EDGARv4.3.1 (Emissions Database for Global Atmospheric Research version v4.3.1; http://edgar.jrc.ec.europa.eu/index.php) is used as the reference inventory of anthropogenic emissions, providing global grid maps of sector-specific historical emission data from 1970 to 2010 for SO2, NOx, CO, total non-methane volatile organic compounds (NMVOCs), NH3, as well as particulate matter compounds, namely PM10, PM2.5, BC, OC. EDGARv4.3.1 relies on international energy balances of IEA (2014) and agricultural statistics of FAO (2012) and regional or national information and assumptions on technology use and emission control standards. EDGARv4.3.1 is one of the few global emission inventories with consistent methodologies to calculate emission time series covering 4 decades for air pollutants with high spatial resolution of 0.1∘ × 0.1∘ and consistent sector-specific breakdowns. Moreover, recent comparisons show the reliability of this emission inventory based on the good agreement between the EDGARv4.3.1 2008 and 2010 emission data and the best estimates provided by official national data merged in the HTAP_v2.2 data set (Janssens-Maenhout et al., 2015). A more detailed comparison between EDGARv4.3.1 and the MACCity data at the country level is documented in Sect. S3 in the Supplement, showing regional differences from a few percent up to 50 %, which is within the range of uncertainties for specific components. The EDGAR data sets are calculated using a consistent bottom-up approach with full time series of the activity data and allow straightforward implementation of scenario assumptions. We start from the calculation of the reference emissions (EM) of a specific pollutant x at time t due to activity data (AD) of sector i with technologies j and end-of-pipe measures k and fuel type f in a country C as follows: EMREF,C,i(t,x)=∑j,k,fADC,i,f(t)×TECHC,i,j,f(t)×EOPC,i,j,k,f(t)×EFC,j,f,i(t,x)×(1-REDC,i,j,f,k(x)), with TECH representing the penetration fraction of a specific technology in a sector, EOP the installed fraction of end-of-pipe measures (EOP), EF the uncontrolled emission factors and RED the emission reduction associated with the end-of-pipe control measures. We treat all 12 sectors of Table 1 separately and for each sector we provide a global grid map per pollutant and per month in 2010 for the three scenarios on http://edgar.jrc.ec.europa.eu/pegasos/; in addition time series (1970–2010) of the reference data are reported as annual emissions by sector for each country on the same website. An overview of 2010 pollutant emissions is also reported in Sect. S2 in the Supplement for the three scenarios for 24 world regions. For the STAG_TECH and STAG_ENERGY scenarios we focus on the main sectors that were effectively targeted by air quality measures imposed by EU policies

In 2010, the EU had 27 member states and is therefore defined as such in this study.

STAG_TECH: we modeled the STAG_TECH scenario by assuming for the three sectors of interest constant 1970 emission factors to all technologies present in the emission database globally (specified mainly regionally, but few also globally). We further assumed in the STAG_TECH scenario no implementation of end-of-pipe (EOP) control measures in Europe. Other regional standards regulating end-of-pipe control, e.g., US power plant standards were not changed, because they fell outside the European scope of this study. Thus in STAG_TECH, European energy production is generated by the power plants of 1970 without additional end-of-pipe measures as shown in Eq. 2 (in other words the emission reduction RED equals the lower limit

This is the technological default of fly ash not passing through the stack.

EMSTAG_TECH,C,i(2010,x)=∑j,k,fADC,i,f(2010)×TECHC,i,j,f(2010)×EOPC,i,j,k,f(1970)×EFC,j,f,i(1970,x)×(1-REDC,i,j,f,k(x)) In the 1970s and 1980s there was a relatively large turnover of power plants (Platts database, 2007, http://www.platts.com/), resulting in a high share of power plants which reached 30–40 years of operational time in 2010, which is weighing strongly in the STAG_TECH scenario results. However, in other world regions, while 1970 EFs were kept constant, the emission reduction factors (RED) were changing over time. Here STAG_TECH results lead to even larger emissions. For the manufacturing industry, the effect of technology stagnation could only be reflected by keeping the emission factors, modeled at regional or global levels, constant. For road transport, the technology stagnation was mainly reflected by not considering the emission reductions from particle filters and catalysts of all the vehicles under EURO standards, mainly present in the fleet inside Europe but also outside

Aside of the American continent (North, Central, South), which implements mainly US Tier standards, other countries show a considerable share of EURO standards even outside Europe, in particular but not only Japan, Korea, etc.

STAG_ENERGY: the STAG_ENERGY scenario was modeled by assuming that the three sectors of interest consumed the same amount of energy (TJ) as in 1970, but with the 2010 fuel mix, energy efficiency (EFF), technologies, and end-of-pipe abatements, as shown in Eq. (3). Since the fuel market is to a large extent global, this scenario was implemented in all countries for the three selected sectors. All power plants, vehicles and industries with the reference 2010 emissions standards consume coal, gas and oil with the 2010 share but at the 1970 energy level (in TJ). In addition to the calibration per sector of the energy consumption level (in TJ), we evaluated the change in energy efficiency by fuel type, sector and region. For the power generation sector we scaled for each country the “main activity producer electricity plants (TJ)” with the 1970 over 2010 ratio of the “electricity output of main activity producer electricity plants (GWh)” from IEA (2014). For the road transport sector we scaled the “fuel consumption for road transport” with a factor composed of the 1970 over 2010 road transport fuel consumption ratio multiplied with the 1970 over 2010 fuel efficiency ratio. The latter was calculated with the macro-regional averaged values of petrol and diesel economies (L 100 km-1) in 2010 and 1975 (because of missing 1970 data) for different type of vehicles (passenger cars, light-duty vehicles, heavy-duty vehicles, busses, mopeds and motorcycles) distinguishing the fuel consumption for petrol and diesel based on the EPA Trends report (EPA, 2013). EMSTAG_ENERGY,C,i(2010,x)=∑fADC,i,f(1970)/EFFC,i,f(1970)∑fADC,i,f(2010)/EFFC,i,f(2010)×EMREF,C,i(2010,x) The comparison of the STAG_ENERGY emissions scenario with the reference emissions REF in 2010 highlights the emission reductions that were not realized because of increased energy consumption (e.g., to generate extra kWh electricity or to drive more distance per vehicle) with the 2010 technology and 2010 end-of-pipe abatement. Compared to the 1970 reference emissions, STAG_ENERGY demonstrates the benefit of all industrial developments towards less energy-intensive and less polluting technologies. It includes not only the technological progress with end-of-pipe measures but also the shifts towards less carbon-intensive fuels (e.g., natural gas instead of coal) and increase of fuel economy and energy efficiency. On the other hand, compared to the REF(2010) data, STAG_ENERGY assesses to what extent emission increases by consumption growth. It should be noted that pre-combustion treatment (cleaning) of fuels, such as coal washing or desulfurization of diesel and heavy residual fuel oil is not part of the STAG_ENERGY scenario but is addressed as a technology effect in the STAG_TECH scenario. Therefore, the fuel quality directives show their emissions savings (mainly on sulfur) in the STAG_TECH scenario while the fuel taxation policies (e.g., preferring diesel over petrol) are present in the STAG_ENERGY scenario.

In the ACPD version of this paper (Crippa et al., 2015), another scenario called STAG_FUEL discussed the combined impact of stagnation of fuel-mix and fuel amount. However, in revising the manuscript we decided to focus on the consumption of energy (TJ) instead of fuel because we consider the fuel mix and efficiency choices as exogenous variables just as the technology progress and end-of-pipe measures. When considering a stagnation of fuel with constant fuel mix and energy since 1970, due to the remaining contributions of relatively dirty fuel, this scenario results in higher emissions than STAG_ENERGY. This scenario is here not further discussed, since the interpretation of results is not adding much to the STAG_ENERGY scenario. The interested reader is referred to the corresponding ACPD paper.

The TM5-FASST model (Fast Scenario Screening Tool, version v4.2.0_2014) is a linearized source–receptor model derived at a receptor resolution of 1∘ × 1∘ from the global chemical transport model TM5-CTM (Tracer model 5, chemistry transport model, Krol et al., 2005) for gaseous and particulate matter atmospheric pollutants. It considers 56 world regions (both as source and receptor regions), with a higher detail over Europe which is represented by 16 regions. Detailed information on the FASST model can be found in dedicated works (Van Dingenen et al., 2009, 2015), while here we summarize its basic working principle and assumptions. The concentration of a substance x at time t, caused by the emission of a precursor l with source strength EM (emission) in source region S1, . . . 56 and received in receptor region R1, . . . 56 is calculated by the addition of a base concentration (BASE) and the contribution of the linearized matrix function for each precursor (Eq. 4): CONCR1…CONCR56t,x=BASER1…BASER56t,x+∑lαR1S1t,x,l…αR1S56t,x,l……αR56S1t,x,l…αR56S56t,x,l×ΔEMS1t,l…ΔEMS56(t,l), where αR1S1, ..., αR56S56 are the source receptor coefficients for precursors l of substance x and ΔEM the difference between the reference emission and the actual data. They have been derived from precursor emission perturbation model runs with TM5 using a reference emission data set for the year 2000 and meteorology fields for the year 2001.

For each source region, the TM5-FASST model requires as input the annual emissions of primary PM2.5, to be specified as BC, primary organic matter (assumed to be 1.3 × OC), and of the precursors (SO2, NOx, NMVOCs and NH3) in order to estimate the corresponding PM2.5 and ozone concentrations in the receptor regions. Making use of source–receptor relationships, it converts the emissions from any source region to pollutant concentrations at any receptor region, emulating underlying meteorological and chemical processes. Only anthropogenic emissions are input to this model and the considered chemical reactions include the formation of secondary inorganic aerosol species (ammonium nitrate and sulfate) from gaseous precursors (SO2, NOx and NH3), while no estimation of anthropogenic SOA (secondary organic aerosols) is performed. We note that natural emissions, e.g., secondary organic aerosol from biogenic sources, lightning NOx, and biogenic sources of VOCs (etc.), are included in the reference simulation following the AEROCOM recommendations in Dentener et al. (2006), however without accompanying source–receptor relationship calculations. Ozone formation is simulated through the reactions involving VOCs and NOx. FASST evaluates the impacts of O3 and PM2.5 concentrations on health and crops and vegetation. Tropospheric ozone and particulate matter negatively affect human health, increasing respiratory and cardiovascular diseases, lung cancer, etc. (WHO, 2013). Through parameterizations relating pollutant concentrations and exposed population (Anenberg et al., 2009; Jerrett et al., 2009; Burnett et al., 2014), TM5-FASST estimates the premature mortality for a population older than 30 years of age, exposed to O3 and PM2.5 concentrations. Moreover, ozone is a toxic compound for plants, reducing crop productivity (especially for wheat) and affecting plant diversity (UNEP/WMO, 2011). Following the procedure developed by Van Dingenen et al. (2009), the TM5-FASST model can quantify the loss yield due to crop exposure to O3 for four types of crops (wheat, maize, rice and soy) at global and regional levels.

Global SO2 emissions from power, industry and road transport sectors (Fig. 3) do not show a significant trend from 1970 (at 68 Tg) to 2010 (75 Tg), because the emission reductions by pre-combustion preparation of fuel (e.g., coal sulfur wash) and the post combustion exhaust treatment of SO2 (e.g., by flue gas desulfurization (FGD) units) were counterbalanced by the increased use of fuel (in particular of coal) worldwide. In both 1970 and 2010 the power sector contributed strongest to total sulfur emissions, and for the STAG_TECH scenario, when no technological progress for sulfur emission reduction would have taken place in the EU, global power SO2 emissions would have been 1.3 times higher than REF. However, for STAG_ENERGY, assuming constant energy consumption as in 1970 for the power, industry and road transport sectors, but with current technology and end-of-pipe measures and 2010 fuel mix and efficiency, global SO2 emissions would be only 24 % of the 2010 emissions. For the other pollutants, like NOx, CO and NMVOCs, we see a significant increase in REF from 1970 to 2010, which reflects mainly the increase in energy consumption in the power sector (e.g., NOx emissions from power generation tripled globally, from 10.5 to 31.4 Tg). Depending on the pollutant, the STAG_ENERGY emissions are lower by 40 to 75 % in 2010 (Fig. 3). The largest impact of STAG_ENERGY in 2010 is seen in the power generation sector emissions (more than 80 % as global average for all pollutants), ranging from ca. 20 % reduction in developing and emerging regions (e.g., Asia, Latin America, etc.) to more than 95 % reduction in industrialized regions (USA, EU, Japan, etc.).

When comparing the STAG_TECH to the REF(2010) case, we find a global reduction of road emissions by 8.5 times for SO2, ca. 1.5 times for NOx, CO and NMVOCs and 2 times for PM, which clearly illustrates the phasing-in of less emitting vehicles due to stricter EURO standards. A shift from diesel to petrol could have caused a further reduction of PM emissions but the opposite took place: diesel vehicles represented only ca. 20 % of the global fleet in 1970 compared to ca. 75 % in 2010. Therefore the PM emission reductions shown in the STAG_ENERGY scenario compared to the REF(2010) is less strong for transport emissions and mainly driven by the power sector. Other sectors, such as residential combustion, also contribute significantly to the total emissions, but are not so easily controlled and moreover there was no comprehensive legislation in place. For the aerosols (PM10, PM2.5 and BC) the increase in REF emission levels from 1970 to 2010 is stronger than for other pollutants (ca. 40 %), although we remark here that the major contribution comes from residential combustion, which was not evaluated in our scenarios. More information on the ratios between each retrospective scenario and the reference case are given in Tables S1.1 and S1.2. In the sections below, we focus on the three sectors separately for the historical trend in Europe with REF(1970) and REF(2010), and then the STAG_TECH and STAG_ENERGY scenario cases.

Figure 9 shows the spatial (0.1∘ × 0.1∘ long-lat) distribution in Europe of the difference in emissions of the STAG_TECH and REF(2010) scenarios for selected pollutants (SO2, NOx, CO, PM10 and BC). These maps can be interpreted as the emissions reduction in Europe due to the implementation of European air quality legislation on the power generation and road transport, together with the change in emission factors and technologies, which also affected the industrial sector. The avoided SO2 emissions are mainly located in western European urban areas (e.g., Paris, Madrid, London, Rome, Berlin and the Benelux region) due to the co-location of several major emitting activities, while many point sources (power plants and industries) are spread over Europe, leading to the reduction in SO2 emissions due to the switch to cleaner fuels (shifting of fuel types and lower sulfur content). A different spatial distribution of emission reductions is found for NOx and CO, where in addition to urban areas, a large reduction is observed for road transport (road tracks are visible for both pollutants). Interestingly, Italy, Germany, the United Kingdom and the Benelux region display strong and more uniform CO emission reductions compared to other European regions (e.g., France and Spain) because of the contributions of the manufacturing industry. PM10 and BC grid maps highlight the effectiveness of EURO standards on road vehicles especially in western European countries, representing a successful example to be followed by eastern European regions. Finally, the implementation of particulate filters on power plants and industries was highly effective in very industrialized areas (e.g., Benelux) and other major conurbations.

Power, industry and road emissions data from the considered scenarios have been used in the TM5-FASST model to derive the corresponding population weighted-average PM2.5 and O3 concentrations for the main world regions. Figure 10 shows the global impact of the historical change in emissions of 1970 compared to 2010 (REF(1970) vs. REF(2010)) on PM2.5 and O3 concentrations, and those of STAG_ENERGY and STAG_TECH relative to the REF(2010) data. Note that delta emissions are calculated as the difference between each scenario and the REF(2010) data. PM2.5 REF concentrations decreased from 1970 to 2010 by 4.7 and 5 µg m-3 in the USA and Europe, respectively, while increased concentrations in 2010 are especially observed for Asian countries (15 and 12.5 µg m-3 for China and India respectively), Africa (0.7 µg m-3) and Latin America (0.5 µg m-3). A similar pattern is also observed for O3 where industrialized countries had higher computed concentrations in 1970 compared to 2010 (delta O3 equal to 3.1 ppb for USA and to 0.4 ppb for Europe), while developing countries increased their concentrations in 2010 by 12.8, 5, 2.7 and 0.9 ppb for Asia, Latin America, Africa and Russia, respectively. Similarly to the change observed from 1970 to 2010, comparing the STAG_ENERGY scenario to the REF(2010) case, PM2.5 and O3 concentrations show the expected opposite patterns for industrialized and emerging countries; however, the largest impact is observed for Asia where the stagnation of energy consumption at 1970 levels would have produced much lower concentrations (annual population-weighted average delta PM2.5= 11.9 µg m-3 and delta O3= 13.7 ppb) compared to the reference 2010 levels. A markedly different impact pattern is observed for the STAG_TECH scenario versus the REF(2010), since the stagnation of technologies at 1970 levels applied to present-day consumption would have produced enhanced emissions for all world regions, especially for Europe (delta PM2.5= 4.9 µg m-3 and delta O3= 2.5 ppb), Asia (delta PM2.5= 1.7 µg m-3 and delta O3= 1.8 ppb) and Russia (delta PM2.5= 0.8 µg m-3 and delta O3= ppb). As shown in more detail in Fig. S6.1 in the Supplement, the implementation of EU air quality legislation for industrial facilities, in particular power plants, and of the EURO standards for road vehicles, coupled with the change in technology and fuel quality, led to on average 4–5 µg m-3 lower concentrations in Europe as PM2.5 (including both the primary and secondary particulate components simulated by TM5-FASST). Power plant PM2.5 concentrations decreased by 1.9 and 2.9 µg m-3 in western and central Europe (with a more coal-fired power industry). However, effects of power plant emission reductions (STAG_TECH scenario) in other regions (Japan, China, USA, India, etc. with delta PM2.5 values between 1.5 and 3.7 µg m-3) are significant. For Japan and China, large contribution in concentration reduction is seen from the road sector (1.3 and 1.1 µg m-3 respectively), and are related to an export of EURO standards via market globalization. The market impacts of this specific sector are studied in more detail separately in Crippa et al. (2016). For the USA, the large reduction in PM2.5 concentration is due to the power generation sector. Although the end-of-pipe measure implementation in US power plants is ascribed to US air quality legislation (and thus not analyzed in the STAG_TECH scenario), both Europe and the USA profited equally from fuel quality improvement. Looking into the detailed chemical composition of the PM2.5 changes for the power, road transport and industrial sectors (Fig. S6.2), we gain further insights into the sectors and processes that contributed to the concentration reductions. Power plant emissions typically consist of aerosol precursor gases and particulate matter (also including fly ash); therefore, its delta PM2.5 chemical composition is mainly formed by secondary inorganic components (nitrates, sulfates and ammonium) as well as other PM2.5. Similar aerosol chemical composition changes are found for industrial sources with less secondary particulate sulfates as heavy residual fuel oil is phased out. A different PM2.5 chemical composition response is found for road transport, which consists of primary organic matter, BC, as well as ammonium nitrate particles formed by the chemical reaction of NOx and NH3 emitted by this sector. The delta particulate SO4 mainly represents the impact of the change in the sulfur fuel content in worldwide regions due to the implementation of EU fuel quality directives as well as international conventions and globally it corresponds to ca. 0.5 µg m-3 less for the year 2010 on average (1.5 for USA, 1.3 for China, 1.1 for central Europe and 0.9 µg m-3 for western Europe). While the impact on PM2.5 concentrations due to technology and emission reductions (STAG_TECH scenario) is mostly found within the source region with emissions change (Europe, Japan, China), longer-range effects are found for ozone. This is an important result because it represents the need of having intercontinental policies for some pollutants. O3 formation is driven by the reaction of the precursors NOx, CO and NMVOCs, derived mostly from the road sector and strongly abated over the past 2 decades with the EURO standards. The avoided annual and regional average O3 concentrations range between 0.5 and 6 ppb, which significantly affects the current O3 levels ranging from 30 to 50 ppb and which are most present in the hot arid regions of North Africa, the Middle East and Turkey.

Modeled surface delta PM2.5 and O3 concentrations are used in TM5-FASST to estimate the impact of the specific scenario assumptions on human health (using the life expectancy standard of GBD, 2010) and crop yield (expressed in tons of crops produced extra or not lost; see Fig. 10). An increase of PM2.5 is reflected by a reduction of the average life expectancy, since enhanced PM concentrations are the main harmful components impacting human health. So, comparing the reference 1970 situation with the 2010 one, globally a loss in average life expectancy is observed mainly for North America (4.7 months for USA and 1.4 months for Canada) and Europe (4.2 and 5 months for central and western Europe, respectively), while increased life expectancy is found for developing countries (4.8 months for Asia; see Fig. 10a). Considering the STAG_ENERGY case, life expectancy increases in most of the world's regions (4.2 and 7.2 months for China and India, 1–2 months for African countries and 0.8 months for Latin America), while a loss in life expectancy is observed for the STAG_TECH scenario versus the REF(2010). As expected, in STAG_TECH a significant negative impact on life expectancy is found for western and central European countries (4–5 months, see also Fig. S6.3), where the impact of emission reduction measures is largest. Coherently, life expectancy decreases also for other industrialized countries, like Japan and USA (3.3 and 2 months) and to a lesser extent in developing and emerging countries. Major health benefits from emission reduction measures are observed in highly populated areas where PM and ozone changes are large. It is obvious that the road transport sector, important in densely populated (urban) areas is contributing significantly to the health impact. O3 concentrations negatively influence crop growth, so the reduction in O3 concentrations observed for the REF(1970) vs. REF(2010) and STAG_ENERGY vs. REF(2010) is reflected in a net positive crop yield (global gain equal to 15.5 and 28 million Mt for the two scenarios, respectively). Conversely, as shown in Fig. 10b and more in detail in Fig. S6.3, also here, the emission control measures on vehicles are mostly responsible for mitigating impacts on crop yields. The introduction of vehicle EURO standards led to reduction of worldwide ozone levels due to its atmospheric transport, corresponding to a crop yield benefit up to 8.3 million Mt of avoided crop loss, representing 0.3 % of world production of maize, wheat, rice and soy. Specifically, the reduction of road transport emissions allowed the production of an additional 2 Mt of crops in China in 2010, 1.4 in western Europe, 1 in India, etc.