Introduction
The AErosol RObotic NETwork project (AERONET; Holben et al., 1998) has
provided significant contributions to remote sensing of aerosols during the
course its 24-year history. Observations have largely been utilized to
validate satellite retrievals of aerosol optical depth (AOD) (e.g., Ichoku et
al., 2002; Kahn et al., 2005; Remer et al., 2002; Sayer et al., 2012),
characterize aerosol absorption and size distributions (e.g., Dubovik et al.,
2002), and evaluate model products (e.g., Kinne et al., 2003; Sessions et
al., 2015) and more recently forecasts through assimilation (e.g., Randles et
al., 2017; Rubin et al., 2017) of aerosol properties. These investigations
have largely been dominated by the highly accurate observations of extensive
properties such as spectral AOD, and as more data became available, the
intensive products retrieved from inversions of the radiative transfer
equation such as a complex index of refraction and particle size distribution
have come to the fore. The accuracy of the ground-based AERONET
quality-assured (Level 2) point observations of aerosol optical depth is very
high and therefore is considered a “ground truth” for most satellite and
model comparison purposes. AOD is a direct measure of a column-integrated
spectral property and can be derived from essentially an instantaneous
measurement. Thus, the only uncertainty arises from calibration and
contamination from outside influences such as optical and digital
contamination in the instrument in some rare cases and cirrus clouds (e.g.,
Chew et al., 2011). Given the accuracy of the calibration (Eck et al., 1999)
and processing algorithms (Smirnov et al., 2000, and manual quality assurance
assessment), the accuracy of Level 2 AOD is estimated to be ∼ 0.01 in
the visible and NIR spectrum for fully calibrated field instruments when pre-
and post-calibrations have been applied. Further, analytic solutions to the
relative contributions of the fine and coarse modes to the AOD are provided
by AERONET through the spectral deconvolution method algorithm (O'Neill et
al., 2003) and verified by Kaku et al. (2014).
The accuracy of the intensive AERONET aerosol properties (single scattering
albedo, particle size distribution, and complex index of refraction) is less
clear due to larger uncertainties in the inversion retrievals and difficulty
in obtaining adequate verification data from other methodologies. These
properties are extinction-weighted atmospheric-column integrated properties
that exhibit different uncertainties than the wide variety of techniques
associated with in situ measurements and estimates. The retrieval
uncertainties of the column-integrated aerosol properties inverted by the
Dubovik and King (2000) algorithm are well discussed in Dubovik et
al. (2000); however, the additional uncertainty ins the measurement techniques
is very difficult to assess due to atmospheric variability during the time
of observations. The uncertainties associated with in situ techniques are
well discussed by Reid et al. (2003, 2008b, 2005, 2006) for the size
distributions of dust, smoke, and sea salt aerosols. Andrews et
al. (2017) found that, provided the AERONET guidelines of only using
absorption or an index of refraction data when 440 nm AOD > 0.4 are adhered to, inversion
products were within the stated uncertainty bounds. The accuracy of the inverted
parameters is predicated upon the atmosphere being stable and spatially
uniform within the measurement space of the sky radiance measuring
radiometer. For example, if we assume that the aerosol is in the lowest 2 km
of the atmosphere and the solar zenith angle is 60∘, the AERONET
observation path would be 4 km long and cover a horizontal distance of
approximately 3.5 km. Thus, for this particular solar zenith angle and layer
height geometry example, AERONET retrievals are assuming relative uniformity
in an atmospheric cylinder of 7 km diameter, 2 km vertically, and a
measurement slant path of 4 km about the surface center point. Quality
assurance algorithms and spatial averaging of measured sky radiance
distributions have been utilized to minimize this uncertainty associated with
the spatial variance of aerosol (Holben et al., 2006).
AERONET and other ground-based remote-sensing systems have the distinct
advantage of the time domain with direct sun measurement frequencies of
seconds to minutes throughout the day and in some instances at night.
Nominally, the AOD sampling frequency for AERONET network measurements is
15 min, and more recently 3 min, intervals for sites with sufficient
communication infrastructure. The measurements of sky radiance used to
retrieve the inversion products are nominally taken hourly for AERONET but in
some instances are taken more frequently such as early in the morning and
late in the afternoon when optical air mass changes rapidly. Other networks
such as the SKYNET network
(http://atmos2.cr.chiba-u.jp/skynet/data.html; Hashimoto et al., 2012)
make almucantar sky scan measurements at 10 min intervals. These high-frequency ground-based remote-sensing measurements allow the opportunity to
assess aerosol properties diurnally and provide a higher probability
of making valid aerosol observations under variable atmospheric conditions,
such as in partial cloud cover and/or spatially or temporally varying
aerosol. The temporal domain may be a powerful ally for assessing transport
processes and in some instances a proxy for the spatial domain.
Individual ground-based systems inherently do not represent the spatial
variation in aerosol properties. Thus, they complement the satellite
retrievals and regional and global model predictions. Typically, a spatial-scale bridge to the ground-based measurements (including in situ) to
satellite and model assessments has been through aircraft observations.
Aircraft flights occur over ground-based point observations from profiles
and various altitude transects extending tens, hundreds and thousands of
kilometers and can provide spatial continuity during intensive field
operations that enables scaling point location observations to the satellite
observations and regional model simulations.
Field campaigns are of limited duration and aircraft flights are often
discontinuous during the measurement campaign. The following question arises:
is there a need for continuous high spatial- and temporal-resolution aerosol
data that neither a single point nor airborne, satellite, or model results
address? Furthermore, is there an approach that will clarify the uncertainty in comparisons of in
situ and remote-sensing aerosol properties? In hindsight and with some
foresight, the answers have proven to be yes and yes.
The series of Distributed Regional Aerosol Gridded Observation Network
(DRAGON) campaigns began in 2011 primarily as a means to encourage
collaboration between remote-sensing and in situ communities to compare
measurements and retrievals of the intensive properties of aerosol particles,
such as single scattering albedo (SSA), particle size distribution, and
complex index of refraction. Note that earlier DRAGON-like campaigns (e.g.,
UAE Unified Aerosol Experiment, UAE2 – Reid et al., 2008a; and TIGERZ
– Giles et al., 2011) were performed to assess spatial and temporal
intensive and extensive aerosol optical properties for comparison to
satellite retrievals and thus provided further motivation for satellite and
model intercomparisons with high-resolution ground-based measurement systems.
We define a DRAGON network as a relatively high spatial density of ground-based sun photometers and other associated measurements of limited duration. Typically, these instruments are in a loose mesoscale grid with
a two-dimensional spacing of tens to hundreds of kilometers for a period of
30 days or more with high-frequency sampling in minutes (typically at 3 min
intervals for AOD) during daylight hours. Contrast a DRAGON network with the
overall AERONET global spatial distribution of hundreds to thousands of
kilometers that has developed out of individual PI (principal investigator)
and institutional contributor needs since 1993. An assessment of the
published AERONET measurements from 1993 through 2011 showed very few in situ
versus remote-sensing comparisons, many of which were of limited
applicability (see Table 1; the measurements also available from the AERONET
website under the various DRAGON campaigns:
https://aeronet.gsfc.nasa.gov/new_web/dragon.html). Indeed, clearer
descriptions of aerosol types beyond these five generic multimodal categories
have been expressed in the more recent literature. Many investigations have
provided clarity on the definition of fine- and coarse-mode aerosols in terms
of the particle size and chemical composition of various aerosol types
particularly from the in situ point of view. From a remote-sensing
perspective aerosol typing remains difficult but progress is moving forward
primarily by the assessment of the fine–coarse partition, single scattering
albedo, Ångström exponent (AE), and absorption AE (AAE) (O'Neill et
al., 2008; Giles et al., 2012; Russell et al., 2014, among others). Table 2
is updated based in large part on contributions to this special issue and
several other important studies using DRAGON data sets.
Principle intensive parameters retrieved by sun and sky scanning
spectral radiometers for five aerosol types. Sixteen published
validations/comparisons of these retrievals against in situ measurements were
made during field campaigns prior to 2010. Ra: Ramanathan et al. (2001); Re:
Remer et al. (1997); H: Haywood et al. (2003); L: Leahy et al. (2007); B:
Bergstrom et al. (2003); Chand et al. (2006); E: Eck et al. (2010); M:
Müller et al. (2010); Mü: Müller et al. (2012); Rp: Reid et
al. (2003); Ru: Reid et al. (2008b); S: Smirnov et al. (2003); Sc: Schafer et
al. (2008); T: Toledano et al. (2011); O: Osborne et al. (2008); and J:
Johnson et al. (2009). Note that most categories are incomplete (–), are not
updated for the current inversion algorithm, and/or not relevant to
total-column ambient retrievals.
Parameter/type
Urban
Biomass burning
Dust
Sea salt maritime
Mixed
SSA (ωo)
Raa,c
Hd,c, L, B, Ca, Sce
T, M, Mü
–
Od, Jd, Ea
Size distribution dV/dlnr, rv
Reb
Hd,c
Rpc, Ru, Mü
Sa,c
Jd
Real index (n)
–
–
–
–
–
Imaginary (k)
–
–
–
–
–
Asymmetry (g)
–
–
–
–
Jd
% sphericity
–
–
–
–
–
a Regional comparisons. b Nakajima
retrievals. c Version 1. d Single point.
e Surface comparison.
The aerosol types detectable from remote-sensing (RS) techniques and
compared with in situ field measurements. We show only those direct
RS–in-situ comparisons. Unlike Table 1, here the aerosol type describes the
properties of the aerosols rather than sources. We acknowledge that aerosol
typing is difficult and still subjective and incomplete. (C: Corrigan et al.,
2008; E: Esteve et al., 2012; Sc: Schafer et al., 2014, 2017). Some studies
appearing below are given in the caption of Table 1.
Parameter/type
Fine
Coarse
Mixed
Inorganic
Organic
Mineral
Organic
NaCl
Hygroscopic
B, C
Br, C
SSA (ωo)
Sc, E, A
C
–
T, M
–
–
Od, Jd
Size distribution dV/dlnr, rv
Sc
–
–
Mü, Rpc, Ru
–
–
Jd
Real index (n)
–
Hd,c
–
–
–
–
–
Imaginary (k)
–
–
–
–
–
–
–
Asymmetry (g)
–
–
–
–
–
–
Jd
The description of the aerosol size distribution is of primary importance as
a first-order physical and optical parameter corresponding to particle size
and the associated concentration of variously sized particles. Coarse-mode
aerosol is sometimes considered to have a particle radius of greater than
1 micron (µm), and the fine mode ranges from 0.05 to 1 µm
(in volume distributions), although definitions vary widely. This type of
classification may be generally applied for remote sensing from sun and sky
scanning radiometers that use inversion schemes to retrieve aerosol
properties (Dubovik and King, 2000; Nakajima et al., 1996, among others).
Different definitions of fine/coarse-mode breakdown of the AOD are applied to
the spectral deconvolution algorithm (O'Neill et al., 2003), while the
Ångström exponent computed from spectral optical depth is a general
scaling of fine/coarse optical influence, although it varies considerably as
a function of wavelength for fine-mode-dominated aerosols (Eck et al., 1999).
Note that the AERONET retrieval scheme of Dubovik and King (2000) reports the
size in terms of particle radius, with the retrieved radius limits of 0.05 to
15 µm. The inflection point defining the upper limit of the
fine-mode-sized particles of a retrieval lies between a 0.44 and
0.99 µm radius in volume distributions that are composed of
discrete particle sizes from a mixture of spheres and spheroids with a fixed
shape distribution (Dubovik et al., 2006).
Generally, natural sources for coarse-mode hygroscopic sea salt aerosol are
breaking waves and associated bursting water bubbles. These particles are
nominally spherical at most ambient relative humidity over the ocean, with
AOD typically dominated by particles larger than 0.5 µm radius.
Dust particles are highly nonspherical airborne mineral soil and typically
have radii on average greater than 1 µm, with numerous electron
micrographs showing particles with lengths exceeding 10 µm, yet
sometimes with a dimension of submicron size. These dust sources from arid
and semiarid regions often originate in dried lake beds and intermittent
waterways (Prospero and Carlson, 1972, among others). Other sources of coarse
particles reported in the literature include diatomaceous earth from the
Bodélé Depression in Chad (Washington et al., 2006; Ben-Ami et al.,
2010), intensive construction in megacities, resulting in localized, highly
variable, and largely unknown particle properties, dust from agricultural
fields, pollen grains which are very large organic particles that quickly
settle from the atmosphere, fly ash from unfiltered coal combustion (WHO,
1999), and ash from episodic volcanic eruptions. Thus, Table 2 has three
categories for coarse-mode aerosol: sea salt, mineral dust (such as particles
that contain hematite, causing absorption in the blue and UV spectrum,
diatomaceous earth, and anthropogenic coarse particles), and pollen
(organic). The chemistry of “dust” particles is highly variable and is
beyond the scope of this discussion; however, it is noteworthy that as
chemical analysis of coarse particles is more geographically studied and
better understood, there will be greater opportunity to assess the response
of remote sensing to the properties of these particles.
The fine-mode (or accumulation-mode) aerosols are sometimes loosely referred
to in the literature as either urban/industrial or biomass burning. These
terms were convenient in the early days of remote sensing but are only a
rough guide to our greater understanding of their diversity and properties.
The range of fine-mode aerosol types that contribute to remote sensing can be
rather daunting and often does not exist in a single type distribution in the
atmosphere. Artaxo et al. (1994) in early work and continuing with Fuzzi et
al. (2007), among many others, have made extensive
investigations of the smoke aerosol generated during the burning season in
the Amazon basin that includes both black carbon particles from flaming-phase
burning and primarily brown carbon particles that are organic and from both
flaming and smoldering combustion (Falkovich et al., 2005). Particle sizes
are generally less than 1 µm in radius in volume distributions,
although a distinct coarse mode of ash aggregates and suspended soils is also
present (Reid et al., 2005). Both have been shown to have very different
absorptive properties from each other and from other types of particles;
thus, we have added black carbon and brown carbon to Table 2, which typically
range in volume median radius from 0.14 to 0.2 µm. Gas to particle
conversion from nitrates, organic compounds, and SO2 can form fine-mode
aerosols. This process is enhanced in the presence of clouds and fog.
Sometimes, hydroxymethanesulfonate (HMS) may form within cloud/fog droplets
when sulfur dioxide is present, whereupon after evaporation it can form large
particles. These have been shown to have a variable modal range but typically
have a mean volume modal radius of ∼ 0.45 µm; see Eck et
al. (2012) and Li et al. (2014). The properties of these aerosol types
require further evaluation. This complexity gives rise to three fine-mode
aerosol types in Table 2: black carbon, brown carbon, and “other”; they can
be distinguished in principle by ground-based sun and sky scanning
radiometers by combinations of size, shape, and/or absorption magnitude.
Table 2 shows those studies that have objectively assessed all of the known
direct comparisons of aerosol properties of AERONET to in situ
measurements.
The DRAGON campaigns
The DRAGON field campaigns were developed in consideration of the spatial and
temporal advantages and disadvantages of remote-sensing systems and in situ
systems for ground-based, aircraft-based, and remote-sensing systems. In the
previous section we described generally the assets available for a typical
AERONET deployment. Table 3 presents an overview of the DRAGON campaigns,
including the dominate aerosol type, the time frame, the approximate range of
aerosol characteristics from a remote-sensing perspective, and the principle
contact for each campaign. We have attempted to provide an exhaustive list up
to the time of this writing, and this table will be maintained and updated on
the AERONET website as new information is received.
The distribution of DRAGON campaigns conducted from 2004 to 2016 are
framed in yellow with red labels. Yellow labels indicate larger campaigns
with dashed frames that include DRAGON networks.
![]()
The method of the DRAGON campaigns was to establish a high density of
ground-based sun and sky scanning spectral radiometers within a local or
mesoscale region to capture small-scale aerosol variations. For this
discussion we present those distributions over tens to hundreds of kilometers
and a time period of weeks to months. Very early studies dating back to the
1950s by Flowers et al. (1969) showed regional to continental-scale
variations across the US, and in the 1980s sun photometry documented regional
Sahelian aerosol loading during the drought (d'Almeida, 1986; Holben et al.,
1991). The 1990s brought AERONET regional measurements to the Amazon Basin
(Holben et al., 1996), the boreal forests in Canada called BOREAS (Markham et al., 1997), and southern Africa,
with two campaigns called ZIBBIE (Eck et al., 2001) and SAFARI2000 (Swap et
al., 2003; Eck et al., 2003). These and other regional
investigations brought tremendous knowledge of aerosol properties over
regions dominated by a single aerosol type; however, they could not address
the variability in small-scale regional aerosol processes. They also came
largely before the massive data collection ushered in by the EOS satellite
era that began with Terra in 2000 and continues today from an expanding
series of spaceborne quantitative Earth monitoring platforms. Figure 1 shows
the location of DRAGON field experiments relevant to this paper.
United Arab Emirates – Unified Aerosol Experiment
(UAE<inline-formula><mml:math id="M53" display="inline"><mml:msup><mml:mi/><mml:mn mathvariant="normal">2</mml:mn></mml:msup></mml:math></inline-formula>)
The UAE2 was established across the northern UAE with 18 AERONET sites
distributed over approximately 150 000 km2 including islands in the
Arabian Gulf (Reid et al., 2008a). The campaign was conducted in August and
September 2004 with the objective to assess the radiative properties of dust
aerosols in a humid coastal environment from ground, airborne, and satellite
perspectives. Sites were selected to provide characterization of Arabian
Gulf, coastal, and interior desert sites from satellite product validation –
especially in locations of consistent changes in the lower boundary condition
(e.g., soil albedo, Case II waters). UAE2 was conducted in concert with
an ongoing weather modification assessment (NCMS/NCAR) in the region.
Although southwest Asia and the Middle East are often thought of as
coarse-mode dust-dominated aerosol environments, fine-mode aerosol particles
from the petroleum industry and urban pollution contribute equally to overall
AOD (Eck et al., 2008). From a product verification point of view, the
UAE2 deployment provided the first conclusive piece of evidence that
dust size retrievals are consistent with in situ measurements (Reid et al.,
2008b) and that dust retrievals including vertical homogeneity can be further
constrained by the inclusion of UV and near-infrared data (O'Neill et al.,
2008).
CALIPSO And Twilight Zone (CATZ)
The CATZ campaign was the first AERONET Intensive Operation Period (IOP) to
support CALIPSO aerosol retrievals. This was temporally synchronized with
CALIPSO over-flights to assess the aerosol variability within the along-track
averaged CALIPSO retrieval. Up to 12 AERONET sites were placed along 230 km
of the daytime Aqua track within the CALIPSO footprint on the Delmarva
Peninsula on seven different dates from late June to mid-August 2007. Very
low to high aerosol loadings occurred, which were all fine mode dominated.
Transects: Indo-Gangetic aERosol Zone (TIGERZ)
The TIGERZ campaign was an effort during the pre-monsoon of May 2008 to
characterize the complex and high loading aerosol environment in the
Indo-Gangetic Plain (IGP) of northern India in support of CALIPSO
satellite-borne lidar validation. The deployment of additional instruments
was centered around the long-term monitoring site on the IIT campus in the
industrial city of Kanpur. The pre-monsoon aerosol environment is
characterized by regional fine-mode haze from fossil fuel emissions, mostly
from coal with episodic dust events both locally generated and regionally
transported from the northwest. The local Kanpur City aerosol plume was
enhanced by a megawatt power plant plume and numerous coal-fired brick kilns
dotting the region. Despite local strong sources, the Kanpur aerosol
properties were similar to a village site 400 km downwind (Giles et al.,
2011). Sites were established specifically to be in and very near the CALIPSO
footprint, and during May, captured the spatial variability and provided
validation of CALIPSO retrievals. Sites were local to the descending CALIPSO track but had a
radius of up to a 300 km around Kanpur.
Seven South East Asian Studies (7-SEAS)
The 7-SEAS interdisciplinary research program has a rich history of
ground-based measurements in Southeast Asia beginning in 2007, including
region-wide deployments of AERONET sites throughout the Maritime Continent
(Indonesia, Malaysia, Philippines, Singapore) and peninsular Southeast Asia
(Laos, Thailand, and Vietnam). Overall AERONET properties can be found in
Reid et al. (2013). Specific to the DRAGON concept, the AERONET program
collaborated with local scientists to develop two DRAGON programs during the
August–September 2012 burning season. These programs were based at the
National University Singapore (NUS) for Singapore and Universiti Sains
Malaysia for Penang, Malaysia.
Penang
Penang Island is mountainous with an eastern coastal plain and lies 2 to
15 km offshore of mainland peninsular NW Malaysia, within the Strait of
Malacca. Its densely populated capital of Georgetown (2 million) is across
the Penang Strait from industrial Butterworth, while the Malacca Strait side
of the island is rural. Anchored ships, industry, and automobile traffic
contribute to fossil fuel emissions, while episodic pulses of biomass burning
aerosols from Riau, Sumatra, Indonesia, added to a background of sea salt
aerosol within the sampled 30 km transect. During September 2012, Universiti
Sains Malaysia staff maintained eight AERONET stations. In addition to
satellite and model validation, research was conducted specific to coastal
areas with these data sets utilized for air quality investigations (see Tan
et al., 2015a).
Singapore
Singapore is a highly industrialized urbanized center on an island at the
southern tip of the Malay Peninsula, with dimensions of approximately 30 km east–west by 20 km north–south . The regional population including Johor
Bahru is well over 5 million. Thus, fossil fuel emissions from cars,
petrochemical industries, and ships constitute a major portion of the aerosol
sources; however, maritime aerosol from the South China Sea and the Straits of
Malacca provide a rather constant but weak background regime. Biomass burning
primarily from Sumatra and Kalimantan imposes an episodic and at times massive
aerosol burden on the region. The September 2012 DRAGON campaign, in
collaboration with NUS' Centre for Remote Imaging, Sensing and Processing
(CRISP), afforded the opportunity to assess the variability in the aerosol
loading in response to local and regional sources from six well-distributed
AERONET sites and a suite of detailed ground-based measurements across the
region.
Deriving Information on Surface Conditions from Column and
VERtically Resolved Observations Relevant to Air Quality (DISCOVER-AQ)
DISCOVER-AQ was a NASA-sponsored Earth Venture Suborbital 4-year campaign
(2011 to 2014) to relate remote-sensing measurements to air quality
assessments at four selected sites across the United States (central
Maryland; Houston, TX; San Joaquin Valley, CA; Denver Front Range Region,
Colorado; https://discover-aq.larc.nasa.gov). For each campaign, this
involved repeated in situ and remote-sensing ground and airborne (NASA's P-3B
and King Air) measurements during most days for the duration of the campaign.
This involved a series of high and low airborne transects, targeted airborne
profiles, high-altitude down-looking lidar profiling, and passive
remote-sensing measurements, combined with in situ ground, ground-based
lidar, ozonesonde releases, and AERONET measurements configured in a
mesoscale grid. As conditions allowed, flights would last for approximately
8 h day-1 on most days through the ∼ 30-day campaign. This
resulted in very detailed 4-D characterizations of meteorology, aerosol, and
trace gas measurements and processes that affect air quality, air quality
forecasts, and their relationship to remote sensing. The AERONET DRAGON
networks established for these campaigns represent the most detailed AERONET
spatial characterizations to date.
Maryland (Greater Baltimore) – July 2011
This campaign selected a highly urbanized and industrial region of the
mid-Atlantic that is subjected to high summertime humidity and periodic
pollution buildup. The studied region was approximately 125 km long,
following the I-95 corridor from the Washington Beltway north to the
Maryland–Delaware state line, and about 40 km wide, encompassing Baltimore,
agricultural fields, suburbs, and the Chesapeake Bay. Forty-three AERONET
sites were established 1 month prior to the campaign and continued
monitoring for approximately 1 month after. The meteorology was classic
mid-Atlantic for July, with daytime temperatures approaching 39 ∘C on
the hottest days and high humidity with daytime dew points sometimes reaching
25 ∘C, combined with nearly stagnant conditions with southerly flow resulting
in AODs exceeding 1.0 at 500 nm on some days and showing considerable
diurnal and day-to-day dynamics. Two cold frontal passages advected the
pollution away from the region (AOD as low as 0.1 at 500 nm), with
subsequent gradual buildup over a period of days. The Ångström exponent
(440–870 nm) during this period was typically greater than 1.5, indicating
fine-mode-dominated aerosols as one would expect in this region and season.
Munchak et al. (2013) utilized DRAGON Maryland AERONET data to assess the
impact of urban surface reflectance variations on the biases in satellite-retrieved AOD from the MODIS Dark Target algorithm. They also determined that the
new 3 km resolution MODIS retrievals could detect AOD gradients better and
make retrievals closer to clouds than the standard 10 km MODIS product.
San Joaquin Valley, California (Bakersfield to Fresno) – mid-January to mid-February 2013
The San Joaquin Valley occupies the southern half of California's Central
Valley which is bounded by the convergence of the high Sierra Nevada range to
the east and a series of coastal mountain chains to the west. The valley is
flat, with intensive irrigated agriculture. The region is notable for the air
quality challenges to its 3 million inhabitants: freeway corridors and
intensive agriculture, including ammonia emissions and fugitive dust that
contributes to particularly strong air pollution in January and February. The
planetary boundary layer (PBL) is typically shallow at ∼ 1 km or less
and adiabatically stable owing to strong radiational cooling at night
resulting in frequent and persistent fog events. This, combined with various
agricultural, fossil fuel, petrochemical, and largely undocumented biomass
burning emissions throughout the valley, creates a complex environment for
aerosol and reactive gas processes that were observed from 20 January to
15 February 2013 by DISCOVER-AQ. A DRAGON deployment of 17 AERONET stations
was established from Fresno in the north to Bakersfield 175 km to the south and to the east
at Porterville near the foothills of the Sierra Nevada to Huron 75 km to the
west. At
the time of the campaign, Porterville was heavily affected by pollutant
buildup from airflow blockage by the mountains to the east. Optical depths at
500 nm at Porterville showed an extreme episodic and diurnal range of AOD
owing to local emissions, hygroscopic growth from high relative humidity in
fog, and the variable PBL height. Measured AOD values at 500 nm ranged from
1.2 during stagnation conditions and post-fog events to 0.1 after the valley
was ventilated from passage of a cold front.
Houston, Texas (Greater Houston/Galveston) – August 2013
Houston is a massively sprawling city with a downtown center approximately
30 km north of Galveston and the Gulf of Mexico. A dense petrochemical
industry borders the ship channel that bisects southern Houston, with
numerous sources of gases and aerosols complemented by automobile emissions
and other industry. Climatology showed that air quality is poorest during
August; thus, like the Maryland campaign, it afforded the best opportunity to
understand the processes relating emissions and air quality issues to remote
sensing. The aircraft tracks were largely square racetrack circuits with six
intensive vertical profiles over ground-based supersites. Seventeen DRAGON
AERONET sites were used to characterize the column aerosol properties for
3 months (July–September). A wide range of aerosol
conditions of mostly fine-mode aerosols with AOD ranging from ∼ 0.1 to
0.7 at 500 nm were measured. On 23–25 August a
Saharan dust intrusion moved into the region, lowering the Ångström
exponent to 0.8. The region during August was characterized by high humidity
and significant afternoon cloud development.
Colorado – July 2014
The northeastern plains of the front range of the Rockies formed the backdrop
for the last DISCOVER-AQ campaign conducted in July 2014. The airborne and
ground-based measurement campaign track included diverse landscapes and aerosol sources from central Denver
to suburban Fort Collins 130 km N and continued 50 km east to rural Greeley feedlots. The track turns south 30 km to Platteville, which is dominated by irrigated crops and intense fossil fuel exploration
and extraction. The track is closed by returning to Denver 40 km to the southeast. High temperatures and intense solar radiation characterized
July 2014. Aerosol optical depths averaged 0.2 at 500 nm and day-to-day
variations were typically small; however, several days of fine-mode aerosol
events elevated the AOD to ∼ 0.4.
DRAGON-NE Asia – Korea, Japan
Northeast Asia faces arguably the most severe air quality issues on the
planet owing to the very high population density coupled with high levels of
industrialization and, additionally, its position downwind of major dust
source regions. These contribute to significant trans-boundary aerosol
transport compounded by emissions from several megacities in the region.
Given the AERONET limitations for retrievals with low uncertainty
(AOD > 0.4 at 440 nm) for complex refractive index retrieval products,
NE Asia routinely experiences aerosol loading that exceeds those limitations
on most days; thus, investigations of the spatial and temporal variations in
single scattering albedo in addition to AOD are possible. The following two
campaigns called DRAGON-KOREA and DRAGON-JAPAN operated from March to June 2012. The NE Asia DRAGON campaigns did not have a
significant airborne component; thus, the emphasis was on assessing the
spatial and temporal variations in aerosol optical properties. Numerous
opportunities occurred for satellite and model validation under a variety of
aerosol gradients.
DRAGON-KOREA
Seoul was the focus for half of the 22 AERONET surface stations deployed from
March to June 2012, including five permanent sites in South Korea with
long-term records. Seoul is a megacity of 25 million (metropolitan region)
spread across a landscape of the Han River plains, hills, and low-elevation
forested mountains. Industry and fossil fuel power generation contribute
emissions to a significant pollution aerosol loading in addition to aerosol
advected from China. South Korea in general is a landscape that is
challenging for satellite retrievals of AOD due to significant variation in
background surface reflectance and a varied topography (∼ 70 %
mountainous, mostly forested) and variability in aerosol properties (fine and
coarse). A decision was made to expand the network in spring 2012 to a
regional or mesoscale network to further assess the impact of transported
aerosols from across the Yellow Sea and from Seoul; this was done by
including sites on the west coast, in the interior and the east and
south. AOD at 500 nm
from regional sites had daily values ranging from ∼ 0.2 to 1.5, while
sites in Seoul varied from ∼ 0.5 to 2.1 during episodic aerosol events.
DRAGON-JAPAN
Osaka, Japan, was the focus of a DRAGON campaign with eight AERONET sites;
this was coincident in time with the DRAGON-KOREA campaign from March through
June 2012. Osaka is a megacity of very dense urban development that is
bounded by low mountains on three sides and Osaka Bay to the south (see Sano et al., 2016). Industry and transportation emissions are
sources of the dominant background aerosol loading, and, as in Seoul, episodic
coarse-mode dust and transported fine-mode industrial aerosols were observed
during the 4-month intensive measurement period. Owing to two nearby
mountain sites, boundary layer assessments were possible and were also facilitated by
a mobile handheld sun photometer.
A second DRAGON network of six AERONET and one SKYNET sites on the small
(326 km2) rural western island of Fukue captured the dynamics of
transported fine-mode aerosol properties, while an airborne campaign measured
in situ gas chemistry from these events (Hatakeyama et al., 2014).
Historically, many researchers have used Fukue Island to identify long-range
transported aerosols (Takami et al., 2013). Sano investigated AOD at the site
in 2003 (Sano, 2004). Measurements showed periodic high-AOD days that may be due to transported anthropogenic aerosols and Asian dust events
from the continent. Part of the DRAGON-Fukue network was maintained until
2013.
Studies of Emissions and Atmospheric Composition, Clouds, and
Climate Coupling by Regional Surveys (SEAC<inline-formula><mml:math id="M70" display="inline"><mml:msup><mml:mi/><mml:mn mathvariant="normal">4</mml:mn></mml:msup></mml:math></inline-formula>RS)
The SEAC4RS mission (Toon et al., 2016) was a combined airborne and
ground-based effort to assess aerosols and trace gas chemistry processes. The
objective necessarily required knowledge of surface and boundary layer
meteorology to assess sources of aerosols and trace gasses. The airborne
implementation was changed from SE Asia (Maritime Continent) to the southeast
US regional assessment of aerosol and trace gas chemistry processes in 2013,
after permission to utilize airfields in SE Asia was not granted. This change
in locations represented a major challenge and a significant scaling up from
a mesoscale to a regional-scale ground-based aerosol network. It also
provided an opportunity to overlap with the Houston DISCOVER-AQ DRAGON
network (12 sites in ∼ 60 × 60 km) with a regional-scale
SEAC4RS network of (30 sites in ∼ 1000 × 2000 km). Both
networks operated at full density from August through October 2013. About
50 % of the SEAC4RS sites remain in operation as of 2017 to provide
long-term context for the program. Toon et al. (2016) provides a detailed
overview of the SEAC4RS program results. The NASA DC-8 with in situ
aerosol sampling instrumentation and the 4STAR airborne sun photometer
provided regional- and continental-scale transects that have been compared to
the ground-based measurements (Reid et al., 2017).
Additionally, another airborne and ground-based field campaign occurred during this time period called Southeast Nexus (SENEX; Warneke et al., 2016)
that emphasized volatile organic compounds (VOCs) and aerosol precursors.
This campaign was focused on Alabama, Georgia, and northern Florida. The regional
network by its size captures the range of aerosol properties one would
expect over the southeast US, including transported dust from west Africa,
biogenic aerosols created from VOCs, fossil fuel emissions, coastal maritime
aerosols, and biomass burning transported from fires in the western US.
KORUS-AQ
Similarly to the DISCOVER-AQ campaigns, a focused airborne campaign called
KORUS-AQ was conducted across South Korea from 1 May to 12 June 2016 by the
National Institute of Environmental Research (NIER) and NASA. In situ and
remote-sensing resources were on board three aircraft flying from the near
surface to ∼ 8500 m profiling the atmosphere in three dimensions for
up to 8 h on approximately 20 days. This campaign was heavily supported by a
DRAGON mesoscale network of 21 advanced AERONET Cimel photometers; most had
solar and lunar AOD retrievals as well as the experimental hybrid sky scans
designed to allow the retrieval of aerosol radiometric and microphysical
optical properties throughout the day. AERONET results for the lunar AOD and
retrievals from hybrid scans are undergoing evaluation at the time of
writing. It is noteworthy that two over-water oceanographic platforms
provided aerosol and normalized water leaving radiances over two sites in the
Yellow Sea during this time in support of ocean color investigations.
Additionally, two ships had Microtops sun photometers that were calibrated at
GSFC to be consistent with AERONET reference instruments. Furthermore,
supporting the KORUS-AQ campaign there was a high spectral-resolution lidar
(HSRL) onboard the DC-8 and ground-based lidars as well as several
contributing SKYNET PREDE sun–sky scanning spectral radiometers.
In addition, a regional-scale campaign of ground-based remote-sensing and
in situ measurements upwind and downwind of South Korea was conducted during
this period. This included the Institute of Remote Sensing and Digital Earth
SONET network, AERONET, and the China Aerosol Remote Sensing NETwork
(CARSNET; Che et al., 2009, 2015) Cimel Sun–sky radiometer networks in NE China that
contributed 20 stations focused eastward from Beijing and south to
Shanghai. In collaboration with the Institute of Remote Sensing and Digital
Earth of the Chinese Academy of Sciences and the University of Maryland, an
airborne in situ aircraft-based study of the chemical composition of the
atmosphere was also conducted during this period.
Coincidentally, an enhanced network of eight AERONET sites was distributed
across Japan from Fukuoka in the south to Sapporo in the north. This network
augments the extensive SKYNET network of sun–sky radiometers in Japan that
provides similar aerosol observations to AERONET but also collocated lidar
profiling and, in some supersite locations, in situ particle observations.
Since there is overlap at some of the AERONET and SKYNET sites in South Korea
and Japan, a unique and comprehensive comparison is planned between the
networks.
The greater KORUS campaigns extensively sampled fine-mode aerosols from
locally and regionally transported industrial and urban sources, biomass
burning from Siberian fires, and regionally transported coarse-mode-dominated
dust that strongly affected all countries on 5 May 2016 and to a lesser
extent on several other days during the campaign. All aerosol types except
for the Siberian biomass burning aerosols were also sampled during research
aircraft flight days. The opportunity to assess accuracies and limitations of
multiple satellite and AERONET retrievals and aerosol model forecasts for a
variety of aerosol types and cloud and humidity conditions is expected to
increase our understanding of the processes that govern air quality issues in
NE Asia.
ObseRvations of Aerosols above Clouds and their intEractionS
(ORACLES)
The NASA venture class suborbital program (ORACLES) is an ongoing airborne
campaign focused on biomass burning aerosol emissions from southern central
Africa transported over the South Atlantic to assess the aerosol–cloud
interaction over the persistent stratocumulus deck from August through
September 2016; it is planned for repeats in 2017 and 2018 (Zuidema et al.,
2016). Approximately 15 AERONET sites from Mozambique, Zambia, Angola,
Namibia, South Africa, St. Helena, and Ascension Island are providing regional context of aerosol
properties from source to receptor sites for the campaign. Additionally, a
tightly focused DRAGON network (seven sites in 20 × 30 km grid) was
set up on the central Namibian coast to assess the impact of aerosols on
coastal fog and quantify any influence fog may play in the aerosol size
distribution in this arid region.
DRAGON campaign summaries. D: dust; FF: fossil fuel;
B: biogenic; BB: biomass burning; M: maritime. Because of the time
period of measurement and the number and location of instruments and variable
aerosol types transported by synoptic-scale meteorology, AOD and particularly
SSA averages are approximate. Most campaigns are referenced at:
https://aeronet.gsfc.nasa.gov/new_web/campaigns.html, where DRAGON data
sets are also available with detailed point of contact (POC) information.
Campaign
Date
Location lat/long
AERONET sites
Aerosol source
∼ AOD440 range
∼ SSA440
POC
UAE2
Aug–Sep 2004
UAE 24∘ × 54∘
16
D, FF
0.1–0.8
0.93
Reid/Holben
CATZ
Jun–Aug 2007
USA 39∘ × -76∘
24
M, B, FF
0.1–0.8
0.96
Holben/AERONET
TIGERZ
May–Jun 2008
India 26∘ × 80∘
8
D, BB
0.3–1.2
0.88
Holben/Tripathi
7-SEAS
PENANG
Jul–Sep 2012
Malaysia 5∘ × 100∘
8
FF
0.3–2.0
0.96
Holben/Lim
Singapore
Aug–Sep 2012
Singapore 1∘ × 104∘
6
FF
0.2–1.5
0.94
Holben/Salinas
DISCOVER-AQ
Crawford
Maryland
Jun–Aug 2011
USA 39∘ × -77∘
43
FF, B
0.1–0.8
0.98
Holben/AERONET
San Joaquin
Jan–Feb 2013
USA 37∘ × -120∘
16
FF
0.1–1.3
NA
Holben/AERONET
Houston
Sep 2013
USA 30∘ × -95∘
18
FF
0.1–0.3
0.NA
Holben/AERONET
Colorado
Jul 2014
USA 40∘ × -105∘
13
FF, BB
0.1–0.3
NA
Holben/AERONET
D-KOREA
Mar–May 2012
South Korea 36∘ × 127∘
22
FF, D
0.1–1.3
0.98
J.Kim/Holben
D-JAPAN
Mar–May 2012
southern Japan
15
FF, M, D
0.1–1.3
0.98
Sano/Holben
SEAC4RS
Aug–Sep 2013
SEUS 33∘ × -87∘
24
FF, B, BB, M
0.1–0.7
0.95
Toon/Holben
KORUS-AQ
Crawford
Korea
May 2016
South Korea
22
D, FF, M
0.2–1.0
0.91
J.Kim/Holben
Japan
May 2016
Japan 35∘ × 135∘
7
FF, D
0.1–0.8
0.94
Sano/Holben
China
May 2016
China 40∘ × 116∘
11
FF, D
0.1–1.2
0.89
Z. Li/Che
ORACLES
Aug–Sep 2016
Namibia -22∘ × 14∘
7
FF, D, BB, M
0.1–0.5
0.84
Holben/Knox
Summary of the special issue contributions
Three important research areas have emerged as a result of the DRAGON
campaigns: (1) in situ and remote-sensing aerosol properties comparisons;
(2) aerosol process studies; and (3) satellite and model validation studies.
The first DRAGON-like campaigns focused in part on in situ versus
remote-sensing comparisons of aerosol optical, radiative, and microphysical
properties. Although some of the associated publications both pre- and
postdate this issue, they do merit a brief discussion. Schafer et al. (2014)
showed an average difference of ∼ 0.01 between in situ SSA from
aircraft profiles compared to AERONET-based retrievals for the DISCOVER-AQ MD
DRAGON data set in July 2011. Sawamura et al. (2014) used the diversity of
airborne and ground-based aerosol observations including the DRAGON
measurements as a reference to intercompare project observations to HSRL
radiative and microphysical properties. They found better agreement within
the specified uncertainties using the remote-sensing techniques compared to
the airborne in situ observations. Schafer et al. (2018) has made comparisons of in situ measured size
distributions from the multiple DISCOVER-AQ airborne profiles to the DRAGON
AERONET sun photometer retrievals. Comparisons of rehydrated in situ
measurements integrated vertically to the ambient retrieved remote-sensing
observations showed relatively good quantitative agreement based on
approximately 40 flights coincident in time and space with the ground-based
measurements. Sawamura et al. (2017) used DRAGON AERONET (California and
Houston) to evaluate HSRL-2 and airborne in situ AOD measurements.
Process studies have also broadened the research horizon possible from these
data sets, some of which appear in this special issue. For example Eck et al. (2014) used the DISCOVER-AQ Maryland DRAGON network observations to study
the effect of non-precipitating cumulus clouds on AOD in adjacent regions on
a horizontal scale of a few kilometers. They found that on some days, the Ångström
exponent and size distribution were relatively constant while AOD was
significantly enhanced (sometimes doubling in less than 1 h) near
moderately sized cumulus clouds. These results were corroborated by airborne
lidar and airborne in situ measurements. This has potential implications for
the need for a better understanding of small-scale high temporal variations in
aerosol–cloud processes and potential particle formation in clouds.
Much of the research activity with the DRAGON campaigns focused on air
quality, relating remote-sensing parameters to surface PM10. Seo et
al. (2015) analyzed the DRAGON-KOREA 2012 database, testing various linear
models that include boundary height and effective radius to surface PM10
measurements in the vicinity of Seoul for the winter and spring and also for long-term measurements. They found the best relationship in the winter, owing
to well-mixed aerosol layers, while the poorest relationships occurred during the
spring when long-range aerosol transport stratified the aerosol profile.
The DRAGON-Asia campaigns were used to broadly describe trans-boundary
advection of aerosols as a DRAGON-scale network in Osaka was imbedded in a
regional-scale network over southern Japan (Sano et al., 2016). This analysis
showed that, during episodic long-range trans-boundary transport, aerosol
loading was highest in the west of Japan but highly variable in space and
time both for fine- and coarse-mode aerosol events. The long-range
trans-boundary aerosols during this period were shown to follow the
NCEP-derived 700 to 850 mb wind vectors. Sano et al. (2016) investigated the
variability in AOD under clean and polluted days in Osaka using DRAGON
network measurements. They also detailed aerosol transportation over the city
using high spatial- and temporal-resolution measurements by DRAGON-Osaka.
Owing to two nearby mountain sites, boundary layer assessments were possible,
facilitated by nearby DRAGON-Osaka and AERONET stations. The DRAGON-Fukue
instruments did not capture the intense 10–11 March fine-mode event due to
cloud contamination. However, the authors successfully measured the event by
judiciously timed, handheld Microtops-II sun photometer observations (Nakata
et al., 2016). The value of AOD at 440 nm was over 2. Takami et al. (2013)
reported a particle composition of less than 1 µm diameter by
Aerodyne's aerosol mass spectrometer and that the most abundant components
were SO42-, NH4+, and OC during the event (Kaneyasu et al.,
2014).
Tan et al. (2015b) investigated the ability to use surface-based measurements
to predict AOD in the cloudy tropics of Penang, Malaysia, where data gaps can
be frequent and persistent. His predictive model had an r2 of 0.68
compared to actual measurements of AOD from the DRAGON network.
By far the largest application of the DRAGON data sets has been in the
validation of satellite data. Most synoptic-scale validation teams assume a
spatial uniformity about a ground-based control point, often citing the
Anderson et al. (2003) nominal scale length of 100 km. Frequently, queries
are made about the spatial representation of AERONET sites, for which there
is no simple answer due to the proximity to aerosol sources and local and
synoptic meteorology. The DRAGON campaigns have provided a better
understanding for some specific circumstances that provide for a better
assessment of the spatial resolution of various satellite products and also
of high- and low-resolution model assessments. Prior to this issue, Munchak
et al. (2013) noted the new Collection 6 MODIS 3 km AOD product could
potentially assess local aerosol gradients missed by the standard 10 km
resolution product. They used the MD DISCOVER-AQ airborne
high spectral-resolution lidar and
MD DRAGON data sets to assess the
fidelity of the 3 km AOD product, finding improvement over the
coarse-resolution product but some additional variability due to the
complexity of urban cover types. Kim et al. (2016) used the DRAGON- NE Asia
networks to refine the single scattering input to a single channel AOD
retrieval model used with the GEO COMS Meteorological Imager (MI). They note
that the surface-based inputs from DRAGON significantly improved the model to
predict AOD, thereby reducing previous over-estimates.
The Ozone Monitoring Instrument (OMI) on board Aqua has been a pioneering
instrument to retrieve SSA and AOD from space in the UV spectrum. Jeong et
al. (2016) have used the DRAGON NE-Asia data set in an optimal-estimation
procedure that provides error estimates while simultaneously retrieving
inversion products. This method was shown to compare better to the
ground-based measurements than the OMI operational retrieval. From this
validation, the authors identified the parameters that most affected the AOD
and SSA retrieval accuracy.
In a comprehensive comparison of the high temporal-resolution Geostationary
Ocean Color Imager (GOCI) and polar orbiting VIIRS and MODIS instruments,
Xiao et al. (2016) used DRAGON NE-Asia and additional AERONET observations in
2013, which encompassed a broad range of conditions from low to high aerosol
loading. Their
analysis suggests that the satellite products do a better job of tracking
aerosol variability on a day-to-day basis than tracking the high-resolution
spatial variability.
Choi et al. (2016) used the DRAGON NE-Asia data sets to evaluate the GOCI AOD
retrievals using the improvements to the GOCI Yonsei Aerosol Retrieval (YAER)
algorithm. The algorithm makes retrievals over the Yellow Sea that often have
Case II waters (highly turbid from sediment) as well as the highly variable
South Korean landmass reflectances during periods with highly variable aerosol
types and concentrations. GOCI YAER AOD correlated very well with AERONET but
showed lower skill with the Ångström exponent, fine-mode fraction, and SSA.
Garay et al. (2017) have assessed the current 17.6 km resolution AOD
products against multiple diverse DRAGON data sets collected around the
world. They found that 75 % of the data fell within 0.05 of the AERONET
surface-based measurements. They document the development and assessment of a
prototype version of high-resolution (4.4 km) retrieval products compared to
the same DRAGON data sets.