Aerosol pollution potential from major population centers

Introduction Conclusions References


Introduction
One of the big challenges of the future is the sustainable growth of urban agglomerations.Since 2008 the majority of human population lives in urban regions, with the prospect that the trend of human inflow into urban areas will continue throughout this century.The largest of these agglomerations, so-called megacities or major population centers (MPCs), have two faces.They often develop much faster than the country in which they are located in and promote human Published by Copernicus Publications on behalf of the European Geosciences Union.

D. Kunkel et al.: Aerosol pollution potential from major population centers
development as well as generation of knowledge, technology, and wealth (Kraas, 2007).Furthermore, they can offer a significant chance for climate mitigation strategies, as the per capita emission is smaller when many people live close together (Parrish et al., 2009).However, these human agglomerations are very heterogeneous with respect to possible lifestyle of different groups, and consequently manifest different awareness in the handling of energy, water, and waste.Moreover, both increased residential housing as well as the clustering of industry in these metropolitan areas lead to the strong localized release of air pollutants, which substantially contributes to a decrease in air quality for local dwellers as well as in regions downwind (Molina and Molina, 2004).The urban growth occurs on less than 3 % of the land surface but with global consequences (Grimm et al., 2008).The enduring urbanization leads to major changes in land cover and land use with subsequent changes in surface mean temperatures (Kalnay and Cai, 2003).
There has been considerable research on these impacts on local to regional (e.g., Parrish et al., 2009Parrish et al., , 2011) ) and global (e.g., Lawrence et al., 2007;Butler and Lawrence, 2009;Butler et al., 2012) scales, as well as several large megacity projects (e.g., Baklanov et al., 2010;Molina et al., 2010).Lawrence et al. (2007, later referred to as L07) studied the outflow characteristics of generic gas-phase tracers with three different lifetimes (1, 10, 100 days) from 36 MPCs to obtain insight into the balance between local pollution buildup and long-range transport.They found that low-level export is strongest in middle and high latitudes, while export into the free troposphere is dominant in the tropics.Moreover, vertical mixing is most responsible for reducing the surface pollution.They provide detailed insight in the outflow characteristics of the gas-phase tracers, but with a limited range of tracer properties.This study picks up where L07 left off and answers the following questions: 1. How robust are the results of L07, or do they depend strongly on the selected model?

2.
To what degree do outflow characteristics depend on the lifetime of the tracer, especially if it varies in space and time, and how does it differ for gaseous and aerosol substances?
3. What more can be learned from tracer distributions and from tracer deposition fields when they are convolved with geographical population or agricultural distributions?
To investigate these aspects we use the global 3-D stateof-the-art atmospheric chemistry general circulation model (AC-GCM) EMAC (ECHAM5-MESSy Atmospheric Chemistry, Jöckel et al., 2006) with generic aerosol tracers of various sizes and different solubility as well as generic gas-phase tracers equal to those used by L07.46 cities around the world (see Table 1) have been chosen as emission hotspots to de-termine the balance between pollutant build-up around the source and pollutant export.
In Kunkel et al. (2012) the same set of MPCs was used to determine the remote aerosol deposition potential, i.e., the deposition beyond various distances.In particular, it was found that on average about half of the emitted mass of small (d ≤ 2.5 µm) aerosol tracers is deposited beyond 1000 km.Several other studies investigated trace gas or aerosol transport from various, mostly larger sources such as countries or even continents on continental scales (e.g., Duncan and Bey, 2004;Park et al., 2005;Wagstrom and Pandis, 2010) or transcontinental scales (e.g., Betzer et al., 1988;Jaffe et al., 1999;Wild and Akimoto, 2001;Stohl et al., 2002Stohl et al., , 2003;;Liu and Mauzerall, 2007;Liu et al., 2009).More studies concerning pollutant dispersion from smaller emission points have been published for the regional scale, commonly focusing on one source point either for different meteorological conditions or with changed emission scenarios (e.g., de Foy et al., 2006;Streets et al., 2007;Wang et al., 2010;Song et al., 2010).In this study, as well as for L07, Butler et al. (2008), and Kunkel et al. (2012), the focus is directly on the comparison between MPCs all over the world.
We structured the paper to first provide a description of the EMAC model system, the metrics to characterize the outflow, and further data on population and agricultural distributions.In Sect. 3 we briefly compare EMAC results with results from L07 to assess the robustness of their analysis before the aerosol dispersion is discussed in Sect. 4. The aerosol tracers are then compared to generic gas-phase tracers in Sect.5, and finally in Sect.6 the aerosol deposition and surface densities are used to assess the impact of MPCs on agriculture and population.The key conclusions and further open questions are summarized in the last section.Republic of the Congo.All MPCs are taken as large emission point sources for generic tracers to study the dispersion in the atmosphere, and in the case of aerosol tracers subsequent deposition onto the surface.For this we apply the AC-GCM EMAC (Jöckel et al., 2006) with ECHAM5 (version 5.3.01, Roeckner et al., 2006) as a base model to calculate the meteorology.The base model is coupled to the Modular Earth Submodel System (MESSy, version 1.9, Jöckel et al., 2005, and http://www.messy-interface.org),which allows the inclusion of different sub-models for physical and chemical processes in the atmosphere and interactions between the atmosphere and biosphere.The model provides a good representation of gas-phase tracers (e.g., Jöckel et al., 2006;Pozzer et al., 2010;Tost et al., 2007) as well as aerosols (e.g., Pringle et al., 2010;Pozzer et al., 2012).Generic aerosol tracers have been used by Burrows et al. (2009) to study the transport of biogenic aerosol particles between different ecosystems and by Lelieveld et al. (2012) to determine the global exposure risk after nuclear accidents.

Meteorological setup
The model resolution was chosen to be T106L31, that is a triangular truncation at wave number 106 for the spectral core of ECHAM5 (T106), and with 31 hybrid sigma-pressure levels in the vertical (L31), mainly distributed in the troposphere from the surface up to 10 hPa.The chosen spectral resolution corresponds to a quadratic Gaussian grid of about 1.125 • × 1.125 • , or ∼ 100 km × 100 km in longitude and latitude, respectively.This is a relatively high spatial resolution for global model simulations, but still a low resolution compared to what would be needed to simulate individual megacities.Therefore, the details of megacities or the concentration D. Kunkel et al.: Aerosol pollution potential from major population centers fields within the megacities are not considered here; rather, the megacities represent large point sources for examining the long-range transport.
The model simulations were initialized in July 2004 and ended in January 2006, with the first half a year used as model spin-up time.Meteorological variables, namely temperature, vorticity, divergence, and the logarithm of surface pressure are weakly nudged towards analysis data from the ECMWF ERA-Interim reanalysis project (Dee et al., 2011).The year 2005 was chosen as analysis time period as it shows no strong annual modes in ENSO or NAO and coincides with the main analysis year of the MEGAPOLI project (Baklanov et al., 2010).

MESSy setup
The present model study comprises the following physical processes: emission, atmospheric transport, and wet and dry deposition of generic aerosol tracers and emission, atmospheric transport, and a radioactive-like decay of generic gasphase tracers equal to those used in L07.For each of the 46 emission points, an independent set of tracers was released at a constant, prescribed emission rate of 1 kg s −1 .The emission was placed in the middle of the lowest model layer in all grid boxes hosting an MPC.These boxes have a mean depth of about 60 m in the vertical and approximately 100 km in the horizontal.Pozzer et al. (2009) found that the emission height affects tropospheric chemistry close to the source if the emissions are within the planetary boundary layer; the overall effect on the chemical composition is rather low.However, potential transport distances might increase with injections into the free troposphere, e.g., from industrial stacks.Thus, by using surface emissions, we obtain a conservative estimate for results regarding long-range transport.
In this study monomodal aerosol tracers represent aerosols with ambient diameters of 0.1, 0.5, 1.0, 2.5, and 10.0 µm with a fixed standard deviation of 1.0.Instead of using a complex, computationally expensive aerosol module to explicitly treat aerosol micro-physics, we use two different aerosol solubility states which differ in the treatment of scavenging by nucleation: in one case all aerosols can undergo this process (further denoted as NS act ) which is tracer size dependent, and in the other case this process is completely inhibited and the aerosol tracers are considered to be insoluble (further denoted as NS inact ), which is summarized in Table 2.These two aerosol states are considered as extrema with respect to aerosol solubility, representing an upper and lower estimate of results with the range in between being representative for most of the real atmospheric aerosol particles and consequently their chemical composition and reactivity.Thus, ten aerosol tracers are released at each emission source point, one for each size and solubility state, all with the same source strength.This allows us to use the various aerosol tracer properties to obtain information on various parts of the aerosol size distribution.In total, we use a set of 460 aerosol tracers (46 source points × 5 aerosol sizes × 2 aerosol solubility states) which do not interact with each other and have no feedbacks on clouds or the radiative balance.Interpretation of the sizes is the same as that described in Kunkel et al. (2012).
Figure 1 shows the different annual atmospheric residence times of the tracers, ordered by the longest lifetime of the NS inact 1 µm aerosol tracers.The aerosol residence time is calculated with a steady state approach by dividing the annual mean aerosol load by the total annual emission (Seinfeld and Pandis, 2006).We obtain residence times for each MPC with an upper limit provided by NS inact and a lower limit by NS act aerosol tracers.As the latter are more susceptible to wet deposition, they have shorter atmospheric residence times.80 to 95 % of the mass of fine aerosols is wet deposited, which is the main sink of fine aerosols in nature and in the model (Monks et al., 2009;Kunkel et al., 2012).All aerosol tracers with diameter d ≤ 2.5 µm are defined as fine aerosols, similar to the definition of PM 2.5 , which is the mass of aerosols with an ambient diameter less than 2.5 µm.For larger aerosols the difference in residence time with respect to aerosol solubility becomes smaller as dry and wet deposition contribute almost equally to the sink processes (Kunkel et al., 2012).Furthermore, the effect of nucleation scavenging is smaller for larger particles as they are removed very efficiently by impaction scavenging (Tost et al., 2006).In the following chapters we will refer to aerosol tracers with 0.1, 0.5, and 1.0 µm diameters as small aerosol tracers, and to those with 10.0 µm diameters as coarse aerosol tracers.

Metrics
We use similar metrics to L07 to distinguish between pollution at low levels and in the upper troposphere, and also between local and remote pollution from MPCs (summarized in Table 2).The equal emission rate for all tracers and MPCs provides the potential to compare the pollution from urban centers only by their location and prevailing meteorology, independent of real emissions.The upper tropospheric pollution potential is characterized by the mass fraction above 5 km altitude (E UT ).The low-level pollution potential is split into a long-range (ELR 1 km ) and a city-scale (RCL 0.5 km ) fraction with ELR 1 km being the mass fraction below 1 km altitude and at least 1000 km away from the source, while RCL 0.5 km is the mass fraction retained in the first 500 km from the emission source points and below 0.5 km altitude.
The first two metrics describe the outflow of an MPC, either vertically (E UT ) or quasi-horizontally (ELR 1 km ), the third one (RCL 0.5 km ) the pollution build-up around the source.To calculate RCL 0.5 km and ELR 1 km , a set of equidistant points around each emission source have to be determined.On a plane they are located on a circle with the emission source as center point; on a graticule (latitude-longitude grid) this becomes an ellipsoid.The corresponding distances from the center points are based on the radii of the WGS84 ellipsoid (National Imagery and Mapping Agency, http://earth-info.nga.mil/GandG/publications/tr8350.2/wgs84fin.pdf,2000 or see L07).In addition, the local pollution build-up can be assessed by adding up the area, or other properties in this area, in which the tracer density exceeds a given threshold A Y , with Y a threshold value in ng m −3 (tracer mixing ratios are provided in mol mol −1 and can be converted to ng g −1 with molar masses for the tracers, here chosen to be 1 g mol −1 , and the molar mass of air 28.97 g mol −1 ; the surface air density, approximately 1 kg m −3 , is then used to convert the mass mixing ratio in the lowest layer to a concentration in ng m −3 ).We further focus on the concentrated deposition patterns, i.e., high relative deposition.We calculate the emission normalized deposition fraction D X (in %), i.e., the annual mean deposition flux DEP accu (in kg m −2 s −1 ) in each model grid box divided by the annual global mean emission flux EM total (in kg m −2 s −1 ) from the corresponding source point: This fraction together with a threshold value X (in %) can be used to characterize the impact of an MPC on different ecosystems.We fold D X with surface properties, namely total surface area, water or land area, cropland, pasture, or forest.Similarly, we use A Y to assess the population number being exposed to a certain surface mixing ratio.Depending on the property, we speak of a cropland, pasture, or forest deposition or human exposure in the following sections.Finally, it is noted that for seasonal analyses, metric results for urban centers from the Southern Hemisphere have been shifted by six months to match the seasons in the Northern Hemisphere.

Agricultural and population data
Gridded population numbers for the year 2005 are used to determine the impact of aerosols on humans (see also Fig. S1 in the Supplement).Data are based on gridded population data of more than 300 000 national and sub-national administrative units to assign population values to grid cells (SEDAC, 2005).
Moreover, the fractions of three different land ecosystems, cropland, pasture, and forest, are used to determine the impact on land ecosystems (see Fig. S1 in the Supplement).Cropland and pasture data are compiled using satellitederived land-cover data and agricultural inventory data (Ramankutty et al., 2008)   All these data sets have been regridded to the model resolution, and together with the metrics D X and A Y are used to assess the human exposure, cropland deposition exposure, pasture deposition exposure, and forest deposition exposure.

Model comparison of outflow characteristics from major population centers
We first summarize the results of the model comparison between the EMAC model and the MATCH-MPIC model, a semi-offline, three-dimensional chemical transport model (e.g., Rasch et al., 1997;Lawrence et al., 1999;Kuhlmann et al., 2003) which was used by L07.A more detailed analysis is given in the electronic Supplement.We use the same 36 major population centers as L07 to test the reproducibility of their results.Total and upper tropospheric column and surface densities agree well between the two simulations and also the resulting metrics give comparable numbers with high coefficients of determination (R 2 ) of about 0.75.In general, the low-level export is stronger when simulated with MATCH-MPIC, as well as the area with a mass threshold exceedance of 10 ng m −3 , while the upper tropospheric pollution is stronger in EMAC (see Table 3).
The main results from L07 are well reproduced when using EMAC, which underlines the robustness of the results.Discrepancies in the metrics mainly result from the different underlying meteorological fields (from NCEP for MATCH-MPIC and from ECMWF for EMAC) and different advective and convective transport schemes which introduce the largest uncertainties in such calculations.

Atmospheric dispersion of aerosols from urban emission hotspots
In the following section we focus on aerosol transport with subsequent deposition which leads to different atmospheric residence times of single tracers.Aerosols are removed by dry and wet deposition processes, thus, aerosol weight, surface properties as well as meteorological conditions determine the distance an aerosol particle can travel in the atmosphere before being deposited to the ground.We utilize RCL 0.5 km to characterize the local low-level pollution buildup, ELR 1 km for the remote low-level pollution, and E UT for upper tropospheric pollution.

Local pollution build-up
In Fig. 2, the top panel shows annual averages of RCL 0.5 km .
The gray shaded area which is bounded by the solid and dashed black lines represents the MPC averages for NS inact and NS act aerosol tracers.The light gray shaded areas contiguous to this area denote the standard deviation, each plotted one-sided for the upper and lower bound.The overall annual mean values range between 10 to 14 % for small tracers and 15 to 19 % for large tracers.Six climate classes are used equal to those in Kunkel et al. (2012), to provide more information about regional differences.As our results are mainly dependent on the prevailing meteorology, the MPCs are grouped by their Köppen-Geiger climate classification (Peel et al., 2007), which depends on temperature and precipitation (see also For small NS inact aerosol tracers, minimum retention of tracer mass occurs in urban centers in arid (B) and temperate dry summer (Cs) regions with RCL 0.5 km of about 7 %, and the maximum for MPCs in tropical (A) and cold (D) regions with values of about 11 to 12 %.The same regions show the minima for NS act aerosol tracers of about 9 %, while tropical and temperate no dry season (Cf) means are the maxima, with values of about 17 %.Large aerosol tracers from urban centers in arid regions reveal minimum values of 9 % (NS inact ) and 10 % (NS act ), respectively, while the maximum values are found for both solubility states in cold regions and also in the soluble case for temperate no dry season regions, with climate class means of 20 % (NS inact ) and 22 % (NS act ).Further, almost all mean values for arid and temperate dry summer regions show, independent of solubility state, smaller mean values than all other climate class means, which can also be seen in almost every season of the year (see Fig. S4 in the Supplement).Only in temperate dry winter regions do low RCL 0.5 km values appear during winter months (DJF).Thus, the relative pollution build-up in these two regions is smaller than in all other climate classes.Furthermore, the pollution build-up in two contrary climate regions, tropical and cold, is highest, although meteorological conditions lead to different atmospheric dispersion (see Sect. 4.2).
The retention of tracer mass is controlled by two mechanisms.On the one hand, RCL 0.5 km anti-correlates with the atmospheric residence time of the tracers.As mentioned before, the residence time is mainly governed by wet removal, and thus aerosol tracers with short residence times often orig-

D. Kunkel et al.: Aerosol pollution potential from major population centers
inate in regions with high precipitation rates, i.e., tropical regions.In contrast, for tracers from arid regions the residence time is relatively long due to missing precipitation.Thus, the relative contribution of the emission to the global burden of a tracer is larger for those with short lifetimes, since a large fraction of the burden is generally closer to the source with higher RCL 0.5 km values.This accounts for about 50 % of the variance in the overall RCL 0.5 km results.On the other hand, meteorological conditions that change the distribution of the tracer mass in the atmosphere, though not directly their residence time, are responsible for the missing variance.For example, surface winds, turbulent mixing, or the depth of the boundary layer height can alter RCL 0.5 km values without affecting the aerosol lifetime.
The influence of precipitation also plays a major role in the seasonal variations in RCL 0.5 km .The minimum climate means stay almost constant throughout the year since the minima are given by mean values from regions which are weakly affected by precipitation.In contrast, the maxima vary with the season, with the difference between RCL 0.5 km minima and maxima being largest in summer (JJA) when precipitation rates are generally larger and consequently the aerosol residence times shorter.

Remote low-level outflow
In contrast to local pollution build-up, the emissions from urban centers also undergo transport which can either be horizontal or vertical.We first assess the horizontal transport focusing on the mass which is transported near the surface or which was first lifted and subsequently descended again to lower altitudes, e.g., in a dry intrusion or simply by aerosol sedimentation.Overall annual means of ELR 1 km vary between 16 and 20 % for small tracers and 12 and 15 % for large tracers, with larger values now given by NS inact aerosol tracers (see middle panel in Fig. 2).Including climate class means, the diversity increases with values ranging from 11 to 27 % for small and from 8 to 19 % for large aerosol diameters.Aerosol tracers from cold regions possess the largest remote low-level mass fraction with more than 25 % of their total mass (averaged over NS inact and NS act tracers) located in the near-surface layer downwind of the source.In contrast to the local build-up, aerosols from temperate dry summer regions show a large downwind pollution potential.The least export occurs in temperate dry winter and tropical regions as upper-level export is favored in these regions (see following Sect.4.2.2).
Similarly, the smallest seasonal means appear for aerosol tracers from tropical regions, except during winter when those from temperate dry winter regions show the least downwind pollution potential due to relatively enhanced transport into the upper troposphere.ELR 1 km means peak for aerosol tracers from either cold or temperate dry summer regions.In general, the highest downwind pollution potential is found during winter months (DJF) with overall means between 21 and 25 %, and the least during summer (JJA) with overall means of about 10 to 12 %.
In winter, atmospheric low-level inversions occur more frequently, and aerosol tracer mass is kept more efficiently within the boundary layer, which is commonly shallower during this time of the year.Together with less convective activity, the ELR 1 km seems to be dominated by the atmospheric stability.Higher stability leads to greater remote low-level transport.
Similar to L07 we analyzed the response of a change in the depth of the near-surface layer and consider a metric height of 500 m.L07 found that the remote low-level export decreased almost linearly with the volume, i.e., only a tenth of the ELR 1 km numbers are found in a layer which is 100 m deep.For aerosols, on annual average we find that about 50 % of the ELR 1 km mass is present in a volume half as deep as our original volume, thus the same relationship is found for aerosols as for gases in L07.We see a small seasonal variation in this number with a maximum in winter (on average 53 %) and a minimum in summer (on average 47 %) due to the same reason as for the seasonal variation in ELR 1 km .Additionally, the ratio is slightly dependent on aerosol solubility, with higher ratios for soluble aerosol tracers but with almost no dependence on aerosol size.

Outflow into the upper troposphere
Aerosol mass can also be lofted out of the source region into the upper troposphere, commonly by convective or frontal lifting depending on location and time of year (bottom panel of Fig. 2).Overall annual means of E UT are between 3 and 4 % for small aerosol tracers and 2.5 and 3.5 % for large aerosol tracers, and thus much smaller than for the other two metrics and especially much smaller than E UT values for gas tracers reported by L07.The order of the climate means is reversed compared to the order found for ELR 1 km , now with maximum pollution for aerosol tracers from temperate dry winter and tropical regions and minimum pollution for those from cold regions.
These findings rely on several facts which need to be discussed.First, the difference between NS inact and NS act means strongly varies between climate classes.For tropical regions which are heavily and regularly affected by convection, the different treatment of nucleation scavenging leads to strong differences in the upper tropospheric pollution with NS act means accounting for only about 60 % of the NS inact means.In contrast, for aerosols from cold regions this ratio is about 90 %, and thus the difference is almost negligible, caused by an inefficient nucleation scavenging of snow and ice crystals according to the model formulation.Second, in L07 the strongest vertical tracer displacement occurred in tropical regions, contrary to our results with maximum vertical aerosol transport in temperate dry winter regions, especially in the With sufficient lifting activity the E UT mass fraction strongly depends on the liquid and/or ice water content of the clouds and their extension, which in turn determine the scavenged mass.From the pure gas-phase tracer simulations, we know that the upward transport is strongest in the tropics, though the moisture present there limits the upward transport of aerosols so strongly that less mass than in the temperate dry winter regions reaches the upper troposphere.In arid regions precipitation hardly affects lifting processes; however, the tracer distribution is affected by the development of a very deep, well-mixed air layer which can reach altitudes up to 5 km, such as the Saharan Air Layer (Hall and Peyrillé, 2007).The aerosols emitted at the surface are already mixed to altitudes of several kilometers in this layer, and consequent lifting above 5 km can occur faster and is often associated with little or no precipitation.Third, the seasonal cycle is also reversed compared to what was reported for ELR 1 km , with minimum values during winter, overall mean of about 2 %, and maximum values during summer, overall mean between 4.5 and 6.5 % (see Fig. S4 in the Supplement).Furthermore, the variability, indicated by the standard deviation and also by the differences between various climate classes, is largest in summer and smallest in winter.This is again attributed to the convective activity, which is stronger in summer, but also shows strong differences between different regions of the world and thus also between different climate classes.Thus, the interplay of atmospheric moisture and lifting activity and their effect on aerosol tracers is the determining factor for transport to the upper troposphere, with relatively small aerosol mass fractions in this domain.

Relation between pollution build-up and outflow
Finally, we want to see whether there are significant correlations between pollution build-up and outflow for aerosols from MPCs.L07 reported that for gas-phase tracers a significant relation was found, and that the upward transport, rather than the low-level transport, is mainly responsible for ventilating surface pollution.Moreover, they found that the trade-off between long-range low-level export and dilution of pollutants in the region surrounding the MPCs becomes more evident for regions with similar vertical transport.
In general, we find an anti-correlation between the horizontal and vertical export (see Table 4).This relationship is expected since aerosol mass is either lifted to higher altitudes or kept at lower levels, being advected rather horizontally.However, the overall anti-correlation is weakened by the weak positive correlation in the tropics.In the tropics, an ambiguous situation is found for different MPCs.For instance, in Jakarta and Manila the opposing relationship between the two export pathways is still assessed, while it breaks down in other MPCs, for instance in Rio de Janeiro, Kolkata, and Lagos.A possible interpretation might be that in the first case convective scavenging dominates the wet deposition, and in the second case more aerosols are scavenged in large-scale clouds.In regions with dominant convective scavenging, more aerosol mass is effectively lifted to higher altitudes but also efficiently removed by scavenging.Less mass is then left over to participate in any long-range transport, especially at low levels.This could result in an opposing relationship between vertical and near-surface long-range transport.Oppositely, more large-scale scavenging indicates that a larger proportion of aerosol mass has possibly undergone farther transport before removal.A correlation between vertical and near-surface long-range transport would then suggest that the outflow occurs at high and low levels.
Between low-level export and low-level retention, the correlation coefficients show a very ambivalent behavior with most values in the range of [−0.5, 0.5].The sign varies between various climate classes and also between aerosol size and solubility.Only in two cases are higher correlations assessed: a positive one in temperate dry summer regions for small and large aerosol tracers and a negative one for small aerosol tracers from cold regions.In the first region, the mass fraction is rather small for RCL 0.5 km but high for ELR 1 km .The positive correlation indicates that the low-level export forms an important path to accumulate mass in the nearsurface region downwind.A positive correlation might be expected for the cold regions with high ELR 1 km and RCL 0.5 km mass fraction.However, as the vertical export is low in this climate region, the strength of surface wind and turbulent mixing within the boundary layer determines whether more of the mass is retained close to the source or advected downwind.
As it is expected, the correlation coefficients between lowlevel retention and vertical export reveal in most cases a negative sign.For small aerosol tracers the correlations are rather weak; only in temperate dry summer regions do we see a strong anti-correlation and in cold regions a strong correlation.The positive correlation coefficients for cold regions reveal that if lifting takes place, it occurs close to the emission source at low levels.On the other hand, an anti-correlation points more to lifting that either takes place farther away from the source or has its origin above 500 m altitude.For most climate regions this is true for the large aerosol tracers as they show stronger anti-correlations, even in cold regions.However, the sampling statistics are small and not equally distributed over the climate classes, thus the results should only be interpreted as a first indication.

Aerosol-gas-phase comparison
Finally, to conclude the analysis of atmospheric dispersion from urban emission sources, we want to show similarities and differences between gas-phase tracers with a fixed atmospheric lifetime and aerosol tracers with various atmospheric residence times.Here, we mainly focus on aerosol tracers with a diameter of 1 µm which have a lifetime comparable to, though generally shorter than, the lifetime of the gasphase tracers, but also include aerosol tracers with a diameter of 10 µm in parts of the discussion.In Sect. 3 we briefly discussed similarities and differences in the dispersion from MPCs with two different model systems.Now, we compare the aerosol results discussed in Sect. 4 with gas-phase tracer results from an EMAC simulation using the same configuration as used for the aerosol simulation.
Figure 3 gives a qualitative overview of how the atmospheric dispersion patterns change with respect to the removal of an atmospheric tracer.The amount of various removal processes acting on the individual tracers decreases from left to right in Fig. 3.In general, the gas-phase tracers in the right column are wider spread from the source due to their longer lifetime.Close to the source points, similar values are found for gas-phase and aerosol tracers, but farther away the discrepancies between the various types of tracers become more apparent.The total column densities reveal higher values in the direct outflow regions of the MPCs in the case of the gas-phase tracers and also in remote areas like the South Pacific with no significant tracer burden present for aerosol tracers.Tracer distributions in the surface layer vary strongly, and the transport to higher latitudes is not the strongest for the gas-phase tracers, but rather for NS inact tracers, especially in Northern Europe.In contrast, the gas-phase tracers predominate in the northern latitude storm tracks where almost no aerosol mass is simulated anymore.The largest differences between aerosol and gas-phase tracers are found for the column densities in the upper troposphere above 5 km.The absolute aerosol tracer peak values are about the magnitude of the background levels of the gas-phase tracers in remote areas.Moreover, location and strength of peak values differ, for instance for Khartoum and Teheran, which both have a much weaker signal in the gas phase compared to the background than for the aerosol tracers.Furthermore, Santiago and Lima are clearly discernible in the aerosol phase but not in the gas phase, and also the outflow from East Asia extends farther northward for the gas-phase tracers.
The question arises then how the pollution potentials differ between gas-phase and aerosol tracers.Figure 4 shows the three pollution potentials used to investigate the similarities and differences in the local pollution build-up and pollution export: ELR 1 km (top panel), E UT (middle panel), and A 10 (bottom panel).Generally, the accordance between aerosol and gas-phase tracers is higher when using the NS inact aerosols.The greatest similarity in pollution potential is found for the low-level pollution potentials.Aerosols show slightly higher ELR 1 km values than gas tracers, with coefficients of determination (R 2 ) higher than 0.9 (NS inact ) and 0.8 (NS act ), respectively.Since the R 2 are a measure of how much of the variance in the correlation can be explained by the source location, the tracer properties seem to be less important to determine the mass fraction found in the lowest layers downwind of the source.Instead, the location of the source along with the prevailing meteorology account for the mass transported downwind.In contrast, the vertical transport shows gas-phase mass fractions up to ten times higher than aerosol mass fractions.The correlation between gas and aerosol tracers is still positive but weak, with an R 2 of about 0.4 (NS inact ) and about 0.15 (NS act ), respectively.The source location is in this case of minor importance compared to the tracer properties, especially the solubility of a tracer, which determines to a great extent the mass fractions above 5 km.In the case of the area mass threshold exceedance, both the source location and tracer properties need to be considered more equally to explain the relation between gases and aerosols as positive correlations are found with R 2 = 0.78 (NS inact ) and R 2 = 0.67 (NS act ) for small aerosols and an averaged R 2 of about 0.53 for large aerosols.Hence, the source location determines the A 10 more in the case of small aerosols, while the size of the aerosol seems to be more important for the large aerosols.
In summary, the findings of Sect. 4 are further supported by the comparison of gas-phase and aerosol tracers, especially that the source location is most important in determining the low-level pollution while tracer properties have a substantial impact on the vertical transport.

MPC impact on ecosystems and humans
Aerosol particles have the ability to potentially influence their environment when they are airborne in several ways.For instance, humans can breathe in small aerosol particles deeply into the lungs, and it is certain that this affects the life expectancy of humans (e.g., Dockery et al., 1993;Pope et al., 2009).Large particles are mainly responsible for decreasing visibility (e.g., Watson, 2002), and generally particles which contain toxic constituents from anthropogenic combustion can harm crops (e.g., Wiman et al., 1990;Burkhardt, 2010) and humans (e.g., Pöschl, 2005;Stanek et al., 2011).In this chapter we assess the potential impact of the urban emitted aerosol particles on the environment by using the metric D X , which is the ratio of the total (= dry and wet) deposition flux to the global mean emission flux (defined in Sect.2.2) and threshold values for X of 1 % and 5 %, and we refer to this as high relative deposition.Although we are aware of the different impacts of dry and wet deposited particles, we discuss only their sum here and leave a further investigation open for following studies.Furthermore, we use the metric A Y to quantify the number of people that are exposed to a certain atmospheric mixing ratio.6.1 Impact on ecosystems

Impact on land and sea surfaces
First, we assess the total surface area which fulfills the criterion of D X .In Fig. 5 the total surface area is illustrated for each MPC for two aerosol diameters, 1.0 µm and 10.0 µm, on the left for D 1 and on the right for D 5 .The solubility states are displayed by different gray shadings with the light gray bar representing the additional area between the two solubility states.In contrast to the atmospheric metrics, the smaller area, indicated by the dark gray bar, is not always determined by the same solubility state for each MPC, but most commonly by the NS inact tracers.If only a dark gray bar is drawn, then there is no difference between the two states, while a single light gray bar indicates that no area is covered at all for one state.By definition, areas which are covered with at least 5 % of the emissions are also covered with at least 1 %.
In the case of D 1 , each MPC covers at least a small area, independent of aerosol size, which is not the case for D 5 .Aerosol tracers from MPCs which are known to undergo substantial long-range transport, e.g., Baghdad, Teheran, Cairo, or Khartoum, are more widely spread and thus aerosol mass is deposited in lower concentrations in each grid cell.However, other MPCs, e.g., London or Tianjin, show also zero area covered, and thus long-range transport is not the only limiting factor for high relative deposition.In contrast, the maximum is found for MPCs which are either influenced by heavy precipitation rates, e.g., Lagos or MPCs in South East Asia, or located in a basin with low ventilation, e.g., Bogotá or Szechuan.On average, aerosol tracers with a diameter of 10.0 µm cover larger areas compared to the small aerosol tracers.The reason is that the large aerosol tracers travel shorter distances and the subsequent deposition occurs on smaller total overall area.Since the total global emitted mass is the same for each tracer, the deposition rates of the large aerosol tracers are in turn higher in the area where deposition occurs.Furthermore, the faster deposition rates for the large aerosols lead to smaller differences between the areas covered by the two solubility states.
Further, we divide the total surface area into land and sea fractions to obtain a more accurate knowledge over the location of high relative deposition.Generally, the low-level outflow of aerosols is similar to the gas-phase outflow reported by L07, with more low-level long-range transport over oceans.Nevertheless, the total deposition still occurs more over continents with about 55 % of the small tracer emissions and 68 % of the large tracer emissions being deposited on land surfaces.For the high relative deposition, we focus only on D 1 since not all of the MPCs cover a surface area in case of D 5 .Since the land-sea fractionation is very similar for all aerosol tracers, independent of size or solubility, we choose the NS act 1.0 µm aerosol tracers to show D 1 in Fig. 6.The dark gray bars represent the land fraction which outweighs the sea fraction in the light gray color bars, although almost  50 % of the MPCs are located at coastlines and several more close to the coast.In several cases the land fraction is close to 100 %, mainly for MPCs with a dominant outflow over continental regions, e.g., Moscow, Rhine-Ruhr area, Teheran, Delhi, and Szechuan.However, there are also some MPCs close to coastlines which do not have high relative deposition over seas, e.g., Santiago and Beijing.The highest marine D 1 values appear for Athens and Rome, with about 80 % sea fraction, but not surprisingly Jakarta and Manila show the highest sea fractions in terms of absolute area since each of these cities is located on a small island in a maritime environment.

Impact on cropland, pasture, and forest
To answer the question which of the three ecosystem -cropland, pasture, or forest -is most exposed to deposition from MPCs, we initially fold the sum of all annual accumulated MPC deposition fields with the relative fractions for cropland, pasture, and forest (see Fig. S1 in the Supplement).The differences between small aerosol tracers (d ≤ 1.0 µm) are minor as well as between the two solubility states.Thus, we focus the discussion on the NS act 1.0 µm and 10.0 µm aerosol tracers, which are shown in Fig. 7.The pure deposition patterns have their maximum around the source location (Kunkel et al., 2012).In Fig. 7 MPCs are in most cases well detectable by the red areas in the six panels.For several MPCs these red areas exceed the MPC size since these are surrounded by large areas of an ecosystem, e.g., for cropland the corn-belt around Chicago, the region between Paris and east of the Rhine-Ruhr area, the region north of Lagos or large areas on the Indian subcontinent and in eastern China.In the case of pastureland, only a few hotspots are detected, for instance the region around Bogotá, the Sahel region in Africa, and again large areas in eastern China.Forests seem to be widely exposed to deposition from MPCs, e.g., in the USA in a belt from New England down to Atlanta, high forest deposition exposure is simulated, as well as in Bogotá and in the region between Rio de Janeiro and São Paulo.In Central Africa the Congo Basin with the African rainforest and in Asia regions in eastern China and also in Japan reveal high exposure.In Europe, the high northern latitudes are affected due to the vast appearance of boreal forest.For the same reason the deposition of Moscow can still be detected north of Mongolia far away from its source.In contrast, one of the most forested areas, the Amazon Basin, seems not to be affected by the deposition of the MPC tracers used in this study because it is not located downwind of the MPCs.
Using D 1 and D 5 , we can calculate the area of each ecosystem which is exposed to more than 1 % or 5 % of the emission.Table 5 shows the averages of small and large aerosol tracers for each ecosystem.Generally, larger areas are found for the soluble tracers and, as discussed before for the total surface area, for larger aerosol tracers, with one exception in the case of forests for D 1 (NS act 1.0 µm).The results for the two aerosol solubility states differ by less than 25 %, in most cases even less than 20 %.Maximum averages are found for forests, followed by croplands with up to 40 % less and pasture with about 50 % less area exposed to high relative deposition compared to forests.More detailed information for each MPC and ecosystem is found in Fig. S5 in the Supplement.Taken together, forests which cover large areas on each continent are most affected by the MPC deposition.In contrast, pasture areas are also widespread, though they are less affected by the MPC deposition than croplands, which are mainly present in a few concentrated regions.However, these regions are nearly all close to or in the outflow regions of MPCs.

Population-based impact
We now assess the impact of MPC aerosols on humans.For this, surface densities are used instead of aerosol deposition fields, since airborne particles are potentially inhaled by humans and consequently cause health issues.Small aerosols are again represented by the 1.0 µm aerosol tracers, and for comparison and as an upper limit for efficiently inhalable aerosols we show also results for 10.0 µm aerosols.
Figure 8 displays surface densities in the upper panels for the NS act tracers, on the left for the small aerosols (also shown in the middle panel in Fig. 3 with another color-scale), and on the right for the large aerosols.In the surface layer the differences between NS act and NS inact are almost negligible, especially in regions close to the MPCs, thus we focus our discussion on the NS act aerosols tracers.The small aerosol tracers are commonly more widely dispersed due to the faster gravitational settling of large aerosols and subsequent removal out of the surface layer.MPCs are easily detectable by the red dots in the figure, even when the surface densities are convolved with the geographical population distribu-tion (lower panels in Fig. 8).However, in addition two large domains are more accentuated: India and (East) China.Together these two countries host about one third of the global human population and thus the high population densities ensure that the highest numbers in the convolved variable of surface density and population distribution are found there.Other densely populated regions like Europe, Japan, or the Eastern US show similar surface densities but lower convolved numbers.
To further investigate and quantify how many people are affected by each MPC, we add up the number of people in the area with a mass threshold exceedance (A Y ).In Fig. 9 this is shown for the 1.0 µm and 10.0 µm aerosol tracers and a mass threshold Y of 10 ng m −3 .The highest values are found independently of aerosol size or solubility in Delhi, Beijing, Tianjin, Karachi, and Shanghai, with more than 1.3 billion people being exposed to more than 10 ng m −3 in and around Delhi.For other MPCs like Sydney, Melbourne, Santiago, or Kinshasa, the lowest population numbers are found, simply due to the relatively sparse population density around these cities.
In conclusion, the area A Y is a measure of the distribution of aerosol mass in the surface layer, especially of the dense part of the outflow plume.However, A Y does not seem to determine how many people are exposed to a certain aerosol mass.Instead, this is more dependent on the population in and around an MPC.This is also underlined when computing the coefficients of determination to see how much of the variance of the exposed population number is explicable by the area itself.For A 10 we find only weak correlations for small (R 2 < 0.2) and large (R 2 < 0.06) aerosol tracers, and also by changing the threshold value to 1 ng m −3 and 100 ng m −3 we ascertain no substantially higher correlations.

Summary and conclusions
This study is an extension of Lawrence et al. (2007) to understand and compare the dispersion and deposition of air pollutants from urban centers.The primary focus is on generic aerosol tracers of five different monomodal sizes between 0.1 µm and 10.0 µm and two solubilities with various atmospheric residence times.In the AC-GCM EMAC ten tracers were released at each of the 46 different emission points, i.e., the major population centers considered in this study.The emission source strength is set to be equal for all tracers to obtain results only depending on prevailing meteorology and tracer properties.
We show that on average about 10-14 % of the emitted mass of aerosol tracers with diameters less than 2.5 µm (15-19 % for aerosols with diameters of 10.0 µm) is retained around the source.This number is highest in regions where pollutants have short atmospheric residence times, which are constrained by meteorological conditions, especially by scavenging in and below clouds, e.g., the tropics.Export from the cities, either by vertical mixing or low-level longrange transport, is determined by the stability of the atmosphere.High stability favors low-level export (on average between 12-20 % for the range of the different aerosol tracers) and is most dominant in cold regions, whereas the verti-cal export (on average 2.5-4 %) is strongest in regions with high lofting potential, i.e., high instability.Furthermore, the export to higher altitudes is dependent on the moisture in the atmosphere and therefore cloud formation and subsequent wet removal processes, which leads to the strongest export in temperate dry winter regions and not in tropical regions, in contrast to insoluble gas-phase tracers.The export at low levels is similar for aerosol and gas-phase tracers but substantially different for transport above 5 km, with gasphase tracer concentrations up to ten times higher than the aerosol tracer concentrations due to the effective removal of aerosols in clouds.Moreover, the dispersion of generic gasphase tracers from MPCs has been compared for two models, one with online calculated meteorology and one with semioffline calculated meteorology as well as different physical parametrizations.These differences account for only about 25 % of the variance in the results; thus a high accordance is shown in vertical and long-range transport of the tracers between the two models.
Aerosol deposition fields convolved with geographical distributions of agricultural data and a metric to account for regions with high relative deposition are used to determine the impact of MPCs on ecosystems.We assess the area in which more than 1 % and 5 % of the emissions are deposited.On average the 10.0 µm aerosol tracers exceed these deposition fractions over larger areas with less variation between single MPCs than the aerosol tracers with smaller diameters.The areas vary more for smaller aerosols due to their longer lifetime and thus wider dispersion.This effect is strongest for MPCs in arid regions where aerosols have the longest atmospheric residence time while regions characterized by high precipitation or unfavorable geographic conditions reveal larger areas with high relative deposition.Most of this deposition occurs over land regions except for some MPCs like Rome, Athens, Jakarta, and Manila, with more deposition onto water surfaces.We further find that forests are more exposed to the aerosol deposition from MPCs than other land ecosystems such as cropland or pasture.Similarly, we use an area of mass threshold exceedance folded with population to assess the impact on humans and find that high population densities favor high human exposure, more so than the size of the area containing high surface mixing ratios.
All our results are solely based on model results and thus dependent on model uncertainties.The EMAC model was extensively evaluated for trace gas and aerosol distributions in many studies.A further improvement would be to represent the aerosol tracers better with properties, size and solubility, which can evolve in time and space and interact with other atmospheric constituents including clouds and atmospheric trace gases.This will be the scope of a followup study which investigates the impacts of different anthropogenic emissions from MPCs on the burdens of various aerosol types.Moreover, the coarse model resolution is a key issue with respect to the sub-grid-scale processes like the introduction of the emissions, vertical mixing, e.g., out of the boundary layer, convection, or scavenging as these physical processes are of major importance for the distribution of tracers in the atmosphere.Higher vertical and horizontal resolution to better resolve small-scale circulation patterns can be provided either by regional-scale models or by Lagrangian dispersion models.However, regional-scale models have the limitation of many simulations being necessary to cover all the major population centers used in this study, while Lagrangian models have the advantage of having no numerical diffusion but lack in the representation of several tropospheric processes necessary to determine the distribution of trace gases and aerosols, e.g., convection, clouds, mixing, and scavenging.Nevertheless, since urban centers will become even more important in the future by serving as habitats for more people than at present, the research activities which have been accomplished so far, e.g., projects like MEGAPOLI (http://www.megapoli.info/,Baklanov et al., 2010), CityZEN (http://www.cityzen-project.eu/)and MI-LAGRO (http://www.eol.ucar.edu/projects/milagro/,Molina et al., 2010), should rather be seen as first initiatives which scratched the surface and provided initial results, but also left several questions open or created new questions.

Fig. 1 .
Fig. 1.Aerosol residence times for all MPCs and all aerosol tracers.The upper value is determined by the NS inact tracers, the lower value by the NS act tracers.

Fig. 3 .
Fig. 3. Total column densities (top panel), surface densities (middle panel), and upper tropospheric column densities (bottom panel) for the sum of all 46 MPC tracer sources from the EMAC simulations.Left and central columns show the NS inact and NS act 1.0 µm aerosol tracers and the right column the gas-phase tracers with a 10 d decay lifetime.Units are 10 −6 kg m −2 for total column densities, 10 −9 kg m −3 for the surface densities, and 10 −9 kg m −2 for the upper tropospheric column densities.Please note the different color-scaling, by a factor of ten, for the upper tropospheric column densities between the aerosol and gas-phase tracers.

Fig. 4 .
Fig. 4. Comparison between gas-phase and aerosol pollution potentials: ELR 1 km (top panel, in %), E UT (middle panel, in %), and A 10 (bottom panel, in 10 6 km 2 ).Colored dots represent different aerosol properties: black = NS inact 1.0 µm, red = NS act 1.0 µm, gray = NS inact 10.0 µm, and blue = NS act 10.0 µm.These aerosol tracers are considered to cover the aerosol spectrum of this study and other tracers are neglected to keep lucidity.The solid black lines represent the bisecting line, the colored dashed lines the regression lines (see equations in the Appendix).

Fig. 5 .
Fig. 5.Total area covered with more than 1 % (left side) and 5 % (right side) of the global mean annual emission.Top panels display the 1.0 µm aerosol tracers, and bottom panels the 10.0 µm aerosol tracers.The dark gray shaded bars show the minimum area covered by either the NS act or NS inact tracers, the light gray shaded bars on top the additional area covered by the tracers with the other nucleation scavenging state.Display order of the MPCs is by longitude, the area shown is in units of 10 3 km 2 .

Fig. 6 .
Fig. 6.Example for land-sea fractionation in % of D 1 for the NS act 1.0 µm aerosol tracers.Dark gray bars represent the land fractions, light gray bars the sea fractions.

Fig. 7 .
Fig. 7. Folded maps of the sum of all MPC annual accumulated deposition fields (in mg m −2 yr −1 ) and the relative fractions of the three ecosystems (see also Fig. S1 in the Supplement).The left side shows the NS act 1.0 µm aerosol tracers, the right side the NS act 10.0 µm aerosol tracers.

Fig. 8 .
Fig. 8. Sum of all MPC surface densities in 10 −9 kg m −3 alone (upper panels) and convolved with the geographical population distribution (lower panels).The left side shows the NS act 1.0 µm aerosol tracers, the right side the NS act 10.0 µm aerosol tracers.

Fig. 9 .
Fig. 9. Population in millions which reside in A 10 for 1.0 µm (top panels) and 10.0 µm (bottom panels) aerosol tracers.Dark and light gray bars are defined as before.

Table 1 .
Name, location, and climate class of all MPCs used as urban point sources.Longitude and latitude correspond to the model grid box with tracer emissions.DRC = Democratic Republic of the Congo, RC = Republic of the Congo, PRC = People's Republic of China, ROK = Republic of Korea.Climate classes from Peel et al. (2007): Am = tropical, monsoon; Aw = tropical, savannah; BWh = arid, desert, hot; BSk = arid, steppe, cold;

Table 2 .
Brief description of symbols used in this study.
for the year 2000.Data is originally available at about 10 km horizontal resolution (http://www.geog.mcgill.ca/landuse/pub/Data/,as of 8 June 2012).The forest data is based on US Geological Survey Global Forest Resource Assessment(FRA 2000)and is based on images from the Advanced Very High Resolution Radiometer (AVHRR) satellite at a resolution of about one kilometer (http://edc2.usgs.gov/glcc/fao/index.php, as of 8 June 2012).

Table 3 .
Differences between EMAC and MATCH-MPIC annual values and corresponding ranks for ELR 1 km , E UT , and A 10 for the 36 MPCs.x, mean difference; x d , corresponding mean absolute deviation; with x = N i=1 x i /N , x d = N i=1 |x i |/N, and x i = x i − x, where x i is the metric value of an MPC and N is the total number of MPCs.

Table 4 .
Pearson correlation coefficients for correlations between E UT , ELR 1 km , and RCL 0.5 km , climate class resolved and for all MPCs (row All) for small (1.0 µm) and large (10.0 µm) aerosol tracers.

Table 5 .
Average area (in 10 3 km 2 ) covered with high relative deposition, D 1 and D 5 .