Observationally-constrained estimates of global small-mode AOD

Introduction Conclusions References

collects data from sparsely distributed stations, AERONET data is inadequate for global or regional mean estimates. The goal of the present study is to give global and regional mean sAOD (as well as AOD) estimates at the level of the AERONET data accuracy. A global estimate of AOD at the AERONET data accuracy was made before by Chung et al. (2005) who integrated MODIS (Moderate Resolution Imaging Spectro-5 radiometer) AOD and AERONET AOD. This study is significant expansion of Chung et al.'s (2005), in that (a) we bring in MISR (Multi-angle Imaging Spectroradiometer) data as well, (b) more importantly we address sAOD as well as AOD, and (c) we look at the trend of global and region mean AOD and sAOD. The trend of global and regional mean sAOD would be an independent and powerful measure of changes in global and 10 regional anthropogenic aerosol emission. Typical air-quality monitoring such as PM (Particulate Matter) mass density measurement relies on unevenly-distributed sites, and so global or regional averages cannot be obtained reliably from these site measurements.
Aerosol data used in this study are described in Sect. 2. The techniques to integrate 15 satellite and ground-based data for AOD and sAOD are discussed in Sects. 3 and 4, respectively. We discuss the trends of global and regional averages of AOD and sAOD in Sect. 5. Conclusion follows in Sect. 6.

Data
Table 2 summarizes the datasets used here. Below, we discuss the datasets in detail. Discussion Paper | Discussion Paper | Discussion Paper | Discussion Paper | over the continents (Kaufman et al., 1997) on the globe. However, the MODIS data and algorithms are not adequate to retrieve AODs over bright surfaces such as deserts and snow surfaces (Kaufman et al., 1997). We downloaded the M3 product of Collection 5.1 data file, which is the latest monthly mean product. This product gives out AOD and fine-mode fraction (FMF) which are then used to calculate sAOD. 5 AOD and sAOD data originally on the 1 • by 1 • spatial resolution are converted into the T42 resolution (approximately 2.8 • by 2.8 • grid) from 2001 to 2010, as follows. Over each T42 gridbox in each month, there are typically many values from the 1 • × 1 • Terra and Aqua data combined. If there are at least five values in a T42 gridbox, we use the median to represent that T42 grid. Median, instead of arithmetic averaging, is a tool we 10 choose in reducing cloud contamination. In a sensitivity test, we replace these median values by average values. We find that the global and decadal (2001)(2002)(2003)(2004)(2005)(2006)(2007)(2008)(2009)(2010) mean AOD differs insignificantly between the two averaging methods. However, the 98th percentile AOD at 550 nm is about 50 % lower in the median method than in the average method. This reduction is due mainly to removing many of extremely-large values in excess 15 of 1.0. Monthly AOD exceeding 1.0 appears quite unrealistic and probably pertinent largely to cloud contamination, and thus eliminating many of these values should be viewed as improving the quality of AOD. TheÅngström exponent (AE) over the ocean is obtained using the T42-resolution AOD at four wavelengths of 470, 550, 660 and 870 nm.

MISR
The Multi-angle Imaging Spectroradiometer (MISR) instrument aboard the Terra satellite (again, launched in December 1999) retrieves key aerosol optical properties, such as AOD, over the oceans and the continents including highly reflective surfaces like deserts. We downloaded the CGAS MIL3MAE.004 product for the study. Out of this 25 monthly product, we extracted AOD, sAOD, mAOD, lAOD and SSA. These parameters originally available on the 0.5 • by 0.5 • spatial resolution are converted into the T42 resolution using the median method that we adopted for the MODIS data processing in 31667 Introduction Sect. 2.1. The effect of using the median method over the arithmetic averaging is assessed again, yielding similar results to that for MODIS AOD. The global and decadal (2001)(2002)(2003)(2004)(2005)(2006)(2007)(2008)(2009)(2010) mean MISR AOD at 550 nm shows a small difference (about 0.01) between the median and arithmetic average methods. In case of the 98th percentile value, AOD is about 40 % lower in the median method than in the average method.

5
Where necessary, we logarithmically interpolate AOD to the desired wavelength. We do the same logarithmical interpolation for other AOD data. MISR AE is computed using the T42-resolution AOD values at 443, 555, 670 and 865 nm.

AERONET
The AErosol RObotic NETwork (AERONET) is a ground-based aerosol network consisting of world-wide located automatic sun-and sky-scanning spectral radiometers (Holben et al., 1998). AERONET collects AOD from the direct sun measurements, and provides the retrieved optical parameters, such as SSA, using an inversion algorithm (Dubovik and King, 2000). sAOD and lAOD are computed from spectral AOD data using the spectral deconvolution algorithm of O'Neill et al. (2003). This algorithm assumes 15 a bimodal aerosol size distribution.
We downloaded the monthly Level 2.0 (cloud-screened and quality-assured data) from AERONET Version 2 product, and the product includes the data from China. Among the various parameters of the dataset, we use AOD, sAOD and SSA fields. AE is obtained using the AOD at three wavelengths of 440, 675 and 870 nm. 20 We calculate AOD and SSA at 550 nm, as follows. The 550 nm AOD is computed using logarithmic interpolation between AOD values at 500 nm and 675 nm. If AOD value is not available at 675 nm, values at 440 nm and 500 nm are extrapolated instead. If AOD value is not available at 500 nm, values at 440 nm and 675 nm are interpolated. If AOD value is absent at two of these three wavelengths at least, AOD is not obtained 25 at 550 nm. In case of SSA, the interpolation is done linearly between SSA values at 439 nm and 673 nm.

GOCART
The Georgia Tech/Goddard Global Ozone Chemistry Aerosol Radiation and Transport (GOCART) model simulates 550 nm AOD for major types of aerosols such as black carbon, organic carbon, sulfate, dust and sea salt (Chin et al., 2002). We use the GO-CART simulation as an interpolation tool to fill up the gaps in aerosol observations. The

5
GOCART model produces a global gridded AOD at the 2.5 • latitude by 2.0 • longitude resolution, and the results are archived on a monthly scale from January 2000 to July 2002. All these products are interpolated onto the T42 resolution, and then averaged for each calendar month yielding climatological seasonal cycle. GOCART SSA for total aerosols is computed using the AOD for each aerosol species, as in Chung et al. (2005), where SSAs of black carbon, organic carbon, sulfate, dust and sea salt are AOD-weighted averaged. Unlike in Chung et al. (2005), we use a black carbon SSA of 0.185 as theoretically obtained by Chung et al. (2012a), an organic carbon SSA of 0.85 as analyzed by , and a dust SSA of 0.96 as in Chung et al. (2012b). These new SSA values reflect recent advances in understanding 15 the optical properties of aerosols. The GOCART AOD is only available for 550 nm, and so we calculate AE using the AOD for each type of aerosols, as below:  Chung et al. (2005). Please note that the GOCART AOD, SSA and AE are to fill up the observation gaps. Thus, the accuracy of these GOCART products is not a major concern in this study.

AOD, SSA, AAOD and AE integration
In this study, we integrate various aerosol datasets to produce the best estimates of AOD and sAOD. We also provide the best estimates of SSA and AAOD. In this section, we describe the integration procedures.

Aerosol optical depth (AOD)
10 First, we compare AERONET AODs with MODIS and MISR AODs in Fig. 2. Such comparison was made before Remer et al., 2002Remer et al., , 2005Kahn et al., 2005), revealing small differences between AERONET and MODIS/MISR data. The identified differences included generally negative biases by MODIS/MISR. Here we use monthly mean AODs for the comparison since the previous comparison was based 15 on instantaneous AODs. The previously identified bias might be greater or smaller for monthly (mean) AOD for many reasons. For instance, AERONET, MODIS and MISR filter out data if deemed contaminated by cloud, and apply different filtering algorithms. This should result in differences in sampling clear and cloudy skies. A satellite study (Chand et al., 2009) showed a positive correlation between aerosol and cloud cover 20 over the southern Atlantic Ocean, meaning that different sampling of clear and cloudy skies should contribute to monthly AOD bias while instantaneous AOD is irrelevant to these sampling differences. This compels that monthly AOD have different biases than instantaneous AOD. Validating monthly AOD has been done very little. Chung et al. (2005;hereafter  in monthly mean AOD between MODIS and AERONET than shown by instantaneous AOD investigation studies. In Chung2005, an older version of MODIS data was examined. In the present study, we use the latest version of MODIS AOD, and also include MISR data.
Here, we choose to validate monthly AODs at 550 nm as done in Chung2005. In the 5 first step, MODIS and MISR AODs (before the T42 resolution conversion) are collocated in time and space with AERONET AODs. Any monthly AOD greater than 1.0 is considered cloud contaminated, and thus removed. Figure Remer et al., 2002Remer et al., , 2005Kahn et al., 2005) identified for instantaneous AOD. MODIS data is in slightly better agreements with AERONET AOD than MISR data is (Fig. 2). The slopes in Fig. 2, on the other hand, correspond to the bias relative to AERONET AOD. The regression slopes are found to be 0.61 for MISR ( Fig. 2a) and 0.66 for MODIS ( Fig. 2b), again in support 15 of the superiority of MODIS AOD over MISR AOD. Previous instantaneous AOD studies reported that the regression slopes of MODIS on AERONET are 1.02±0.05 for 660 nm AOD on a global scale , and 0.86 for 660 nm AOD over land . Remer et al. (2005) found the regression slope of 0.78 for 550 nm AOD over land and 0.94 for 550 nm AOD over ocean. For instantaneous MISR AOD, Kahn 20 et al. (2005) showed that the regression slopes of MISR versus AERONET AODs at 555 nm are 0.896 ± 0.003 for biomass burning aerosols, 0.681 ± 0.004 for continental aerosols, 0.693 ± 0.004 for dusty aerosols, and 1.381 ± 0.013 for maritime aerosols. By comparing our results with instantaneous AOD study results we conclude that monthly AODs by MODIS/MISR tend to have greater biases than instantaneous AODs. 25 We also compare the global and decadal ( be 0.26. The most salient difference between the results here and those of Chung2005 is that the difference between AERONET and MODIS AOD is sharply reduced here. Another noticeable difference is that MODIS and MISR now underestimate AODs compared to AERONET in contrast with the overestimated AOD by MODIS in Chung2005.
The differences between Chung2005's MODIS AOD and the MODIS AOD here are 5 more than the period and the retrieval algorithm. In this study, we use median values in the conversion of MODIS and MISR data from the original resolution into the T42 resolution, and this median method significantly reduces the effect of cloud contamination while this sophisticated scheme was not adopted by Chung2005.
In view of the aforementioned errors of monthly MODIS/MISR AOD, it would be ideal 10 to use AERONET AOD alone. Because AERONET collects data from sparsely distributed measurement sites over the globe, the AEROENT AOD gridded at the T42 resolution is dominated by gaps. To compensate for the data gaps, we combine the MODIS/MISR AOD with the AERONET AOD as in Chung2005. Chung2005 developed a methodology of retaining the spatial patterns of MODIS AOD and adjusting them with 15 the AERONET AOD values. We apply the overall approach of Chung2005, but deviate from Chung2005 in minor aspects. We generate the combined AERONET + MODIS + MISR 550-nm AOD as follows. The data combination is conducted for each month from 2001 to 2010. We use GO-CART simulations to fill the gaps in MISR or MODIS AOD. The MISR, MODIS, and 20 AERONET AODs are assimilated so that the order of influence is AERONET > MODIS > MISR > GOCART. MODIS seems to be more accurate than MISR in view of Fig. 2, and this is the basis for designing greater influence of MODIS than of MISR AOD. The assimilation takes place in three steps: (i) The MISR AOD bias of 0.0286, compared to the AERONET AOD, is added to the MISR AOD at each grid. This bias is the 25 global and decadal average of the difference between collocated AERONET and MISR AODs. We then fill the gaps of these MISR AODs with GOCART AODs using the interative difference-successive correction method developed by Cressman (1959). (ii) The MODIS AOD bias of 0.0110 is added to the MODIS AOD at each grid. Then, the gaps ACPD 12,2012 Observationallyconstrained estimates of global small-mode AOD in the MODIS AODs are filled with the MISR + GOCART AOD again using Cressman's (1959) approach. (iii) The spatial pattern in the MODIS + MISR + GOCART AOD is coupled with the sparsely distributed AERONET AOD values, using Chung2005's technique, as below.
where N AOD j is the adjusted new value of the AOD at grid j ; AERONET j ,i is an AERONET AOD at station location i nearby grid j ; d j ,i is the distance between j and i ; and MMG AOD j , i is the MODIS + MISR + GOCART AOD at the grid of AERONET j ,i . Final AOD (i.e. N AOD ) matches AERONET AOD wherever AERONET AOD exists.  15 We generate global SSA only with GOCART SSA and AERONET SSA. MODIS does not provide SSA data and the SSA from MISR appears to be quite unrealistic. GOCART SSA is for 550 nm, and is subsequently adjusted by 550 nm AERONET SSA as in Chung2005. The final SSA is for each month from 2001 to 2010, and the decadal average is shown in Fig. 3b. In computing the decadal average SSA, the averaging is 20 AOD weighted. The lowest SSA, which points to very absorbing aerosols, is found over the biomass burning areas in Africa. Areas with heavy industry such as eastern China do not show low SSA, probably because scattering anthropogenic aerosols such as sulfate are plentiful over there.

Absorption aerosol optical depth (AAOD)
We obtain AAOD using the integrated AOD and SSA at 550 nm. The relationship between AOD, AAOD and SSA is given by Eq. (3).

3.4Ångström exponent for AOD (AE)
As we integrate various datasets for AOD, we combine MODIS, MISR, AERONET and GOCART AEs for our best estimate of AE, except that we combine MODIS and MISR 10 data as follows. We only use the MODIS AE over the ocean and use the MISR AE over the land. We choose to do this, because the MODIS AOD over the land is only available at three wavelengths of 470, 550 and 660 nm while over the ocean the AOD is additionally offered at 870 nm. Because we will adjust the MODIS AE with the AERONET AE and the AERONET AE is computed using the AOD over 440-870 nm, we only use 15 the MODIS AOD over the ocean to compute the MODIS AE for consistency. Figure 4b shows the MODIS AE; in this figure the land AE is computed using the AODs at 470, 550 and 660 nm. As shown in Fig. 4b, there is clear discontuity between the MODIS AE over the land and the AE over the ocean. The MISR AE (Fig. 4a) on the other hand does not suffer the discontuity. 20 Our best AE estimate, which integrates AERONET, MODIS and MISR data, is shown in Fig. 4c. The overall pattern does not show the aforementioned discontuity, and is very consistent with the spatial distribution of average particle size. AE is inversely proportional to particle size, and should be small over the ocean and deserts where aerosols are generally large. As shown in Fig. 4c

Small-mode aerosol optical depth (sAOD) integration
For our best estimates of sAOD, we integrate AERONET, MODIS and MISR data and GOCART simulation as we did for AOD. Applying the exactly same integration method will not work, because AERONET, MODIS and MISR use different particle cut-off sizes to define sAOD (see Table 3). For example, MODIS sAOD represents smaller particles 5 than AERONET sAOD does, and thus MODIS sAOD would have a low bias relative to AERONET sAOD. To overcome this problem, we design two methods. In addition, the simple integration method that was used for AOD is applied for the comparison. In the following subsections, we discuss and evaluate the three methods.

10
The overall strategy of the first method is to use AE to derive sAOD. AE was computed using AODs, and MODIS, MISR and AERONET AODs do not have the bias issue that sAODs have. AE is known to give information about aerosol size (King et al., 1978;Nakajima et al., 1986;Kaufman, 1993;Eck et al., 1999;O'Neill et al., 2001aO'Neill et al., , b, 2003Schuster et al., 2006). Thus, it is conceivable that AE is related to sAOD, and yet the 15 relationship between the two has not been investigated to our knowledge. Here, we attempt to investigate this relationship, but a direct comparison does not work well since sAOD is proportional to aerosol amount while AE is not. Therefore, to better quantify the relationship between sAOD and AE, we replace sAOD by FMF (the finemode fraction) which is the ratio of sAOD to AOD. 20 We analyze the relationship between AE and 500 nm FMF, using monthly AERONET data, as shown in Fig. 5. AE is correlated remarkably well with FMF! We find that this positive relationship is also reproduced by running a Mie scattering code. We use a Mie scattering code to produce additional AE/FMF data when AE > 2.0, since there are so few data with AE > 2.0 in Fig. 5. Then, we divide Fig. 5  The sAOD estimate is obtained, as follows. (i) These fitting equations are applied to compute global FMF using the integrated AE. (ii) The obtained FMF is multiplied by the integrated AOD to compute sAOD. (iii) This sAOD is adjusted by AERONET sAOD, using Eq. (2). These 3 steps constitute the first method. Figure 6a depicts the sAOD obtained in this fashion. The most noticeable feature in Fig. 6a is that sAOD is largest over eastern China. Compared to total AOD in Fig. 3a, sAOD is particularly large in China. This contrast between AOD and sAOD is in our view the most unexpected aspect of our obtained sAOD. Based on this, we propose that small (in size) aerosols are a major component of aerosols in eastern China. As will be shown later, this method is our best method, and gives more accurate data than 10 MODIS or MISR data alone. Our finding that aerosols in China are mostly made up of small particles cannot be reached using an AERONET data analysis alone, since there are so few AERONET stations in China. On the other hand, MISR and MODIS give a global picture of sAOD. In Fig. 6 we compare the integrated sAOD ( Fig. 6a) with MISR sAOD (Fig. 6b) and 15 MODIS sAOD (Fig. 6c) In MISR or MODIS data, sAOD is not particularly large over eastern China. Furthermore, the MODIS sAOD is not available over much of the land.

Method 2: FMF integration
The second method is based exclusively on FMF. The concept of the method is that differing definitions of sAOD create biases in FMF and we compute these biases for 20 the correction.
First, we compare AERONET FMF with MODIS/MISR FMF. Here, we define MISR FMF as the ratio of sAOD + mAOD to tAOD, since MISR sAOD + mAOD explains a range of particle size similar to that of AERONET/MODIS sAOD (see Table 3). When monthly AERONET data are compared to MODIS or MISR data after collocation in time and space, we find that MISR FMF is better correlated with AERONET FMF than MODIS. In view of this, we allow greater influence of MISR FMF than MODIS FMF. Introduction The second method is conducted in five steps: (i) Collocated AERONET and MODIS AOD and sAOD from 2001 to 2010 are globally averaged, showing that the AERONET FMF is 40 % larger than the MODIS FMF on the average. We multiply the MODIS FMF at each grid point by 1.4 (i.e. 140 %) to bring it on a par with the AERONET FMF. We then fill the gaps of this adjusted MODIS FMFs with GOCART FMFs using Cressman's 5 (1959) method. (ii) The gaps in MISR FMFs are filled with the MODIS + GOCART FMFs using again Cressman's (1959) approach. The MISR FMF is multiplied by 0.98 at each grid to be on a par with the AERONET FMF. (iii) The spatial pattern in the MISR + MODIS + GOCART FMF is coupled with the sparsely distributed AERONET FMFs, using Eq. (2). (iv) The combined AERONET + MISR + MODIS + GOCART FMF is multiplied by our best AOD estimate to obtain sAOD. (v) This sAOD is adjusted by AERONET sAOD again using Eq. (2).

Method 3: simple sAOD integration
In this method, we integrate MODIS, MISR and AERONET sAODs without regard to the different sAOD definitions, using the exactly same integration method as for AOD. This 15 method needs to be implemented, in order to quantify the performance of the previous methods. Like AOD, we find that MODIS sAOD is better correlated with AERONET sAOD than MISR sAOD.
The integration takes place as follows: (i) The MISR sAOD + mAOD bias of 0.015412, compared to the AERONET sAOD, is added to the MISR sAOD + mAOD 20 at each grid. This bias is the global and decadal average of the difference between collocated AERONET sAOD and MISR sAOD + mAOD. We then fill the gaps of these MISR sAOD + mAOD with GOCART sAOD (sum of AODs for black carbon, organic carbon, and sulfate) using Cressman's (1959) method. (ii) The MODIS sAOD bias of 0.051476 is added to the MODIS sAOD at each grid. Then, the gaps in the MODIS 25 sAOD are filled with the MISR + GOCART sAOD again using Cressman's (1959) approach. (iii) The spatial pattern in the MODIS + MISR + GOCART sAOD is coupled with the sparsely distributed AERONET sAOD using Eq. (2)

Evaluating the three methods
We evaluate the three methods by applying each method with fewer AERONET data and then comparing the estimated sAOD with the unused AERONET sAOD data. This evaluation process is repeated with (randomly-chosen) different choices of the unused AERONET data. These different choices are spatially as well as temporally based. The 5 errors are then averaged in space and time and also over the difference choices of the unused data. We find that the best performing method is the first method which is AE based. An overall error for the first method is 13 ± 2 %, and this is significant improvement over the simple use of MISR sAOD which gives an overall error of 62 %. ±2 % represents 10 the uncertainty due to sampling the unused AERONET data. The second and third methods yield an overall error of 31 ± 3 % and 89 ± 18 %, respectively. forest fire and indoor wood combustion, can be regarded as less dependent on overall economy. Turning to regional averages, AOD and sAOD tended to decrease over the 10 yr in US/Canada sector, Western Europe sector and Eastern Europe/Soviet Union sector (Fig. 7). For Western Europe and US/Canada combined, AOD and sAOD were down 5 20 % in the decade (Table 4a). However, sAOD had a strong increase in China and India sectors. In India and eastern China collectively, AOD decreased slightly while sAOD was up 43 %. sAOD could only increase against a backdrop of decreasing AOD, if large-mode AOD decreases significantly. If large-mode AOD, which mostly represents natural aerosols, did not change in this sector of the world, the level of PM would have 10 gone up huge in this region.

Linear trend in AOD and sAOD
Using AERONET data alone (Table 4b), sAOD was only up 16 % in India and eastern China combined. There are an insufficient number of AERONET sites in this region to properly represent the region, and the difference between Table 4a and b corroborates this. The area-average sAOD estimates are the novelty of the present study. 15 The decreasing tendency in AOD and sAOD in the west (from Fig. 7 and Table 4a)  Table 4a also shows that while sAOD increased in India and eastern China combined, AAOD decreased. Not all the aerosol species absorb solar radiation, and the primary absorbing species are dust, black carbon and organic carbon (Chung et al., 2012b). Since large-mode AOD was down significantly in this region, a very large re-25 duction in dust amount is likely a major explanation for the AAOD reduction. A reduction in carbonaceous aerosols is not a good explanation for the AAOD decrease because such a reduction is not supportive of the sAOD increase. A global pattern of the linear trend in AOD and sAOD is displayed in Fig. 8. sAOD increases are only pronounced over India and eastern China. The reduction in AOD or sAOD is widespread in the west. These results are consistent with the identified dimming and brightening in various regions around the world. Wild et al. (2009) showed a brightening phenomenon for the period from 2000 to 2005 in Europe and USA, and 5 a dimming phenomenon in India and China/Mongolia. Long et al. (2009) also reported a widespread brightening for 12 yr (1995)(1996)(1997)(1998)(1999)(2000)(2001)(2002)(2003)(2004)(2005)(2006)(2007) over the continental United State.

Discussion
Through this study, we have sought to provide the best estimates of global AOD, sAOD and AAOD by integrating various aerosol observations. Our estimates have a variety of uses. In one use, our product can be used to validate aerosol simulations. Satellitebased AODs have been commonly used for such validation (e.g. Kinne et al., 2006) and in some cases (e.g. Myhre et al., 2009) AERONET AODs were used. The AOD we have offered here is global and should be considered more accurate than pure satellite-based values, and moreover we have also offered global sAOD. In another 15 use, the ratio of sAOD to AOD can be used to roughly determine the anthropogenic fraction of total aerosol. In the previous aerosol forcing estimation studies (e.g. Chung et al., 2005;Bellouin et al., 2005Bellouin et al., , 2008, the anthropogenic fraction was derived entirely or partially from aerosol simulations. Our global sAOD product can make it possible to derive the anthropogenic fraction without using a model. 20 In this study, we have also demonstrated the usefulness of our product by analyzing the area averages of sAOD. The principal findings of the present study are from the sAOD area averages. sAOD area averages are a powerful and independent measure of anthropogenic aerosol emission. A sAOD area average by itself does not provide the amount of emission, but the time series of sAOD area averages can be used to quantify Introduction uncertainties. For black carbon emission alone, there is a factor of 2 or more uncertainty in the estimates (Bond et al., 2004(Bond et al., , 2007. Given these uncertainties, the time series of our sAOD area averages is particularly valuable. Lastly, the time series of our AOD and sAOD area averages quantifies the overall trend of aerosols. Direct measurement of aerosol concentrations in the ambient atmo-Introduction  Tanré, D., Kaufman, Y. J., Holben, B. N., Chatenet, B., Karnieli, A., Lavenu, F., Blarel, L., Dubovik, O., Remer, L. A., and Smirnov, A.: Climatology of dust aerosol size distribution and optical properties derived from remotely sensed data in the solar spectrum, J. Geophys. Res., 106, 18205-18217, doi:10.1029/2000JD900663, 2001. Wåhlin, P.: Measured reduction of kerbside ultrafine particle number concentrations in Copen- 12,2012 Observationallyconstrained estimates of global small-mode AOD          Here, we use our best estimates of AOD (sAOD). See Fig. 1 for the division of the world.