To study variations in historical anthropogenic aerosols in the major source regions, E3SMv1 developed by the US Department of Energy (DOE) (Golaz et al., 2019) is used in this study. E3SMv1 is updated on the basis of the Community Atmosphere Model version 5 (CAM5) in order to explore several key emerging issues in the field of environment and climate studies and is a branch of the widely used Community Earth System Model (CESM) (Rasch et al., 2019). E3SMv1 consists of atmosphere, land surface, ocean, sea ice, and river model components. It features numerous upgrades to aerosol, turbulence, chemical, and cloud-related processes, offering multiple spatial resolution options. The model can run simulations for decades or more at higher resolution to help understand past, present, and future changes in Earth's behavior and to explore how the atmosphere interacts with other components of the Earth system. Aerosol microphysics and interactions with stratiform clouds are treated with the four-mode Modal Aerosols Module (MAM4) (Liu et al., 2016), which predicts the mass and number concentrations of sulfate, black carbon (BC), primary organic matter (POM), secondary organic aerosol (SOA), marine organic aerosol, mineral dust, and sea salt (Wang et al., 2020). EAMv1 applies the “Morrison and Gettelman version 2” (MG2) two-moment bulk microphysics parameterization for stratiform clouds (Gettelman and Morrison, 2015). It allows for aerosol–cloud interactions in all stratiform and shallow convective clouds but neglects them in deep convective clouds (Rasch et al., 2019). Liquid cloud drop activation is based on Abdul-Razzak and Ghan (2000), and ice nucleation parameterization based on classical nucleation theory for the heterogeneous ice formation in mixed-phase clouds follows Y. Wang et al. (2014). Hygroscopicity is specified for soluble aerosols to calculate the particle size based on relative humidity. The aerosols are assumed to mix internally in the same aerosol mode and externally between modes when calculating the aerosol optical properties (Ghan and Zaveri, 2007). The model has been applied to investigate the variations in anthropogenic and natural aerosols related to air–sea interactions (Yang et al., 2022b; Zeng et al., 2021). Compared to the regional model, the model of E3SMv1 with an aerosol-tagging tool introduced in this study is more suitable for the simulation of transboundary and intercontinental transport of aerosols across the globe. In this study, the model is configured at its standard horizontal spatial resolution of approximately 1° with 72 vertical layers.

Global emissions of aerosols and precursor gases used in the simulations are obtained from the CMIP6 (the Coupled Model Intercomparison Project Phase 6) datasets (Hoesly et al., 2018; van Marle et al., 2017). However, the anthropogenic emissions in China are replaced with the MEIC (Multi-resolution Emission Inventory for China) inventory, which fully considers the implementation of clean air actions over China since the 2010s (Gao et al., 2022, 2023; Li et al., 2021; Zheng et al., 2018). Following previous studies (Ren et al., 2021; Yang et al., 2018a), the near-surface concentrations of PM2.5 here are estimated as the sum of sulfate, BC, POM, and SOA concentrations. Effective radiative forcing (ERF) refers to the change in the net radiative flux at the top of the atmosphere (TOA) after the external forcing is applied. In this study, ERF due to aerosol–radiation interactions (ERFari) for the individual tagged source regions is derived as the difference in TOA net radiative fluxes from a pair of diagnostic radiation calculations with and without the particular tagged aerosols from the source regions for the all-sky condition following Ghan (2013).

This study focuses on the variations in the source region and sector contributions in four major emission regions of the world (North America, Europe, East Asia, and South Asia) during the three key historical periods of emission changes (1850–1980, 1980–2010, and 2010–2017). Four simulations with monthly anthropogenic emissions of aerosols and precursors fixed at the 1850, 1980, 2010, and 2017 levels, respectively, are conducted. All simulations are performed for 1 year following 6-month model spin-up. Greenhouse gas concentrations, solar insolation, sea surface temperature, and sea ice extent are prescribed at the 2000 levels. The meteorological fields including 3-dimensional temperature, specific humidity, and winds are nudged toward the MERRA-2 (Modern-Era Retrospective analysis for Research and Applications, Version 2) reanalysis (Gelaro et al., 2017) in 2017 at a 6-hourly relaxation timescale.

Source apportionment aims to quantify the contributions to aerosols from specific sources. To examine the source–receptor relationships of aerosols, we implemented the Explicit Aerosol Source Tagging (EAST) system in E3SMv1, which plays a critical role in attributing aerosol concentrations to their respective emission sources. EAST follows the BC-source-tagging technique introduced in H. Wang et al. (2014), the sulfate-source-tagging method used in Yang et al. (2017), and other carbonaceous aerosol tagging applied in Yang et al. (2018a), which was previously implemented in the Community Atmosphere Model version 5 (CAM5-EAST). This tagging system is different from the traditional emission sensitivity method that zeroes out or perturbs emissions from a given source region or sector in sensitivity simulations along with a baseline simulation, which has to assume a linear response to emission perturbation and requires many additional simulations for estimating the contributions from multiple sources (H. Wang et al., 2014). EAST independently considers all aerosol physical, chemical, and dynamical processes for each tagged source through introducing additional aerosol-related variables within one simulation, and it does not rely on a linear response to emission perturbations. These capabilities make it physically more accurate and time efficient than the sensitivity experiments. This tagging method has previously been adopted in regional models and has now implemented in the global E3SMv1 model to better understand intercontinental transport from sources outside the regional domain.

In this study, in total 18 tags are set for each anthropogenic species of aerosols and precursors. Each of the four major source regions, including North America (NAM), Europe (EUR), East Asia (EAS), and South Asia (SAS), has four tags for energy transformation and extraction (ENE); industrial combustion and processes (IND); residential, commercial, and other (RCO); and the rest of the anthropogenic-emission sectors (RST). One tag is assigned to the anthropogenic emissions from rest of the world (ROW), and the last tag is allocated for all the natural/biogenic sources including open biomass burning, volcanic emissions, and oceanic emissions.

In order to evaluate the performance of E3SMv1 in reproducing the aerosol concentrations, Fig.  compares the simulated near-surface PM2.5 concentrations with observations from the Interagency Monitoring of Protected Visual Environments (IMPROVE) over the US, the European Monitoring and Evaluation Programme (EMEP) over Europe, US embassies and consulates in India, and the China National Environmental Monitoring Center (CNEMC) over China in 2017. The model successfully reproduces the spatial distribution of PM2.5 concentrations, with relatively high concentrations in eastern China and India and low concentrations in the US and Europe. The spatial correlation coefficient (R) between the E3SMv1 PM2.5 concentration simulations and observations is +0.80. The model reproduces the PM2.5 concentrations in the US well with normalized mean biases (NMB) of -11 %. However, it largely underestimates the PM2.5 concentrations in China and Europe, which has also been revealed in several studies (e.g., Gao et al., 2018; Gao et al., 2023; Navinya et al., 2020; Zeng et al., 2021). This discrepancy is partly because E3SMv1 only considers limited aerosol species (BC, POM, SOA, and sulfate) without including nitrate and ammonium aerosols in the aerosol module. On the other hand, the overestimated wet scavenging at mid-latitudes and high latitudes and underestimated gas-to-particle conversion can also lead to the low bias (Zeng et al., 2021). The evaluation in 2010 also shows similar high correlation and biases (Fig. S3).

In order to evaluate the model performance in reproducing the historical changes in aerosol concentrations during the important periods of emission changes, the variations in near-surface PM2.5 concentrations are compared with observations (Fig. S4) and MERRA-2 reanalysis (Fig. S5). The model reproduces the decreases in PM2.5 concentrations in the eastern US and Europe and the increases in East Asia and South Asia during 1980–2010 well, with a spatial R value of 0.78 between model results and MERRA-2 data. The model also simulates the aerosol decline in North America, Europe, and East Asia and a continuous increase in South Asia during 2010–2017 well, with an R value of 0.81 between model results and observational data. However, the model simulation does not capture the increases in PM2.5 in southwestern Canada and eastern Russia. This is because the wildfire emissions were kept unchanged during the simulation, while wildfires occurred more frequently in these regions during the analyzed time period (Jolly et al., 2015; Goss et al., 2020), leading to the increases in PM2.5 in observations.

Figure  shows the variations in anthropogenic emissions of sulfur dioxide (SO2) during the three key periods of historical emission changes from the tagged source regions. Since industrialization, anthropogenic SO2 emissions had rising trends during 1850–1980, especially in Europe and North America. Due to the implementation of clean air actions in Western countries, SO2 emissions in North America and Europe declined considerably during 1980–2010, while the emissions in East Asia and South Asia continued to increase. After 2010, China issued several clean air policies, which led to significant decreases in anthropogenic SO2 emissions in East Asia, while the SO2 emissions in South Asia kept increasing during 2010–2017. The changes in anthropogenic BC and organic carbon (OC) emissions are similar to those of SO2 (shown in Figs. S1 and S2).

The changes in near-surface mass concentrations (Fig. ) and column burdens (Fig. ) of anthropogenic PM2.5 contributed by the tagged source regions during the focused three historical time periods follow the corresponding changes in anthropogenic emissions. Column burden refers to the concentration of aerosols contained in the air column above a unit area with a top at 60 km, which can better reflect the aerosol transport within the air column and is more related to the aerosol radiative effect. The near-surface concentration of aerosols represents the concentration of aerosols in the air near the surface (from the surface to 997 hPa for the model layer), which is more related to air quality and human health. Local anthropogenic-emission changes drove the PM2.5 level to reach its peak in 1980 in North America and Europe and then to decrease. The maximum PM2.5 level appeared in 2010 in East Asia, and the anthropogenic PM2.5 level continued growing in South Asia during 1850–2017, consistent with previous studies (Dey et al., 2020; Guttikunda and Nishadh, 2022; Singh et al.,2023).

To explore which aerosol species contributed the most to the changes in PM2.5 during the focused three historical time periods, Figs.  and  illustrate the relative contributions of individual aerosols to the simulated changes in near-surface PM2.5 mass concentrations and column burdens, respectively, in four major emission regions. In general, the historical changes in anthropogenic PM2.5 were primarily driven by the changes in sulfate. In North America, the contribution of sulfate to near-surface PM2.5 concentration (column burden) rose from 7 % (11 %) in 1850 to 67 % (81 %) in 1980 and then dropped to 52 % (67 %) in 2017. In Europe, sulfate contribution changed from 24 % (34 %) in 1850 to 71 % (85 %) in 1980 and then decreased to 50 % (68 %) in 2017. In East Asia, sulfate contribution changes from 2 % (6 %) in 1850 to 51 % (71 %) in 1980 and then decreased to 33 % (56 %) in 2017. It is interesting that the PM2.5 levels in East Asia increased during 1980–2010 but the sulfate contribution decreased in this time period. It is because the carbonaceous aerosols increased remarkably, which reduced the fractional contribution of sulfate. The sulfate contribution to the PM2.5 concentration (column burden) increased throughout the period of 1850–2017, from 2 % (5 %) to 42 % (62 %) in South Asia. Note that the carbonaceous aerosols, especially POM, dominated PM2.5 in all four targeted regions in 1850, resulting from the high heating demand from the residential sector.

Figure  shows the relative contributions from local and remote anthropogenic sources to the near-surface concentrations and column burdens of PM2.5 in the four targeted regions in 2010. Local sources dominated the near-surface anthropogenic PM2.5 concentrations over the high-emission regions including eastern China, the eastern US, and the Indo-Gangetic Plain, with local contributions being higher than 90 %. In the regions with low emissions, such as the Tibetan Plateau, the anthropogenic PM2.5 concentrations were largely contributed by the long-range transport of aerosols. The spatial distributions of burden contribution are similar to those of the corresponding contribution near the surface, but the long-range transport contributed more to the column burden than to the near-surface contribution due to the more efficient pollutant transport in the free troposphere than within the boundary layer. The long-range transport contributes 30 %–35 % of the PM2.5 burden in East Asia and South Asia and 50 %–55 % in North America and Europe, much higher than the 10 %–25 % for the near-surface concentrations over the four targeted regions.

Since both local and remote emissions can contribute to anthropogenic PM2.5, it is valuable to know the historical changes in these contributions, especially by the local sources from individual emission sectors and by remote sources from major emission regions. Figure  illustrates changes in the local source contributions from major emission sectors to the near-surface concentrations and column burdens of PM2.5 during 1850–2017. In North America and Europe, the historical changes in anthropogenic PM2.5 were largely induced by changes in emissions from the local energy (ENE) sector, followed by the industry (IND) sector, which increased before 1980 and decreased afterward. In East Asia, ENE, IND, and residential (RCO) sector emissions all had significant contributions to the increases in PM2.5 concentration and burden from 1850 to 1980. Then the contribution from the local IND sector showed the largest increases from 1980 to 2010 and decreases from 2010 to 2017. In South Asia, increases in anthropogenic PM2.5 from 1850 to 1980 were mainly attributed to the RCO emission increases. After that, increases in ENE emissions dominated the rising PM2.5 levels in South Asia during 1980–2017.

Figure  presents changes in remote emission contributions from the tagged source regions to the column burdens of PM2.5 during 1850–2017. The contributions from long-range transport to the near-surface concentrations and their historical variations over the four major emissions regions are relatively small, which were also reported in previous studies (e.g., Yang et al., 2018b) and are not discussed here. During 1850–1980, the long-range transport contributions to the PM2.5 burdens show increases and the contributions from ROW increased the most among the tagged source regions over all four targeted receptor regions. Note that aerosol emissions from North America contributed to the increase in the European PM2.5 burden by 1.7 mgm-2 and sources from Europe were also responsible for the PM2.5 burden increase by 1.0 mgm-2 in East Asia and 1.1 mgm-2 in South Asia. During 1980–2010, the long-range transport from North America and Europe decreased but that from East Asia and South Asia increased. East Asia contributed 0.4–0.6 mgm-2 to the PM2.5 burden increases in North America, while Europe and South Asia contributed about 0.3 mgm-2. In East Asia, 1.6 mgm-2 of the PM2.5 burden increase was attributed to South Asian sources, and 0.8 mgm-2 of the PM2.5 burden increase in South Asia during this time period was due to increases in East Asian emissions. From 2010 to 2017, owing to the clean air actions in China, contributions from East Asia to the PM2.5 burdens in the other three targeted regions decreased by 0.3–0.6 mgm-2. However, due to the continuous increases in South Asian emissions, South Asia still contributed to the PM2.5 burden increase in East Asia by 0.4 mgm-2 during this short time period.

The variation in aerosols can have a significant impact on ERF through aerosol–radiation and aerosol–cloud interactions. Figure  shows changes in ERF due to aerosol–radiation interactions (ERFari) at the top of the atmosphere (TOA) that can be attributed to changes in anthropogenic emissions from the tagged regions in the three key periods during 1850–2017. Due to the increases in aerosols from 1850 to 1980, a large negative ERFari was located over the major source regions and their downwind areas, with maximum ERFari changes being larger than 2 Wm-2 over the eastern US, Europe, and eastern China. In 2010, there were positive ERFari changes (warming effect) by a maximum of 2 Wm-2 in North America and Europe compared to 1980, which were due to the decreases in anthropogenic aerosols in these two regions. The positive ERFari changes due to the decrease in North American aerosols extended across the North Atlantic and caused an increase in incoming radiation by up to 0.5 Wm-2 in Europe. Similarly, the decrease in aerosols from Europe also led to a positive ERFari change by up to 0.5 Wm-2 in the downwind regions including Russia and northern China during 1980–2010. This is also revealed by previous studies that aerosol reduction has caused fast warming in downwind regions (Urdiales-Flores et al., 2023). Increases in aerosols in China during 1980–2010 and decreases during 2010–2017 produced negative (cooling) and positive (warming) changes in ERFari, respectively, over eastern China and the North Pacific, which largely contradicted each other. The continuously growing aerosols in South Asia induced negative ERFari changes (cooling) in South Asia, Southeast Asia, and southern China during both 1980–2010 and 2010–2017. Note that in this study we only quantify ERFari for the major emission regions based on the source-tagging technique. The quantification of ERF due to aerosol–cloud interactions (ERFari) requires additional simulations, which could be further examined in future studies.