The global IAM community has developed a family of scenarios that describe a variety of possible socioeconomic futures (the SSPs). The formation, qualitative, and quantitative aspects of these scenarios have been discussed widely in the literature . We briefly summarize here relevant narratives of the baseline SSPs concerning socioeconomic development (see, e.g., Fig. ), energy systems , land use , greenhouse gas (GHG) emissions , and air pollution .

SSP1 and SSP5 describe worlds with strong economic growth via sustainable and fossil fuel pathways, respectively. In both scenarios, incomes increase substantially across the globe and inequality within and between countries is greatly reduced; however, this growth comes at the expense of potentially large impacts from climate change in the case of SSP5. Demand for energy- and resource-intensive agricultural commodities such as ruminant meat is significantly lower in SSP1 due to changes in behavior and advances in energy efficiency. In both scenarios, pollution controls are expanded in high-income economies with other nations catching up relatively quickly with the developed world, resulting in reductions in air pollutant emissions. SSP2 is a so-called middle-of-the-road scenario with moderate population growth and slower convergence of income levels across countries. In SSP2, food consumption, especially for resource-intensive livestock-based commodities, is expected to increase and energy generation continues to rely on fossil fuels at approximately the same rates as today, resulting in continued growth of GHG emissions. Efforts at curbing air pollution continue along current trajectories with developing economies ultimately catching up to high-income nations, resulting in an eventual decrease in pollutant emissions. Finally, SSP3 and SSP4 depict futures with high inequality between countries (i.e., “regional rivalry”) and within countries, respectively. Global gross domestic product (GDP) growth is low in both scenarios and concentrated in currently high-income nations, whereas population increase is focused in low- and middle-income countries. Energy systems in SSP3 see a resurgence of coal dependence, whereas reductions occur in SSP4 as the high-tech energy and economy sectors see increased developments and investments leading to higher diversification of technologies . Policy making (either regionally or internally) in areas including land-use regulation, air pollution control, and GHG emissions limits is less effective. Thus policies vary regionally in both SSPs with weak international institutions, resulting in the highest levels of pollutant and aerosol emissions and potential effect on climate outcomes .

A matrix of socioeconomic–climate scenarios relevant to the broad scientific community was created with SSPs on one axis and climate policy futures (i.e., mitigation scenarios) delineated by end-of-century (EOC) RF on the other axis (see Fig. ). The scenarios selected for inclusion in ScenarioMIP, shown in Table , are comprised of both baseline and mitigation cases, in which long-term climate policies are lacking or included, respectively. They are divided into Tier-1 scenarios, which span a wide range of uncertainty in future forcing and are utilized by other MIPs, and Tier-2 scenarios, which enable more detailed studies of the effect of mitigation and adaptation policies which fall between the Tier-1 forcing levels. Each scenario is run by a single model within ScenarioMIP, comprised of the AIM/CGE, GCAM4, IMAGE, MESSAGE-GLOBIOM, and REMIND-MAgPIE modeling teams. We provide a short discussion here on their selection and refer the reader to Sect. 3.2.2 for fuller discussion of the experimental design.

The Tier-1 scenarios include SSP1-2.6, SSP2-4.5, SSP3-7.0, and SSP5-8.5, designed to provide a full range of forcing targets similar in both magnitude and distribution to the RCPs as used in CMIP5. Each EOC forcing level is paired with a specific SSP, which is chosen based on the relevant experimental coverage. For example, SSP2 is chosen for the 4.5 Wm-2 experiment because of its high relevance as a reference scenario to IAV communities as a scenario with intermediate vulnerability and climate forcing and its median positioning of land use and aerosol emissions (of high importance for DAMIP and DCPP), whereas SSP3 is chosen for the 7.0 Wm-2 experiment as it allows for quantification of avoided impacts (e.g., relative to SSP2) and has significant emissions from near-term climate forcing (NTCF) species such as aerosols and methane (also referred to as short-lived climate forcers, or SLCF).

The Tier-2 scenarios include SSP1-1.9, SSP3-LowNTCF, SSP4-3.4, SSP4-6.0, and SSP5-3.4-Overshoot (OS), chosen to both complement and extend the types of scenarios available to climate modelers beyond those analyzed in CMIP5. SSP1-1.9 provides the lowest estimate of future forcing matching the most ambitious goals of the Paris Agreement (i.e., “pursuing efforts to limit the [global average] temperature increase to 1.5 ∘C above pre-industrial levels”). The SSP3-LowNTCF scenario provides an important experimental comparison to scenarios with high NTCFs for use in AerChemMIP contrasting with SSP3-7.0 (see Appendix  for more detail on differences in assumptions between SSP3-7.0 and SSP3-LowNTCF). Both SSP4 scenarios fill gaps in Tier-1 forcing pathways and allow investigations of impacts in scenarios with relatively strong land-use and aerosol climate effects but relatively low challenges to mitigation. Finally, SSP5-3.4-OS allows for the study of a scenario in which there is large overshoot in RF by mid-century followed by the implementation of substantive policy tools to limit warming in the latter half of the century. It is specifically designed to be twinned with SSP5-8.5, following the same pathway through 2040, and support experiments examining delayed climate action.

We construct a common dataset of historical emissions for the year 2015

For sulfur emissions in China, we include values up to 2017, due to a drastic reduction in these emissions in the most recently available datasets.

Emissions harmonization is defined as a procedure designed to match model results to a common set of historical emissions trajectories. The goal of this process is to match a specified base-year dataset while retaining consistency with the original model results to the best extent possible while also providing a smooth transition from historical trajectories. This non-disjoint transition is critical for global climate models when modeling projections of climate futures which depend on historical model runs, guaranteeing a smooth functional shape of both emissions and concentration fields between the historical and future runs. Models differ in their 2015 data points in part because the historical emissions datasets used to calibrate the models differ (e.g., PRIMAP – ; EDGAR – ; CEDS – ). Another cause of differences is that 2015 is a projection year for all of these models (the original scenarios were originally finalized in 2015).

Harmonization can be simple in cases where a model's historical data are similar to the harmonization dataset. However, when there are strong discrepancies between the two datasets, the choice of harmonization method is crucial for balancing the dual goals of accurate representation of model results and reasonable transitions from historical data to harmonized trajectories.

The quantity of trajectories requiring harmonization increases the complexity of the exercise. In this analysis, given the available sectoral representation of both the historical data and models, we harmonize model results for 14 individual emissions species and 13 sectors as described in Table . The majority of emissions–sector combinations are harmonized for every native model region

Further information regarding the model region definitions is available via the IAMC Wiki at https://www.iamcdocumentation.eu (last access: 8 April 2019) and .

Downscaling, defined here as distributing aggregated regional values to individual countries, is performed for all scenarios in order to improve the spatial resolution of emissions trajectories, and as a prelude to mapping to a spatial grid (discussed in Appendix ). We developed an automated downscaling routine that differentiates between two classes of sectoral emissions: those related to AFOLU and those related to fuel combustion and industrial and urban processes. In order to preserve as much of the original model detail as possible, the downscaling procedures here begin with harmonized emissions data at the level of native model regions and the aggregate sectors (Table ). Here we discuss key aspects of the downscaling methodology and refer the reader to the downscaling documentation (https://github.com/iiasa/emissions_downscaling/wiki, last access: 8 April 2019) for further details.

AFOLU emissions, including agricultural waste burning, agriculture, forest burning, peat burning, and grassland burning, are downscaled using a linear method. Linear downscaling means that the fraction of regional emissions in each country stays constant over time. Therefore, the total amount of open-burning emissions allocated to each country will vary over time as economies evolve into the future, following regional trends from the native IAM. However there is no subregional change in the spatial distribution of land-use related emissions over time. This is in contrast to other anthropogenic emissions, where the impact, population, affluence, and technology (IPAT) method is used to dynamically downscale to the country level as discussed above. Note that peat burning emissions were not modeled by the IAMs and are constant into the future.

All other emissions are downscaled using the IPAT method developed by , where population and GDP trajectories are taken from the SSP scenario specifications . The overall philosophy behind this method is to assume that emissions intensity values (i.e., the ratio of emissions to GDP) for countries within a region will converge from a base year, ti (2015 in this study), over the future. A convergence year, tf, is specified beyond 2100, the last year for the downscaled data, meaning that emissions intensities do not converge fully by 2100. The choice of convergence year reflects the rate at which economic and energy systems converge toward similar structures within each native model region. Accordingly, the SSP1 and SSP5 scenarios are assigned relatively near-term convergence years of 2125, while SSP3 and SSP4 scenarios are assigned 2200, and SSP2 is assigned an intermediate value of 2150.

The downscaling method first calculates an emissions intensity, I, for the base and convergence years using emissions level, E, and GDP. 1It=EtGDPt

An emissions intensity growth factor, β, is then determined for each country, “c”, within a model region, “R”, using convergence year emissions intensities, IR,tf, determined by extrapolating from the last 10 years (e.g., 2090 to 2100) of the scenario data. 2βc=IR,tfIc,ti1tf-ti

Using base-year data for each country and scenario data for each region, future downscaled emissions intensities and patterns of emissions are then generated for each subsequent time period. 3Ic,t=βcIc,t-1 4Ec,t*=Ic,tGDPc,t

These spatial patterns are then scaled with (i.e., normalized to) the model region data to guarantee consistency between the spatial resolutions, resulting in downscaled emissions for each country in each time period. 5Ec,t=ER,t∑c′∈REc′,t*Ec,t*

For certain countries and sectors the historical dataset has zero-valued emissions in the harmonization year. This would result in zero downscaled future emissions for all years. Zero emissions data occur largely for small countries, many of them small island nations. This could be due to either lack of actual activity in the base year or missing data on activity in those countries. In order to allow for future sectoral growth in such cases, we adopt, for purposes of the above calculations, an initial emissions intensity of one-third the value of the lowest country in the same model region. We then allocate future emissions in the same manner discussed above, which is consistent with our overall convergence assumptions. Note that we exclude the industrial sector (Table ) from this operation as it might not be reasonable to assume the development of substantial industrial activity in these countries.

Finally, some scenarios include negative CO2 emissions at some point in the future (notably from energy use). For CO2 emissions, therefore, we apply a linear rather than exponential function to allow a smooth transition to negative emissions values for both the emissions intensity growth factor and future emissions intensity calculations. In such cases, Eqs. () and () are replaced by Eqs. () and (), respectively. 6βc=(IR,tfIc,ti-1)1tf-ti 7Ic,t=(1+βc)Ic,t-1

The nine ScenarioMIP scenarios were selected to provide a robust experimental design space for future climate studies as well as IAV analyses with the broader context of CMIP6. Chief among the concerns in developing such a design space are both the range and spacing of the global climate response within the portfolio of scenarios . Prior work for the RCPs studied a range of climate outcomes between ∼2.6 and 8.5 Wm-2 at EOC. Furthermore, recent work finds that statistically significant regional temperature outcomes (>5 % of half the land surface area) are observable with a minimum separation of 0.3 ∘C, which is approximately equivalent to 0.75 Wm-2 . Given the current policy context, notably the recent adoption of the UN Paris Agreement, the primary design goal for the ScenarioMIP scenario selection is thus twofold: span a wider range of possible climate futures (1.9–8.5 Wm-2) in order to increase relevance to the global climate dialogue and provide a variety of scenarios between these upper and lower bounds such that they represent statistically significant climate variations in order to support a wide variety of CMIP6 analyses.

We find that the selected scenarios meet this broad goal, as shown in Fig. , by using the simple climate model MAGICC6 with central climate-system and gas-cycle parameter settings for all scenarios to calculate pathways of both RF and the resulting response of global mean temperature (see Appendix Table  for a listing of all EOC RF values).

We also present illustrative global mean temperature pathways. EOC temperature outcomes span a large range, from 1.4 ∘C at the lower end to 4.9 ∘C for SSP5-8.5, the scenario with the highest warming emissions trajectories. Notably, two scenarios (SSP1-1.9, which reaches 1.4 ∘C by EOC, and SSP1-2.6, reaching 1.7 ∘C) can be used for studies of global outcomes of the implementation of the UN Paris Agreement, which has a desired goal of “[h]olding the increase in the global average temperature to well below 2 ∘C above pre-industrial levels and pursuing efforts to limit the temperature increase to 1.5 ∘C above pre-industrial levels” Article 2.1(a). The difference between scenario temperature outcomes is statistically significant in nearly all cases, with a minimum difference of 0.37 ∘C (SSP1-1.9 and SSP1-2.6) and maximum value of 0.77 ∘C (SSP3-7.0 and SSP5-8.5). The EOC difference between SSP4-3.4 and SSP5-3.4-OS is not significant (0.07 ∘C); however global climate outcomes are likely sensitive to the dynamics of the forcing pathway .

A subset of four scenarios (SSP1-2.6, SSP2-4.5, SSP4-6.0, and SSP5-8.5) was also designed to provide continuity between CMIP5 and CMIP6 by providing similar forcing pathways to their RCP counterparts assessed in CMIP5. We find that this aspect of the scenario design space is also met by the relevant scenarios. SSP2-4.5 and SSP5-8.5 track RCP4.5 and RCP8.5 pathways nearly exactly. We observe slight deviations between SSP1-2.6 and RCP2.6 as well as SSP4-6.0 and RCP6.0 at mid-century due largely to increased methane emissions in the historic period (i.e., methane emissions broadly follow RCP8.5 trajectories after 2000, resulting in higher emissions in the harmonization year of this exercise; see Fig.  below).

The remaining five scenarios were chosen to “fill gaps” in the previous RCP studies in CMIP5 and enhance the potential policy relevance of CMIP6 MIP outputs . SSP3-7.0 was chosen to provide a scenario with relatively high vulnerability and land-use change with associated near-term climate forcing (NTCF) emissions resulting in a high RF pathway. We find that it reaches an EOC forcing target of ∼7.1 Wm-2 and greater than 4 ∘C mean global temperature increase. While contributions to RF from CO2 in SSP3-7.0 are lower than that of SSP5-8.5, methane and aerosol contributions are considerably higher (see, e.g., , for a discussion on the effect of shortwave forcing on methane's contribution to overall RF). A companion scenario, SSP3-LowNTCF, was also included in order to study the effect of NTCF species in the context of AerChemMIP. Critically, emissions factors of key NTCF species are assumed to develop similar to an SSP1 (rather than SSP3) scenario. SSP3-LowNTCF sees substantially fewer contributions to EOC forcing from NTCF emissions (notably SOx and methane), resulting in a forcing level of 6.3 Wm-2 and global mean temperature increase of 3.75 ∘C by the end of the century. This significant reduction is largely due to updating emissions coefficients for air pollutants and other NTCFs to match the SSP1 assumptions. SSP4-3.4 was chosen to provide a scenario at the lower end of the range of future forcing pathways. Reaching a EOC mean global temperature between SSP2-4.5 and SSP1-2.6 (∼2.25 ∘C), it is an ideal scenario for scientists to study the mitigation costs and associated impacts between forcing levels of 4.5 and 2.6 Wm-2.

The final two scenarios, SSP1-1.9 and SSP5-3.4-OS, were chosen to study policy-relevant questions of near- and medium-term action on climate change. SSP1-1.9 provides a new low end to the RF pathway range. It reaches an EOC forcing level of ∼1.9 Wm-2 and an associated global mean temperature increase of ∼1.4 ∘C (with temperature peaking in 2040), in line with the goals of the Paris Agreement. SSP5-3.4-OS, however, is designed to represent a world in which action towards climate change mitigation is delayed but vigorously pursued after 2050, resulting in a forcing and mean global temperature overshoot. A peak temperature of 2.5 ∘C above pre-industrial levels is reached in 2060 after which global mitigation efforts reduce EOC warming to ∼2.25 ∘C. In tandem, and including SSP2-4.5 (which serves as a reference experiment in ScenarioMIP; ), these scenarios provide a robust experimental platform to study the effect of the timing and magnitude of global mitigation efforts, which can be especially relevant to science-informed policy discussions.

Emissions contributions to the global climate system are myriad but can broadly be divided into contributions from greenhouse gases (GHGs) and aerosols. The models used in this analysis explicitly represent manifold drivers and processes involved in the emissions of various gas species. For a fuller description of these scenario results see the original SSP quantification papers . Here, we focus on emissions species that most strongly contribute to changes in future mean global temperature and scenarios with the highest relevance and uptake for other MIPs within CMIP6, namely the Tier-1 scenarios SSP1-2.6, SSP2-4.5, SSP3-7.0, and SSP5-8.5. Where insightful, we provide additional detail on results from other scenarios; however results for all scenarios are available in Appendix .

CO2 emissions have a large span across scenarios by the end of the century (-20 to 125 Gt yr-1), as shown in Fig. . Scenarios can be categorized based on characteristics of their trajectory profiles: those that have consistent downward trajectories (SSP1, SSP4-3.4), those that peak in a given year and then decrease in magnitude (SSP2-4.5 in 2040 and SSP4-6.0 in 2050), and those that have consistent growth in emissions (SSP3). SSP5 scenarios, which model a world with fossil-fuel-driven development, have EOC emissions which bound the entire scenario set, with the highest CO2 emissions in SSP5-8.5 peaking in 2080 and the lowest CO2 emissions in SSP5-3.4-OS resulting from the application of stringent mitigation policies after 2040 in an attempt to stabilize RF to 3.4 Wm-2 after overshooting this limit earlier in the century. A number of scenarios exhibit negative net CO2 emissions before the end of the century. SSP1-1.9, the scenario with the most consistent negative emissions trajectory, first reports net negative emissions in 2060 with EOC emissions of -14 Gt yr-1. SSP5-3.4-OS, SSP1-2.6, and SSP4-3.4 each cross the zero-emissions threshold in 2070, 2080, and 2090, respectively.

Global emissions trajectories for CO2 are driven largely by the behavior of the energy sector in each scenario, as shown in Fig. . Positive emissions profiles are also greatly influenced by the industry and transport sectors, whereas negative emissions profiles are driven by patterns of agriculture and land-use as well as the means of energy production. In SSP1-2.6, early to mid-century emissions continue to be dominated by the energy sector with substantial contributions from industry and transport. Negative emissions from land use are observed as early as 2030 due to large-scale afforestation while net negative emissions from energy conversion first occur in 2070. Such net negative emissions are achieved when carbon dioxide removal from bioenergy from carbon capture and storage (CCS) exceeds residual fossil CO2 emissions from the combustion of coal, oil, and gas. Emissions contributions from the transport sector diminish over the century as heavy- and light-duty transport fleets are electrified. Emissions from industry peak and the decrease over time such that the residential and commercial sector (RC) provides the majority of positive CO2 emissions by the end of the century. SSP2-4.5 experiences similar trends among sectors but with smaller magnitudinal changes and temporal delays. Negative emissions, for example, are experienced in the land-use sector for the first time in 2060 and are not experienced in the energy sector until the end of the century. Energy-sector CO2 emissions continue to play a large role in the overall composition until 2080, at which point the industrial sector provides the plurality of CO2. Emissions from the transport sector peak at mid-century, but are still a substantive component of positive CO2 emissions at the end of the century. Finally, the SSP5-8.5 scenario's emissions profile is dominated by the fossil-fueled energy sector for the entirety of the century. Contributions from the transport and industrial sectors grow in magnitude but are diminished as the share of total CO2 emissions, CO2 emissions from the AFOLU sector, decrease steadily over time. By the end of the century, the energy sector comprises almost 75 % of all emitted CO2 in this scenario relative to 50 % today.

Methane (CH4) is an emissions species with substantial contributions to potential future warming mainly due to its immediate GHG effect, but also because of its influence on atmospheric chemistry, as a tropospheric ozone precursor, and its eventual oxidation into CO2 in the case of CH4 from fossil sources . At present, approximately 400 Mt yr-1 of CH4 is emitted globally, and the span of future emissions developed in this scenario set range from 100 to nearly 800 Mt yr-1 by the end of the century. Global emissions of methane in SSP1 scenarios follow similar trajectories to CO2, with large emissions reductions; SSP2 follows suit, with emissions peaking in 2030 and then reducing throughout the rest of the century; in SSP3's baseline scenario, emissions continue to grow while in the NTCF scenario they are reduced drastically as policies are implemented to reduce forcing from short-lived emissions species; SSP4 is characterized by growing (SSP4-6.0) or mostly stable (SSP4-3.4) CH4 emissions until the middle of the century which peak in 2060 and then decline; and finally SSP5's baseline scenario sees a plateauing of CH4 emissions between 2050 and 2070 before their eventual decline, while the overshoot scenario has drastic CH4 emissions reductions in 2040 corresponding to significant mid-century mitigation efforts in that scenario.

Historically, CH4 emissions are dominated by three sectors: energy (due to fossil fuel production and natural gas transmission), agriculture (largely enteric fermentation from livestock and rice production), and waste (i.e., landfills). In each scenario, global emissions of CH4 are largely dominated by the behavior of activity in each of these sectors over time. For example, in the SSP1 scenarios, significant reductions in energy emissions are observed as energy supply systems shift from fossil to renewable sources while agriculture and waste-sector emissions see only modest reductions as global population stabilizes around mid-century. In the SSP2 scenario, emissions from the energy sector peak in 2040 as there is continued reliance on energy from natural gas but large expansions in renewables in the future; however, emissions from the agricultural and waste sectors are similar to today's levels by the end of the century. Finally, CH4 emissions in SSP5's baseline scenario are characterized by growth in the energy sector from continued expansion of natural gas and a peak and reduction in agricultural emissions resulting in 20 % higher emissions at the end of the century relative to the present as population grows in the near term before contracting globally.

GHG emissions are broadly similar between the main scenarios in CMIP5 (RCPs) and CMIP6 (SSPs). Notably, we observe that the SSPs exhibit slightly lower CO2 emissions in the 2.6 Wm-2 scenarios and higher emissions in the 8.5 Wm-2 scenarios due to lower and higher dependence on fossil fuels relative to their RCP predecessors. CH4 emissions are largely similar at EOC for 2.6 and 4.5 Wm-2 scenarios between the RCPs and SSPs, with earlier values differing due to continued growth in the historical period (RCPs begin in 2000 whereas SSPs begin in 2015). The 8.5 Wm-2 scenario exhibits the largest difference in CH4 emissions between the RCPs and SSPs because of the SSP5 socioeconomic story line depicting a world which largely develops out of poverty in less-developed countries, reducing CH4 emissions from waste and agriculture. This contrasts with a very different story line behind RCP8.5 .

In nearly all scenarios, aerosol emissions are observed to decline over the century; however, the magnitude and speed of this decline are highly dependent on the evolution of various drivers based on the underlying SSP story lines, resulting in a wide range of aerosol emissions, as shown in Fig. . For example, sulfur emissions (totaling 112 Mt yr-1 globally in 2015) are dominated at present by the energy and industrial sectors. In SSP1, where the world transitions away from fossil-fuel-related energy production (namely coal in the case of sulfur), emissions decline sharply as the energy sector transitions to non-fossil-based fuels and end-of-pipe measures for air pollution control are ramped up swiftly. The residual amount of sulfur remaining at the end of the century (∼10 Mt yr-1) is dominated by the industrial sector. SSP2-4.5 sees a similar transition but with delayed action: total sulfur emissions decline due primarily to the decarbonization of the energy sector. SSP5 also observes declines in overall sulfur emissions led largely by an energy mix that transitions from coal dependence to dependence on natural gas, as well as strong end-of-pipe air pollution control efforts. These reductions are similarly matched in the industrial sector, where natural gas is substituted for coal use as well. Thus, overall reductions in emissions are realized across the scenario set. Only SSP3 shows EOC sulfur emissions equivalent to the present day, largely due to increased demand for industrial services from growing population centers in developing nations with a heavy reliance on coal-based energy production and weak air pollution control efforts.

Aerosols associated with the burning of traditional biomass, crop, and pasture residues, as well as municipal waste, such as black carbon (BC) and organic carbon (OC, see Appendix Fig. ), are affected most strongly by the degree of economic progress and growth in each scenario, as shown in Fig. . For example, BC emissions from the residential and commercial sector comprise nearly 40 % of all emissions in the historical time period with a significant contribution from mobile sources. By the end of the century, however, emissions associated with crop and pasture activity comprise the plurality of total emissions in SSP1, SSP2, and SSP5 due to a transition away from traditional biomass usage based on increased economic development and population stabilization and emissions controls on mobile sources. Only SSP3, in which there is continued global inequality and the persistence of poor and vulnerable urban and rural populations, are there continued quantities of BC emissions across sectors similar to today. OC emissions are largely from biofuel and open burning and follow similar trends: large reductions in scenarios with higher income growth rates with a residual emissions profile due largely to open-burning-related emissions. Other pollutant emissions (e.g., NOx, carbon monoxide, CO; and volatile organic carbon, VOC) also see a decline in total global emissions at rates depending on the story line .

Harmonization, by definition, modifies the original model results such that base-year values correspond to an agreed-upon historical source, with an aim for future values to match the original model behavior as much as possible. Model results are harmonized separately for each individual combination of model region, sector, and emissions species. In the majority of cases, model results are harmonized using the default methods described in Sect. ; however, it is possible for models to provide harmonization overrides in order to explicitly set a harmonization method for a given trajectory.

We assess the impact that harmonization has on model results by analyzing the harmonized and unharmonized trajectories. Figure  shows global trajectories for each scenario of a selected number of emissions species. Qualitatively, the CO2 and sulfur emissions trajectories match relatively closely to the magnitude of model results due to general agreement between historical sources used by individual models and the updated historical emissions datasets. This leads to convergence harmonization routines being used by default. In the case of CH4 and BC, however, there is larger disagreement between model results and harmonized results in the base year. In such cases, Aneris chooses harmonization methods that match the shape of a given trajectory rather than its magnitude in order to preserve the relationship between driver and emissions for each model.

We find that across all harmonized trajectories the difference between harmonized and unharmonized model results decreases over the modeled time horizon. Panels (e–h) in Fig.  show the distribution of all 15 954 trajectories (unharmonized and less harmonized result) for the harmonization year (2015) and two modeled years (2050 and 2100). Each emissions species data population exhibits the same trend of reduced difference between modeled and harmonized results. Not only does the deviation of result distributions decrease over time, but the median value also converges toward zero in all cases.

The trajectory behavior for a number of important emissions species is dominated by certain sectors, as shown in Appendix Fig. . Notably, the energy sector tends to dominate behavior of CO2 emissions, agriculture dominates CH4 emissions trajectories, the industrial sector largely determines total sulfur emissions, and emissions from the residential and commercial sectors tend to dominate BC emissions across the various scenarios. Accordingly, we further analyzed the harmonization behavior of these sector–species combinations. Importantly, we again observe an overall trend towards convergence of results at the end of the century; thus harmonized results largely track unharmonized results for these critical emissions sectors. The deviation of distributions of differences consistently decreases with time for all scenarios, and nearly all medians converge consistently towards zero, save for energy-related CO2 SSP5-8.5, which has a higher growth rate than convergence rate, thus larger differences in 2050 than 2015. Overall, we find the harmonization procedure successfully harmonized results' historical base year and closely matches model results across the scenarios by EOC.

The extent to which reductions or growth of emissions are distributed regionally varies greatly among scenarios. The regional breakdown of primary contributors to future warming potential, CO2 and CH4, is shown in Fig. . While present-day CO2 emissions see near-equal contributions from the Organization for Economic Cooperation and Development (OECD) and Asia, future CO2 emissions are governed largely by potential developments in Asia (namely China and India). For SSP1-2.6, in which deep decarbonization and negative CO2 emissions occur before the end of the century, emissions in Asia peak in 2020 before reducing to zero by 2080. Mitigation efforts occur across all regions, and the majority of carbon reduction is focused in the OECD; however, all regions have net negative CO2 emissions by 2090. Asian CO2 emissions in SSP2-4.5 peak in 2030, and most other regions see overall reductions except Africa, in which continued development and industrialization results in emissions growth. Notably, Latin America is the only region in which negative emissions occur in SSP2-4.5 due largely to increased deployment of biomass-based energy production and carbon sequestration. Sustained growth across regions is observed in SSP5-8.5, where emissions in Asia peak by 2080, driving the global emissions peaking in the same year. Other scenarios (see Appendix Fig. ) follow similar trends with future CO2 emissions driven primarily by developments in Asia.

CH4 emissions, resulting from a mix of energy use, food production, and waste disposal, show a different regional breakdown across scenarios. In SSP1-2.6, CH4 emissions are reduced consistently across regions as energy systems transition away from fossil fuel use (notably natural gas) and the husbandry of livestock is curtailed globally. CH4 emissions in other scenarios tend to be dominated by developments in Africa. In SSP5-8.5, for example, emissions in Africa begin to dominate the global profile by mid-century, due largely to expansion of fossil-fuel-based energy production. SSP3 and SSP4 see continued growth in African CH4 emissions across the century, even when global emissions are reduced as in the case of SSP4 scenarios.

CO2 and CH4 are well-mixed climate forcers and thus their spatial variation has a higher impact from a political rather than physical perspective. Aerosols, however, have substantive spatial variability which directly impacts both regional climate forcing via scattering and absorption of solar radiation and cloud formation as well as local and regional air quality. Thus in order to provide climate models with more detailed and meaningful datasets, we downscale emissions trajectories from model regions to individual countries. In most cases, models explicitly represent countries with large shares of emissions (e.g., USA, China, India). MESSAGE-GLOBIOM and REMIND-MAGPIE are notable exceptions; however, their regional aggregations are such that these important countries comprise the bulk of emissions in their aggregate regions (e.g., the MESSAGE-GLOBIOM North American region comprises the USA and Canada). For regions constituted by many countries, country-level emissions are driven largely by bulk region emissions and country GDP in each scenario (per Sect. ). Afterwards, country-level emissions are subsequently mapped to spatial grids . We here present global maps of two aerosol species with the strongest implications on future warming, i.e., BC in Fig.  and sulfur in Fig. . We highlight three cases which have relevant aerosol emissions profiles: SSP1-2.6, which has significantly decreasing emissions over the century, SSP3-7.0, which has the highest aerosol emissions, and SSP3-LowNTCF, which has socioeconomic drivers similar to those of the SSP3 baseline but models the inclusion of policies which seek to limit emissions of near-term climate forcing species.

At present, BC has the highest emissions in China and India due largely to traditional biomass usage in the residential sector and secondarily to transport-related activity. In scenarios of high socioeconomic development and technological progress, such as SSP1-2.6, emissions across countries decline dramatically such that by the end of the century, total emissions in China, for example, are equal to those of the USA today. In almost all countries, BC emissions are nearly eradicated by mid-century while emissions in southeast Asia reach similar levels by the end of the century. In SSP3-7.0, however, emissions from southeast Asia and central Africa increase until the middle of the century as populations grow while still depending on fossil-fuel-heavy energy supply technologies, transportation, and cooking fuels. By the end of the century in SSP3-7.0, global BC emissions are nearly equivalent to the present day (see, e.g., Fig. ), but these emissions are concentrated largely in central Africa, southeast Asia, and Brazil while they are reduced in North America, Europe, and central Asia. By enacting policies that specifically target near-term climate forcers in SSP3-LowNTCF, the growth of emissions in the developing world is muted by mid-century and is cut by more than half of today's levels (∼9 vs. ∼4 Mt yr-1) by the end of the century. These policies result in similar levels of BC emissions in China as in SSP1-2.6, while most of the additional emissions are driven by activity in India and central Africa due to continued dependence on traditional biomass for cooking and heating.

The spatial distribution of sulfur emissions varies from that of BC due to large contributions from energy and industrial sectors, and is thus being driven by a country's economic size and composition, as opposed to household activity. Emissions today are largely concentrated in countries having large manufacturing, industrial, and energy supply sectors with heavy reliance on coal, such as China, India, the USA, Russia, and some parts of the Middle East. Again, we observe in SSP1-2.6 a near elimination of sulfur emissions by the end of the century with some continued reliance on sulfur-emitting technologies in India and China in the middle of the century. In SSP3-7.0, although global sulfur emissions over the course of the century peak slightly before reducing to below current levels, increased emissions in southeast Asia offset reductions in emissions elsewhere due to an expanding industrial sector with continued reliance on coal. Notably, emissions in India peak around mid-century before reducing to a magnitude lower than emissions levels today. In the SSP3-LowNTCF scenario, NTCF policies have the added effect of reducing sulfur emissions, resulting in more RF but fewer potential health impacts due to sulfur pollution. By the end of the century in SSP3-LowNTCF, only India, China, and Brazil have nontrivial quantities of emissions at significantly lower magnitudes than today.