.0 but a non-negligible contribution from methane for several regions. Hence, in medium- and low-income regions, SLCFs, and in particular methane, are projected to play a continued important role for future temperature change.

Clearly, and as expected, the largest difference in SLCF contributions to future temperature response is between SSP1-1.9 and SSP3-7.0. To see where the largest additional climatic benefit can be gained from mitigating SLCF emissions in line with SSP1-1.9, relative to SSP3-7.0, we show the difference in temperature between these two scenarios in 2030, 2050, and 2100 in Fig. 5. Results are shown by region and sector, for all combinations where the temperature difference is greater than ±0.01 ∘C. For comparison, the CMIP6 mean difference in projected surface temperature between SSP3-7.0 and SSP1-2.6 (which is close to SSP1-1.9 in emissions) is around 0.5 ∘C in 2050 and 2 ∘C in 2100 when accounting for all global emissions (Tokarska et al., 2020). As seen from Figs. 4 and S3, CO2 is the key driver of this long-term temperature difference between the scenarios for most sectors and regions. However, as seen in Fig. 5, there are also important SLCF contributions, most notably from the large sources of methane: agriculture, energy, and waste management. Furthermore, 9 of the 12 top contributions are from regions in Africa, South Asia, or South and Central America, again demonstrating the importance of the development in low- and middle-income countries for future levels of SLCFs. Figure 5 also shows how the strong SLCF mitigation in SSP1-1.9, relative to SSP3-7.0, can result in a net warming contribution to climate for some region–sector combinations, as exemplified by the industry sector in East Asia and South Asia. As shown by the data on the right-hand side of Fig. 5, for most sector–region combinations, around 10 % of the avoided (or added) warming from strong mitigation would already be realized by 2030, and around 40 %–50 % would be realized by 2050.

The AGTP is a well-established framework that has been applied in several studies of attribution of temperature impacts to emission sources and scenarios (e.g., Collins et al., 2013; Lund et al., 2017; Sand et al., 2015; Stohl et al., 2015; Aamaas et al., 2019). Here we have also consistently included the carbon–climate feedback in the AGTP for all species. This increases the non-CO2 AGTPs, however, less than initially suggested by Myhre et al. (2013) as discussed by Gasser et al. (2017). Figure S5 shows the global mean net temperature response to total emissions under six of the nine SSP–RCPs, with and without the feedback. By the end of century, there is a 5 %–9 % difference depending on scenario.

A key strength of the AGTP framework is that allows us to investigate the effects of individual species, sources, and scenarios, which would be confounded by the low signal-to-noise ratio in fully coupled models, in a transparent manner. However, there are also caveats. Importantly, the AGTP metric is linear, while in reality the radiative efficiency can have nonlinear dependencies on the background atmospheric conditions. In this study, we account for one part such nonlinearities by using radiative efficiencies for the aerosols and ozone precursors that vary with emission location to calculate region-specific AGTPs. The part of the nonlinearities caused by changing background levels of pollutants over time is, however, not included. For the well-mixed greenhouse gases CO2, CH4, and N2O, the radiative efficiency (RE) is reduced with increasing atmospheric background concentrations. Previous literature suggests that the sensitivity to emission scenario is small, and the relationship between emissions and temperature response more linear, for CO2 (Caldeira and Kasting, 1993). However, the same has not been shown for methane (and N2O, which is not considered here). We therefore perform an additional sensitivity test where we calculate an AGTP(t) that is adjusted to the global atmospheric concentrations over time (using the equation from Etminan et al. (2016) and global concentrations for each SSP–RCP from the IIASA SSP database (IIASA, 2020; Riahi et al., 2017). Figure S5 shows the resulting temperature response, compared to the temperature response calculated with and without the CCf. As expected, using a dynamically adjusted RE results in a lower warming in the high-emission scenarios and a slightly higher temperature response under low emissions. In the case of extreme scenario SSP5-8.5, the effect is of the same order of magnitude as that from adding the CCf but is of opposite sign. For aerosols and ozone precursors, potential saturation effects involve complex, spatially heterogeneous chemistry, cloud, and climate interactions that require detailed chemistry–climate simulations to be resolved, and even then may not be fully captured due to, e.g., the coarse resolution of current models. We emphasize that the absolute magnitude of temperature changes quantified with the AGTP framework should be interpreted with care, as this method is primarily designed to study relative importance and relationships between individual emissions and sources.

Our analysis reflects best-estimate input data to the extent possible, but results have considerable uncertainty in emissions, RF, and climate sensitivity. As shown in Fig. 2a, we estimate, due to uncertainty in RF alone, a 1 SD range in the total net temperature response on the 10-year time horizon of ±0.01 ∘C, about 38 % of the net temperature response of 0.03 ∘C (the range is considerably lower on the 100-year timescale as the RF of SLCFs is much more uncertain than that of CO2). Uncertainties in emission inventories are difficult to quantify, but are generally considered lowest for CO2 and SO2 emissions and high for carbonaceous aerosols (Hoesly et al., 2018). The level of uncertainty also differs across regions and sectors, with emissions from nature-related emissions (e.g., agriculture, landfills) more uncertain than emissions in the fossil fuel sector (Amann et al., 2013; Jonas et al., 2019). Moreover, recent studies point to emission trends that are not accurately represented in the global inventory, such as SO2 and NOx in China (Zheng et al., 2018) and fossil fuel CH4 emissions (Hmiel et al., 2020). However, due to high spatiotemporal variability and lack of consistent data, a comprehensive uncertainty analysis at the regional and sectoral level is challenging. The equilibrium climate sensitivity (ECS) inherent in the climate response in IRF used in the present analysis is 0.885 K (W m-2)-1. This is in the upper range reported by Bindoff et al. (2013) but lower than many recent estimates (Forster et al., 2019; Zelinka et al., 2020). While emission uncertainties can have a strong spatiotemporal character, changes in the ECS mostly act to scale estimates for all sectors and regions but is less important for their relative ranking.

Our analysis is limited to temperature change as a measure of climate impacts. SLCFs, and in particular aerosols, also play a key role in shaping local and regional hydrology and dynamics. Comparing the SSP3-7.0 and SSP3-lowNTCF scenarios, Allen et al. (2020) recently found a significant precipitation increase due to removal of aerosols, with the strongest moistening trends over Asia. An increase in the Asian summer monsoon precipitation in scenarios with strong air pollution reductions was also recently found by Wilcox et al. (2020). Hence, further studies using coupled models are needed to fully capture the effects of the SLCFs under SSPs on local climate and environment.