applications over land in an urban/suburban region€¦ · remer et al. (2013) describes how pixels...

Atmos. Meas. Tech., 6, 1747–1759, 2013www.atmos-meas-tech.net/6/1747/2013/doi:10.5194/amt-6-1747-2013© Author(s) 2013. CC Attribution 3.0 License.

Atmospheric Measurement

TechniquesO

pen Access

MODIS 3 km aerosol product: applications over land in anurban/suburban regionL. A. Munchak1,2, R. C. Levy1, S. Mattoo1,2, L. A. Remer3, B. N. Holben1, J. S. Schafer1, C. A. Hostetler4, andR. A. Ferrare41Earth Science Division, NASA Goddard Space Flight Center, Greenbelt, MD 20771, USA2Science Systems and Applications, Inc., Lanham, MD 20709, USA3Joint Center for Earth Systems Technology (JCET), University of Maryland Baltimore County, Baltimore MD, 21228, USA4NASA Langley Research Center, Hampton, VA 23681, USA

Correspondence to: L. A. Munchak ([email protected])

Received: 31 January 2013 – Published in Atmos. Meas. Tech. Discuss.: 14 February 2013Revised: 14 June 2013 – Accepted: 17 June 2013 – Published: 23 July 2013

Abstract. MODerate resolution Imaging Spectroradiometer(MODIS) instruments aboard the Terra and Aqua satelliteshave provided a rich dataset of aerosol information at a 10 kmspatial scale. Although originally intended for climate appli-cations, the air quality community quickly became interestedin using the MODIS aerosol data. However, 10 km resolutionis not sufficient to resolve local scale aerosol features. Withthis in mind, MODIS Collection 6 includes a global aerosolproduct with a 3 km resolution. Here, we evaluate the 3 kmproduct over the Baltimore–Washington D.C., USA, corri-dor during the summer of 2011 by comparing with spatiallydense aerosol data measured by airborne High Spectral Res-olution Lidar (HSRL) and a network of 44 sun photometers(SP) spaced approximately 10 km apart, collected as part ofthe DISCOVER-AQ field campaign. The HSRL instrumentshows that AOD can vary by over 0.2 within a single 10 kmMODIS pixel, meaning that higher resolution satellite re-trievals may help to better characterize aerosol spatial dis-tributions in this region. Different techniques for validating ahigh-resolution aerosol product against SP measurements areconsidered. Although the 10 km product is more statisticallyreliable than the 3 km product, the 3 km product still per-forms acceptably with nearly two-thirds of MODIS/SP col-locations falling within an expected error envelope with highcorrelation (R > 0.90), although with a high bias of ∼ 0.06.The 3 km product can better resolve aerosol gradients andretrieve closer to clouds and shorelines than the 10 km prod-uct, but tends to show more noise, especially in urban ar-eas. This urban degradation is quantified using ancillary land

cover data. Overall, we show that the MODIS 3 km productadds new information to the existing set of satellite derivedaerosol products and validates well over the region, but dueto noise and problems in urban areas, should be treated withsome degree of caution.

1 Introduction

For over 12 yr, the MODerate resolution Imaging Spectro-radiometer (MODIS) instruments, flying aboard the Terraand Aqua satellites, have been collecting data concerningthe global distribution of climate related variables. Severalof these variables describe aerosols, which have been shownto be one of the largest sources of uncertainty in the climatesystem (IPCC, 2007). MODIS provides a suite of informa-tion about aerosol properties, including aerosol optical depth(AOD) and aerosol sizing parameters, that have been usedextensively for climate studies (e.g. Kaufman et al., 2002;Bellouin et al., 2005; Quaas et al., 2008). Currently, threealgorithms are used operationally to provide products to thepublic – the dark target algorithm over land (Kaufman et al.,1997a; Levy et al., 2007b, 2010), the dark target algorithmover ocean (Tanre et al., 1997; Remer et al., 2005, 2008) andthe deep blue algorithm over land (Hsu et al., 2004, 2006).All produce data at a 10 km nominal spatial scale, which isan appropriate scale for climate studies (Remer et al., 2013).Soon after launch, it was realized that the retrieved aerosol

parameters could be used as proxies for estimating surface

Published by Copernicus Publications on behalf of the European Geosciences Union.

https://ntrs.nasa.gov/search.jsp?R=20140011189 2020-04-30T14:10:06+00:00Z

1748 L. A. Munchak et al.: MODIS 3 km aerosol product: applications over land

air quality conditions (e.g. Hutchison, 2003; Wang andChristopher, 2003; Chu et al., 2003; Engels-Cox et al., 2004;Hutchison et al., 2004). The MODIS instruments are particu-larly appealing for air quality applications because the broad(2330 km) swath of the instrument allows most global loca-tions to be monitored on a nearly daily basis. However, the10 km resolution of the MODIS aerosol products is insuffi-cient to resolve small-scale aerosol features, including pointsources in urban areas (C. C. Li et al., 2005) and small fireplumes (Lyapustin et al., 2011). Although several researchtechniques exist to retrieve aerosols at higher spatial scales(100m to 1 km) from MODIS observations (e.g. Lyapustinet al., 2011; Li et al., 2012), these have not been producedglobally in an operational environment.In response to clear scientific needs for aerosol observa-

tions at a higher spatial resolution, MODIS Collection 6 willinclude a global aerosol product at nominal 3 km scale, inaddition to the standard 10 km resolution. This 3 km productwill include retrievals based on both dark target algorithms(land and ocean). Remer et al. (2013), describe the 3 kmproduct in detail, and analyze a six-month test database tocompare the performance of the 3 km and 10 km products ona global scale. The 3 km product can show more spatial de-tail than the 10 km product, but agrees less well with ground-based sun photometers on a global scale. However, the 3 kmproduct may be better at characterizing aerosol distributionson local scales, and can only be tested through comparisonwith dense observations on a small spatial scale. We wishto examine the 3 km aerosol product on a regional scale todetermine the additional information content that the higherresolution retrieval provides, and the possible degradation indata quality introduced by moving to a higher resolution.In this work, we focus primarily on AOD, because it has

been used for surface air quality applications (van Donkelaaret al., 2006) and can easily be compared with lidar and sunphotometer data. Specifically, we can evaluate AOD prod-ucts at both 10 km and 3 km resolutions by comparing withsurface- and airborne-based AOD measurements collectedduring the DISCOVER-AQ field campaign conducted duringthe summer of 2011 over the Baltimore–Washington D.C.corridor of the United States. We first assess the similaritiesand differences in 10 km and 3 kmMODIS AOD images, andevaluate whether the 3 km product provides information thatis not captured by the 10 km product. We look critically atthe best method to validate the higher resolution aerosol re-trieval against sun photometer measurements, and validatethe 10 km and 3 km products for the campaign duration. Theability of the MODIS product at both 10 km and 3 km resolu-tions to capture spatial variability in AOD is also addressed.Due to the limited validation data for the over ocean vari-ables, we will only address the land variables in this study.

2 The MODIS 3 km land algorithm

The retrieval method of the global MODIS 3 km aerosolproduct is thoroughly detailed in Remer et al. (2013). Forthe sake of completeness, we will briefly describe the landalgorithm here. The 3 km algorithm emerges from the “darktarget” retrieval methodology (Kaufman et al., 1997a; Tanreet al., 1997; Remer et al., 2005; Levy et al., 2007b, 2013),based on the concept that in the visible wavelengths, aerosolsare bright and vegetated surfaces tend to be dark. The spec-tral contrast between aerosols and the surface can be used toretrieve quantitative information about aerosol properties.To increase signal-to-noise, the MODIS algorithm re-

trieves aerosol parameters at a lower spatial resolution thanthe nominal (at nadir) 500m top of atmosphere (TOA) re-flectance measurements. The algorithm creates N by N “re-trieval boxes” of pixels, in order to filter out the subset that isnot desirable for aerosol retrieval. Thus, the 10 km algorithmworks with 20×20 pixel retrieval boxes (400 pixels), whereasthe 3 km algorithm works with 6×6 pixel retrieval boxes (36pixels). Remer et al. (2013) describes how pixels are maskedfor cloud (Martins et al., 2002), sediments in water (Li et al.,2003), snow and ice (R. Li et al., 2005) and surfaces that aretoo bright for retrieval. The remaining pixels are sorted bytheir 0.66 µm reflectance, and the brightest 50% and darkest20% of the remaining pixels are discarded. This means thatin the 10 km (3 km) algorithm, there are at most 120 pixels(11 pixels) remaining from which to do aerosol retrieval. Thereflectances of these remaining pixels are averaged, result-ing in a set of spectral reflectance values to drive the aerosolretrieval. These spectral reflectance values are further “cor-rected” for gas absorption (e.g. Levy et al., 2013).The expected “quality” of the retrieval is determined by

the number of pixels that remain after all masking and fil-tering. If at least 51 pixels remain (out of 120) for 10 km or5 pixels remain (out of 11) for 3 km, the retrieval is initiallyexpected to be of “high” quality. The 10 km retrieval will beattempted only if 12 pixels remain (out of 120), but it is ex-pected to be of low quality. There is no corresponding lowquality attempt for the 3 km retrieval (Remer et al., 2013).The over land algorithm performs the retrieval by match-

ing the observed TOA spectral reflectance with lookup tablesof pre-computed TOA spectral reflectance. The largest un-certainty is in characterizing the contribution of the surfaceto the TOA reflectance, for which errors of 0.01 can leadto errors of 0.1 in retrieved AOD. Kaufman et al. (1997b)showed that the surface contribution in the 0.47 and 0.66 µmbands can be related to the surface contribution in the2.1 µm shortwave-IR (SWIR) band. The assumed relation-ship between the visible and SWIR was refined by Levy etal. (2007b), and is used as a constraint in the aerosol retrieval.As described by Levy et al. (2007b), the current retrieval usesthree channels (0.47 µm, 0.66 µm, and 2.1 µm) to determineboth the amount of aerosol (total AOD) and the relative ra-tio of fine and coarse aerosol models. The selection of the

Atmos. Meas. Tech., 6, 1747–1759, 2013 www.atmos-meas-tech.net/6/1747/2013/

L. A. Munchak et al.: MODIS 3 km aerosol product: applications over land 1749

fine and coarse aerosol models are prescribed by season andlocation. The resulting primary retrieved aerosol parametersover land are total AOD at 0.55 µm, the fractional contribu-tion of the fine-dominated aerosol type, the constrained sur-face reflectance, and fitting error (Levy et al., 2013). All ofthese assumptions (aerosol lookup tables, assumed surfacereflectance relationship, and methodology of the inversion)are identical for both 10 km and 3 km retrievals (e.g. Levy etal., 2013; Remer et al., 2013).Previously, the MODIS 10 km product has been thor-

oughly evaluated against sun photometers on global scales.The global expected error (EE) of the 10 km AOD at 0.55 µmproduct over land is ±0.05± 0.15AOD (Levy et al., 2010),and a testbed of 10 km data produced with the Collection6 MODIS algorithm shows that, globally, 69% of MODIS-Aqua land AOD retrievals fall within this range when col-located and compared to ground-based sun photometer mea-surements (Levy et al., 2013). Remer et al. (2013) has com-pared the same testbed of global 3 km AOD data with sunphotometers and show that it tends to validate less well thanthe 10 km. Specifically, since only 63% of global colloca-tions fell within expected error over land, a new expected er-ror of ±0.05± 0.20AOD has been established for the 3 kmland product (Remer et al., 2013). Although a new 3 km EEhas been established, we choose to continue to use the morestringent 10 km EE in this paper. Therefore, in this paper,when “EE” is referred to, the ±0.05± 0.15AOD definitionis employed, unless otherwise stated.

3 Data sources

3.1 MODIS products

This analysis uses five minute time length sections ofthe MODIS orbits, termed “granules”, collected between20 June 2011 to 31 July 2011 from both the Aqua and Terrasatellites. The granules were created outside of the MODISoperational framework using the finalized MODIS aerosolCollection 6 algorithm and Level 1B inputs. Therefore, smalldifferences between these granules and the publicly releasedgranules are not expected, but may exist. Additionally, theMODIS land surface cover product (MCD12Q1) is used tocharacterize surface type. The land cover product uses atrained algorithm to characterize land surface at a 500mresolution using 5 different classification schemes (Freidl etal., 2010). In this work, we only consider the 17-class In-ternational Geosphere-Biosphere Programme (IGBP) system(Loveland and Belward, 1997). Data from both the Terra andAqua satellites are included in both the training dataset andthe product.

3.2 AERONET/DRAGON

The AErosol RObotic NETwork (AERONET) (Holben etal., 1998) has 5 permanent stations operating in the study

region. These stations were supplemented with an additional39 temporary stations, distributed in a roughly 10 km by10 km grid, termed the Distributed Regional Aerosol Grid-ded Observation Networks (DRAGON). This network pro-vided an unprecedented opportunity to validate satellite de-rived aerosol properties at a high spatial resolution. Each sta-tion is equipped with a Cimel sun photometer (SP) measuringat eight spectral bands between 340 nm and 1020 nm. AODis calculated at 550 nm using a quadratic log-log fit (Eck etal., 1999). The AOD retrieval is expected to be accurate towithin ±0.015 (Eck et al., 1999; Schmid et al., 1999).In this work, we use the level 1.5 AERONET/DRAGON

SP data, which are cloud screened (Smirnov et al., 2000) butnot quality controlled. However, comparisons between level1.5 and level 2.0 AOD have a correlation of 0.99, a slope of 1and no offset, and the level 1.5 data contains more measure-ments.

3.3 High Spectral Resolution LIDAR (HSRL)

The NASA Langley Research Center airborne HSRL in-strument calculates aerosol optical depth at 532 nm usingindependent measurements of vertically resolved aerosolbackscatter and extinction, and is therefore expected to bemore accurate than a standard backscatter lidar, which typi-cally requires additional data and/or assumptions to produceextinction profiles, and therefore AOD, from the backscatter(Hair et al., 2008). The NASA Langley King Air B200 air-plane flew 25 flights with the HSRL instrument aboard on 13days between 1 July 2011 and 29 July 2011.

4 Results

4.1 Case studies

Because much of the value of the 3 km resolution retrieval isbetter at resolving smaller scale aerosol features, we first lookat two case studies from differing days in the DISCOVER-AQ study. Figure 1 shows the 10 km (panels a and c) and3 km (panels b and d) resolution MODIS AOD at 0.55 µmobserved by MODIS-Aqua on 21 July 2011 at 18:30UTC.On this day, there is a strong AOD gradient, with a changefrom 0.15 to 0.45AOD over ∼ 50 km, and a local AOD en-hancement reaching to over 0.75. HYSPLIT back trajecto-ries, using winds from the NAM model at 12 km resolution,show that the lower AOD air mass originated in the Midwest-ern United States, while the higher AOD air mass was com-ing from southern Virginia and North Carolina, which hadactive biomass burning nearby, and also passed over moreurbanized regions. The two MODIS products both resolvethe AOD gradient, and, in general, show the same aerosolsituation. Both products resolve the SP measured AOD max-imum of 0.75, although the 3 km product shows more highAOD pixels which may or may not be noise. The 3 km prod-uct is able to retrieve closer to clouds, and therefore shows

www.atmos-meas-tech.net/6/1747/2013/ Atmos. Meas. Tech., 6, 1747–1759, 2013


Fig. 1. AOD at 0.55 µm observed by MODIS-Aqua on 21 July 2011 at 18:30UTC is plotted at the 10 km resolution (a and c) and 3 kmresolution (b and d). MODIS/SP collocations are plotted in the circles. In (a–d), the inner circle is the SP temporally averaged AOD, whichis an average of ≥ 2 SP measurements within 30min of MODIS overpass. In (a) and (b), the outside circle is the AOD of the MODIS pixelcontaining the SP site. In (c) and (d), the outer circle is the spatial average of a MODIS AOD in a 5×5 pixel box around the SP station. Onlyland pixels with a QA= 3 are used in the collocation. Washington D.C. is shown with the large black star and Baltimore, MD is shown withthe small black star. The true color image, created from the MODIS red, green and blue bands, is shown in the background.

more detail and retrieves more area than the 10 km product.The 3 km product is also able to retrieve over bodies of wa-ter which are too narrow for 10 km retrievals, and has morepixels near the coastline that are unable to be retrieved bythe 10 km product. The ability to retrieve closer to coastlineis particularly valuable for air quality applications becausemany major cities, including some of the largest and mostpolluted, are located on coasts. This granule not only high-lights the value of a high resolution satellite product, but alsoshows that AOD can vary significantly over small distances.Plotted overtop of the MODIS AOD images in the circles

are MODIS/SP collocations. In all of the panels (a–d), theinner circle shows the SP temporal mean, which is calcu-lated by averaging all SP 0.55 µm AOD measurements at astation within 30min of the MODIS overpass time, with aminimum of 2 observations. In the top row (panels a and b),the outer circle shows the 0.55 µm AOD of the MODIS pixel

in which the SP station is located. In the bottom row (pan-els c and d), the outer circle shows the MODIS spatial mean,which is calculated by averaging all high quality (QA= 3)land pixels within a 5 pixel by 5 pixel box surrounding theSP station requiring at least 5 out of a possible 25 pixels(Ichoku et al., 2002). Therefore, the spatial averaging boxfor the 3 km product is 15 km by 15 km, and the spatial aver-aging box for the 10 km product is 50 km by 50 km. The sin-gle pixel MODIS collocations in the top row show excellentagreement between the MODIS retrieval and the SP mea-surements for both the 10 and 3 km products. However, theMODIS spatial average shown in the bottom row gives theimpression that the 10 and 3 km products provide differentanswers; in this case, the smaller averaging box creates betterMODIS/SP agreement for the 3 km product, while the largeraveraging box of the 10 km product smoothes out the gradi-ent and causes disagreement. The spatial averaging technique



Fig. 2. AOD at 0.55 µm observed by MODIS-Terra on 1 July 2011 at 15:35UTC is plotted at the 10 km resolution (a and c) and 3 kmresolution (b and d). MODIS/SP collocations are plotted in the circles. In (a–d), the inner circle is the SP temporally averaged AOD, whichis average of ≥ 2 SP measurements within 30min of MODIS overpass. In (a) and (b), the outside circle is the AOD of the MODIS pixelcontaining the SP site. In (c) and (d), the outer circle is the spatial average of a MODIS AOD in a 5×5 pixel box around the SP station. Onlyland pixels with a QA= 3 are used in the collocation. Washington D.C. is shown with the large black star and Baltimore, MD is shown withthe small black star. The true color image, created from the MODIS red, green and blue bands, is shown in the background.

does increase the number of MODIS-SP collocations as com-pared to the single pixel technique, adding an extra collo-cation for each product. This granule shows that the singlepixel collocation technique better characterizes AOD at theSP site, but limits the number of collocations and thereforethe statistical robustness of the validation.Figure 2 shows the 10 km and 3 km resolution MODIS

AOD observed by MODIS-Terra on 1 July 2011 at15:35UTC, which has a markedly different aerosol distri-bution than that shown in Fig. 1. On this day, AOD is low(0.12 as measured by the SPs) and is homogeneous acrossthe region. As in Fig. 1, the large-scale view of the twoMODIS products agree, but there are nuanced differences.Some noisy pixels are present in the 3 km product, and areless apparent in the 10 km product. Both the 10 km and3 km products have elevated AOD along the New Jersey andDelaware coastline, which are plausibly cloud contaminated

from subpixel clouds. The 10 km has fewer contaminatedpixels, but they extend further from the coastline; the 3 kmproduct has more contaminated pixels, but their spatial ex-tent is limited to right along the coastline. The 3 km productalso shows a small amount of striping that results from dif-ferences in the MODIS instrument detectors. As in Fig. 1, the3 km product has more coverage over the water areas in thisscene. The MODIS/SP collocation circles are as they were inFig. 1, with MODIS single pixel collocations shown in theouter circles of the top row, MODIS spatial average colloca-tions in the outer circles of the bottom row, and SP temporalaverages in the inner circle of all panels.In this aerosol situation, the spatial averaging of the 10 km

product AOD increases the agreement with the SP measure-ments because the averaging decreases the impact of noisypixels. Twenty-eight out of the 31 collocations shown inFig. 2c fall within expected error, and the average retrieved



0.55 µm AOD at the SP stations is 0.15, only 0.03 higher thanthe SP determined AOD. The MODIS 3 km spatially aver-aged collocations show an AOD enhancement of 0.1 to 0.3over Baltimore City that is not observed by the SP measure-ments; 9 out of the 26 MODIS 3 km spatial averages shownin Fig. 2d have an above expected error. The largest differ-ence between the MODIS 3 km and SP measurements wasobserved at the sun photometer station in Essex, Maryland,13 km east of downtown Baltimore, where the SP AOD is0.11, and the AOD of the 3 km spatial average is 0.38. Asimilar AOD enhancement is seen over Washington D.C., al-though there are no SP measurements to confirm that thisenhancement is a retrieval artifact. The single MODIS pixelcollocations in the Fig. 2b do not capture this important re-trieval problem due to the coincidences of where the SP sta-tions happen to be located and if the exact pixel where thestation is located is retrieved. This case study suggests thatthe spatial average collocation technique better character-izes the retrieval performance over the larger region, and thesingle pixel collocation technique may misrepresent productperformance because it is much more sensitive to the sitingof the SP stations.The airborne HSRL measurements provide an opportu-

nity to look at aerosol extinction and AOD spatial variabil-ity with a nominal 1min resolution, which corresponds toa nearly 6 km horizontal resolution. Figure 3 shows an-other high aerosol loading day, observed by MODIS-Terraon 29 July 2011 at 16:00UTC. HSRL 532 nm columnar AODalong the flight track between 15:30 and 16:30UTC is shownin the thick line, plotted atop of the 3 km resolution 0.55 µmAOD. The sections of the flight track where the plane wasflying lower than 7 km, flying above clouds, or was turn-ing are shown in grey. SP AOD interpolated to 0.55 µm isshown in the circles. The SP measurements are not tem-porally averaged; the nearest measurement to 16:00UTC isused, provided the measurement was taken between 15:30and 16:30UTC.The SP AOD measurements range from 0.33 to 0.58, with

lower values towards the south and west, and higher val-ues towards the north and east. The HSRL AOD measure-ments have a wider range from 0.35 to 0.67. There is excel-lent agreement between HSRL and SP measurements at theAERONET stations, which indicates that the broader rangeof AOD values from HSRL is likely due to the HSRL instru-ment being able to measure a larger area than the sun pho-tometers and not due to inaccuracies in either measurement.The MODIS 3 km product generally captures the spatial pat-tern of AOD as observed by both the HSRL instrument andthe sun photometers, although as in Fig. 2, the AOD is over-estimated in the urban areas and there are noisy retrievals.Outside of the urban areas, the MODIS image shows AODnear 0.45 for much of the southern and western portion ofthe granule, increasing to 0.65–0.75 in the northeastern por-tion of the granule that also has HSRL and SP measurements.

V�� V�� V�� V�� V��

��

��

��

��

��

V��

Fig. 3. A 3 km resolution AOD at 0.55 µm observed by MODIS-Terra on 29 July 2011 at 16:00UTC is plotted in the background.HSRL 532 nm columnar AOD along the flight track between 15:30and 16:30UTC is shown in the thick line. SP AOD interpolatedto 0.55 µm is shown in the circles. The SP measurements are nottemporally averaged; the nearest measurement to 16:00UTC isshown, provided the measurement was taken between 15:30 and16:30UTC. Washington D.C. is shown with the large black star andBaltimore, MD is shown with the small black star. The true colorimage, created from theMODIS red, green and blue bands, is shownin the background.

Overall, the three instruments show a similar aerosol situa-tion albeit with varying levels of detail.The HSRL AOD measurements reflect the more general

pattern observed by the sun photometers, but also showmanysmall-scale enhancements that cannot be observed even withthe high spatial density of the DRAGON network. The west-ern portion of the flight track in Fig. 4 shows that the HSRLinstrument measured an increase in AOD from 0.41 to 0.59over ∼ 4min between 16:00:08 and 16:04:28UTC, and trav-eling south by 29.8 km and west by 5.3 km in this time pe-riod. This change takes place over three 10 km boxes and ten3 km boxes. TheMODIS 3 km product better retrieves the ab-solute change in AOD, although the gradient along the flighttrack is not well reproduced because of missing and noisypixels. The maximum AOD is overestimated in the 10 kmproduct, and is more accurately retrieved in the 3 km prod-uct, although a neighboring pixel has an anomalously highAOD. It is probable that there is a source of contamination inthe noisy 3 km pixel, which was included in the 10 km pixelalong with the good pixels, resulting in an overestimation inthe 10 km pixel. From this comparison, it is clear that whilethe 3 km product contains more noisy pixels that are likelycontaminated, the smaller pixel size limits the spatial extent



��

V !�!" V !�!� V !�� V !�## V !�$!

$"�"�

$"�%�

$%��

$%��

��$��

&��

V�� $� ��#� �� !� ��

Fig. 4. Close up view of the area enclosed in the white box in Fig. 3. (a) MODIS 10 km 0.55 µm AOD with 532 nm HSRL AOD plotted ontop. (b)MODIS 3 km 0.55 µm AOD with 532 nm HSRL AOD plotted on top.

of the contamination and allows nearby uncontaminated pix-els to retrieve the correct AOD. In the 3 km product, the ef-fects of cloud or surface contamination are larger within asingle pixel, but at the same time are also spatially limited.As highlighted in Fig. 4, because the HSRL instrument

makes several measurements within a single MODIS pixel, itprovides an opportunity to assess AOD variability across thestandard MODIS retrieval resolution. A measure of subpixelAOD variability is shown in Fig. 5. For each 10 km resolutionMODIS pixel, the minimum HSRL measured AOD is sub-tracted from the maximum HSRL AOD, provided the mea-surements are within half an hour of the MODIS overpass.Data from all flights that have valid data within half an hourof a MODIS overpass are included, and only MODIS pix-els that contain more than 5 HSRL observations are ana-lyzed to ensure that two fully independent HSRL measure-ments are obtained within the MODIS pixel. In all, 482MODIS pixels are available for analysis for subpixel AODvariations. Within a single MODIS 10 km pixel, the maxi-mum HSRL measured AOD change is 0.25, increasing from0.36 to 0.61AOD. The mean 0.55 µm AOD variation withina 10 km pixel is 0.037 while the median variation within a10 km pixel is 0.023. Fewer than 25% of pixels containedless than 0.01AOD variation. The 10 km retrieval assumesuniformity of the AOD across the retrieval box, and that issimply not the case. The assumption is likely to hold bet-ter across the 3 km retrieval box; however, the nominal res-olution of the HSRL instrument for this study is larger thanthe 3 km retrieval box. For this case, decreasing the HSRLnominal horizontal resolution for AOD from 6 km to 1 kmincreased the noise to an unacceptably high level such that

0.00 0.05 0.10 0.15 0.20HSRL AOD Range

0.00

0.05

0.10

0.15

0.20

0.25

Nor

mal

ized

Fre

quen

cy 10 km

Fig. 5. The maximum difference between HSRL 0.55 µm AODmeasurements observed within a MODIS 10 km pixel, normalizedby the total number of analyzed grid boxes. Only MODIS pixelsthat contain 6 or more HSRL measurements are included.

the AOD variation within the MODIS 3 km pixels could notbe quantified.The case study comparisons with ground- and airborne-

based AOD measurements show that the MODIS 3 km prod-uct contains additional aerosol information as compared tothe 10 km product. Figure 1 shows a situation in which a fewhigh AOD pixels in the 10 km product could be interpretedas noise, but the higher resolution of the 3 km product pro-vides many more pixels, and SP data confirms that a true highaerosol loading event was identified. Additionally, the 3 kmretrieval is able to retrieve over small bodies of water andclose to coastline. The subpixel AOD variability observedby HSRL suggests that a 10 km product will frequently miss



��V��

�� μ��

��

��

��

��

��

��

��

��μ�

��

��V��

�� μ��

��

��

��

��

��

��

��

��μ�

��

�

�

�

�

�

!�"#��$%

�$��

�� μ��

V��

V��

��

��

��

��

V��

��μ�

��

�� !&'#$(�� !&'#$(

�'��(��)�!��V��

�� μ��

��

��

��

��

��

��

��

��μ�

��

��(��)�!��V��

�� μ��

��

��

��

��

��

��

��

��μ�

��

�

�

�

�

� !�"#��$%

�*��(��)�!��

�� μ��

V��

V��

��

��

��

��

V��

��μ�

��

�� !&'#$(�� !&'#$(

Fig. 6. (a) Density scatter plot of MODIS/SP collocations for 10 km AOD at 0.55 µm, using the single pixel collocation technique. Expectederror over land (±0.05± 0.15AOD) is shown in the dashed lines, the best fit line is shown by the solid blue line, and the one-to-one line isshown in the solid black line. (b) Same as in (a), except for the 3 km product. (c)Average difference betweenMODIS single pixel collocationsand SP 0.55 µm AOD, as a function of SP 0.55 µmAOD. A standard deviation of±1 is shown for the 3 km product. (d–f) As in (a–c), exceptthe MODIS 5× 5 pixel spatial average collocation is used.

small changes in AOD over an urban/suburban region. How-ever, retrieving with a higher resolution reduces the statisticalrobustness of the aerosol product, resulting in more noise andinaccurate retrievals over urban areas.

4.2 MODIS 3 km validation

The case studies highlight the strengths and weaknesses ofthe 3 km product; however, the sun photometer data from theentire campaign provides the most comprehensive dataset tocharacterize its performance. The SP data also allows a crit-ical look at how the different collocation techniques demon-strated in Figs. 1 and 2 lead to different conclusions on thevalidity on the MODIS AOD data in this region for both the10 and 3 km resolutions.In Fig. 6a, all 10 km MODIS/SP 0.55 µm AOD colloca-

tions are shown in a density scatter plot, using the MODISsingle pixel collocation technique employed in Figs. 1a and2a, and the same is shown in Fig. 6b for the 3 km product. InFig. 6c, these single pixel MODIS/SP collocations are binnedby the SP AOD, and the mean MODIS-SP difference for thebin is plotted for both the 10 and 3 km products. This processis repeated in Fig. 6d–f except that the 5× 5 pixel MODISspatial collocation technique is employed in the MODIS/SP

collocations. The collocation statistics for both techniquesfor both products is shown in Table 1.For both MODIS/SP collocation techniques, the bulk of

the 3 kmMODIS retrievals are within the±0.05± 0.15AODexpected error (63.6% for the single pixel technique and68% for the spatial average technique). The satellite retrievalis highly correlated with the ground-based measurements,although with a small negative offset (0.011 for the singlepixel technique and 0.013 for the spatial average technique)and a significant slope (1.26 for single pixel, 1.22 for spatialaverage). Accordingly, the MODIS 3 km AOD agrees wellwith SP AOD at low aerosol loadings, but is generally biasedhigh at higher AOD. The primary difference between the twocollocation techniques is that the spatial average has nearlytwice as many collocations (817 as opposed to 451, for the3 km product); therefore, the spatial averaging technique maybe more representative of the region at large.The binned MODIS-SP differences in Fig. 6c and f agree

well with the impression given by the scatter plots. The biasof the MODIS 3 km product is near zero at AOD below 0.1,then is biased high above 0.1. The single pixel collocation islikely to best compare the 10 km and 3 km products, becausethe spatial averaging technique samples a different area for



Table 1. Statistics from the 4 different MODIS/SP collocations shown in Fig. 6a,b and d–e.N = number of valid collocations, R = correlationcoefficient, Slope = slope of regression line (shown in blue in Fig. 6), Offset = y intercept of the regression line, bias =MODIS average AODminus SP average AOD, RMSE= root mean square error of collocations. EE for all cases is ±0.05± 0.15AOD.

Product Averaging % above % within % belowresolution method N R Slope Offset Bias RMSE EE EE EE

10 km single pixel 499 0.925 1.235 0.004 0.049 0.093 26.7 69.3 4.010 km spatial average 848 0.935 1.134 0.008 0.039 0.075 21.5 75.9 2.63 km single pixel 451 0.914 1.263 0.011 0.062 0.109 33.7 63.6 2.73 km spatial average 817 0.928 1.228 0.013 0.057 0.096 30.7 68.0 1.3

(a) 10 km resolution

0.0 0.2 0.4 0.6 0.8 1.0HSRL 532 nm AOD

0.0

0.2

0.4

0.6

0.8

1.0

MO

DIS

0.5

5 μ

m A

OD

% within EE = 53.56% above EE = 46.00% below EE = 0.43N = 1613; R = 0.88Y= 1.01x+ 0.07

(b) 3 km resolution

0.0 0.2 0.4 0.6 0.8 1.0HSRL 532 nm AOD

0.0

0.2

0.4

0.6

0.8

1.0

MO

DIS

0.5

5 μ

m A

OD

% within EE = 56.05% above EE = 42.04% below EE = 1.92N = 1199; R = 0.82Y= 0.94x+ 0.07

1

2

3

4

5

6

7

8

9

10

Frequency

Fig. 7. Density scatter plot of MODIS/HSRL collocations for AOD at 0.55 µm for MODIS and 532 nm for HSRL for the 10 km MODISproduct (a) and the 3 km MODIS product (b). The MODIS AOD is sampled along the HSRL flight path; no spatial or temporal averaging isperformed. Expected error over land (±0.05± 0.15AOD) is shown in the dashed lines, the best fit line is shown by the solid blue line, andthe one-to-one line is shown in the solid black line.

the different products. The 3 km product single pixel aver-ages shown in Fig. 6c show that the average AOD of the 3 kmproduct is frequently, but not always, higher than the 10 kmproduct.The retrieved AOD from the HSRL airborne instrument

provides a different perspective on validation of the 3 kmproduct. Because the plane is moving quickly in space,no spatial or temporal averaging is used to make theMODIS/HSRL collocations. The MODIS data are sampledalong the HSRL flight track, provided that the HSRL ob-servation is made within the geographic region mapped inFigs. 1 and 2, and is within 30min of the MODIS overpass.Shortening the time requirement to either 15 or 5min doesnot significantly affect the collocation statistics, but providesfar fewer collocations. The HSRL and MODIS AOD are notinterpolated to the same wavelength, which is expected tointroduce between 2–4% error (Kittaka et al., 2011). It isexpected that the AOD retrieved by the HSRL instrumentwould be lower than the AOD retrieved byMODIS by at least0.01 to 0.02 because the HSRL instrument does not measurethe entire column, and therefore neglects the contribution ofabove the HSRL profile (i.e. above 7 km) to total columnarAOD.

Figure 7 shows that the both products agree with HSRL,but with significant high bias (0.093 for 10 km, 0.086 for3 km). The increased bias of the 10 km product could bedue to the pixel contamination effect shown in Fig. 4 – al-though the 3 km product contains more contaminated pixels,the spatial extent of those pixels is limited, allowing morecorrect pixels as well. The 10 kmMODIS/HSRL comparisonalso shows the necessity of a higher resolution retrieval; thestripes in the HSRL/MODIS comparison show the change inHSRL AOD within a single pixel; the pattern that is quanti-fied in Fig. 5.From both SP and HSRL comparisons, we consider the

3 km product to be validated in this region. Using eithercollocation technique, more than 63% of MODIS/SP col-locations are within the expected error of the 10 km prod-uct, which is more stringent than the defined EE of the 3km product. If the more lax ±0.05± 0.20AOD expected er-ror definition for the 3 km product (Remer et al., 2013) isemployed, then the percent of collocations within EE in-creases to 67.1% for the single pixel collocation technique,and 73.1% for the spatial collocation technique. However,the increased noise and urban problems give the 3 km producta larger high bias than the 10 km product as compared to SP



V �� V �� V �� V �� V ��"�

�"��

�"�"�

�%��

�%��

��

��

��

��

Fig. 8. (a) Percent of 3 km spatially averaged MODIS/SP collocations of 0.55 µm AOD within expected error (±0.05± 0.15AOD) at eachSP station. (b) Correlation coefficient between 3 km spatially averaged MODIS 0.55 µm AOD and SP temporally averaged 0.55 µm measure-ments at each station. Land identified as urban/built up by the MODIS land cover product (MCD12Q1) is plotted in grey in both panels.

measurements. The single pixel and spatial average MODIScollocation techniques show quantitative but not qualitativedifferences.

4.3 Sources of uncertainty

Figures 2 and 3 show an overestimation of AOD in urbanareas. In order to study this further, we separate the 3 kmMODIS/SP spatially averaged collocations by SP station,and compute agreement statistics for each station. We choosethe spatial averaging collocation technique because it pro-vides more collocations. The percent of comparisons withinexpected error for each station is shown in Fig. 8a, and thecorrelation coefficient (R) between MODIS and SP 0.55 µmAOD is shown in Fig. 8b. In each panel, land classified as “ur-ban” by the MODIS Land Cover type product (MCD12Q1)is plotted in grey.Figure 8a shows remarkable variability in the percent of

collocations within EE for each SP station, ranging between16% to 95%. The stations with the smallest percent withinEE are clustered around the Baltimore urban center, andthe more urbanized I-95 corridor between Washington D.C.and Baltimore. However, the stations with small percentagesof collocations within EE still have correlation coefficientsabove 0.85, suggesting that the error is a systematic bias thatcould be accounted for in future retrievals, and not a randomerror.We wish to quantify the urban overestimation of the 3 km

product using the dense spatial distribution of the sun pho-tometer sites in the DRAGON network. Using the same15 km averaging box as the MODIS spatial collocation, thepercent of pixels within the averaging box identified as ur-ban by the MCD12Q1 product is calculated. In Fig. 9, thisquantity is plotted against the percent of collocations aboveEE for each SP site. The percent of urban land within theMODIS averaging box is well correlated with the percent

0 20 40 60 80 100Urban landcover percentage

0

20

40

60

80

100

Perc

ent o

f co

loca

tions

abo

ve E

E Y = 0.93x + 12.83R = 0.91

Fig. 9. Percent of 3 km spatially averaged MODIS/SP 0.55 µmAODcollocations above expected error (+0.05 + 0.15AOD) at each SPstation, plotted against the percentage of pixels within the 15 km by15 km collocation box identified as urban by the MODIS land coverproduct.

of MODIS/SP collocations that are above the expected er-ror. This confirms that the largest source of uncertainty inthis study is related to the urban areas. The cause is sus-pected to be improper land surface reflectance characteri-zation, considering that aerosol sources are not expected tovary significantly within the region of study and that it hasbeen shown that the surface reflectance in the 0.66 µm chan-nel over some urban surfaces is ∼ 0.7 of the 2.13 reflectance(Castanho et al., 2008; Oo et al., 2010) which is higher thanthe ratio used in the MODIS operational retrievals. In mixed-use urban/suburban regions like the DISCOVER-AQ studyregion, it is likely that the brighter urban surface 500m pix-els were selectively discarded in the pixel selection processof the 10 km product. However, because the resolution of the3 km retrieval is smaller, more urban pixels remain within



Fig. 10. (a) 3 km AOD at 2.1 µm observed by MODIS-Aqua on 21 July 2011 at 18:30UTC. (b) 3 km AOD at 2.1 µm observed by MODIS-Terra on 1 July 2011 at 15:35UTC. Only land pixels with a QA= 3 are shown. Washington D.C. is shown with the large black star andBaltimore is shown with the small black star. The true color image, created from the MODIS red, green and blue bands, is shown in thebackground.

the retrieval box after pixel selection, causing the surfacereflectance in the visible wavelengths to be underestimated,leading to an overestimation in AOD.Ideally, the noisy pixels could be flagged as poor quality,

and therefore treated with some degree of caution. One ofthe products produced operationally is the AOD at 2.1 µm.This is not a validated product, but is used as a diagnos-tic to the retrieval. It may hold information that can identifypoorer quality 3 km retrievals. In Fig. 10a, the 2.1 µm AODon 21 July 2011 at 18:30 from the 3 km product is plotted,the same granule as that shown in Fig. 1b, and in Fig. 10b;the 2.1 µm AOD on 1 July 2011 at 15:35UTC from the 3 kmproduct is plotted, the same granule as that shown in Fig. 2b.The scene-wide 2.1 µm AOD is 0.046 in the left panel, andis 0.065 in the right panel. Both granules contain pixels witha 2.1 µm AOD over 0.2, more than 3 to 5 times higher thanthe scene-wide average. The error is not due to the wrongaerosol model being picked; in both scenes, all of the pix-els were assigned the fine mode dominated urban aerosolmodel (Levy et al., 2007a). It appears that the same pix-els that have anomalously high 2.1 µm AOD retrievals alsotend to have inaccurate 0.55 µm AOD retrievals. For exam-ple, the pixels with high (> 0.55) AOD in Fig. 1b that agreewith SP observations have retrieved 2.1 µm AOD values thatdo not stand out from the background. However, the pixelswith high AOD in Fig. 2b that do not agree with SP observa-tions have retrieved 2.1 µm AOD values significantly higherthan the background. Although the 2.1 µm AOD retrieval isan unvalidated product, it may serve a purpose as a consis-tency check in this fine mode dominated, urban environment.This consistency check is likely to be inapplicable in dusty orother coarse mode dominated regimes.

Using anomalous 2.1 µm AOD pixels as a proxy for poorquality 0.55 µm pixels, and therefore throwing them out, re-sults in a reduction of the both the negative bias at low AODand the positive bias at high AOD. If the 3 km product isfiltered such that pixels with a 2.1 µm AOD over 0.2 are ex-cluded, and MODIS/SP are collocated using the spatial av-eraging technique, the percent within EE rises from 68% to78%, while the number of collocations falls from 817 to 727,and the R rises from 0.92 to 0.95. However, this screeningis not globally applicable, and is not recommended for usewithout closer examination.It appears that a primary source of error in the 3 km

MODIS aerosol product is the incorrect estimation of surfacereflectance over urban regions. This error is not as apparent inthe 10 km product because brighter urban surfaces are pref-erentially discarded during the pixel selection process. How-ever, because the 3 km product has a smaller footprint, it islikely that not all urban pixels are discarded, and the surfacereflectance is underestimated, leading to an overestimation inAOD. The 2.1 µmAOD shows some potential as an imperfectfilter for these poor quality pixels.

5 Summary and conclusions

The MODIS dark target aerosol algorithms were designedwith climate applications in mind, and the spatial scale ofthe retrieval reflects that design. However, as the applica-tions of the aerosol product have evolved, the proper scale atwhich to observe small-scale aerosol events came into ques-tion. In response to the needs for aerosol observations at ahigher resolution than 10 km, an operational aerosol product



produced at a 3 km resolution will be available in MODISCollection 6.This new product was evaluated against ground and air-

borne data over the Baltimore–Washington D.C., USA cor-ridor over the summer of 2011 to understand the poten-tial added information from the higher resolution retrievaland the statistical reliability of the product. The 10 km and3 km granules generally resolve similar large scale patterns inaerosol distribution but differ in the details. The smaller foot-print of the 3 km product retrieves closer to clouds and oversmall bodies of water, but also makes the product more sus-ceptible to cloud contamination and other sources of noise.Subpixel comparisons against HSRL show that AOD canvary significantly within a 10 km pixel, and a 3 km resolutionis a more appropriate scale for studying aerosol distributionsin urban and suburban settings.The dense network of sun photometers deployed during

the DISCOVER-AQ campaign provided an unprecedentedopportunity to validate aerosol products at a high spatialscale. Given the new resolution of both the satellite data andthe ground validation network, the MODIS spatial-temporalcollocation techniques for validation were revisited. It wasfound that using a spatial average of MODIS pixels around asun photometer station, or using only the pixel where the SPstation is located resulted in a quantitative, but not qualita-tive, difference. We conclude with good confidence that theMODIS 3 km validates in this region, but is biased high atAOD values greater than 0.1.There is evidence that a significant source of the bias ob-

served in the 3 km product results from improper characteri-zation of urban surfaces. The method for attributing the sur-face contributions to TOA measured reflectances has beendocumented to underestimate the surface reflectances in ur-ban areas (Castanho et al., 2008; Oo et al., 2010), whichin turn overestimates AOD. Indeed, we found a strong cor-relation between the amount of urban surface near a sunphotometer station, and the percent of collocations that areabove the expected error. While case studies show that the10 km product also has one or two pixels that are affected,the smaller pixel size of the 3 km product causes it to be moresusceptible to contamination.The poor performance of the 3 km product over urban sur-

faces is clearly a limitation in terms of air quality applica-tions. How to best address this problem in an operational en-vironment remains an open question. However, despite thislimitation, the 3 km aerosol product has new capabilities forstudying aerosol on local scales, including resolving smallscale AOD gradients and point sources, retrieving aerosolin patchy cloud fields, and retrieving closer to coastlines.We expect that the added information of the 3 km resolutionproduct will complement the existing MODIS 10 km reso-lution aerosol product and products from other passive andactive sensors to provide a more complete view of globalaerosol properties from space.

Acknowledgements. The authors thank Bill Ridgway and theMODAPS team for facilitating iterative testing needs. We alsothank the University of Wisconsin PEATE team for the productionof the Terra L1B data used in previous iterations of this study. Werecognize the entire AERONET team for their efforts to deployand maintain 44 functioning AERONET stations. HSRL operationswere funded by the NASA DISCOVER-AQ program. The authorsalso thank the NASA Langley King Air flight crew for theiroutstanding work supporting these flights.

Edited by: M. King

References

Bellouin, N., Boucher, O., Haywood, J., and Reddy, M. S.: Globalestimate of aerosol direct radiative forcing from satellite mea-surements, Nature, 438, 1138–1141, 2005.

Castanho, A. D., Martins, J. V., and Artaxo, P.: MODIS AerosolOptical Depth Retrievals with high spatial resolution over an Ur-ban Area using the Critical Reflectance, J. Geophys. Res., 113,D02201, doi:10.1029/2007JD008751, 2008.

Chu, D. A.: Global monitoring of air pollution over land fromthe Earth Observing System-Terra Moderate Resolution Imag-ing Spectroradiometer (MODIS), J. Geophys. Res., 108, 4661,doi:10.1029/2002JD003179, 2003.

Eck, T. F., Holben, B. N., Reid, J. S., Dubovik, O., Smirnov, A.,O’Neill, N. T., Slutsker, I., and Kinne, S.: Wavelength depen-dence of the optical depth of biomass burning, urban and desertdust aerosols, J. Geophys. Res., 104, 31333–31350, 1999.

Engel-Cox, J. A., Holloman, C. H., Coutant, B. W., and Hoff, R. M.:Qualitative and quantitative evaluation of MODIS satellite sensordata for regional and urban scale air quality, Atmos. Environ., 38,2495–2509, 2004.

Friedl, M. A., Sulla-Menashe, D., Tan, B., Schneider, A., Ra-mankutty, N., Sibley, A., and Huang, X.: MODIS Collection 5global land cover: Algorithm refinements and characterization ofnew datasets, Remote Sens. Environ., 114, 168–182, 2010.

Hair, J. W., Hostetler, C. A., Cook, A. L., Harper, D. B., Ferrare, R.A., Mack, T. L., Welch, W., Izquierdo, L. R., and Hovis, F. E.:Airborne high spectral resolution LIDAR for profiling aerosoloptical properties, Appl. Optics, 47, 6734–6752, 2008.

Hutchison, K. D.: Applications of MODIS satellite data and prod-ucts for monitoring air quality in the state of Texas, Atmos. Env-iron., 37, 2403–2412, 2003.

Hutchison, K. D., Smith, S., and Faruqui, S.: The use of MODISdata and aerosol products for air quality prediction, Atmos. Env-iron., 38, 5057–5070, 2004.

Holben, B. N., Eck, T. F., Slutsker, I., Tanre, D., Buis, J. P., Set-zer, A., Vermote, E., Reagan, J. A., Kaufman, Y. J., Nakajima,T., Lavenu, F., Janowiak, I., and Smirnov, A.: AERONET – Afederated instrument network and data archive for aerosol char-acterization, Remote Sens. Environ., 66, 1–16, 1998.

Hsu, N. C., Tsay, S. C., King, M. D., and Herman, J. R.:Aerosol Properties Over Bright-Reflecting Source Regions, IeeeT. Geosci. Remote, 42, 557–569, 2004.

Hsu, N. C., Tsay, S. C., King, M. D., and Herman, J. R.: Deep blueretrievals of Asian aerosol properties during ACE-Asia, IEEE T.Geosci. Remote, 44, 3180–3195, 2006.



Ichoku, C., Chu, D. A., Mattoo, S., Kaufman, Y. J., Remer,L. A., Tanre, D., Slutsker, I., and Holben, B. N.: A spatio-temporal approach for global validation and analysis of MODISaerosol products, Geophys. Res. Lett., 29, MOD1.1–MOD1.4,doi:10.1029/2001GL013206, 2002.

IPCC1: AR4 WG, Climate Change 2007: The Physical Science Ba-sis, Contribution of Working Group I to the Fourth AssessmentReport of the Intergovernmental Panel on Climate Change, editedby: Solomon, S., Qin, D., Manning, M., Chen, Z., Marquis, M.,Averyt, K. B., Tignor, M., and Miller, H. L., Cambridge Univer-sity Press, ISBN 978-0-521-88009-1, 2007.

Kaufman, Y. J., Tanre, D., Remer, L., Vermote, E., Chu, A., and Hol-ben, B.: Operational remote sensing of tropospheric aerosol overland from EOS moderate resolution imaging spectroradiometer,J. Geophys. Res., 102, 17051–17067, 1997a.

Kaufman, Y. J., Wald, A., Remer, L., Gao, B., Li, R., and Flynn,L.: The MODIS 2.1 µm channel – Correlation with visible re-flectance for use in remote sensing of aerosol, IEEE T. Geosci.Remote, 35, 1286–1298, 1997b.

Kaufman, Y. J., Tanre, D., and Boucher, O.: A satellite view ofaerosols in the climate system, Nature, 419, 215–223, 2002.

Kittaka, C., Winker, D. M., Vaughan, M. A., Omar, A., and Remer,L. A.: Intercomparison of column aerosol optical depths fromCALIPSO and MODIS-Aqua, Atmos. Meas. Tech., 4, 131–141,doi:10.5194/amt-4-131-2011, 2011.

Levy, R. C., Remer, L. A., and Dubovik, O.: Global aerosol opti-cal properties and application to Moderate Resolution ImagingSpectroradiometer aerosol retrieval over land, J. Geophys. Res.-Atmos., 112, D13210, doi:10.1029/2006JD007815, 2007a.

Levy, R. C., Remer, L. A., Mattoo, S., Vermote, E. F., and Kauf-man, Y. J.: Second-generation operational algorithm: Retrieval ofaerosol properties over land from inversion of Moderate Resolu-tion Imaging Spectroradiometer spectral reflectance, J. Geophys.Res.-Atmos., 112, D13211, doi:10.1029/2006JD007811, 2007b.

Levy, R. C., Remer, L. A., Kleidman, R. G., Mattoo, S., Ichoku,C., Kahn, R., and Eck, T. F.: Global evaluation of the Collection5 MODIS dark-target aerosol products over land, Atmos. Chem.Phys., 10, 10399–10420, doi:10.5194/acp-10-10399-2010, 2010.

Levy, R. C., Mattoo, S., Munchak, L. A., Remer, L. A., Sayer, A.M., and Hsu, N. C.: The Collection 6 MODIS aerosol productsover land and ocean, Atmos. Meas. Tech. Discuss., 6, 159–259,doi:10.5194/amtd-6-159-2013, 2013.

Li, C. C., Lau, A. K. H., Mao, J. T., and Chu, D. A.: Retrieval, val-idation, and application of the 1-km aerosol optical depth fromMODIS measurements over Hong Kong, IEEE T. Geosci. Re-mote, 43, 2650–2658, 2005.

Li, R.-R., Kaufman, Y. J., Gao, B.-C., and Davis, C. O.: Remotesensing of suspended sediments and shallow coastal waters,IEEE T. Geosci. Remote, 41, 559–566, 2003.

Li, R., Remer, L., Kaufman, Y., Mattoo, S., Gao, B., and Vermote,E.: Snow and ice mask for the MODIS aerosol products, IEEEGeosci. Remote Sens. Lett., 2, 306–310, 2005.

Li, Y., Xue, Y., He, X., and Guang, J.: High resolution aerosol re-mote sensing retrieval over urban areas by synergetic use of HJ-1CCD and MODIS data, Atmos. Environ., 46, 173–180, 2012.

Loveland, T. R. and Belward, A. S.: The IGBP-DIS global 1km landcover data set, DISCover: first results, Int. J. Remote Sens., 18,3289–3295, 1997.

Lyapustin, A., Wang, Y., Laszlo, I., Kahn, R., Korkin, S., Remer,L., Levy, R., and Reid, J. S.: Multiangle implementation of atmo-spheric correction (MAIAC): 2. Aerosol algorithm, J. Geophys.Res.-Atmos., 116, D03211, doi:10.1029/2010JD014986, 2011.

Martins, J. V., Tanre, D., Remer, L. A., Kaufman, Y. J., Mattoo,S., and Levy, R.: MODIS Cloud Screening for Remote Sens-ing of Aerosol over Oceans using Spatial Variability, Geophys.Res. Lett., 29, MOD4.1–MOD4.4, doi:10.1029/2001GL013252,2002.

Oo, M. M., Jerg, M., Hernandez, E., Picon, A., Gross, B. M.,Moshary, F., and Ahmed, S. A.: Improved MODIS aerosol re-trieval using modified VIS/SWIR surface albedo ratio over urbanscenes, IEEE T. Geosci. Remote, 48, 983–1000, 2010.

Quaas, J., Boucher, O., Bellouin, N., and Kinne, S.: Satellite-basedestimate of the direct and indirect aerosol climate forcing, J. Geo-phys. Res., 113, D05204, doi:10.1029/2007JD008962, 2008.

Remer, L. A., Kaufman, Y. J., Tanre, D., Mattoo, S., Chu, D. A.,Martins, J. V., Li, R. R., Ichoku, C., Levy, R. C., Kleidman, R. G.,Eck, T. F., Vermote, E., and Holben, B. N.: The MODIS aerosolalgorithm, products, and validation, J. Atmos. Sci., 62, 947–973,2005.

Remer, L. A., Kleidman, R. G., Levy, R. C., Kaufman, Y. J.,Tanre, D., Mattoo, S., Martins, J. V., Ichoku, C., Koren, I., Yu,H., and Holben, B. N.: Global aerosol climatology from theMODIS satellite sensors, J. Geophys. Res. Atmos., 113, D14S07,doi:10.1029/2007JD009661, 2008.

Remer, L. A., Mattoo, S., Levy, R. C., and Munchak, L.: MODIS3 km aerosol product: algorithm and global perspective, Atmos.Meas. Tech. Discuss., 6, 69–112, doi:10.5194/amtd-6-69-2013,2013.

Schmid, B., Michalsky, J., Halthore, R., Beauharnois, M., Harrison,L., Livingston, J., Russell, P., Holben, B., Eck, T., and Smirnov,A.: Comparison of aerosol optical depth from four solar radiome-ters during the fall 1997 ARM intensive observation period, Geo-phys. Res. Lett., 26, 2725–2728, 1999.

Smirnov, A., Holben, B. N., Eck, T. F., Dubovik, O., and Slutsker,I.: Cloud screening and quality control algorithms for theAERONET database, Remote Sens. Environ, 73, 337–349, 2000.

Tanre, D., Kaufman, Y. J., Herman, M., and Mattoo, S.: Remotesensing of aerosol properties over oceans using the MODIS/EOSspectral radiances, J. Geophys. Res.-Atmos., 102, 16971–16988,1997.

van Donkelaar, A., Martin, R. V., and Park, R. J.: Estimat-ing ground-level PM2.5 with aerosol optical depth determinedfrom satellite remote sensing, J. Geophys. Res., 111, D21201,doi:10.1029/2005JD006996, 2006.

Wang, J. and Christopher, S. A.: Intercomparison between satellite-derived aerosol optical thickness and PM2.5 mass: Implica-tions for air quality studies, Geophys. Res. Lett., 30, 2095,doi:10.1029/2003GL018174, 2003.


AMTD6, 1683–1716, 2013

Applications overland in an

urban/suburbanregion

L. A. Munchak et al.

Title Page

Abstract Introduction

Conclusions References

Tables Figures

� �

� �

Back Close

Full Screen / Esc

Printer-friendly Version

Interactive Discussion

Discussion

Pa

per|

Discussion

Paper

|D

iscussionP

aper|

Discussion

Paper

|

Atmos. Meas. Tech. Discuss., 6, 1683–1716, 2013www.atmos-meas-tech-discuss.net/6/1683/2013/doi:10.5194/amtd-6-1683-2013© Author(s) 2013. CC Attribution 3.0 License.

Atmospheric Measurement

Techniques

Open A

ccess

Discussions

This discussion paper is/has been under review for the journal Atmospheric MeasurementTechniques (AMT). Please refer to the corresponding final paper in AMT if available.

MODIS 3 km aerosol product: applicationsover land in an urban/suburban region

L. A. Munchak1,2, R. C. Levy1,2, S. Mattoo1,2, L. A. Remer3, B. N. Holben1,J. S. Schafer1,4, C. A. Hostetler5, and R. A. Ferrare5

1Earth Science Division, NASA Goddard Space Flight Center, Greenbelt MD 20771, USA2Science Systems and Applications, Inc., Lanham MD 20709, USA3Joint Center for Earth Systems Technology, University of Maryland Baltimore County,Baltimore MD 21228, USA4Sigma Space Corporation, Lanham MD 20706, USA5NASA Langley Research Center, Hampton VA 23681, USA

Received: 31 January 2013 – Accepted: 1 February 2013 – Published: 14 February 2013

Correspondence to: L. A. Munchak ([email protected])

Published by Copernicus Publications on behalf of the European Geosciences Union.

1683

AMTD6, 1683–1716, 2013




Title Page



Tables Figures

� �

� �

Back Close

Full Screen / Esc



Discussion

Pa

per|

Discussion

Paper

|D

iscussionP

aper|

Discussion

Paper

|

Abstract

MODerate resolution Imaging Spectroradiometer (MODIS) instruments aboard theTerra and Aqua satellites have provided a rich dataset of aerosol information at a 10 kmspatial scale. Although originally intended for climate applications, the air quality com-munity quickly became interested in using the MODIS aerosol data. However, 10 km5

resolution is not sufficient to resolve local scale aerosol features. With this in mind,MODIS Collection 6 is including a global aerosol product with a 3 km resolution. Here,we evaluate the 3 km product over the Baltimore/Washington D.C., USA, corridor dur-ing the summer of 2011, by comparing with spatially dense data collected as part of theDISCOVER-AQ campaign; these data were measured by the NASA Langley Research10

Center airborne High Spectral Resolution Lidar (HSRL) and a network of 44 sun pho-tometers (SP) spaced approximately 10 km apart. The HSRL instrument shows thatAOD can vary by up to 0.2 within a single 10 km MODIS pixel, meaning that higher res-olution satellite retrievals may help to characterize aerosol spatial distributions in thisregion. Different techniques for validating a high-resolution aerosol product against SP15

measurements are considered. Although the 10 km product is more statistically reliablethan the 3 km product, the 3 km product still performs acceptably, with more than two-thirds of MODIS/SP collocations falling within the expected error envelope with highcorrelation (R >0.90). The 3 km product can better resolve aerosol gradients and re-trieve closer to clouds and shorelines than the 10 km product, but tends to show more20

significant noise especially in urban areas. This urban degradation is quantified usingancillary land cover data. Overall, we show that the MODIS 3 km product adds newinformation to the existing set of satellite derived aerosol products and validates wellover the region, but due to noise and problems in urban areas, should be treated withsome degree of caution.25

1684

AMTD6, 1683–1716, 2013




Title Page



Tables Figures

� �

� �

Back Close

Full Screen / Esc



Discussion

Pa

per|

Discussion

Paper

|D

iscussionP

aper|

Discussion

Paper

|

1 Introduction

For over twelve years, the MODerate resolution Imaging Spectroradiometer (MODIS)instruments, flying aboard the Terra and Aqua satellites, have been collecting data onthe global distribution of climate related variables. Several of these variables describeaerosols, which have been shown to be one of the largest sources of uncertainty in5

the climate system (IPCC, 2007). MODIS provides a suite of information about aerosolproperties, including aerosol optical depth (AOD) and aerosol sizing parameters, thathave been used extensively for climate studies (e.g. Kaufman et al., 2002; Bellouinet al., 2005; Quaas et al., 2008). Currently, three algorithms are used operationally toprovide products to the public – the dark target algorithm over land (Kaufman et al.,10

1997a; Levy et al., 2007b, 2010), the dark target algorithm over ocean (Tanre et al.,1997; Remer et al., 2005, 2008) and the deep blue algorithm over land (Hsu et al.,2004, 2006). All produce data at a 10 km nominal spatial scale, which is an appropriatescale for climate studies (Remer et al., 2013).

Soon after launch, it was realized that the retrieved aerosol parameters could be15

used as proxies for estimating surface air quality conditions (e.g Hutchison, 2003;Wang and Christopher, 2003; Chu et al., 2003; Engels-Cox et al., 2004; Hutchisonet al., 2005). The MODIS instruments are particularly appealing for air quality applica-tions because the broad (2330 km) swath of the instrument allows most global locationsto be monitored on a nearly daily basis. However, the 10 km resolution of the MODIS20

aerosol products is insufficient to resolve small-scale features relevant to local air qual-ity, including point sources in urban areas (C. C. Li et al., 2005) and small fire plumes(Lyapustin et al., 2011). Although several research techniques exist to retrieve aerosolsat higher spatial scales (100 m to 1 km) from MODIS observations (e.g., Li et al., 2012;Lyapustin et al., 2011), they have not been produced globally in an operational envi-25

ronment.In response to requests from the air quality community, MODIS Collection 6 will

include a global aerosol product at nominal 3 km scale, in addition to the standard

1685

AMTD6, 1683–1716, 2013




Title Page



Tables Figures

� �

� �

Back Close

Full Screen / Esc



Discussion

Pa

per|

Discussion

Paper

|D

iscussionP

aper|

Discussion

Paper

|

10 km resolution. This 3 km product will include retrievals based on both dark targetalgorithms (land and ocean). Remer et al. (2013), describe the 3 km product in detail,and analyzed a six-month test database to compare the performance of the 3 km and10 km products on a global scale. The 3 km product can show more spatial detail thanthe 10 km product, but agrees less well with ground based sun photometers on a global5

scale. Accordingly, the 10 km product is still the best choice for studies concerningaerosol effects on climate. However, the 3 km product may be better at characterizingaerosol distributions on local scales, and can only be tested by comparing to denseobservations on a small spatial scale. We wish to examine the 3 km aerosol producton a regional scale to determine the additional information content that the higher10

resolution retrieval provides, and the possible degradation in data quality introducedby moving to a higher resolution.

In this work, we focus primarily on AOD, because it has been used for surface airquality applications (van Donkelaar et al., 2006), and can easily be compared withsunphotometer data. Specifically, we can evaluate AOD products at both 10 km and15

3 km resolutions by comparing with surface and airborne-based AOD measurementscollected during the DISCOVER-AQ field campaign conducted during the summer of2011 over the Baltimore-Washington D.C. corridor of the United States. We first assessthe similarities and differences in 10 km and 3 km MODIS AOD images, and evaluatewhether the 3 km product provides information that is not captured by the 10 km prod-20

uct. We look critically at the best method to validate the higher resolution aerosol re-trieval against sun photometer measurements, and validate the 10 km and 3 km prod-ucts for the campaign duration. The ability of the MODIS product at both 10 km and3 km resolutions to capture spatial variability in AOD is also addressed. Due to the lim-ited validation data for the over ocean variables, we will only address the land variables25

in this study.

1686

AMTD6, 1683–1716, 2013




Title Page



Tables Figures

� �

� �

Back Close

Full Screen / Esc



Discussion

Pa

per|

Discussion

Paper

|D

iscussionP

aper|

Discussion

Paper

|

2 The MODIS 3 km land algorithm

The retrieval method of the global MODIS 3 km aerosol product is thoroughly detailedin Remer et al. (2013). For the sake of completeness, we will briefly describe the landalgorithm here. The 3 km algorithm emerges from the “dark target” retrieval methodol-ogy (Kaufman et al., 1997a; Tanre et al., 1997; Remer et al., 2005; Levy et al., 2007b,5

2013), based on the concept that in the visible wavelengths, aerosols are bright andvegetated surfaces tend to be dark. The spectral contrast between aerosols and thesurface can be used to retrieve quantitative information about aerosol properties.

To increase signal-to-noise, the MODIS algorithm retrieves aerosol parameters ata lower spatial resolution than the nominal (at nadir) 500 m top of atmosphere (TOA)10

reflectance measurements. The algorithm creates N by N “retrieval boxes” of pixels,in order to filter out the subset that is not desirable for aerosol retrieval. Thus, the10 km algorithm works with 20×20 pixel retrieval boxes (400 pixels), whereas the 3 kmalgorithm works with 6×6 pixel retrieval boxes (36 pixels). Remer et al. (2013) describeshow pixels are masked for cloud (Martins et al., 2002), sediments in water (Li et al.,15

2003), snow and ice (R. Li et al., 2005) and surfaces that are too bright for retrieval.The remaining pixels are sorted by their 0.66 μm reflectance, and the brightest 50 %and darkest 20 % of the remaining pixels are discarded. This means that in the 10 km(3 km) algorithm, there are at most 120 pixels (11 pixels) remaining from which to doaerosol retrieval. The reflectances of these remaining pixels are averaged, resulting20

in on set of spectral reflectance values to drive the aerosol retrieval. These spectralreflectance values are further “corrected” for gas absorption (e.g. Levy et al., 2013).

The expected “quality” of the retrieval is determined by the number of pixels thatremain after all masking and filtering. If at least 51 pixels remain (out of 120) for 10 kmor 5 pixels remain (out of 11) for 3 km, then the retrieval is initially expected to be “high”25

quality. The 10 km retrieval will be attempted if only 12 pixels remain (out of 120), butit is expected to be low quality. There is no corresponding low quality attempt for the3 km retrieval (Remer et al., 2013).

1687

AMTD6, 1683–1716, 2013




Title Page



Tables Figures

� �

� �

Back Close

Full Screen / Esc



Discussion

Pa

per|

Discussion

Paper

|D

iscussionP

aper|

Discussion

Paper

|

The over land algorithm performs the retrieval by matching the observed TOA spec-tral reflectance with lookup tables of pre-computed TOA spectral reflectance. Thelargest uncertainty is in characterizing the contribution of the surface to the TOA re-flectance, for which errors of 0.01 can lead to errors of 0.1 in retrieved AOD. However,Kaufman et al. (1997b) showed that the surface contribution in the 0.47 and 0.66 μm5

bands can be related to surface contribution in the 2.1 μm shortwave-IR (SWIR) band.The assumed relationship between the visible and SWIR was refined by Levy etal. (2007b), and is used as a constraint in the aerosol retrieval. As described by Levyet al. (2007b), the current retrieval uses the three channels (0.47 μm, 0.66 μm, and2.1 μm) to determine both the amount of aerosol (total AOD) and the relative ratio of10

fine and coarse aerosol models. The selection of the fine and coarse aerosol modelsare prescribed by season and location. The resulting primary retrieved aerosol parame-ters over land are total AOD at 0.55 μm, the fractional contribution of the fine-dominatedaerosol type, the constrained surface reflectance, and fitting error (Levy et al., 2013).All of these assumptions (aerosol lookup tables, assumed surface reflectance relation-15

ship, and methodology of the inversion) are identical for both 10 km and 3 km retrievals(e.g. Levy et al., 2013; Remer et al., 2013).

Previously, the MODIS 10 km product has been thoroughly evaluated against sunphotometers on global scales. The global expected error (EE) of the 10 km AOD at0.55 μm product over land is 0.05±0.15 AOD (Remer et al., 2005), and a testbed20

of 10 km data produced with the Collection 6 MODIS algorithm shows that, globally,70.6 % of MODIS-Aqua land AOD retrievals fall within this range when collocated andcompared to ground based sun photometer measurements (Levy et al., 2013). Re-mer et al. (2013) has compared the global 3 km MODIS product with sunphotometersand shown that it tends to validate less well than the 10 km. Specifically, since only25

62 % of global collocations fell within expected error over land, a new expected error of0.05±0.25 AOD has been established for the 3 km land product (Remer et al., 2012).Although a new 3 km EE has been established, we choose to continue to use the more

1688

AMTD6, 1683–1716, 2013




Title Page



Tables Figures

� �

� �

Back Close

Full Screen / Esc



Discussion

Pa

per|

Discussion

Paper

|D

iscussionP

aper|

Discussion

Paper

|

stringent 10 km EE in this paper. Therefore, in this paper, when “EE” is referred to, the0.05±0.15 AOD definition is employed.

3 Data sources

3.1 MODIS products

This analysis uses five minute sections of the MODIS orbits, termed “granules”, col-5

lected between 20 June 2011 to 31 July 2011 from both the Aqua and Terra satellites.The level 1B calibrated reflectances for Terra used in the aerosol retrieval were notfinalized for collection 6, and may be slightly different from the operational collection 6level 1B. However, the new calibration technique employed in MODIS Collection 6 isfollowed in the creation of these level 1B granules.10

Additionally, the MODIS land surface cover product (MCD12Q1) is used to charac-terize surface type. The land cover product uses a trained algorithm to characterizeland surface at a 500 m resolution using 5 different classification schemes (Freidl et al.,2010). In this work, we only consider the 17-class International Geosphere-BiosphereProgramme (IGBP) system (Loveland and Belward, 1997). Data from both the Terra15

and Aqua satellites is included in both the training dataset and the product.

3.2 AERONET/DRAGON

The AErosol RObotic NETwork (AERONET) (Holben et al., 1998) has 5 permanentstations operating in the study region. These stations were supplemented with an ad-ditional 39 temporary stations, distributed in a roughly 10 km by 10 km grid, termed the20

Distributed Regional Aerosol Gridded Observation Networks (DRAGON). This networkprovided an unprecedented opportunity to validate satellite derived aerosol propertiesat a high spatial resolution. Each station is equipped with a Cimel sun photometer (SP)measuring at eight spectral bands between 340 nm and 1020 nm. AOD is calculated at

1689

AMTD6, 1683–1716, 2013




Title Page



Tables Figures

� �

� �

Back Close

Full Screen / Esc



Discussion

Pa

per|

Discussion

Paper

|D

iscussionP

aper|

Discussion

Paper

|

550 nm using a quadratic log-log fit (Eck et al., 1999). The AOD retrieval is expected tobe accurate to within ±0.015 (Eck et al., 1999; Schmid et al., 1999)

In this work, we use the level 1.5 AERONET/DRAGON SP data, which are cloudscreened (Smirnov et al., 2000) but not quality controlled. However, comparisons be-tween level 1.5 and level 2.0 AOD have a correlation of 0.99, a slope of 1 and no offset,5

and the level 1.5 data contains more measurements.

3.3 High Spectral Resolution LIDAR (HSRL)

The NASA Langley Research Center airborne HSRL instrument calculates aerosol op-tical depth at 532 nm using independent measurements of vertically resolved aerosolbackscatter and extinction, and is therefore expected to be more accurate than a stan-10

dard backscatter lidar, which typically requires additional data and/or assumptions toproduce extinction profiles, and therefore AOD, from the backscatter (Hair et al., 2008).25 flights were conducted on 13 days between 1 July 2011 and 29 July 2011 on theNASA Langley King Air B200 airplane.

4 Results15

4.1 Case studies

Because much of the value of the 3 km resolution retrieval is in better resolving smallerscale aerosol features, we first look at two case studies from differing days in theDISCOVER-AQ study. Figure 1 shows the 10 km (panels a and c) and 3 km (panelsb and d) resolution MODIS AOD at 0.55 μm observed by MODIS-Aqua on 21 July 201120

at 18:30 UTC. On this day, there is a strong AOD gradient, with a change from 0.15 to0.45 AOD over ∼50 km, and a local AOD enhancement reaching to over 0.75. HYS-PLIT back trajectories, using winds from the NAM model at 12 km resolution, showsthat the lower AOD air mass originated in the Midwest, while the higher AOD air mass

1690

AMTD6, 1683–1716, 2013




Title Page



Tables Figures

� �

� �

Back Close

Full Screen / Esc



Discussion

Pa

per|

Discussion

Paper

|D

iscussionP

aper|

Discussion

Paper

|

was coming from southern Virginia and North Carolina, which had active biomass burn-ing nearby, and also passed over more urbanized regions. The two MODIS productsboth resolve the AOD gradient, and, in general, show the same aerosol situation. Bothproducts resolve the SP measured AOD maximum of 0.75, although the 3 km productshows more high AOD pixels which may or may not be noise. The 3 km product is5

able to retrieve closer to clouds, and therefore shows more detail and retrieves morearea than the 10 km product. The 3 km product is also able to retrieve over water bod-ies which are too narrow for 10 km retrievals, and has more pixels near the coastlinethat are unable to be retrieved by the 10 km product. The ability to retrieve closer tocoastline is particularly valuable for air quality applications because many major cities,10

including some of the largest and most polluted, are located on coasts. This granulenot only highlights the value of a high resolution satellite product, but also shows thatAOD can vary significantly over small distances.

Plotted overtop of the MODIS AOD images in the circles are MODIS/SP collocations.In all of the panels (a–d), the inner circle shows the SP temporal mean, which is cal-15

culated by averaging all SP 0.55 μm measurements at a station within 30 min of theMODIS overpass time, with a minimum of 2 observations. In the top row (panels a andb), the outer circle shows the AOD of the MODIS pixel in which the SP station is located.In the bottom row (panels c and d), the outer circle shows the MODIS spatial mean,which is calculated by averaging all high quality (QA=3) land pixels within a 5 pixel by20

5 pixel box surrounding the SP station requiring at least 5 out of a possible 25 pixels(Ichoku et al., 2002). Therefore, the spatial averaging box for the 3 km product is 15 kmby 15 km, and the spatial averaging box for the 10 km product is 50 km by 50 km. Thesingle pixel MODIS collocations in the top row show excellent agreement between theMODIS retrieval and the SP measurements for both the 10 and 3 km products. How-25

ever, the MODIS spatial average shown in the bottom row gives the impression that the10 and 3 km products provide different answers; in this case, the smaller averaging boxdata creates better MODIS/SP agreement for the 3 km product, while the larger aver-aging box of the 10 km product smoothes out the gradient and causes disagreement.

1691

AMTD6, 1683–1716, 2013




Title Page



Tables Figures

� �

� �

Back Close

Full Screen / Esc



Discussion

Pa

per|

Discussion

Paper

|D

iscussionP

aper|

Discussion

Paper

|

The spatial averaging does increase the number of MODIS-SP collocations as com-pared to the single pixel technique, adding 3 collocations for the 3 km product and 2 forthe 10 km product. This granule shows that the single pixel collocation technique bettercharacterizes AOD at the SP site, but limits the number of collocations and thereforethe statistical robustness of the validation.5

Figure 2 shows the 10 km and 3 km resolution MODIS AOD observed by MODIS-Terra on 1 July 2011 at 15:35 UTC, which has a markedly different aerosol distributionthan that shown in Fig. 1. On this day, AOD is low (0.11 as measured by the SPs) andis homogeneous across the region. As in Fig. 1, the large-scale view of the two MODISproducts agree, but there are nuanced differences. Some noisy pixels are present in10

the 3 km product, and are less apparent in the 10 km product. Both the 10 km and 3 kmproducts have elevated AOD along the New Jersey and Delaware coastline, which areplausibly cloud contaminated from sub pixel clouds. The 10 km has fewer contaminatedpixels, but they extend further from the coastline; the 3 km product has more contam-inated pixels, but their spatial extent is limited to right along the coastline. The 3 km15

product also shows some striping that results from differences in the MODIS instru-ment detectors. As in Fig. 1, the 3 km product has more coverage over the water areasin this scene. The MODIS/SP collocation circles are as they were in Fig. 1, with MODISsingle pixel collocations shown in the outer circles of the top row, MODIS spatial aver-age collocations in the outer circles of the bottom row, and SP temporal averages in20

the inner circle of all panels.In this aerosol situation, the spatial averaging of the 10 km product AOD increases

the agreement with the SP measurements because the averaging decreases the im-pact of noisy pixels. 30 out of the 31 collocations shown in Fig. 2c fall within expectederror, and the average retrieved 0.55 μm AOD at the SP stations is 0.10, only 0.01 lower25

than the SP determined AOD. The MODIS 3 km spatially averaged collocations showan AOD enhancement of 0.1 to 0.3 over Baltimore City that is not observed by theSP measurements; 6 out of the 28 MODIS 3 km spatial averages are above expectederror. The largest difference between the MODIS 3 km retrieved AOD & SP measured

1692

AMTD6, 1683–1716, 2013




Title Page



Tables Figures

� �

� �

Back Close

Full Screen / Esc



Discussion

Pa

per|

Discussion

Paper

|D

iscussionP

aper|

Discussion

Paper

|

AOD was observed at the AERONET/DRAGON station in Essex, Maryland, 13 km eastof downtown Baltimore, where the SP AOD is 0.11, and the AOD of the 3 km spatialaverage is 0.41. A similar AOD enhancement is seen over the Washington D.C., al-though there are no SP measurements to confirm that this enhancement is a retrievalartifact. The single MODIS pixel collocations in the Fig. 2b do not capture this impor-5

tant retrieval problem due to the coincidences of where the AERONET stations happento be located and if the exact pixel where the station is located is retrieved. This casestudy suggests that the spatial average collocation technique better characterizes theretrieval performance over the larger region, and the single pixel collocation techniquemay misrepresent product performance because it is much more sensitive to the siting10

of the SP stations.The airborne HSRL measurements provide an opportunity to look at aerosol extinc-

tion and AOD spatial variability with an nominal 1 min resolution, which correspondsto a nearly 6 km horizontal resolution. Figure 3 shows another high aerosol loadingday, observed by MODIS-Terra on 29 July 2011 at 16:00 UTC. HSRL 532 nm columnar15

AOD along the flight track between 15:30 and 16:30 UTC is shown in the thick line,plotted atop of the 3 km resolution 0.55 μm AOD. The sections of the flight track wherethe plane was flying lower than 7 km, flying above clouds, or was turning are shown ingrey. SP AOD interpolated to 0.55 μm is shown in the circles. The AERONET measure-ments are not temporally averaged; the nearest measurement to 16:00 UTC is used,20

provided the measurement was taken between 15:30 and 16:30 UTC.The SP AOD measurements range from 0.33 to 0.58, with lower values towards

the south and west, and higher values towards the north and east. The HSRL AODmeasurements have a wider range from 0.35 to 0.67. There is excellent agreementbetween HSRL and SP measurements at the AERONET stations, which indicates that25

the broader range of AOD values from HSRL than from AERONET is likely due to theHSRL instrument being able to measure a larger area and not due to inaccuracies ineither measurement. The MODIS 3 km product generally captures the spatial pattern ofAOD as observed by both the HSRL instrument and the AERONET stations, although

1693

AMTD6, 1683–1716, 2013




Title Page



Tables Figures

� �

� �

Back Close

Full Screen / Esc



Discussion

Pa

per|

Discussion

Paper

|D

iscussionP

aper|

Discussion

Paper

|

as in Fig. 2, the AOD is overestimated in the urban areas and there are noisy retrievals.Outside of the urban areas, the MODIS image shows AOD near 0.4 for much of thesouthern and western portion of the granule, increasing for 0.65 in the northeasternportion of the granule that also has HSRL and SP measurements. Overall, the threeinstruments show a rather similar aerosol situation albeit with varying levels of detail.5

The HSRL AOD measurements reflect the more general pattern observed byAERONET, but also show many small-scale enhancements that cannot be observedeven with the high spatial density of the DRAGON network. The western portion ofthe flight track in Fig. 4 shows that the HSRL instrument measured an increase inAOD from 0.41 to 0.59 over ∼ 4 min between 16:00:08 and 16:04:28 UTC, and travel-10

ing south by 29.8 km and west by 5.3 km in this time period . This change takes placeover three 10 km boxes and ten 3 km boxes. The MODIS 3 km product successfullyreplicates the absolute change in AOD, although the gradient along the flight track isnot exactly reproduced because of missing and noisy pixels. The maximum AOD isoverestimated in the 10 km product, and is retrieved exactly in the 3 km product, al-15

though a neighboring pixel has an anomalously high AOD. It is probable that there is asource of contamination in the noisy 3 km pixel, which was included in the 10 km pixelalong with the good pixels, resulting in an overestimation in the 10 km pixel. From thiscomparison, it is clear that while the 3 km product contains more noisy pixels that arelikely contaminated, the smaller pixel size limits the spatial extent of the contamination20

and allows nearby uncontaminated pixels to retrieve the correct AOD. In the 3 km prod-uct, the effects of cloud or surface contamination are larger within a single pixel, but atthe same time are also spatially limited.

As highlighted in Fig. 4, because the HSRL instrument makes several measure-ments within a single MODIS pixel, it provides an opportunity to assess AOD variability25

across the standard MODIS retrieval resolution. A measure of subpixel AOD variabilityis shown in Fig. 5. For each 10 km resolution MODIS pixel, the minimum HSRL mea-sured AOD is subtracted from the maximum HSRL AOD, provided the measurementsare within half an hour of the MODIS overpass. Data from all flights that have valid data

1694

AMTD6, 1683–1716, 2013




Title Page



Tables Figures

� �

� �

Back Close

Full Screen / Esc



Discussion

Pa

per|

Discussion

Paper

|D

iscussionP

aper|

Discussion

Paper

|

within half an hour of a MODIS overpass are included, and only MODIS pixels that con-tain more than 5 HSRL observations are analyzed to ensure that two fully independentHSRL measurements are obtained within the MODIS pixel. 410 MODIS pixels are ableto be analyzed for subpixel AOD variations. Within a single MODIS 10 km pixel, themaximum HSRL measured AOD change is 0.25, increasing from 0.36 to 0.61 AOD. The5

mean 0.55 μm AOD variation within a 10 km pixel is 0.037 while the median variationwithin a 10 km pixel is 0.023. Fewer than 25 % of pixels contained less than 0.01 AODvariation. The 10 km retrieval assumes uniformity of the AOD across the retrieval box,and that is simply not the case. The assumption is likely to hold better across the 3 kmretrieval box; however, the nominal resolution of the HSRL instrument for this study is10

larger than the 3 km retrieval box. For this case, decreasing the HSRL nominal horizon-tal resolution for AOD from 6 km to 1 km increased the noise to an unacceptably highlevel such that the AOD variation within the MODIS 3 km pixels could not be quantified.

The case study comparisons with ground and airborne based AOD measurementsshow that the MODIS 3 km product contains additional aerosol information as com-15

pared to the 10 km product. Figure 1 shows a situation in which a few high AOD pix-els in the 10 km product could be interpreted as noise, but the higher resolution ofthe 3 km product provides many more pixels, and SP data confirms that a true highaerosol loading event was identified. Additionally, the 3 km retrieval is able to retrieveover small water bodies and close to coastline. The sub-pixel AOD variability observed20

by HSRL suggests that a 10 km product will frequently miss small changes in AODover an urban/suburbam region. However, retrieving with a higher resolution reducesthe statistical robustness of the aerosol product, resulting in more noise and inaccurateretrievals over urban areas.

4.2 MODIS 3 km validation25

The case studies highlight the strengths and weaknesses of the 3 km product; however,the sun photometer data from the entire campaign provides the most comprehensivedataset to characterize the performance of the 3 km product. The SP data also allows

1695

AMTD6, 1683–1716, 2013




Title Page



Tables Figures

� �

� �

Back Close

Full Screen / Esc



Discussion

Pa

per|

Discussion

Paper

|D

iscussionP

aper|

Discussion

Paper

|

a critical look at how the different collocation techniques demonstrated in Figs. 1 and 2lead to different conclusions on the validity on the MODIS AOD data in this region forboth the 10 and 3 km resolutions.

In Fig. 6a, all 3 km MODIS/SP 0.55 μm AOD collocations are shown in a densityscatterplot, using the MODIS single pixel collocation technique employed in Figs. 1b5

and 2b. In Fig. 6b, these single pixel MODIS/SP collocations are binned by the SPAOD, and the mean MODIS-SP difference for the bin is plotted for both the 10 and3 km products. This process is repeated in Fig. 6c and d except that the 5×5 pixelMODIS spatial collocation technique is employed in the MODIS/SP collocations.

For both MODIS/SP collocation techniques, the bulk of the 3 km MODIS retrievals10

are within the ±0.05±0.15 AOD expected error (67 % for the single pixel techniqueand 73 % for the spatial average technique). The satellite retrieval is highly correlatedwith the ground based measurements, although with a small negative offset (−0.02 forthe single pixel technique and −0.01 for the spatial average technique) and a significantslope (1.26 for single pixel, 1.22 for spatial average). Accordingly, the MODIS 3 km AOD15

agrees well with SP AOD at low aerosol loadings, but is generally biased high at higherAOD. The primary difference between the two collocation techniques is that the spatialaverage has nearly twice as many collocations (962 as opposed to 560). Therefore,the spatial averaging technique may be more representative of the region at large.

The binned MODIS-SP differences in Fig. 6b and d agree well with the impression20

given by the scatterplots. The MODIS 3 km product is biased slightly low at AOD below0.1, then is biased high above 0.1. The single pixel collocation is likely best to comparethe 10 km and 3 km products, because the spatial averaging technique samples a dif-ferent area for the different products. The 3 km product single pixel averages shown inFig. 6b show that the average AOD of the 3 km product is frequently, but not always,25

higher than the 10 km product for bins above 0.1 AOD.The retrieved AOD from the HSRL airborne instrument provides a different perspec-

tive on validation of the 3 km product. Because the plane is moving quickly in space,no spatial or temporal averaging is used to make the MODIS/HSRL collocations. The

1696

AMTD6, 1683–1716, 2013




Title Page



Tables Figures

� �

� �

Back Close

Full Screen / Esc



Discussion

Pa

per|

Discussion

Paper

|D

iscussionP

aper|

Discussion

Paper

|

MODIS data are sampled along the HSRL flight track, provided that the HSRL obser-vation is made within the geographic region mapped in Figs. 1 and 2, and is within30 min of the MODIS overpass. Shortening the time requirement to either 15 or 5 mindoes not significantly affect the collocation statistics, but provides far fewer colloca-tions. The HSRL and MODIS AOD are not interpolated to the same wavelength, which5

is expected to introduce between 2–4 % error (Kittaka et al., 2011). It is expected thatthe AOD retrieved by the HSRL instrument would be lower than the AOD retrieved byMODIS by at least 0.01 to 0.02 because the HSRL instrument does not measure theentire column, and therefore neglects the contribution of above the HSRL profile (i.e.above 7 km) to total columnar AOD.10

Figure 7 shows that the 3 km product agrees well with HSRL, although less wellthan the 10 km product. Although both MODIS products are biased high in this regionduring the study period, the 3 km product is moreso; the bias of the 10 km product is0.050 while the bias of the 3 km product is 0.056. The increased bias of the 3 km isprimarily attributed to the addition of noisy pixels that the 10 km product avoids during15

its pixel selection process. Although the 10 km product has better statistical agreementwith HSRL, the 10 km MODIS/HSRL comparison also shows the necessity of a higherresolution retrieval; the stripes in the HSRL/MODIS comparison show the change inHSRL AOD within a single pixel – the pattern that is quantified in Fig. 5.

From both AERONET and HSRL comparisons, we consider the 3 km product to be20

well validated in this region. Using either collocation technique, more than two thirdsof MODIS/SP collocations are within the expected error of the 10 km product, which ismore stringent than the defined EE of the 3 km product. However, the increased noiseand urban problems give the 3 km product a larger high bias than the 10 km product.The single pixel and spatial average MODIS collocation techniques show quantitative25

but not qualitative differences.

1697

AMTD6, 1683–1716, 2013




Title Page



Tables Figures

� �

� �

Back Close

Full Screen / Esc



Discussion

Pa

per|

Discussion

Paper

|D

iscussionP

aper|

Discussion

Paper

|

4.3 Sources of uncertainty

Figures 2 and 3 show an overestimation of AOD in urban areas. In order to study thisfurther, we separate the 3 km MODIS/SP spatially averaged collocations by SP station,and compute agreement statistics for each station. We choose the spatial averagingcollocation technique because it provides more collocations. The percent of compar-5

isons within expected error for each station is shown in Fig. 8a, and the correlationcoefficient (R) between MODIS and SP 0.55 μm AOD is shown in Fig. 8b. In eachpanel, land classified as “urban” by the MODIS Land Cover type product (MCD12Q1)is plotted in grey.

Figure 8a shows remarkable variability in the percent of collocations within EE for10

each SP station, ranging between 16 % to 95 %. The stations with the smallest percentwithin EE are clustered around the Baltimore urban center, and the more urbanizedI-95 corridor between Washington D.C. and Baltimore. However, the stations with smallpercentages of collocations within EE still have correlation coefficients above 0.85,suggesting that the error is a systematic bias that could be accounted for in future15

retrievals, and not a random error.We wish to quantify the urban overestimation of the 3 km product using the dense

spatial distribution of the SP sites. Using the same 15 km averaging box as the MODISspatial collocation, the percent of pixels within the averaging box identified as urbanby the MCD12Q1 product is calculated. In Fig. 9, this quantity is plotted against the20

percent of collocations above EE for each SP site. The percent of urban land withinthe MODIS averaging box is well correlated with the percent of MODIS/SP collocationsthat are above the expected error. This confirms that the largest source of uncertaintyin this study is related to the urban areas. The cause is suspected to be improper landsurface reflectance characterization, considering that aerosol sources are not expected25

to vary significantly within the region of study and that it has been shown that thesurface reflectance in the 0.66 um channel over some urban surfaces is ∼ 0.7 of the2.13 reflectance (Castanho et al., 2008; Oo et al., 2010) which is higher than the ratio

1698

AMTD6, 1683–1716, 2013




Title Page



Tables Figures

� �

� �

Back Close

Full Screen / Esc



Discussion

Pa

per|

Discussion

Paper

|D

iscussionP

aper|

Discussion

Paper

|

used in the MODIS operational retrievals. In mixed-use urban/suburban regions likethe DISCOVER-AQ study region, it is likely that the brighter urban surface 500 m pixelswere selectively discarded in the pixel selection process of the 10 km product. However,because the resolution of the 3 km retrieval is smaller, more urban pixels remain withinthe retrieval box after pixel selection, causing the surface reflectance in the visible5

wavelengths to be underestimated, leading to an overestimation in AOD.Ideally, the noisy pixels could be flagged as poor quality, and therefore treated with

some degree of caution. One of the products produced operationally is the AOD at2.1 μm. This is not a validated product, but is used as a diagnostic to the retrieval.It may hold information that can identify poorer quality 3 km retrievals. In Fig. 10a, the10

2.1 μm AOD on 21 July 2011 at 18:30 from the 3 km product is plotted, the same granuleas shown in Fig. 1b, and in Fig. 10b, the 2.1 μm AOD on 1 July 2011 at 15:35 from the3 km product is plotted, the same granule as shown in Fig. 2b. The scene-wide 2.1 μmAOD is 0.042 in the left panel, and is 0.03 in the right panel. Both granules containpixels with a 2.1 μm AOD over 0.2, more than 6 to 10 times higher than the scene-wide15

average. The error is not due to the wrong aerosol model being picked; in both scenes;all of the pixels were assigned the fine mode dominated urban aerosol model (Levy etal., 2007a). It appears that the same pixels that have anomalously high 2.1 μm AODretrievals also tend to have inaccurate 0.55 μm AOD retrievals. For example, the pixelswith high (>0.55) AOD in Fig. 1b that agree with SP observations have retrieved 2.1 μm20

AOD values that do not stand out from the background. However, the pixels with highAOD in Fig. 2b that do not agree with SP observations, exhibit retrieved 2.1 μm AODvalues significantly higher than the background. Although the 2.1 μm AOD retrieval is anunvalidated product, it may serve a purpose as a consistency check in this fine modedominated, urban environment. This consistency check is likely to be inapplicable in25

dusty or other coarse mode dominated regimes.Using anomalous 2.1 μm AOD pixels as a proxy for poor quality 0.55 μm pixels, and

therefore throwing them out, results in a reduction of the both the negative bias atlow AOD and the positive bias at high AOD. If the 3 km product is filtered such that

1699

AMTD6, 1683–1716, 2013




Title Page



Tables Figures

� �

� �

Back Close

Full Screen / Esc



Discussion

Pa

per|

Discussion

Paper

|D

iscussionP

aper|

Discussion

Paper

|

pixels with a 2.1 μm AOD over 0.2 are excluded, and MODIS/SP are collocated usingthe spatial averaging technique, the percent within EE rises from 73 % to 80 %, whilethe number of collocations falls from 963 to 903, and the R rises from 0.92 to 0.95.However, this screening is not globally applicable, and is not recommended for usewithout closer examination.5

It appears that a primary source of error in the 3 km MODIS aerosol product is the in-correct estimation of surface reflectance over urban regions. This error is not as appar-ent in the 10 km product because brighter urban surfaces are preferentially discardedduring the pixel selection process. However, because the 3 km product has a smallerfootprint, it is likely that not all urban pixels are discarded, and the surface reflectance10

is underestimated, leading to an overestimation in AOD. The 2.1 μm AOD shows somepotential as an imperfect filter for these poor quality pixels.

5 Summary and conclusions

The MODIS dark target aerosol algorithms were designed with climate applicationsin mind, and the spatial scale of the retrieval reflects that design. However, as the15

applications of the aerosol product have evolved to include many facets of air qualitystudies, the proper scale at which to observe small-scale aerosol events came intoquestion. In response to the needs of the air quality community, an operational aerosolproduct produced at a 3 km resolution will be available in MODIS Collection 6.

This new product was evaluated against ground and airborne data over the20

Baltimore-Washington D.C., USA corridor over the summer of 2011 to understand thepotential added information from the higher resolution retrieval and the statistical relia-bility of the product. The 10 km and 3 km granules generally resolve similar large scalepatterns in aerosol distribution but differ in the details. The smaller footprint of the 3 kmproduct retrieves closer to clouds and over small bodies of water, but also makes the25

product more susceptible to cloud contamination and other sources of noise. Sub-pixelcomparisons against HSRL shows that AOD can vary significantly within a 10 km pixel,

1700

AMTD6, 1683–1716, 2013




Title Page



Tables Figures

� �

� �

Back Close

Full Screen / Esc



Discussion

Pa

per|

Discussion

Paper

|D

iscussionP

aper|

Discussion

Paper

|

and a 3 km resolution is an appropriate scale for studying aerosol distributions in urbanand suburban settings.

The dense network of AERONET sun photometers deployed during the DISCOVER-AQ campaign provided an unprecedented opportunity to validate aerosol products ata high spatial scale. Given the new resolution of both the satellite data and the ground5

validation network, the MODIS spatial-temporal collocation techniques for validationwere revisited. It was found that using a spatial average of MODIS pixels around a sunphotometer station, or using only the pixel where the SP station is located resulted in aquantitative, but not qualitative, difference. We conclude with good confidence that theMODIS 3 km validates within the expected error of the MODIS aerosol products in this10

region, but is biased high at AOD values greater than 0.1.There is evidence that a significant source of the bias observed in the 3 km product

results from improper characterization of urban surfaces. The method for attributing thesurface contributions to TOA measured reflectances has been documented to underes-timate the surface reflectances in urban areas (Castanho et al., 2008; Oo et al., 2010),15

which in turn overestimates AOD. Indeed, we found a strong correlation between theamount of urban surface near an AERONET station, and the percent of collocationsthat are above the expected error. While case studies show that the 10 km productalso has one or two pixels that are affected, the smaller pixel size of the 3 km productcauses it to be more susceptible to contamination.20

The poor performance of the 3 km product over urban surfaces is clearly a limita-tion in terms of air quality applications. How to best address this problem in an op-erational environment remains an open question. However, despite this limitation, the3 km product has many qualities making it a suitable choice for air quality studies, in-cluding resolving small scale AOD gradients and point sources, the ability to retrieve in25

patchy cloud fields, and the ability to retrieve closer to coastlines. We expect that theadded information of the 3 km resolution product will complement the existing MODIS10 km resolution aerosol product and products from other passive and active sensorsto provide a more complete view of global aerosol properties from space.

1701

AMTD6, 1683–1716, 2013




Title Page



Tables Figures

� �

� �

Back Close

Full Screen / Esc



Discussion

Pa

per|

Discussion

Paper

|D

iscussionP

aper|

Discussion

Paper

|

Acknowledgements. The authors thank Bill Ridgeway and the MODAPS team for facilitatingiterative testing needs. We also thank the University of Wisconsin PEATE team for the pro-duction of the Terra L1B data used in this study. We recognize the entire AERONET team fortheir efforts to deploy and maintain 44 functioning AERONET stations. HSRL operations werefunded by the NASA DISCOVER-AQ program. The authors also thank the NASA Langley King5

Air King Air flight crew for their outstanding work supporting these flights.

References

Bellouin, N., Boucher, O., Haywood, J., and Reddy, M. S.: Global estimate of aerosol directradiative forcing from satellite measurements, Nature, 438, 1138–1141, 2005.

Castanho, A. D., Martins, J. V., and Artaxo, P.: MODIS Aerosol Optical Depth Retrievals with10

high spatial resolution over an Urban Area using the Critical Reflectance, J. Geophys. Res.,113, D02201, doi:10.1029/2007JD008751, 2008.

Chu, D. A.: Global monitoring of air pollution over land from the Earth Observing System-TerraModerate Resolution Imaging Spectroradiometer (MODIS), J. Geophys. Res., 108, 4661,doi:10.1029/2002JD003179, 2003.15

Drury, E., Jacob, D. J., Wang, J., Spurr, R. J., and Chance, K.: Improved algorithm for MODISsatellite retrievals of aerosol optical depths over western North America, J. Geophys. Res.,113, D16204, doi:10.1029/2007JD009573, 2008.

Eck, T. F., Holben, B. N., Reid, J. S, Dubovik, O., Smirnov, A., O’Neill, N. T., Slutsker, I., andKinne, S.: Wavelength dependence of the optical depth of biomass burning, urban and desert20

dust aerosols, J. Geophys. Res., 104, 31333–31350, 1999.Engel-Cox, J. A., Holloman, C. H., Coutant, B. W., and Hoff, R. M.: Qualitative and quantitative

evaluation of MODIS satellite sensor data for regional and urban scale air quality, Atmos.Environ., 38, 2495–2509, 2004.

Evans, K. and Stephens, G.: A new polarized atmospheric radiative-transfer model, J. Quant.25

Spectrosc. Ra., 46, 413–423, 1991.

1702

AMTD6, 1683–1716, 2013




Title Page



Tables Figures

� �

� �

Back Close

Full Screen / Esc



Discussion

Pa

per|

Discussion

Paper

|D

iscussionP

aper|

Discussion

Paper

|

Friedl, M. A., Sulla-Menashe, D., Tan, B., Schneider, A., Ramankutty, N., Sibley, A., and Huang,X.: MODIS Collection 5 global land cover: Algorithm refinements and characterization of newdatasets, Remote Sens. Environ., 114, 168–182, 2010.

Hair, J. W., Hostetler, C. A., Cook, A. L., Harper, D. B., Ferrare, R. A., Mack, T. L., Welch, W.,Izquierdo, L. R., and Hovis, F. E.: Airborne high spectral resolution LIDAR for profiling aerosol5

optical properties, Appl. Optics, 47, 6734–6752, 2008.Holben, B. N., Eck, T. F., Slutsker, I., Tanre, D., Buis, J. P., Setzer, A., Vermote, E., Reagan, J.

A., Kaufman, Y. J., Nakajima, T., Lavenu, F., Janowiak, I., and Smirnov, A.: AERONET – Afederated instrument network and data archive for aerosol characterization, Remote Sens.Environ., 66, 1–16, 1998.10

Hsu, N. C., Tsay, S. C., King, M. D., and Herman, J. R.: Aerosol Properties Over Bright-Reflecting Source Regions, IEEE T. Geosci. Remote, 42, 557–569, 2004.

Hsu, N. C., Tsay, S. C., King, M. D., and Herman, J. R.: Deep blue retrievals of Asian aerosolproperties during ACE-Asia, IEEE T. Geosci. Remote, 44, 3180–3195, 2006.

Hutchison, K. D.: Applications of MODIS satellite data and products for monitoring air quality in15

the state of Texas, Atmos. Environ., 37, 2403–2412, 2003.Hutchison, K. D., Smith, S., and Faruqui, S.: The use of MODIS data and aerosol products for

air quality prediction, Atmos. Environ., 38, 5057–5070, 2004.Ichoku, C., Chu, D. A., Mattoo, S., Kaufman, Y. J., Remer, L. A., Tanre, D., Slutsker, I., and Hol-

ben, B. N.: A spatio-temporal approach for global validation and analysis of MODIS aerosol20

products, Geophys. Res. Lett., 29, 1.1–1.4, doi:10.1029/2001GL013206, 2002.IPCC: AR4 WG1, Climate Change 2007: The Physical Science Basis, Contribution of Work-

ing Group I to the Fourth Assessment Report of the Intergovernmental Panel on ClimateChange, edited by: Solomon, S., Qin, D., Manning, M., Chen, Z., Marquis, M., Averyt, K. B.,Tignor, M., and Miller, H. L., Cambridge University Press, ISBN 978-0-521-88009-1, 2007.25

Kaufman, Y. J., Tanre, D., Remer, L., Vermote, E., Chu, A., and Holben, B.: Operational remotesensing of tropospheric aerosol over land from EOS moderate resolution imaging spectrora-diometer, J. Geophys. Res., 102, 17051–17067, 1997a.

Kaufman, Y. J., Wald, A., Remer, L., Gao, B., Li, R., and Flynn, L.: The MODIS 2.1 μm channel– Correlation with visible reflectance for use in remote sensing of aerosol, IEEE T. Geosci.30

Remote, 35, 1286–1298, 1997b.

1703

AMTD6, 1683–1716, 2013




Title Page



Tables Figures

� �

� �

Back Close

Full Screen / Esc



Discussion

Pa

per|

Discussion

Paper

|D

iscussionP

aper|

Discussion

Paper

|

Kaufman, Y. J., Tanre, D., and Boucher, O.: A satellite view of aerosols in the climate system,Nature, 419, 215–223, 2002.

Levy, R. C., Remer, L. A., and Dubovik, O.: Global aerosol optical properties and applicationto Moderate Resolution Imaging Spectroradiometer aerosol retrieval over land, J. Geophys.Res.-Atmos., 112, D13210, doi:10.1029/2006JD007815, 2007a.5

Levy, R. C., Remer, L. A., Mattoo, S., Vermote, E. F., and Kaufman, Y. J.: Second-generationoperational algorithm: Retrieval of aerosol properties over land from inversion of ModerateResolution Imaging Spectroradiometer spectral reflectance, J. Geophys. Res.-Atmos., 112,D13211, doi:10.1029/2006JD007811, 2007b.

Levy, R. C., Remer, L. A., Tanre, D., Mattoo, S., and Kaufman, Y. J.: Algorithm for remote10

sensing of dark targets from MODIS: Collections 005 and 051: Revision 2; February 2009,MODIS Algorithm Theoretical Basis Document, 2009.

Levy, R. C., Remer, L. A., Kleidman, R. G., Mattoo, S., Ichoku, C., Kahn, R., and Eck, T. F.:Global evaluation of the Collection 5 MODIS dark-target aerosol products over land, Atmos.Chem. Phys., 10, 10399–10420, doi:10.5194/acp-10-10399-2010, 2010.15

Levy, R. C., Mattoo, S., Munchak, L. A., Remer, L. A., Sayer, A. M., and Hsu, N. C.: The Col-lection 6 MODIS aerosol products over land and ocean, Atmos. Meas. Tech. Discuss., 6,159–259, doi:10.5194/amtd-6-159-2013, 2013.

Li, C. C., Lau, A. K. H., Mao, J. T., and Chu, D. A.: Retrieval, validation, and application of the1-km aerosol optical depth from MODIS measurements over Hong Kong, IEEE T. Geosci.20

Rem., 43, 2650–2658, 2005.Li, R., Remer, L., Kaufman, Y., Mattoo, S., Gao, B., and Vermote, E.: Snow and ice mask for

the MODIS aerosol products, IEEE Geosci. Remote S., 2, 306–310, 2005.Li, R.-R., Kaufman, Y. J., Gao, B.-C., and Davis, C. O.: Remote sensing of suspended sediments

and shallow coastal waters, IEEE T. Geosci. Remote, 41, 559–566, 2003.25

Li, Y., Xue, Y., He, X., and Guang, J.: High resolution aerosol remote sensing retrieval overurban areas by synergetic use of HJ-1 CCD and MODIS data, Atmos. Environ., 46, 173–180, 2012.

Loveland, T. R. and Belward, A. S.: The IGBP-DIS global 1km land cover data set, DISCover:first results, Int. J. Remote Sens., 18, 3289–3295, 1997.30

Lyapustin, A., Wang, Y., Laszlo, I., Kahn, R., Korkin, S., Remer, L., Levy, R., and Reid, J.S.: Multiangle implementation of atmospheric correction (MAIAC): 2. Aerosol algorithm, J.Geophys. Res.-Atmos., 116, D03211 doi:10.1029/2010JD014986, 2011.

1704

AMTD6, 1683–1716, 2013




Title Page



Tables Figures

� �

� �

Back Close

Full Screen / Esc



Discussion

Pa

per|

Discussion

Paper

|D

iscussionP

aper|

Discussion

Paper

|

Martins, J. V., Tanre, D., Remer, L. A., Kaufman, Y. J., Mattoo, S., and Levy, R.: MODIS CloudScreening for Remote Sensing of Aerosol over Oceans using Spatial Variability, Geophys.Res. Lett., 29, 4.1–4.4, doi:10.1029/2001GL013252, 2002.

Oo, M. M., Jerg, M., Hernandez, E., Picon, A., Gross, B. M., Moshary, F., and Ahmed, S. A.:Improved MODIS aerosol retrieval using modified VIS/SWIR surface albedo ratio over urban5

scenes, IEEE T. Geosci. Remote, 48, 983–1000, 2010.Quaas, J., Boucher, O., Bellouin, N., and Kinne, S.: Satellite-based estimate of the direct and

indirect aerosol climate forcing, J. Geophys. Res., 113, D05204, doi:10.1029/2007JD008962,2008.

Remer, L. A., Kaufman, Y. J., Tanre, D., Mattoo, S., Chu, D. A., Martins, J. V., Li, R. R., Ichoku,10

C., Levy, R. C., Kleidman, R. G., Eck, T. F., Vermote, E., and Holben, B. N.: The MODISaerosol algorithm, products, and validation, J. Atmos. Sci., 62, 947–973, 2005.

Remer, L. A., Kleidman, R. G., Levy, R. C., Kaufman, Y. J., Tanre, D., Mattoo, S., Martins, J. V.,Ichoku, C., Koren, I., Yu, H., and Holben, B. N.: Global aerosol climatology from the MODISsatellite sensors, J. Geophys. Res. Atmos., 113, D14S07, doi:10.1029/2007JD009661,15

2008.Remer, L. A., Mattoo, S., Levy, R. C., and Munchak, L.: MODIS 3 km aerosol product: algorithm

and global perspective, Atmos. Meas. Tech. Discuss., 6, 69–112, doi:10.5194/amtd-6-69-2013, 2013.

Rogers, R. R., Hair, J. W., Hostetler, C. A., Ferrare, R. A., Obland, M. D., Cook, A. L., Harper, D.20

B., Burton, S. P., Shinozuka, Y., McNaughton, C. S., Clarke, A. D., Redemann, J., Russell, P.B., Livingston, J. M., and Kleinman, L. I.: NASA LaRC airborne high spectral resolution lidaraerosol measurements during MILAGRO: observations and validation, Atmos. Chem. Phys.,9, 4811–4826, doi:10.5194/acp-9-4811-2009, 2009.

Schmid, B., Michalsky, J., Halthore, R., Beauharnois, M., Harrison, L., Livingston, J., Russell,25

P., Holben, B., Eck, T., and Smirnov, A.: Comparison of aerosol optical depth from four solarradiometers during the fall 1997 ARM intensive observation period, Geophys. Res. Lett., 26,2725–2728, 1999.

Tanre, D., Kaufman, Y. J., Herman, M., and Mattoo, S.: Remote sensing of aerosol proper-ties over oceans using the MODIS/EOS spectral radiances, J. Geophys. Res.-Atmos., 102,30

16971–16988, 1997.

1705

AMTD6, 1683–1716, 2013




Title Page



Tables Figures

� �

� �

Back Close

Full Screen / Esc



Discussion

Pa

per|

Discussion

Paper

|D

iscussionP

aper|

Discussion

Paper

|

van Donkelaar, A., Martin, R. V., and Park, R. J.: Estimating ground-level PM2.5 with aerosoloptical depth determined from satellite remote sensing, J. Geophys. Res., 111, D21201,doi:10.1029/2005JD006996, 2006.

Wang, J. and Christopher, S. A.: Intercomparison between satellite-derived aerosol opticalthickness and PM2.5 mass: Implications for air quality studies, Geophys. Res. Lett., 30, 2095,5

doi:10.1029/2003GL018174, 2003.

1706

AMTD6, 1683–1716, 2013




Title Page



Tables Figures

� �

� �

Back Close

Full Screen / Esc



Discussion

Pa

per|

Discussion

Paper

|D

iscussionP

aper|

Discussion

Paper

|

��

��

��

��

��

��

� �� !�� "�� !��

#$%��μ

�� &� ��

Fig. 1. AOD at 0.55 μm observed by MODIS-Aqua on 21 July 2011 at 18:30 UTC is plotted atthe 10 km resolution (a, c) and 3 km resolution (b, d). MODIS/SP collocations are plotted inthe circles. In (a)–(d), the inner circle is the SP temporally averaged AOD, which is averageof ≥2 SP measurements within 30 min of MODIS overpass. In (a) and (b), the outside circleis the AOD of the MODIS pixel containing the SP site. In (c) and (d), the outer circle is a thespatial average of a MODIS AOD in a 5×5 pixel box around the SP station. Only land pixelswith a QA=3 are used in the collocation. Washington D.C. is shown with the large black starand Baltimore, MD is shown with the small black star. The true color image, created from theMODIS red, green and blue bands, is shown in the background.

1707

AMTD6, 1683–1716, 2013




Title Page



Tables Figures

� �

� �

Back Close

Full Screen / Esc



Discussion

Pa

per|

Discussion

Paper

|D

iscussionP

aper|

Discussion

Paper

|

��

��

��

��

��

��

� �� !�� "�� !��

#$%��μ

�� &� ��

Fig. 2. AOD at 0.55 μm observed by MODIS-Terra on 1 July 2011 at 15:35 UTC is plotted atthe 10 km resolution (a, c) and 3 km resolution (b, d). MODIS/SP collocations are plotted inthe circles. In (a)–(d), the inner circle is the SP temporally averaged AOD, which is averageof ≥2 SP measurements within 30 min of MODIS overpass. In (a) and (b), the outside circleis the AOD of the MODIS pixel containing the SP site. In (c) and (d), the outer circle is a thespatial average of a MODIS AOD in a 5×5 pixel box around the SP station. Only land pixelswith a QA=3 are used in the collocation. Washington D.C. is shown with the large black starand Baltimore, MD is shown with the small black star. The true color image, created from theMODIS red, green and blue bands, is shown in the background.

1708

AMTD6, 1683–1716, 2013




Title Page



Tables Figures

� �

� �

Back Close

Full Screen / Esc



Discussion

Pa

per|

Discussion

Paper

|D

iscussionP

aper|

Discussion

Paper

|

-77.40 -77.00 -76.60 -76.20 -75.80

38.40

38.80

39.20

39.60

AOD

-0.05 0.05 0.15 0.25 0.35 0.45 0.55 0.65 0.75

Fig. 3. 3 km resolution AOD at 0.55 μm observed by MODIS-Terra on 29 July 2011 at 16:00 UTCis plotted in the background. HSRL 532 nm columnar AOD along the flight track between 15:30and 16:30 UTC is shown in the thick line. SP AOD interpolated to 0.55 μm is shown in the circles.The SP measurements are not temporally averaged; the nearest measurement to 16:00 UTC isshown, provided the measurement was taken between 15:30 and 16:30 UTC. Washington D.C.is shown with the large black star and Baltimore, MD is shown with the small black star. The truecolor image, created from the MODIS red, green and blue bands, is shown in the background.

1709

AMTD6, 1683–1716, 2013




Title Page



Tables Figures

� �

� �

Back Close

Full Screen / Esc



Discussion

Pa

per|

Discussion

Paper

|D

iscussionP

aper|

Discussion

Paper

|

��

��

��

��

��

��

��

��

��

Fig. 4. Close up view of the area enclosed in the white box in Fig. 3. (a) MODIS 10 km 0.55 μmAOD with 532 nm HSRL AOD plotted on top. (b) MODIS 3 km 0.55 μm AOD with 532 nm HSRLAOD plotted on top.

1710

AMTD6, 1683–1716, 2013




Title Page



Tables Figures

� �

� �

Back Close

Full Screen / Esc



Discussion

Pa

per|

Discussion

Paper

|D

iscussionP

aper|

Discussion

Paper

|

0.00 0.05 0.10 0.15 0.20HSRL AOD Range

0.00

0.05

0.10

0.15

0.20

0.25

Nor

mal

ized

Fre

quen

cy

10 km

Fig. 5. The maximum difference between HSRL 0.55 μm AOD measurements observed withina MODIS 10 km pixel, normalized by the total frequency. Only MODIS pixels that contain 6 ormore HSRL measurements are included.

1711

AMTD6, 1683–1716, 2013




Title Page



Tables Figures

� �

� �

Back Close

Full Screen / Esc



Discussion

Pa

per|

Discussion

Paper

|D

iscussionP

aper|

Discussion

Paper

|

��(�)��*�+�)

�� & �� μ,��

��

��

��&

��

��

��

� ��

��

��μ

,��

��/�:��(��;<��);/��+��

�

�

�

&

�

�=�>

?�(�

@

��(�)��*�+�)

�� & �� μ,��

��

��

��

��

��

� ��

��

��μ

,��

��B,�*=;�?�:��B,�*=;�?�:

��*�:��)��<�=��

�� & �� μ,��

��

��

��&

��

��

��

� ��

��

��μ

,��

��/�:��(��;<��);/��+��

�

�

�

&

�

�=�>

?�(�

@

��*�:��)��<�=��

�� & �� μ,��

��

��

��

��

��

� ��

��

��μ

,��

��B,�*=;�?�:��B,�*=;�?�:

Fig. 6. (a) Density scatterplot of MODIS/SP collocations for AOD at 0.55 μm, using the sin-gle pixel collocation technique. Expected error over land (±0.05+0.15 AOD) is shown in thedashed lines, the best fit line is shown by the solid blue line, and the one-to-one line is shownin the solid black line. (b) Average difference between MODIS single pixel collocations and SP0.55 μm AOD, as a function of SP 0.55 μm AOD. ±1 standard deviation is shown for the 3 kmproduct. (c) As in (a), but the MODIS 5×5 pixel spatial average collocation is used. (d) As in(b), but the MODIS 5x5 spatial average collocation used.

1712

AMTD6, 1683–1716, 2013




Title Page



Tables Figures

� �

� �

Back Close

Full Screen / Esc



Discussion

Pa

per|

Discussion

Paper

|D

iscussionP

aper|

Discussion

Paper

|

(a) 10 km resolution

0.0 0.2 0.4 0.6 0.8 1.0HSRL 532 nm AOD

0.0

0.2

0.4

0.6

0.8

1.0

MO

DIS

0.5

5 μ

m A

OD

% within EE = 68.61% above EE = 29.64% below EE = 1.75N = 1717; R = 0.89Y 0.97x+ 0.04

(b) 3 km resolution

0.0 0.2 0.4 0.6 0.8 1.0HSRL 532 nm AOD

0.0

0.2

0.4

0.6

0.8

1.0

MO

DIS

0.5

5 μ

m A

OD

% within EE = 61.14% above EE = 35.07% below EE = 3.79N = 1557; R = 0.83Y 0.93x+ 0.05

1

2

3

4

5

6

7

8

9

10

Frequency

Fig. 7. Density scatterplot of MODIS/HSRL collocations for AOD at 0.55 μm for MODIS and532 nm for HSRL for the 10 km MODIS product (a) and the 3 km MODIS product (b). TheMODIS AOD is sampled along the HSRL flight path; no spatial or temporal averaging is per-formed. Expected error over land (±0.05+0.15 AOD) is shown in the dashed lines, the best fitline is shown by the solid blue line, and the one-to-one line is shown in the solid black line.

1713

AMTD6, 1683–1716, 2013




Title Page



Tables Figures

� �

� �

Back Close

Full Screen / Esc



Discussion

Pa

per|

Discussion

Paper

|D

iscussionP

aper|

Discussion

Paper

|

Fig. 8. (a) Percent of 3 km spatially averaged MODIS/SP collocations of 0.55 μm AOD withinexpected error (±0.05+0.15 AOD) at each SP station. (b) Correlation coefficient between 3 kmspatially averaged MODIS 0.55 μm AOD and SP temporally averaged 0.55 μm measurementsat each station. Land identified as urban/built up by the MODIS land cover product (MCD12Q1)is plotted in grey in both panels.

1714

AMTD6, 1683–1716, 2013




Title Page



Tables Figures

� �

� �

Back Close

Full Screen / Esc



Discussion

Pa

per|

Discussion

Paper

|D

iscussionP

aper|

Discussion

Paper

|

0 20 40 60 80 100Urban landcover percentage

0

20

40

60

80

100

Perc

ent o

f co

loca

tions

abo

ve E

E Y = 0.72x + 4.41R = 0.84

Fig. 9. Percent of 3 km spatially averaged MODIS/SP 0.55 μm AOD collocations above ex-pected error (+0.05+0.15 AOD) at each SP station, plotted against the percentage of pixelswithin the 15 km by 15 km collocation box identified as urban by the MODIS land cover product.

1715

AMTD6, 1683–1716, 2013




Title Page



Tables Figures

� �

� �

Back Close

Full Screen / Esc



Discussion

Pa

per|

Discussion

Paper

|D

iscussionP

aper|

Discussion

Paper

|

Fig. 10. (a) 3 km AOD at 2.1 μm observed by MODIS-Aqua on 21 July 2011 at 18:30 UTC.(b) 3 km AOD at 2.1 μm observed by MODIS-Terra on 1 July 2011 at 15:35 UTC. Only land pix-els with a QA=3 are shown. Washington D.C. is shown with the large black star and Baltimoreis shown with the small black star. The true color image, created from the MODIS red, greenand blue bands, is shown in the background.

1716

applications over land in an urban/suburban region€¦ · remer et al. (2013) describes how pixels...

Documents