Aerosol Spectral Absorption in the Mexico City Area: Results from Airborne Measurements during Milagro/intex B

This paper presents estimates of the spectral solar absorption due to atmospheric aerosols during the 2006 MILAGRO/INTEX-B (Megacity Initiative-Local And Global Research Observations/Phase B of the Intercontinen-tal Chemical Transport Experiment) field campaign. The aerosol absorption was derived from measurements of the spectral solar radiation and the spectral aerosol optical depth made on the J31 aircraft flying over the Gulf of Mexico and over Mexico City. We present the spectral single scattering albedo (SSA) and aerosol absorption optical depth (AAOD) for two flights over the Gulf of Mexico and three flights over Mexico City for wavelengths from 350 to approximately 1650 nm. The spectral aerosol optical properties of each case are different and illustrate the variability of the aerosol optical properties in the Mexico City area. The results can be described in terms of three different wavelength regions: The 350–500 nm region where the aerosol absorption often falls off sharply presumably due to organic carbonaceous particles and windblown dust; the 500–1000 nm region where the decrease with wavelength is slower presumably due to black carbon; and the near infrared spectral region (1000 nm to 1650 nm) where it is difficult to obtain reliable results since the aerosol absorption is relatively small and the gas absorption dominates. However, there is an indication of a small and somewhat wavelength independent absorption in the region beyond 1000 nm. For one of the flights over the Gulf of Mexico near the coastline it appears that a cloud/fog formation and evaporation led to an increase of absorption possibly due to a water shell remaining on the particles after the cloud/fog had dissipated. For two of the Mexico City cases, the single scattering albedo is roughly constant between 350–500 nm consistent with other Mexico City results. In three of the cases a single absorption Angstrom exponent (AAE) fits the aerosol absorption optical depth over the entire wavelength range of 350 to 1650 nm relatively well (r 2 > 0.86).

Abstract. This paper presents estimates of the spectral solar absorption due to atmospheric aerosols during the 2006 MILAGRO/INTEX-B (Megacity Initiative-Local And Global Research Observations/Phase B of the Intercontinental Chemical Transport Experiment) field campaign. The aerosol absorption was derived from measurements of the spectral solar radiation and the spectral aerosol optical depth made on the J31 aircraft flying over the Gulf of Mexico and over Mexico City. We present the spectral single scattering albedo (SSA) and aerosol absorption optical depth (AAOD) for two flights over the Gulf of Mexico and three flights over Mexico City for wavelengths from 350 to approximately 1650 nm. The spectral aerosol optical properties of each case are different and illustrate the variability of the aerosol optical properties in the Mexico City area.
The results can be described in terms of three different wavelength regions: The 350-500 nm region where the aerosol absorption often falls off sharply presumably due to organic carbonaceous particles and windblown dust; the 500-1000 nm region where the decrease with wavelength is slower presumably due to black carbon; and the near infrared spectral region (1000 nm to 1650 nm) where it is difficult to obtain reliable results since the aerosol absorption is relatively small and the gas absorption dominates. However, there is an indication of a small and somewhat wavelength independent absorption in the region beyond 1000 nm.
For one of the flights over the Gulf of Mexico near the coastline it appears that a cloud/fog formation and evapora-Correspondence to: R. W. Bergstrom (bergstrom@baeri.org) tion led to an increase of absorption possibly due to a water shell remaining on the particles after the cloud/fog had dissipated. For two of the Mexico City cases, the single scattering albedo is roughly constant between 350-500 nm consistent with other Mexico City results. In three of the cases a single absorption Angstrom exponent (AAE) fits the aerosol absorption optical depth over the entire wavelength range of 350 to 1650 nm relatively well (r 2 > 0.86).

Introduction
One of the largest climate uncertainties continues to be the radiative forcing due to atmospheric aerosols. A substantial fraction of that uncertainty is associated with the scattering and absorption of solar radiation by aerosols in cloud-free conditions, the so-called direct aerosol effect. In particular, the spectral absorption of solar radiation by atmospheric aerosols has been difficult to quantify. Use of an Angstrom Absorption Exponent (AAE, defined as the negative of the slope of a log-log plot of the aerosol absorption optical depth (AAOD) versus wavelength) has had some success in describing atmospheric aerosol absorption for certain aerosol types (Bergstrom et al., 2007). In general, black carbon (BC or light absorbing carbon, LAC) has an AAE near 1.0 while organic carbon (OC or  This paper discusses the spectral aerosol absorption measured during the MILAGRO campaign in March 2006 and is one of a series of papers (Coddington et al., 2008;Livingston et al., 2009;Redemann et al., 2009;and Schmidt et al., 2009) based on the data taken by the Solar Spectral Flux Radiometer (SSFR) and the 14 channel Ames Airborne Tracking Sunphotometer (AATS) instruments aboard the J31 aircraft. Coddington et al. (2008) presented results for the surface albedo in Mexico City and compared them with MODIS retrieved surface albedos.  compared the aerosol optical depth measurements measured by the AATS with satellite retrievals. Redemann et al. (2009) compared J31 AATS measurements of AOD and related aerosol properties to results from MODIS-Aqua and MODIS-Terra, with emphasis on differences between the older MODIS Collection 4 (C4) and the new Collection 5 (C5) data set. Schmidt et al. (2009) present a new method of determining the aerosol radiative forcing and values for the aerosol radiative forcing above sea and land surfaces.
Mexico City is a very large urban area and has a wide array of aerosol sources producing particle types that can absorb solar radiation including wind-blown mineral dust, LAC or BC from biomass burning and transportation sources, and a significant amount of organic matter (OM). Previous studies have shown that the aerosol in the Mexico City atmosphere is a complex and highly variable mixture that presents a difficult challenge to the determination and interpretation of the aerosol optical properties (Barnard et al., 2008;Marley et al., 2009a, b;Corr et al., 2009;Adachi and Buseck, 2008;Shinozuka, et al., 2009).
In this paper we use the term absorption to mean the amount of solar irradiance that is absorbed in a particular layer of the atmosphere. The fractional absorption is then the absorption divided by the solar irradiance incident on the top of the layer. The aerosol absorption optical depth (AAOD) of the layer is the aerosol extinction (absorption + scattering) optical depth (often termed just optical depth) multiplied by the co-albedo (which is 1 minus the single scattering albedo, SSA) of the layer.

Aircraft measurements
During MILAGRO in March 2006, the Jetstream 31 aircraft (J31) flew 13 missions from Veracruz, Mexico. The flights were either over the Gulf of Mexico or over the Mexico City area. The J31 daily mission summaries are at http: //www.espo.nasa.gov/intex-b/flightplanningJ31.cgi and a table summarizing flights is in Molina et al. (2010). The SSFR and the AATS were mounted on the J31.

SSFR spectral solar radiant flux measurements
The SSFR is a moderate resolution (8-12 nm) spectrometer that spans the wavelength range from 350-2100 nm. The downward flux (or irradiance) was corrected for the changing aircraft attitude with respect to the horizontal plane and for the angular response of the cosine-weighting integrating sphere (the optical collector for the SSFR). The cosine response was measured before each experiment in the laboratory to correct for non-linearities. The upward flux was corrected with the cosine-weighted response integrated over the lower hemisphere.
Pre-and post-mission, the SSFR was radiometrically calibrated against a NIST traceable 1000 W lamp. Field calibrations were performed to monitor the stability of the SSFR over the experiment using a 200 W LI-COR Field Calibrator. Spectral calibration was achieved by referencing lines from a Hg lamp. The SSFR RMS uncertainty is 3-5% over the SSFR spectral range. Both Coddington et al. (2008) and Schmidt et al. (2009) discuss the SSFR measurements of downwelling and upwelling solar irradiance made during the MILAGRO campaign.

AATS optical depth measurements
The AATS was also integrated on the J31 and measured aerosol optical depth (AOD) from flight level to the top of the atmosphere at 13 solar wavelengths (in the region of 354-2139 nm) and one wavelength for columnar water vapor (CWV) . Vertical differentiation of AOD and CWV data obtained during J31 vertical profiles yields vertical profiles of multi-wavelength aerosol extinction and water vapor concentration, respectively. AOD uncertainties, calculated for each AATS data point, include four error sources: calibration, gas subtraction, detector output measurement, and airmass.

Analysis
The data from the SSFR and the AATS were combined and a radiative transfer model was used to determine the spectral aerosol absorption properties (Bergstrom, et al., 2003(Bergstrom, et al., , 2004 3.1 Radiative transfer model Coddington et al. (2008) and Coddington (2009) describe the recent improvements to a 1-D radiative transfer model (Bergstrom et al., 2003) designed for use in conjunction with the SSFR measurements. The major improvement is expanding the model from 140 bands of 10 nm width covering Atmos. Chem. Phys., 10, 6333-6343, 2010 www.atmos-chem-phys.net/10/6333/2010/ R. W. Bergstrom et al.: Aerosol spectral absorption in the Mexico City area 6335 300-1700 nm to 2201 bands of 1 nm sampling resolution that cover a wavelength range of 300-2500 nm. The model uses: 1. Correlated k-distributions for oxygen, ozone, carbon dioxide, water vapor, and methane for the molecular absorption coefficients.
2. Rayleigh optical depth for an atmosphere containing 370 ppm CO 2 calculated by numerical approximation.

SSFR slit functions.
The details of the absorption coefficient generation, sorting into distribution functions, and the accounting for the spectral resolution of the instrument filter are described in Coddington et al. (2008) and Coddington et al. (20010). The significant absorbing gas species (including overlapping species) in the spectral regions are listed in Table 2 of Coddington et al. (2008). For this study we added the NO 2 absorption coefficients at each 1 nm band between 350 and 500 nm.

Flux divergence method
The flux divergence method to determine aerosol absorption can be described simply but is difficult to accomplish in practice. The net solar flux (downward minus upward) at the top and bottom of an aerosol layer are measured. The difference in the net solar flux across the layer is then the absorption in the layer (Chandrasekhar, 1960). Basically, this is just applying conservation of energy to a vertical layer in the atmosphere. There are a number of difficulties with this method, such as the horizontal inhomogeneity of the aerosol layer. Also, in the Mexico City area it was not possible to fly completely below the Mexico City plume (due to flight restrictions) so that the plane flew at some distance above the surface inside the aerosol layer. Therefore the results for Mexico City are for the upper part of the urban plume.

Measurement uncertainty
The SSFR instrument accuracy is discussed in Coddington et al. (2008) and Schmidt et al. (2009) and the AATS instrument accuracy is discussed in e.g., Russell et al. (2007) and Livingston et al. (2009). Bergstrom et al. (2003) present a more detailed analysis of the flux divergence method. They derive an equation for the error in the single scattering albedo for the case of single scattering and small surface albedo. The following equation relates the error in the SSA, ω to the uncertainty in the fractional absorption (the absorption divided by the incident solar flux), α, and the extinction optical depth, τ : where µ 0 is the cosine of the solar zenith angle. In general, the uncertainty estimates in the measured fractional absorption are about 0.01 (about 10% of the typical aerosol absorption in the shorter wavelengths) and the typical uncertainty in AATS-measured AOD is also ∼0.01. However, as the fractional absorption becomes small the uncertainty in the single scattering albedo becomes large. Equation (1) was used to estimate the error in the single scattering albedo. The error in the AAOD is then the sum of the error in the SSA and the AOD. Coddington et al. (2008) discuss the uncertainties in the new radiative transfer computer program. Since the flux divergence is only a weak function of the asymmetry factor (Bergstrom et al., 2003), we use the method to determine aerosol single scattering albedo (SSA). The asymmetry factor was assumed to be 0.7 at 500 nm.

Model uncertainty
In the region from 500 nm to 2000 nm errors in the modeled amount of absorbing gases have an effect on the determination of the aerosol absorption properties. Figure 1a shows the absorption computed for 10 March case (discussed below) and the absorption computed for the same case with a 10% increase in the absorbing gases: water (H 2 O), carbon dioxide (CO 2 ) and ozone (O 3 ). The difference in the absorption is difficult to see in Fig. 1a. However, Fig. 1b shows the percentage change in the absorption. Surprisingly, in the 500-2000 nm region there are very few wavelengths that are independent of the uncertainties in the absorbing gases. Even in the window regions between the water vapor bands, there is significant dependence on the gas amount, apparently related to the so-called continuum absorption of water vapor (where the strong lines of the absorption band influence the region between the bands) and the overlap of other gases. We computed the aerosol absorption properties only at the wavelengths shown in Fig. 1 to attempt to minimize the effect of uncertainties in the gas amounts. As a result the analysis had very narrow wavelength spacing (1 nm) in the 350 to 557 nm region and much larger spacing from 557 to 1622 nm. Another way to eliminate the effects of differences in measured and modeled gas amounts is to solve for the gas amounts using all the wavelength information and minimize the least squared error between the measurements and predictions. We are exploring this approach and if possible will use it in the future.
For flights over Mexico City, the amount of O 3 and nitrogen dioxide (NO 2 ; a prominent industrial air pollutant) are important inputs to the model (Barnard et al., 2008  For NO 2 we used the OMI NO 2 column values as an estimate and then scaled the NO 2 amount to have the same vertical profile as the aerosol optical depth and assumed that all the NO 2 was in the urban plume. In general, for aerosol optical depths greater than about 0.1 the uncertainty in the single scattering albedo is about 0.02. However, for low optical depths (particularly in the region beyond 1000 nm) the uncertainty in the single scattering albedo can be as high as 0.1. The uncertainty in AAOD is roughly 0.01 and in the region beyond 1000 nm, AAOD values are often below 0.01 making accurate estimates difficult.

Results
We present results from five separate flights (2 flights over the Gulf of Mexico -10 and 13 March and 3 flights over Mexico City -6, 15, and 19 March.  Fig. 2b shows that the cloud/fog had cleared and a residual part of it had moved somewhat further off shore. Starting at the location marked A the J31 flew at a low altitude over the water, then ascended through the residual cloud/fog, then descended to just above the water, and then continued in a northeasterly direction. At the location marked B the plane spiraled up and flew back toward the coast at an altitude of 5 km (above the aerosol layer). What appears in Fig. 2b as a single line between A an B is actually two lines, one beneath the other representing the lower and upper passes. SSFR and AATS data were obtained above and below the aerosol layer along the region between A and B. We hypothesize that the region from point A to the location of the cloud/fog remnant in Fig. 2b had been affected by the cloud/fog layer shown in Fig. 2a and that the region east of the cloud remnant in Fig. 2b was unaffected by the cloud. The layer average absorption, SSA and AAOD for the flight path west of the Fig. 2b cloud remnant and east of it are shown in Fig. 3a-d. Figure 3a shows that there is slightly more absorption west of the Fig. 2b cloud remnant than east of that cloud. However, the single scattering albedo is higher west of the cloud than east of the cloud (Fig. 4b) meaning that there was relatively more scattering and higher optical depth for the aerosol that was previously in the cloud/fog of Fig. 2a. We hypothesize that this is due to water remaining on or in the aerosol droplets. Such a water coating would increase absorption coefficient (e.g., Redemann et al., 2001;Schwarz et al., 2008) in the flight path west of the Fig. 2b cloud remnant (as indeed was observed and is shown in Fig. 3c). The increase in absorption (about 30%) is consistent with recent results of Schwarz et al. (2008) for coated black carbon particles. Both the fact that the increase in scattering is larger than the increase in absorption and that the ratio of the two is a function of wavelength (shorter wavelengths are affected more than longer wavelengths) suggests these results are consis-tent with shell-core calculations (Bond et al., 2006;Redemann et al., 2001). Although there were no measurements of the amount of water actually on the aerosol particles, the results are interesting.

Gulf cases: 10 and 13 March
Fitting a straight line (constant AAE) to the entire spectrum shown in Fig. 3c does not do a good job of describing the absorption coefficient, particularly in the 350 nm region (r 2 =0.74). The wavelength dependence of the absorption optical depth appears to be different within three distinct regions: 350 to roughly 500 nm, 500 to 1000 nm, and 1000 to 1700 nm. Figure 3d shows the AAE fit for the 350 to 500 nm region (r 2 =0.96) and the 500 to 1000 nm (r 2 =0.94). These r 2 values (also listed in Table 1) are notably larger than the r 2 of 0.74 obtained for a single-AAE fit for the entire wavelength range, indicating a much better fit. The spectral AAOD behavior shown in Fig. 3 is similar to many recent results where the aerosol is composed of OM and BC (e.g. Martins et al., 2009 andBarnard et al., 2008) where the AAE is larger in the UV region and then near 1.0 in the visible to near IR region. As shown in Table 1, the extinction Angstrom exponents for this day were 1.4 and 1.1 west and east, respectively, of the remnant cloud in Fig. 2b, indicating relatively large particles (i.e. dust) so that the aerosol is perhaps a mixture of dust, OM and BC. It is important to keep in mind that deriving a constant AAE over a particular wavelength region is a simplification and is useful in that the AAE appears to be an indicator of the absorbing aerosol chemical species. This does not mean that a constant AAE is the best fit for a particular set of data. Higher order fits may do better in reducing the error in fitting the data. Also, it remains to be seen if a constant AAE is the best way to parameterize the aerosol absorption coefficient in global aerosol models.
The results for the other Gulf case (10 March) are shown in Fig. 4a and b and Table 1. The top of the layer was at about 4 km above the surface. For this case the SSA increases with wavelength and a single AAE fit is relatively good with value of 2.6 (r 2 =0.89). However, the AAOD values between 1200 and 1600 nm are significantly less than 0.01 and exhibit a large spread. The extinction Angstrom exponent for this day was 0.8 indicating fairly large particles. Comparing the 13 March case with the 10 March case shows the difficulty in making generalizations about the atmospheric aerosol given only a few cases. In the 13 March case the SSA generally decreases with wavelength while for the 10 March case the SSA increases with wavelength. In the 13 March case the AAE is not constant with wavelength while for the 10 March case the AAE is relatively constant with wavelength.
Both Figs. 3c and 4b show an absorption feature at 480 nm. This appears to be a water vapor band that is not well characterized by the model.

Mexico City area cases: 6, 15 and 19 March
We estimate the aerosol spectral absorption and absorption optical depth values for three days when the plane flew over the Mexico City area, 6, 15 and 19 March. Each day's results are somewhat different.
To show the variation in absorption with the thickness of the aerosol layer above the plane, the fractional spectral absorption for March 6th is shown in Fig. 5 for differences in net flux (irradiance) between 3367 m above the surface and the heights (250, 550 and 1850 m above the surface). As the plane flew closer to the surface the absorption due to both aerosol and water vapor increased. However, the fractional absorption due to water vapor over Mexico City (Fig. 5) is less than that over the Gulf (Fig. 3a) since, at 2240 m above sea level, there is less water vapor in the Mexico City atmosphere and also because the plane was well above the surface.
The SSA and AAOD for the three cases over Mexico City are shown in Fig. 6a-f. The March 6th results for two of the aerosol layers are shown in Fig. 6a and b. Figure 6a shows the SSA is relatively constant or even decreases between 350 and 500 nm then increases after 500 nm. This spectral behavior is similar to other results for Mexico City reported by Barnard et al. (2008), Marley et al. (2009b) and Corr et al. (2009), who point to absorption by organic material as the most likely explanation. (Another possibility is dust combined with BC (Jeong and Sokolik, 2008).) The error bars are quite large for wavelengths greater than 700 nm since the optical depth was relatively small as the plane was in the upper part of the urban plume. The aerosol absorption optical depths in Fig. 6b also show a change in slope at about 500 nm with a steeper slope for wavelengths >500 nm. The absorption optical depth past 1000 nm is below 0.01 and somewhat constant. In spite of the aforementioned slope change, a constant AAE fits the spectrum relatively well (AAE= 2.2, r 2 =0.97; and AAE 1.9, r 2 =0.97).
The 15 March case in Fig. 6c and d is derived from the flux difference between the net flux at 3 km and the net flux at 250 m over T0,.The results are similar to the 6 March case in that the SSA is relatively constant from 350 to 500 nm. However, the SSA in this case is larger and decreases at wavelengths >700 nm, but again with large error bars at those wavelengths. The aerosol absorption optical depth falls off at a relatively constant slope between 350 and 1000 nm. The AAE value is 1.4, which is somewhat larger than expected from BC only suggesting that the aerosol is mostly BC with some OM.
The 19 March flight was interesting and is discussed in detail in Livingston et al. (2009). The winds over Mexico City were quite strong and resulted in a large amount of windblown dust, particularly at T2 (a rural area 63 km northeast of T0). The visibility near T2 was very poor and the plane flew at about 420 m above ground level (AGL) near T2. The atmosphere was somewhat clearer at T0 (it was a Sunday) and the plane flew at 600 m above T0. The SSA and AAOD results shown in Fig. 6e and f are for the flux differences between the net flux at 4 km above the surface and the net flux at 400 m above T2 and 600m above T0. Despite the difference in altitude the optical depths above the plane for the aerosol layers above T2 and T0 were similar ). For both the aerosol layers above T0 and T2 the SSA increases between 350 and 500 nm and then is fairly constant. The AAOD decreases rapidly between 350 nm and about 700 nm and the AAE for T2 is larger than T0 indicating perhaps the effect of BC combining with the dust in the urban center at T0. The difference in the magnitude of the AAOD is due to the smaller SSA (more absorption) at T0 as compared to T2.

Summary of the results
The results for the AAE for 350-500 nm and 500-1000 nm, the increase or decrease of SSA with wavelength, and the extinction Angstrom exponent (EAE) are shown in Table 1 Bergstrom et al. (2007) and Russell et al. (2010). For the 19 March case the constant AAE extends to 600 nm indicating that a division at 500 nm is simply an approximation. X X X X X X X X X X X X X X X X X X X X X X X X X X X X X X X X X X X X X X X X X X X X X X X X X X X X X X X X X X X X X X X X X X X X X X X X X X X X X X X X X X X X X X X X X X X X X X X X X X X X X X X X X X X X X X X X X X X X X X X X X X X X X X X X X X X X X X X X X X X X X X X X X X X X X XX X X X X X X X X X X X X X X XX X X X X X X X X X X X X X X X X X X X X X X X X X X X X X X X X X

Discussion and comparison to other results
We present results from five separate flights (2 flights over the Gulf of Mexico -10 and 13 March and 3 flights over Mexico City -6, 15, and 19 March).

Mexico City results
As mentioned above, previous studies of the aerosol radiative properties in the Mexico City area (Barnard et al., 2008;Marley et al., 2009a, b;Corr et al., 2009) Marley et al. (2009a, b). Barnard et al. (2008) find AAE values that are strongly dependent on wavelength pair (3.2 to 5.1 for 300-500 nm; 1.9 to 2.6 for 300-870 nm). Marley et al. (2009b) find a range of AAE (0.63 to 1.5 including data from an earlier campaign in 2003 and the 2006 results) over the wavelength range 370 to 950 nm. It should be noted that most of the data used by Barnard et al. (2008) was from 2003 when there were forest fires in the Yucatan giving more OM sources than during 2006 (Marley, et al., 2009b).
Our results compare reasonably well with the other results in that the SSA over Mexico City is constant in the 350-400 nm range. We do not see the sharp increase between 400 and 500 nm seen by Barnard et al. (2008). However, this difference could be due to the increased forest fires in 2003. Our AAE values have a similar range as Barnard et al. (2008) but a larger range than Marley et al, (2009b

Other recent results
The idea that individual aerosol components have a unique AAE opens up the possibility of determining the amount of the absorbing species. Recently, Yang et al. (2009) used the wavelength dependence of various aerosol components to apportion the aerosol absorption in Northern China for black carbon, brown carbon (OM), and dust, attributing the high AAE values of dust to the presence of ferric oxides. In one of the MILAGRO studies discussed above, Barnard et al. (2008) subtracted a BC-type (AAE=1) absorption from the total absorption and attributed the remainder to OM. Martins et al. (2009) compared the filter sample absorption results for two urban areas, Sao Paulo, Brazil and Greenbelt, Maryland. They showed that Sao Paulo had increased shortwave absorption presumably due to organics.
Another method of analyzing absorption data is to plot the AAE on one axis and the EAE on the other axis. This has been used by a number of investigators recently (Yang et al., 2009, Fig. 5 andRussell et al., 2010, Fig. 5).  plotted AAE versus the organic mass fraction and SSA while Mielonen et al. (2009) categorized the aerosol by SSA and EAE. These techniques rely on the idea that BC, OM and dust have different locations on the figures. Plots of the data do tend to group in this manner, however there is a large amount of scatter and some of the groups overlap (see Yang et al., 2009, Fig. 5). Yang et al. (2009) identified the specific aerosol type by doing a back trajectory to source locations.
It is difficult to compare our single case results with average values over many cases (Yang et al., 2009) or long-term averages . However, their results can be compared to our dust case (19 March). Yang et al., (2009) have an average of the dust cases at AAE = 1.89 and EAE = 0.59. Russell et al. (2010) have multi-year AERONET-derived averages over three desert dust-influenced sites of AAE = 2.2 and EAE = 0.6. Our dust case has a larger AAE values (3.3 for T2 350-500 nm) and smaller EAE values (0.0 for T2) than the average values but are within the ranges seen by both other studies.
The other cases are more problematic to compare. Yang et al. (2009) have averages of AAE and EAE for biomass burning, fresh plume and coal pollution with very similar values (1.5 and 1.5; 1.35 and 1.49; and 1.46 and 1.39). Similarly the Russell et al. (2010) multi-year AERONET averages are 1.3 and 1.9 for biomass burning and 1.0 and 1.8 for urban pollution. Our non-dust results range from an AAE of 0.96 to 3.3 and an EAE of 0.7 to 1.8. In general then, while our results fit into the range of values seen by other investigators, it is difficult to identify the source of the absorbing material (other than dust) from an AAE versus EAE plot (or the SSA trend with wavelength) without more information.

Conclusions
The results for the spectral single scattering albedo and absorption optical depth for the Mexico City area show a great deal of variability most likely due to aerosol mixtures of varying amounts of organic carbon, black carbon and dust. The aerosol spectral absorption optical depth appears to generally fall into three different wavelength regions: 350-500 nm region where AAOD often falls off sharply towards longer wavelengths due to organic carbonaceous particles and windblown dust; 500-1000 nm region where the AAOD decrease with wavelength is slower most likely due to black carbon; and 1000 nm to 2000 nm where it was difficult for us to obtain reliable results since the AAOD was relatively small and the gas absorption dominated. However, there was an indication of a small and somewhat wavelength independent absorption in this region.
Of the five cases, two had a single, constant absorption Angstrom exponent (AAE) that fit the absorption optical depth over the entire wavelength range of 350 to 1700 nm. The other three cases required separate fits for the 350 to 500 nm and 500 to 1000 nm region.