Analysis of aerosol absorption properties and transport over North Africa and the Middle East using AERONET data

In this paper particle categorization and absorption properties were discussed to understand transport mechanisms at different geographic locations and possible radiative impacts on climate. The long-term Aerosol Robotic Network (AERONET) data set (1999–2015) is used to estimate aerosol optical depth (AOD), single scattering albedo (SSA), and the absorption Ångström exponent (αabs) at eight locations in North Africa and the Middle East. Average variation in SSA is calculated at four wavelengths (440, 675, 870, and 1020 nm), and the relationship between aerosol absorption and physical properties is used to infer dominant aerosol types at different locations. It was found that seasonality and geographic location play a major role in identifying dominant aerosol types at each location. Analyzing aerosol characteristics among different sites using AERONET Version 2, Level 2.0 data retrievals and the Hybrid Single Particle Lagrangian Integrated Trajectory model (HYSPLIT) backward trajectories shows possible aerosol particle transport among different locations indicating the importance of understanding transport mechanisms in identifying aerosol sources.


Introduction
Natural and anthropogenic aerosols suspended in the atmosphere are characterized by their diverse sources, varying particle dynamics, lifetimes, interactive mechanisms, and surface and column distributions.Identifying mixtures containing multiple aerosol types like dust, carbon, sea salt, sulfate, or nitrogen is challenging for spaceborne and in situ observations (Chin et al., 2002;Farahat, 2016).Optical techniques are used to estimate particle sizes; however, determining particle types requires more information regarding aerosol sources, trajectories, regional topography, and atmospheric conditions.As for radiative impacts, aerosol characterization and types are major input parameters for obtaining precise climate predictions (Ramanathan et al., 1989;Satheesh and Moorthy, 2005;Farahat et al., 2015).For instance, particle absorption characteristics are used in assessing aerosol emission sources, types, and interaction phases (El-Askary et al., 2015).Aerosol categorization also helps identify dominant aerosol types over a certain geographic location, while climatology effects due to aerosols' spatial and temporal distribution is carried out by examining aerosol sizes along with their absorption characteristics using modeling, satellites, and ground-based measurements (El-Askary, 2006;Kaskaoutis et al., 2012;Aboel Fetouh et al., 2013;Vukovic et al., 2014;Sprigg et al., 2014).It is noteworthy that aerosol optical and microphysical properties can provide significant information to categorize aerosol types.For example, parameters like aerosol optical depth (AOD), the Ångström exponent (AE), aerosol loading, and particle effective radius and different viewing angles are used to depict information regarding aerosol dominant types, emission sources, dust layers, and trajectories (Reid et al., 1999;Kaskaoutis and Kambezidis, 2006;Agarwal et al., 2007;Gobbi et al., 2007;Kalapureddy and Devara, 2008;Russell et al., 2010).Analyzing the derivatives of the AE and particle effective radius is also used to attain information regarding aerosol sizes and types since the particle types are directly correlated with sizes and optical properties (Gerasopoulos et al., 2003(Gerasopoulos et al., , 2011;;Alados-Arboledas et al., 2003).The Aerosol Robotic Network (AERONET) data retrievals have been used widely to determine dominant aerosol types and categorization within mixing scenarios through investigating particle size distribution and optical and microphysical properties (Kaufman et al., 1994;Holben et al., 1998;Omar et al., 2005;Qin and Mitchell, 2009;Russell et al., 2010).These studies helped partition major aerosol types including dust, anthropogenic industrial pollution, and mixed and biomass burning aerosols.The absorption Ångström exponent (AAE), defined as the dependence of aerosol optical depth on wavelength as well as single scattering albedo (SSA), defined, as the ratio of scattering to extinction efficiency of aerosol particles, is used to categorize aerosol types.For instance, SSA for dust particles containing hematite and clay results in small absorption over the visible to near-infrared band and strong absorption around 440 nm (Sokolik and Toon, 1999).On the other hand, for organic carbon (OC), the SSA increases with increasing wavelength where the OC exhibits strong absorption in the visible and ultraviolet range, while for aerosols composed of black carbon (BC) particles, the SSA is inversely proportional to the wavelength.Sulfate and other hygroscopic aerosols do not show significant SSA spectral dependence (Dubovik et al., 2002).Therefore, mixed aerosol containing sulfates, BC, and OC could produce an indistinct SSAwavelength dependence due to the varying spectral effects of aerosol mixtures in the atmosphere (Dubovik et al., 2002).
In this study we investigate aerosol characteristics over North Africa and the Middle East using eight AERONET sites in the region, namely, Solar Village (24 , and Ben Salem (35 • N, 9 • E) (Fig. 1).The analysis was based on data availability (at least 4 years of successive data measurements from 1999 to 2015) except for the Ben Salem site where the analysis was based on 2013-2015 data only.Previous studies (Table 1) classified dominant aerosol types at some of these sites based on their geographic locations into clear, dust, pollution, and mixed; however, the effect of seasonality and aerosol transport was not fully investigated for those sites.Dust and biomass particles could be dominating over some sites; however, other aerosol types could also exist over the site during the analysis period (Takemura et al., 2000;El-Askary and Kafatos, 2008).Although previous studies categorized sites based on their most dominant aerosol particles, they did not explicitly explore the variability in aerosol types over each site.As seasonal changes, major dust events, and human activities could largely affect dominant aerosol types (El-Metwally et al., 2008).Each location could have more than one dominating particle category during the year.
In this work, we analyze aerosol volume size parameters and compare them with single scattering albedo at each site.The absorption parameters are calculated at various wavelengths.Last, aerosol variability and possible transport from one site to another is investigated.This study highlights this variability in aerosol categorization over each site based on these factors.It also shows possible aerosol transport between sites.
2 Instrumentation, data, and methodology AERONET is an array of sun photometers globally distributed to measure columnar spectral AOD and water vapor in a wavelength range of 340 to 1640 nm and temporal resolution of 600 to 900 s.The network also retrieves columnar optical aerosol properties (e.g., aerosol size distribution, volume mean radius, volume concentration, and multi wavelength single scattering albedo at 440, 675, 870, and 1020 nm (Holben et al., 1998)).This takes place by fitting measurements of the spectral AOD and sky radiances to radiative transfer calculations (Dubovik and King, 2000).AERONET data retrievals comprise 1-2 % estimated uncertainty (Eck et al., 1999) with the highest uncertainty near UV wavelengths (Holben et at., 1998).We used the AERONET Version 2, Level 2.0 products that contain retrievals for 116 different aerosol parameters including, aerosol volume size distribution (AVSD; dV (r) / dlnr (µm 3 / µm 2 )) retrieved in 22 logarithmically equidistant radial bins spanning the range of particle radii 0.05 µm ≤ r ≤ 15 µm, the real and imaginary parts of the refractive index (CRI-R(λ), CRI-I(λ)), and the optical parameters (AOD (λ) and SSA) centered at four wavelengths (440, 675, 870, and 1020 nm;Taylor et al., 2014).Version 2 inversion products also provide the mean geometric radii of the fine and coarse modes, their standard deviation, and their volume concentrations.This is of particular importance for our analysis owing to the mixing scenarios observed over North Africa and the Gulf region.
The AERONET inversion code provides aerosol optical properties by measuring spectral direct beam and diffuse solar radiation.Water vapor and columnar spectral AOD characteristics can be retrieved from the AERONET direct-sun measurements with a temporal resolution of ∼ 600-900 s and 1-2 % typical AOD uncertainty (Holben et al., 1998;Eck et al., 1999).Depending on the site, the data are typically provided at a wavelength range of 340 to 1640 nm, together with Ångström exponent α over certain wavelength ranges.AERONET pre-version 1 (pre-release of Version 1, 2003) data have been applied to categorize aerosols based on their optical characteristics (Dubovic et al., 2002, andRussell et al., 2010).The AERONET Version 2 data, 2006, provided more precise calculations of reflectance over bright surfaces compared to Version 1.For example, in Saudi Arabia, Bahrain, and the United Arab Emirates, Version 2 provided more consistent SSA over small islands vs. vast bright desert, with an SSA difference of less than 0.01 compared to 0.03 for Version 1.
AERONET data retrievals Version 2, Level 2.0 (Smirnov et al., 2000(Smirnov et al., , 2002) ) are used to derive the extinction Ångström exponent (α ext ) using the measured aerosol optical properties and AOD spectral dependence (τ ext ) with wavelength (λ) A. Farahat (Ångström, 1964): Fine-and coarse-mode particles can be identified by calculating the linear regression of ln (τ ext ) vs. ln (λ), where values close to 2 indicate fine mode and 0 represents coarse mode (Nakajima et al., 1989).The absorption Ångström exponent (α abs ) can also be calculated using Aerosol particle physical characteristics affect α abs , for example α abs varies from less than 1.0 and up to 1.6 for large black carbon particles (r > 0.1 µm) based on their core size and coating structure (Kondo et al., 2009).Black carbon particles are characterized by their small radius and spherical shapes, which results in α abs values of ∼ 1.0 (Petzold et al., 1997).Dubovik et al. (2002) also showed that α abs varies between 1.2 and 3.0 for dust, 1.2 and 2.0 for pollution (biomass only), and 0.75 and 1.3 for mixed aerosols.

Aerosol categorization using SSA
Aerosol absorption properties over the eight locations under investigation are represented in Fig. 2  The small SSA variability at those three sites is attributed to physical (size and shape) similarities between particle grains produced by the vast sand area surrounding the sites.The SSA standard deviations calculated for this study are 0.004 lower and 0.008 greater than Dubovik et al. (2002) for the Solar Village and Bahrain sites, respectively.The difference between Table 2 and Dubovik et al. (2002) for these two sites is due to applying improved Level 2.0 AERONET data retrievals and utilizing a larger data set.Pure dust could be dis- tinguished for values of α ext less than 0.2 (Kim et al., 2011).Solar Village, Bahrain, and Mezaira have α ext ≥ 0.3, indicating a possible incursion by other aerosols.This is more recognized at the Bahrain site due to the presence of fine particles produced from industrial activities similar to other locations (Gathman, 1983;Gras, 1995;Kalapureddy and Devara, 2008;Park et al., 2015).Cairo is known for its widespread biomass burning activities observed during fall when episodes of ash burning from agriculture waste take place (El-Askary and Kafatos, 2008; Marey et al., 2010Marey et al., , 2011)); however, other factors could also contribute to pollution, such as traffic and industry.Compared to other particle types, the biomass particles are known to have the largest SSA variability due to various combustion phases and fuel types (Eck et al., 2003); however, only a small SSA difference (∼ 0.013) is observed between Cairo and the Solar Village, Bahrain, and Mezaira site.This is an indication of a possible mixing pattern between biomass and dust over Cairo (El-Askary and Kafatos, 2008;El-Metwally et al., 2008).
The similarity in the SSA characteristics between the Saada, Tamanrasset, and Ben Salem sites, on the one hand, and Solar Village, Bahrain, and Mezaira (Fig. 2) with a strong absorption at 440 nm, on the other hand, is an indication of the dust dominance contribution to the Saada, Taman-rasset, and Ben Salem sites.The increased absorption at these three sites could also be due to increased hematite percentage in the dust that could lead to an increased absorption in the blue to near-infrared wavelength band (Sokolik and Toon. 1999).The average SSA values for all locations are ∼ 0.91 (440 nm), 0.94 (675 nm), 0.95 (870 nm), and 0.95 (1020 nm) and reveal insignificant changes with respect to the SSA for all sites (Fig. 2, and Table 2).This is a sign of possible mixing scenarios through transporting aerosols between different sites.
Figure 2d-f show that both small and large particles exist with large particles dominating , as is shown by the two peaks at ∼ 0.1 and ∼ 4 µm with high dust domination over the Solar Village, Bahrain, and Mezaira locations.The high concentration of small size particles with ∼ 0.1 µm over Cairo is an indication of high pollution over this location.
The above results can give a general indication regarding aerosol categorizations at different locations and possible particle transport mechanism; however, they do not show the seasonality effect on particle dominance over a certain location.Long-range frequency of occurrence of clean, mixed, dust, and pollution aerosol categories along with their AOD and Ångström exponent measurements over the Solar Village, Bahrain, Mezaira, Sedé Boqer, Cairo, Saada, Tamanrasset, and Ben Salem stations can provide detailed information of the aerosol categorization and transport patterns among sites.Tables 3-6 and Fig. 4a-c show the effect of seasonality in determining the dominant aerosol particles at each location.Table 3 shows that natural dust particles dominate over the Solar Village, Bahrain, and Mezaira sites during March-June where dust storms events are active during these months.Anthropogenic aerosols dominate over the sites mostly during September-December (Table 4) due to common windstorms (Said and Kadry, 1994) blowing during these months, which help in transporting pollutants and other anthropogenic aerosols.A mixed aerosol pattern is observed all year around over the three sites but mostly during July and August (Table 5), and the air is mostly clean during November and December (Table 6).AOD values and Ångström exponents played a major role in defining the abovementioned four aerosol categories represented in Tables 3-6.Data showed that high pollution over Cairo during the autumn months (Table 4, Fig. 4e) as a result of yearly events where farmers burn their leftover rice straw, causing severe pollution and 2 to 3 months of potential complications for respiratory and heart disease patients (Marey et al., 2010).It is interesting to observe similar pollution patterns over

Pollution
Dust particles dominate over the Saada site (Table 3, Fig. 4f) during July and August, which is not the usual sand storm season (February-April) over Morocco (Goudie and Middleton, 2001).However, this could be attributed to the effect of single dust events sweeping off the coast of Morocco, like the ones on 15 August 2005 and on 7 August 2015.
March is the month with the most frequent dust activity as major dust storms frequently blow during this time of the year.This would also indicate a higher probability of dust transportation within Saudi Arabia and the Arabian Peninsula during March.The major dust event that took place in March 2009 over Saudi Arabia could have contributed to the high AOD observed during this period, with a widespread heavy atmospheric dust load which was reported in Alharbi et al. (2013) and Farahat et al. (2016), who also suggested that the major plume of the March 2009 dust outbreak originated from several dust source areas extending across two regions -the Qasīm region lying some 500 km northwest of Riyadh within an active dust source region (Alharbi, 2009) and the Ad Dibdibah and Aş Şummān Plateau region, which is a major source of frequent dust storms in Saudi Arabia.Parts of it are in Iraq, and the south of Kuwait covers the northeastern part of the Ad Dahnā Desert.Alharbi et al. (2013) and Farahat et al. (2016) also reported that this storm was associated with an increase in AOD, wind speed, and a reduction in temperature and visibility, for few days following the storm.Similar AOD pattern was observed over the Bahrain and Mezaira stations with peak AOD observed in March 2012 (0.59) and June 2010 (0.49) over Mezaira and in May 2006 (0.45) over Bahrain.It is important to mention that the Bahrain station did not produce any data after October 2006.
Based on the available data, dust dominates over Cairo during March and April, which is coincident with khamsin storms, in which major dust plumes are transported from the Sahara to eastern Europe.It is clear that most of the year pollution episodes took place over Cairo with peak pollution during November-January with another peak (α ext ) observed in January 2006(1.35), November 2005(1.30), and December 2004 (1.29).This could be attributed to the activity of Egyptian farmers burning rice straw and other agricultural waste during this time of the year.As a result pollution events during these months could be contributing to direct and indirect radiative forcing by anthropogenic aerosols.These results agree with El-Askary et al. (2009) where peaks of biomass burning were found to occur in September (2001 and 2002) and in October (2003 and2004) over the greater Nile Delta region as observed from the MODIS Terra satellite.El-Askary et al. (2009) also indicated that the long-range pollution observed over Cairo could be attributed to transport from southeast Europe.
Pollution dominates over the Solar Village, Bahrain, and Mezaira sites from October to December with α ext during October 2012 (1.12), October 2007 (1.06), December 2010 (1.04), and December 2006 (1.02) over the Solar Village;December 2004(1.29), October 2005 (1.22), December (1.21), and November 2005 over Bahrain;and November 2010(1.39), January 2010(1.35), November 2008 (1.25), and January 2014 (1.24) over Mezaira.Pollution over the three sites is in general attributed to a boom in the oil industry accompanied by large infrastructure requirements, the establishment of major cement industry in the area, an unprecedented high economic growth rate, low energy costs subsidized by the government, and a population increase.These all contribute to Saudi Arabia, Bahrain, and UAE environmental pollution (Farahat, 2016).Winter pollution dominates from October to December due to a reduction in dust blowing events over these months, which is an indicator of less pollution transport during that period.Clean scenarios are observed over the Solar Village, Bahrain, and Mezaira mostly during the summer season; this is probably due to less traffic pollution as schools are off and many locals travel to cooler places to escape high summer temperatures in this  2008 (1.45), andDecember 2011 (1.26) over Saada; and August 2013 (1.04) over Ben Salem.Microparticle transport plays a major factor in changing aerosol types at different AERONET sites.In this article, we try to shed light on a possible aerosol transport between two AERONET sites, namely Cairo and Sedé Boqer.The two sites are chosen based on their geographic locations to demonstrate aerosol transport from North Africa to the Middle East.Aerosols volume size distribution on 13 November 2011 over the Cairo site and on 14 November 2011 over the Sedé Boqer site (24 h later) clearly shows aerosol transport between the two sites (Fig. 5).The Hybrid Single Particle Lagrangian Integrated Trajectory (HYSPLIT) back trajectory model confirms the possible aerosol transport from Cairo to Sedé Boqer.Gridded meteorological data, at regular time intervals, are used in the calculation of air mass trajectories.For the back-trajectories, data are obtained from existing archives.A complete description of input data, methodology, equations involved, and sources of error for the calculation of the air mass trajectory is presented by Draxler and Hess (1997).Figure 5a shows the air mass back trajectory at Sedé Boqer during the black cloud local pollution episode over the Delta Region at different vertical elevations to highlight the vertical extent of local transportation (Aboel Fetouh et al., 2013;Prasad et al., 2010).El-Askary and Kafatos (2008) found that an inversion layer contributes to the trapping of aerosols and pollutants, resulting in an increase of their concentration and, hence, creating a permanent haze that can develop into a health hazard.Figure 5b and c show the SSA characteristics over both locations, implying the transport from Cairo location to Sedé Boqer.SSA values at the finer radius size were high on the 13th over Cairo and low over Sedé Boqer; then SSA values increased on the 14th over Sedé Boqer while showing a drop on the 15th over Cairo.This behavior suggests aerosol transport from Cairo to Sedé Boqer during the period under investigation.The preliminary results here demonstrate a possible aerosol transport between different sites; however, a complete spatial and temporal analysis is needed to understand the effect of this transport on aerosol types.

Conclusions
A combination of ground-based aerosol-related parameters was used in this analysis to study different sources of aerosol loadings over eight different cities in North Africa and the Gulf region.Our analysis involved a study of the AOD and SSA, as well as other derived parameters and trajectory models.Natural vs. anthropogenic aerosols are well distinguished using our derived Ångström exponent.We also identified possible mixed and clean atmospheric conditions using the calculated α675/440 values.It is clear that dust episodes dominate the spring season, as reflected by the high values of AOD associated with sharp drops in the Ångström exponent.Back trajectory analysis shows agreement with these findings, thereby confirming that the Sahara is the major source of such aerosols during the spring season.Furthermore, the sharp peaks in the Ångström exponent at the Cairo station suggests a possible long-range transport of pollutants from Europe during the summer season as previously discussed by El-Askary et al. (2009).Local pollution is most prominent during the fall and winter seasons, as confirmed by the data presented in the tables and associated figures.In summary, it is clear that the region under investigation is exposed to the impacts of aerosols from various sources throughout the year.This is becoming a serious health hazard because of the persistence of high pollution throughout the year or for prolonged time periods.

Data availability
All data used in this study are publicly available at the Aerosol Robotic Network (AERONET, 2016) homepage: http://aeronet.gsfc.nasa.gov/.

Table 1 .
Previous studies categorizing aerosol types over some of the AERONET sites presented in this work.
Dubovik et al. (2002)ectral behavior agrees withGiles et al. (2012)andDubovik et al. (2002)but with an average decrease of 0.02 at the Solar Village and Bahrain sites.The averaged optical properties, with aerosol data listed and volume size distribution at each site (Table2), shows there is a lower SSA variability at Solar Village, Mezaira, and Bahrain compared to the other sites.

Table 2 .
Aerosols absorption and physical properties categorized by aerosol types, for sample stations, using AERONET Version 2, Level 2.0 data retrievals.Some locations may experience other aerosol types during different seasons (Cairo, for instance).The spectral single scattering albedo (SSA) is listed first followed by the standard deviations based on monthly values.EMA: Egyptian Meteorology Authority.

Table 3 .
Seasonal dust domination over study locations.EMA: Egyptian Meteorology Authority.

Table 4 .
Seasonal pollution domination over study locations.