Introduction
The NASA-CALIPSO (Cloud–Aerosol Lidar with Orthogonal Polarization)
mission offers unprecedented observations of aerosol global optical
properties profiles , vital for
aerosol–radiation–cloud interaction studies to understand their
climatic role. The most recent CALIPSO satellite data product, the
so-called CL3 aims to provide a climatology of the global aerosol
distribution including seasonal and interannual variations. The
product consists of monthly gridded extinction profiles separated into
a daytime and nighttime segment. According to the study of
, the CL3 data appear to be realistic and very well
capture the most important aerosol transport pathways, such as the
westward motion of dust particles originating from the Sahara Desert,
or the smoke-laden plumes in the South Atlantic due to the African
biomass burning season.
As with any satellite product, it is important to quantitatively
evaluate the accuracy of CALIPSO retrievals in comparison with
independent measurements. CALIPSO products have been extensively
evaluated using columnar aerosol optical depth (AOD) data sets from
passive spaceborne measurements e.g., or the well-established AERONET (Aerosol
Robotic Network) measurements e.g.,. However, CALIOP, onboard CALIPSO, is firstly and foremost
a profiling instrument; therefore it is particularly interesting to
compare with ground-based profiling data. EARLINET (the European Aerosol
Research Network) is playing an important role in the validation and
full exploitation of the lidar data that CALIPSO continuously provides
since April 2006. In the frame of the network, several studies have
investigated the CALIPSO Level 1 products e.g.,. , and also
provided validation efforts of the CALIPSO Level 2 aerosol backscatter and
extinction profiles, showing promising results.
Currently, EARLINET space-related activities focus on the CALIPSO mission,
but nonetheless the network's goal is the provision of a long-term
ground-based support for the spaceborne lidar in order to homogenize
observations obtained with different instruments. The planned ESA
(European Space Agency) ADM-Aeolus Atmospheric Dynamics Mission
– Aeolus; and the joint ESA/JAXA (Japan Aerospace
Exploration Agency) EarthCARE Earth, Clouds, Aerosols and
Radiation Explorer; missions will succeed CALIPSO
in observing aerosols and clouds with active remote sensing
techniques. The Atmospheric Doppler Lidar Instrument (ALADIN) onboard
ADM-Aeolus and the Atmospheric Lidar (ATLID) of the EarthCARE
satellite will make use of the high-spectral-resolution-lidar (HSRL)
technique in the UV. In addition to the differences in the techniques
employed in relation to CALIOP, the ALADIN and ATLID will operate at
different wavelengths and will deliver extinction and backscatter
coefficient profiles, independently retrieved. EARLINET aims to
contribute also to the homogenization of the current and future
spaceborne lidar data sets by delivering aerosol and
cloud-type-dependent wavelength conversion factors. These parameters
will facilitate the development of a multi-decadal vertical structure
profile climatology .
So far, few studies about the CL3 data set have been
published. have compared the extinction values
retrieved by CALIOP against the simultaneous measurements of the HSRL
lidar onboard NASA B200 aircraft during CALIPSO underflights
. This comparison showed that the CALIOP retrieval
in the upper troposphere are underestimated due to the instrument
detection limits and to the decreasing aerosol load. Next,
compared CL3 AOD against MODIS (Moderate Resolution
Imaging Spectroradiometer) and found that CL3 demonstrated good
seasonal variability and in overall lower AOD values. Further, the
study showed significantly lower values for CALIPSO compared to MODIS
over deserts, with maximum difference of 0.3 over the Sahara Desert,
and the opposite when biomass burning particles are prevalent, with
maximum difference of 0.25 over South Africa. Owing to the varying
properties of dust on the lidar ratio, examined
the potential improvement of CL3 when introducing a new value of lidar
ratio for the dust. The increased agreement of CL3 when compared to
multi-platform and dust model products highlighted the improvement of
the dust extinction retrieval.
In this paper we present the first study to take full advantage of
long-term aerosol measurements acquired by the EARLINET ground-based
lidar network to critically evaluate CALIPSO climatological products
such as the aerosol optical properties reported in the CL3 data
product. Extinction retrievals from CALIOP, an elastic backscatter
lidar, are inextricably linked to the extinction-to-backscatter ratios
(i.e., lidar ratios) that characterize the CALIPSO aerosol models and
to the performance of the aerosol type identification
module. Therefore, while the CL3 files report only spatially and
temporally averaged extinction profiles, an in-depth validation of
these data must also examine the companion backscatter profiles that,
together with the lidar ratios, are used to create the CL3 extinction
profiles. Hence, we used the CALIPSO Level 2 data to create a modified
version of the CL3 data, hereafter denoted as CL3*, wherein we derive
averaged profiles of CALIPSO extinction and backscatter. Quality
assurance protocols for filtering the Level 2 data followed
established techniques previously reported in the scientific
literature see. The CL3* data set is compiled over
a smaller spatial domain than the standard CL3 data, and is closely
tied to the locations of the individual EARLINET stations. This
additional attention to spatial and temporal matching helps to
minimize differences identified in the previously performed
EARLINET–CL3 comparison (not reported) that could be attributed to
spatial variability over the CL3 grid box.
The data and methodology are presented in Sect. . The
results are reported and discussed in
Sect. . Specifically, Sects. and
focus on the comparison of the extinction
coefficient, backscatter coefficient, and lidar ratio profiles for each
station; further aerosol typing data are also intercompared. In
Sect. , the mean EARLINET type-related lidar ratio
values are confronted with the CALIPSO modeled values. Additionally,
it explores instead the effect of the extinction retrieval
optimization by using the EARLINET estimated lidar ratio
values. Finally, Sect. closes with
our conclusions.
Data
CALIPSO
CALIPSO is a joint NASA/CNES (Centre National d'Études Spatiales)
satellite designed to study aerosols and clouds. Its aim is to provide
profiling information at a global scale for improving our knowledge
and understanding the role of the aerosol in the atmospheric
processes. The main instrument, CALIOP, is a dual-wavelength (532 and
1064 nm) elastic backscatter lidar with the capability of
polarization-sensitive observations at 532 nm
. The high-resolution profiling ability
coupled with accurate depolarization measurements make CALIPSO an
indispensable tool to monitor dust aerosols .
The optical properties retrieval is
based on the successful cooperation of three modules, that have the
main goal to produce the CL2 data. The first module identifies the
features within the lidar signals aerosol, cloud, surface
returns;. Afterwards, this information is passed to
the second module, to determine the type of each feature i.e.,
cloud, aerosol, surface, or stratospheric;. Given this
selection, the module can type further those identified aerosol layers
i.e., clean marine, dust, polluted continental, clean
continental, polluted dust, smoke;, a procedure which is
called the aerosol subtyping. In this stage, also, CALIOP determines
the cloud phase . Finally, the third module
retrieves aerosol extinction and backscatter profiles assuming lidar
ratio values according to subtyping .
The climatological CL3 product is a monthly gridded data set
consisting of CL2 data. The main outputs are the aerosol extinction
coefficient at 532 nm and mean column aerosol optical depth (AOD).
The CL3 product, in which the CL2-532 nm aerosol extinction product is
aggregated, are mapped onto a global 2∘× 5∘
latitude–longitude grid. The output altitude ranges from -0.5 to
12 km above mean sea level with a vertical resolution of
60 m. CALIOP retrieves aerosol below optically thin clouds, in
clear skies and above clouds. Monthly mean-extinction profiles are
computed for four conditions: all-sky, cloud-free, above clouds and
combined (cloud-free and above clouds). In addition, several quality
control flags contained in the CL2 files are used to screen the data
prior to averaging. A detailed summary of the methodology used for the
generation of the CL3 product is provided in the Appendix of
.
EARLINET
EARLINET was established in 2000 (;
https://earlinet.org) as a research project, providing data
concerning the aerosol vertical distribution on a continental
scale. Currently, 27 active stations participate in the network. The
contributing stations have been performing correlative measurements
since CALIPSO started its life cycle, based on a schedule established
before the satellite mission. EARLINET has been an important
contributor to CALIPSO validation studies e.g.,. The strategy followed by the member stations is as
follows: the observations occur during the satellite overflight within
100 km distance of the satellite ground-track from the
station, and are performed for at least 60 min.
In this kind of measurement, the atmospheric variability both in
time and space is a fundamental point. The impact of the distance
on EARLINET–CALIPSO comparison was investigated for
different stations in devoted papers e.g.,. At network level we found that at distance below 100 km
the discrepancies in the signal (CALIPSO Level 1 data) are below 5 %. Moreover,
for cases of long-range transported aerosol like Saharan dust, it was
found that a horizontal distance of 100 km corresponds to high
correlation among the two profiles .
Figure illustrates CALIPSO's overflight
that triggers the measurements of the EARLINET station of
Potenza. Additionally, simultaneous measurements are predicted in
order to study the aerosol temporal variability, or in the case of special
events to study specific aerosol types and to investigate the
geographical representativity of the observations
.
Example showing CALIPSO's ground track that passes the EARLINET
measurement site at Potenza at a distance of less than 100 km.
EARLINET data quality is assured by strictly quality assurance
procedures established within network, firstly on systems and
retrieval processes . Further, data quality check is performed, also, on
the products . The EARLINET database related to
the CALIPSO overpasses is published to the CERA database
. The data are freely available at the EARLINET
web site and ACTRIS (http://www.actris.eu/) and CERA data
portals (http://cera-www.dkrz.de/WDCC/ui/Index.jsp).
Analysis setup
Comparison methodology
The CALIPSO measurements that result in the CL3 data are aggregated in
a 2∘ × 5∘ grid cell, whereas for EARLINET the
measurements can be considered as point. Furthermore, the constituting grid
cell overflights are not closely tied to the locations of the individual
EARLINET sites. For the reasons mentioned, the CL3 and EARLINET data sets are
not comparable in number and spatial representativity, and as a consequence
an ad hoc procedure for obtaining statistically comparable data sets is
necessary. In particular, only CALIPSO data segments corresponding to
EARLINET measurements were selected. The comparison of matched observations
reduces uncertainties from spatial and temporal differences, but greatly
reduces the number of the samples.
To produce the CL3* monthly profiles, we use the CL2 Version 3.01 Aerosol
Profile product, which includes aerosol extinction and backscatter
coefficient profiles at 532 nm. The spatial domain onto which the CL2
data are mapped is nearly 2∘ × 2∘ and contains the
EARLINET sites. This means that the longitudinal resolution is smaller owing
to the distance of CALIPSO overpasses (⩽ 100 km) from the
EARLINET measuring site. The six-step methodology to quality assure the CL3
profiles is modified by adjusting an existing metric
according to the rubric used by :
Extinction_Coefficient_Uncertainty_532 ⩽ 10 km-1.
The lower boundary, here, is set to a smaller value, whereas within
CALIPSO procedure, retrievals deemed unstable are set to 99.9 km-1.
In this case, samples that meet this condition are removed as well as
samples at lower altitudes. Prior to averaging, samples are excluded
where the screening criteria are invoked and moreover, for samples
that represent clear air a value 0.0 km-1 is assigned, although clear-air samples over the surface are ignored from the
averaging process in the case that the base of the lowest aerosol
layer in the profile is below 2.5 km.
Characteristics of CALIPSO and EARLINET data considered for this analysis.
CALIPSO
EARLINET
Quantity
Extinction_Coefficient_532 from L2-AProf 5 km
Particle extinction from the e files
Backscatter_Coefficient_532 from L2-AProf 5 km
Particle backscatter from the b files
Coverage
Nighttime
Nighttime
Comments
≥2 profiles/month create monthly profile
In this analysis, CALIPSO extinction profiles at 532 nm are
directly compared to corresponding EARLINET correlative measurements
for the period 2006–2011, considering only the nighttime segment of
the CALIPSO data set. We calculate the monthly average only when at
least two measurements are available within the considered month. Only
EARLINET cloud-free and below-cirrus-clouds profiles and CALIPSO
cloud-free and above-cloud data are used to calculate the averaged
profiles. As an additional benefit, the reprocessing gives the
opportunity to compare also CALIPSO with EARLINET aerosol backscatter
coefficient and to correlate with the extinction comparisons. The same
screening rubric used for the extinction coefficient is applied to the
backscatter data as well. The characteristics of the data considered
are reported in Table . We also take advantage of the
couple of optical properties to examine the lidar ratio, in accordance
with the findings of the aerosol subtyping scheme of the two
platforms.
For CALIPSO, aerosol classification is a key input to the aerosol
retrieval and must be inferred, therefore the CALIPSO aerosol
classification is compared against EARLINET typing data.
CALIPSO aerosol classification
As was noted in Sect. , CALIPSO retrieval classifies
aerosol layers into six subtypes, a crucial selection on which is based the
aerosol optical properties retrieval. That is due to the absence of
independent optical depth measurements ; therefore the
aerosol lidar ratio inference is required prior to retrieval. The
classification makes use of the aerosol location, aerosol height, the
integrated attenuated backscatter, the approximate particle depolarization
ratio and the surface type in order to type the
layers. Regarding the surface type, clean marine particles are only permitted
over water bodies; therefore the overland flow of marine particles is not
considered in the scheme. The assigned types have been previously identified
from cluster analysis based on AERONET data . Each aerosol
subtype is characterized by a set of lidar ratios for 532 and 1064 nm
wavelengths. Table shows the values set in the
CALIPSO classification scheme for each of the aerosol subtypes.
CALIPSO aerosol subtypes and the associated lidar ratio at
532 nm used in the aerosol optical properties retrieval. CM stands
for clean marine, D for dust, CC for clean continental, PC for polluted
continental, PD for polluted dust, and S for smoke.
Aerosol type
CM
D
CC
PC
PD
S
Lidar ratio at 532 nm [sr]
20
40
35
70
55
70
EARLINET aerosol classification
Aerosol features from EARLINET are typed according to methods already
consolidated within the network . Briefly, the lidar data evaluation is a three-step procedure:
the feature finding and cloud-aerosol discrimination,
the identification of the boundary location of the aerosol
layer, and
the aerosol layer typing by means of investigation of intensive
optical properties (Ångström exponent, lidar ratios, linear
particle depolarization ratio), model outputs, backward trajectory
analyses, and ancillary instruments data if available.
The aerosol layers, identified as above, are typed with respect to the
CALIPSO aerosol subtyping (Table ). The
EARLINET layers, therefore, fall into six subtypes: marine, dust,
polluted continental, clean continental, polluted dust, and smoke. In
order to achieve this, we had to compromise the comparison for the
maritime particles. Since pure marine layers are rarely observed over
the considered stations, typically mixtures of marine and other
aerosol types are measured in the lidar signals, and the clean marine
CALIPSO type is directly compared with the EARLINET marine type. We
will hereafter use the marine notation for both CALIPSO and EARLINET
subtyping. Note that a significant discrepancy of the existing typing
schemes concerns the polluted dust subtype. This subtype represents
a mixed aerosol situation: in the CALIPSO algorithm the subtype takes
into account mixtures of dust with smoke or pollution, while in the
EARLINET classification the dusty mixtures also include maritime
particles.
Number of CALIPSO overflights and EARLINET correlative observations along with the produced monthly profiles, and the minimum distance between the satellite ground track and the EARLINET station.
EARLINET station
CALIPSO overpasses
Monthly profiles
Minimum distance [km]
Évora
15
5
63.6
Granada
20
8
66.8
Leipzig
20
10
51.4
Naples
26
11
64.0
Potenza
33
13
67.9
Total
114
47
63.5
Selected sites
The EARLINET data related to CALIPSO overpasses, spanning the period from
June 2006 to December 2011, consist of 7554 particle backscatter and
extinction profiles . The particle extinction profiles
are 1047, of which 478 correspond to 355, 498 to 532 nm,
and the rest to other wavelengths. The stations, therefore, providing the
largest data set are Évora, Granada, Leipzig, Naples and Potenza, all
equipped with multi-wavelength Raman lidars. Apart from the data redundancy,
the stations were also selected with respect to their range resolution. The
analysis is based on the precise layer location, which can be accomplished by
using a resolution finer or comparable to the CALIPSO one (60 m in the
lower troposphere). Figure shows the geographical distribution
of the sites (yellow squares) – in the West, Évora (293 ma.s.l.)
and Granada (680 ma.s.l.); in Central Europe, Leipzig
(90 ma.s.l.); and in central Mediterranean, Naples
(118 ma.s.l.) and Potenza (760 ma.s.l.). The original CL3
grids linked to the EARLINET sites are reported as blue boxes. The red boxes
embedded in the standard CL3 grid cells correspond to the CL3* data grids.
The CL3* cells for Naples and Potenza exceed the CL3 borders and even
overlap as both site locations lie close to the CL3 borders and are
separated by ∼100 km. The CL3* cell latitudinal edges are kept
the same as for CL3, whilst the longitudinal edges are dictated by the
EARLINET correlative measurements scheme (ca. 1∘ to the west and to
the east from the site's location). The number of available EARLINET
correlative observations and CALIPSO grid overflights that were used to
produce the mean profiles are summarized in
Table . Moreover, the table reports the
mean minimum distance between the satellite ground track and the EARLINET
stations – the total mean minimum distance was found to be 63.5 km. To
ensure that the same air volumes were sampled, the HYSPLIT model
in backward mode was used. The model was initiated for
each CALIPSO measurement and its EARLINET counterpart and the corresponding
trajectories were visually inspected. Each model run was set in the range of
0.5–6 km and for constant height increments, independently of the
existence of aerosol layers. For all the cases related to this study, the
model analyses indicated that the ground-based and satellite lidars sampled
the same air mass.
Spatial boundaries of the CALIPSO data that are related to the five
EARLINET sites. The alternative CL3* domain reflects the finer spatial
resolution with regard to the CL3 domain. The CL3* grid cell is dictated by the
correlative measurements schedule (measurements are triggered when the
satellite's ground track is within 100 km distance from the station);
the latitude borders of the grid are kept equal to the CL3 grid.
The majority of the observations were performed during summer and spring
months (25 and 13 monthly profiles respectively) owing to the favorable
conditions and do not permit to assess the seasonal behavior (eight autumn and
one winter mean profiles). The larger number of available comparisons for the
warmer months, indeed, influences our results to some extent. The analyzed
data set is highly affected by dust/smoke presence which typically occurs
during these months e.g.,and references therein. Clean conditions are less represented, here, but since they
contribute less to the total AOD their influence is less important. However,
it should be noted that the influence of lidar ratio increases with the layer
AOD so it is more relevant for the dust/smoke plumes in general.
AOD differences in the range 0–10 km over 1 km height intervals for the five EARLINET stations.
AODCALIPSO–AODEARLINET
Height range [km]
Évora
Granada
Leipzig
Naples
Potenza
Total
9–10
<0.001
<0.001
-0.001
<0.001
<0.001
<0.001
8–9
0.001
<0.001
-0.001
-0.001
<0.001
<0.001
7–8
0.002
0.003
-0.002
<0.001
0.001
0.001
6–7
0.004
-0.004
-0.002
0.003
0.002
<0.001
5–6
-0.002
-0.003
-0.001
-0.003
0.001
-0.001
4–5
-0.002
-0.003
<0.001
0.001
<0.001
<0.001
3–4
-0.003
<0.001
0.010
<0.001
<0.001
0.002
2–3
0.013
-0.017
0.008
-0.001
-0.008
<0.001
1–2
-0.018
-0.017
-0.001
-0.019
-0.037
-0.018
0–1
n/a
n/a
n/a
-0.026
n/a
n/a
Total
0.038
-0.046
-0.002
-0.052
-0.041
-0.046
Results
Figures , , and show, respectively,
the mean particle extinction, backscatter coefficient, and lidar ratio at
532 nm comparison of EARLINET (red line) and CL3* (blue line) as
a function of height. The monthly mean profiles, shown in
Table , are averaged for the five grids
and presented, here, along with their standard deviation (shaded error bars).
The panels from left to right refer to the five EARLINET grid cells and are
sorted alphabetically. The integral of the extinction coefficient at
1 km range increments was calculated for both profiles, and the
corresponding AOD differences are reported in Table . The
plots of Figs. – represent the typing of the EARLINET measurements (left
panels) and the corresponding CALIPSO overpasses (right panels) for the five
grid cells. The probed altitude range was partitioned into 1 km bins
and the percentage of layers identified within each bin is reported.
Therefore, according to the boundary location, layers can be present in more
than one height bin range. However, the first bin is associated with the
lowest altitude point retrieved by EARLINET, thus the range can be smaller
than 1 km. For this comparison, the same distance was used for both
EARLINET and CALIPSO typing. For the sake of visual consistency, the height
bins are kept equidistant for all the plots.
Extinction coefficient at 532 nm for CL3* (blue line) and
for EARLINET (red line). From left to right: (a) Évora,
(b) Granada, (c) Leipzig, (d) Naples, and
(e) Potenza.
Backscatter coefficient at 532 nm for CL3* (blue line) and
for EARLINET (red line). From left to right: (a) Évora,
(b) Granada, (c) Leipzig, (d) Naples, and
(e) Potenza.
CALIPSO level 3* comparison
Évora
Évora is situated in southern Portugal, and lies 100 km east
of the industrial area of Lisbon . The station is
a rural site and consequently is appropriate for the study of aerosols from
different sources. In Fig. a, the Évora EARLINET monthly
particle extinction coefficient decreases steeply up to 2 km and then
gradually continues to decrease up to 6.5 km. On the other hand,
CALIPSO profile yields a different behavior both in aerosol layering and
extinction values. CALIPSO reported a strong aerosol feature around
2 km not observed by the EARLINET station and did not affect the
resulting mean profile. The feature that caused the discrepancy in the
profiles was flagged by CALIPSO as dust and its mean extinction value was
0.14 km-1. Between 2.5 and 5 km the profiles are in good
agreement. Further, above 5 km height the situation changes as the
ground-based lidar yields zero values, while CALIPSO identifies aerosol
layers. The total AOD difference (Table ) for the whole
range is 0.038. The situation for the backscatter coefficient comparison
(Fig. a) shows better agreement around 2 km, yet the
CALIPSO backscatter values in that specific layer and above exceed the
EARLINET ones. The lidar ratio (Fig. a) within the errors is in
good agreement, though the EARLINET standard deviation is higher than the CALIPSO
one. This is probably the result of the aerosol mixing and difference in the
volumes sampled. The mean EARLINET lidar ratio is 55±10 sr and
the corresponding CALIPSO value is 51±7 sr. Specifically, for
the area of discrepancy around 2 km there is an altered situation
where the CALIPSO lidar ratio is 55±3 sr while EARLINET yields 46±6 sr.
Lidar ratio at 532 nm for CL3* (blue line) and for EARLINET
(red line). From left to right: (a) Évora, (b) Granada,
(c) Leipzig, (d) Naples, and (e) Potenza.
Figure a presents the situation as observed by the
ground-based lidar. Polluted continental and polluted dust showed the
most pronounced impact on the aerosol loading. Typically, air masses
flow from the west and prior to arriving at Évora cross the
polluted area of Lisbon, creating the polluted mixtures. Oddly, pure
dust particles were not detected during the measurements. Marine
particles have a strong influence for the first range bin. On the
other hand, Fig. b reports the particle classification
delivered by the CALIPSO typing module. Polluted dust displayed the
highest and constant frequency for all the height bins. Dust, by
contrast to EARLINET, plays an important role and has increased
frequency rate in higher altitudes. Polluted continental samples
decrease with height, but have a significant contribution in the
first height range. Smoke and marine particles had a minor frequency
throughout the range.
Évora: (a) EARLINET and (b) CALIPSO typing
bar-plots for 1 km range increment. M stands for marine, D for dust,
PC for polluted continental, CC for clean continental, PD for polluted dust,
and S for smoke subtype.
Granada
The Granada EARLINET station is located in the south of Spain and is
situated in a natural basin surrounded by mountains of variable height from 1
to 3.5 kma.s.l. The main contributors to the local aerosol load are
the mineral dust from North Africa and anthropogenic pollution from Europe
. The mean aerosol extinction profiles
(Fig. b) yielded higher values for EARLINET up to
3 km, above that range both profiles showed a good
agreement. The mean AOD difference, reported in
Table , is -0.046. The backscatter comparison
(Fig. b) revealed the same characteristics with enhanced
discrepancy in the lowermost part of the profile, as expected due to
the complex topography of the region .
Despite the observed differences in both extinction and backscatter
coefficient profiles, the agreement on lidar ratio is in general good
(Fig. b). The EARLINET retrieved lidar ratio is 45±3 sr and the calculated CALIPSO lidar ratio is 46±4 sr.
In Fig. a, the ground-based lidar retrieval identified
polluted dust and dust as the most frequent observed particle
subtypes. Polluted dust shows the highest frequency for the first two
height bins and dust for the rest. Dust is present everywhere and
increases its contribution gradually as a function of height. Polluted
continental particles are found as high as 4 km and contribute
significantly in the aerosol load for the lowest altitudes. Marine
particles were observed for the first four height bins; these
particles are transported from the Atlantic Ocean and the
Mediterranean as well. Smoke particles highly affect the lidar signals
over 5 km. For CALIPSO algorithms (Fig. b), as
was the case for EARLINET, polluted dust and dust showed
a complementary behavior, with polluted dust affecting more in the
first height bins and dust higher up. Both smoke and clean continental
particles weakly influence the lidar signals at high
altitudes. No contribution was found for marine and a minor contribution
from polluted continental particles. Overall, the CALIPSO and
EARLINET aerosol typing indicate dust and polluted dust as the major
aerosol types over the Granada grid. Once more, the dusty components
identification is well captured.
Granada: (a) EARLINET and (b) CALIPSO typing
bar-plots for 1 km range increment. M stands for marine, D for dust,
PC for polluted continental, CC for clean continental, PD for polluted dust,
and S for smoke subtype.
Leipzig
The Leipzig EARLINET site is the sole continental location and
presents different characteristics with respect to the other examined
grid cells. Free tropospheric layers are due to advection from North
America, pollution from areas north of 70∘ and East and
Southeast Europe and Russia, as well as, even if more rare, dust
intrusions from the Sahara . In Fig. c,
the extinction profiles indicate aerosols up to 4 km. The
Leipzig station reports aerosol also for higher altitudes although
with rather low extinction values. Two distinct layers, one in the
range 1.8–2.6 km and a second in 2.9–3.6 km, were
captured by CALIPSO, but not observed at Leipzig station. The total
AOD difference is -0.002 (Table ). The particle
backscatter comparison for 532 nm, as shown in
Fig. c, improves significantly in the lowermost part of
the profile. In Fig. c, the mean CALIPSO lidar ratio is
60±4 sr and it is rather constant with height. On the
other hand, the EARLINET lidar ratio is separated into two distinct
regions – in the first region (around 1.8 km) the mean value is
76±10 sr indicating the fine, absorbing particles
located near the surface. The second region (1.8–3 km)
coincides with the calculated mean CALIPSO lidar ratio, and exhibits
a mean value of 62±2 sr.
The Leipzig ground-based observations indicated as the most important
component of the local aerosol load the polluted continental for all
height intervals, as shown in Fig. a. Polluted
dust, smoke, and dust follow in frequency of identification. Dust along
with smoke particles have a stronger influence in the higher
range. Clean continental particles lie in the first two height
bins. The CALIPSO typing, shown in Fig. b, for the height
interval 1–2 km identifies smoke and polluted continental
equally; for the same range polluted dust contributes the most. Smoke
particulates keep a rather constant identification frequency for the
next height increments, whereas polluted dust showed a decreasing
frequency with height. Dust has a slightly increasing frequency with
height and reflects very well the EARLINET identification rate. Clean
continental subtype becomes important in the range 3–4 km and
competes in identification frequency with the dust and smoke subtypes.
Naples
The urban area of Naples is characterized by high aerosol content, mainly
located in the planetary boundary layer (PBL), originating from both natural sources and anthropogenic
activities . Looking at Fig. d,
the strong deviation of the EARLINET and CALIPSO extinction mean profiles
below 2 km (mean extinction bias -0.05 km-1) is evident. This
behavior can be attributed to the local aerosol content of the area of
Naples, which is a densely populated and highly polluted city, and to the
grid on which the CALIPSO profiles are mapped; consisting mostly of maritime
area (see Fig. ). For the upper altitude level the difference
diminished and the agreement is satisfactory (mean extinction difference
<0.001 km-1). The mean AOD difference
(Table ) is -0.052 if we consider the whole range, and
-0.022 for altitudes above 1 km. Nonetheless, the strong
anthropogenic impact around the area of Naples influences the comparison. In
Fig. d, the particle backscatter comparison shows a significant
improvement as the discrepancy in the lowermost part of the profile is
reduced. The retrieved lidar ratio, shown in Fig. d, yields
larger values below 2 km (PBL plus adjoining regions),
Saer=72±9 sr, because of the strong influence of
small absorbing particles. The PBL is capping local anthropogenic aerosols from
combustion, industrial activities, and traffic. In the region of
2–3 km there is good agreement between the two platforms with mean
lidar ratio values of Saer=44±4 sr for Naples station
and Saer=44±2 sr for CALIPSO. In the upper level the
EARLINET lidar ratio fluctuates, owing mainly to the low signal-to-noise ratio. A lidar ratio
almost constant in the 0–2 km range is assumed in the CALIPSO
retrieval with values of 41±3 sr, indicative of dust particles
(Saer=40 sr), and 46±3 sr above
2 km.
Leipzig: (a) EARLINET and (b) CALIPSO typing
bar plots for 1 km range increment. M stands for marine, D for dust,
PC for polluted continental, CC for clean continental, PD for polluted dust,
and S for smoke subtype.
The EARLINET (Fig. a) typing scheme for the first height
bin identifies stronger anthropogenic pollution, that decreases with
height but still represents an important contribution to the aerosol
situation. Dust and polluted dust particles reveal a stable behavior
over the different height intervals. Smoke plumes lie in the higher
altitudes of the profiles. The first two height bins are influenced by
marine particles, that typically for the Naples site are mixed with
the local aerosol content. Figure b indicates the
influence of dust and polluted dust particles in CALIPSO data over the
Naples grid; their vertical distribution is rather constant. These
subtypes have the most profound impact on this grid cell. Marine
particles expectedly lie in the lowest range of the profile, while
polluted continental particles are almost nonexistent. This mismatch
for the polluted continental subtype indicates the large deviation of
the extinction coefficients in the lower part of the profiles. The
clean continental type becomes important in the higher parts of the
profile as well as the smoke category but to a lesser extent. The
agreement, once more, for the dust and polluted dust category is very
good, taking into account the variations of the aerosol field and the
surface type.
Naples: (a) EARLINET and (b) CALIPSO typing
bar plots for 1 km range increment. M stands for marine, D for dust,
PC for polluted continental, CC for clean continental, PD for polluted dust,
and S for smoke subtype.
Potenza
In contrast to the neighboring Naples, the Potenza station is located
at a mountainous, rural site. The relatively low local aerosol content
makes the observations particularly interesting for long-transported
particle plumes . In Fig. e,
the discrepancy in the profiles below 2 km is significantly
high (mean extinction bias -0.05 km-1). The differences
are reduced in the upper levels (mean extinction bias
<-0.01 km-1). The lower-level disparity typically is
weakened during summer months, and is intensified in winter, yet
the sample size is too small to quantify the periodicity of this
discrepancy. The integral of the extinction coefficients over constant
height ranges was calculated, as shown in Table ,
with a total mean AOD bias of -0.041. Figure e shows
that the “gap” in the extinction profiles near the ground disappears
for the backscatter profiles. That might suggest a wrong a priori
selection or inference of lidar ratio in the CALIPSO
retrieval. Therefore the lidar ratio profile for each month is
estimated and directly compared to averaged unconstrained EARLINET
lidar ratio profile. The CALIPSO lidar ratio, in Fig. e,
is kept for the whole altitude range slightly below 50 sr,
Saer = 49 ± 3 sr. On the other hand EARLINET
measured lidar ratios exhibit higher values in the range
1.5–2.7 km, Saer = 62 ± 3 sr, most likely
because of the influence of absorbing particles. In the height range
2.7–5 km, the CALIPSO lidar ratio values agree well with the
EARLINET mean value of 50 ± 5 sr. The obtained lidar ratio
values agree with the findings of , and suggest the
existence of dust particles in the height range 2.7–5 km.
Figure a gives an outlook of the aerosol types observed
by the EARLINET station; polluted continental particles affect the
most in the first height bin and decrease significantly as a function
of height. Polluted dust and dust affect the area around the site,
dust identification frequency is increasing with height, while for
polluted dust the frequency is rather stable. Smoke particles have
a range-invariant character up to 4 km. For CALIPSO,
Fig. b, dust and polluted dust prevail over the
grid. Smoke is present in the range 1–4 km; some polluted
continental is in the first height bin, and clean continental resides
in the higher altitudes. As far as marine particles, they slightly
affect the study area.
Potenza: (a) EARLINET and (b) CALIPSO typing
bar plots for 1 km range increment. M stands for marine, D for dust,
PC for polluted continental, CC for clean continental, PD for polluted dust,
and S for smoke subtype.
General findings and discussion
Figure displays the relative difference of the
extinction and backscatter comparison for each examined station.
First, we calculate the height below which the 90 % of the
columnar AOD is confined using the extinction profiles. Next,
the relative biases are estimated as (xCALIPSO-xEARLINET)/xEARLINET, where x is the
extinction or backscatter profile. For most of the stations,
the backscatter comparison at 532 nm suggests better
performances of the CALIPSO backscatter with respect
to the extinction. Hence, using the CALIPSO backscatter coefficient,
the comparison improves the relative mean biases when compared to the
CALIPSO extinction coefficient. In particular, the mean relative
difference for the averaged backscatter profiles was found to be 18 %
whereas for the extinction profiles it was 25 %. The better
agreement in terms of backscatter has to be ascribed to the higher
influence of lidar ratio assumption on extinction rather than on
backscatter. Nevertheless, this outcome should be treated with care as
the differences are mainly located in the lower troposphere where
typing and subsequent lidar ratio inference is complicated due to
the complexity of the scenes.
Relative difference of extinction and backscatter coefficient for
each considered site.
For what concerns aerosol typing, CALIPSO identifies successfully the
dust component. This is expected as the Saharan dust outbreaks are the
main source of particles in the free troposphere over the considered
sites, and their role is established in the local aerosol loading
e.g.,. More
importantly, CALIPSO's depolarization measurements facilitate the
discrimination of irregular-shaped particles. The polluted dust is
also effectively identified, yet it is overused in the lowest height
bins by contrast to the EARLINET identification frequency (for the
Évora, Granada, and Naples sites). Regarding this situation, a bug
has been identified and documented by and
, which stems from the CALIPSO retrieval code
causing an overestimation of the polluted dust subtype. This
overestimation increases with increasing AOD above a layer and hence
will be most prominent in the lowest altitude regions, as was observed
in this study. The marine layers are surface dependent for the CALIPSO
retrieval codes and are not considered over continental grid cells,
whereas the stations in the Mediterranean are obviously affected by
mixtures of marine particles. Besides, CALIPSO underestimated the
outflow of anthropogenic pollution from coastal sites towards the sea,
as these aerosols are wrongly flagged as marine if observed over the
sea. This situation was observed for the grid cell of Naples and is in
agreement with the outcome of .
Lidar ratio investigation
The choice of lidar ratio values in the CALIPSO retrievals can be
a significant reason for the discrepancies observed in the aerosol
extinction profiles. To investigate this, the mean EARLINET lidar
ratio for each subtype is calculated and then compared with the
corresponding CALIPSO modeled values (see
Table ). The EARLINET subtype layers were
considered in the statistics only when there was an exact
identification of the same subtype by CALIPSO. In many cases the
complexity of the CALIPSO scene makes it almost impossible to assign one
aerosol type to each height bin, though in the case of strong features, such as
dust and polluted dust, the assignment is easier. In the case of complex
aerosol scenes, we simply omitted the profiles when more than one
subtype was identified with the same frequency. Keeping this
prerequisite of simultaneous identification, the number of available
samples was reduced.
Mean lidar ratio at 532 nm for the different aerosol
subtypes as measured by EARLINET sites and corresponding statistical
parameters. The last column refers to the lidar ratio values assumed by
CALIPSO and their associated lidar ratio distributions (mean plus standard
deviation). M stands for marine, MM for mixed marine, D for dust, PC for
polluted continental, CC for clean continental, PD for polluted dust, and S
for smoke subtype. Note that, here, the M subtype corresponds to pure marine
particles
EARLINET
CALIPSO
Aerosol type
Mean ± SD [sr]
Range [sr]
Median [sr]
# Samples
Mean ± SD [sr]
M
23±3
21–24
22
5
20±6
MM
33±5
25–38
34
8
-
D
51±10
41–73
48
16
40±20
PC
62±10
51–78
61
14
70±25
CC
47±4
44–52
46
4
35±16
PD
53±14
35–78
49
13
55±22
S
67±10
54–80
65
11
70±28
The EARLINET mean lidar ratio for the selected types is summarized in
Table along with the corresponding
lidar ratio values (rightmost column) used by CALIPSO
e.g.,. For the smoke
subtype the mean EARLINET measured lidar ratio value is 67 ± 10 sr and it compares well with the assignment made by the
CALIPSO classification scheme, which is Saer = 70 ± 28 sr. The marine lidar ratio is 23 ± 3 sr and
agrees also well with the Saer = 20 ± 6 sr of the
CALIPSO scheme. In this case, only pure marine layers over the
stations are considered, so that the agreement is expected. This study,
also, estimated a mean lidar ratio for mixed marine particles of 33±5 sr, which is consistent with values reported in the literature
e.g.,. The CALIPSO typing
scheme does not incorporate marine mixtures in a separate subtype
as denoted in Sect. ,
therefore a comparison is not feasible. The clean continental
subtype assignment is not a straightforward procedure for the EARLINET
sites, as the aerosol layer classification depends strongly on the
rejection of the other types . The mean EARLINET
lidar ratio is 45 ± 4 sr and deviates from the assumed
CALIPSO Saer = 35 ± 16 sr. For interpreting these
results, one should take into account that the clean continental type
in the CALIPSO scheme is intended as the background aerosol and as
a consequence, deemed not to be influenced by urban pollution. However,
these conditions are probably not realistic for the European
continent. The EARLINET lidar ratio values measured for these cases
seem to indicate that the cases flagged as clean continental are
affected by absorbing particles of anthropogenic nature. For the
polluted continental, the mean EARLINET value is 62 ± 10 sr, and is in fair agreement with the CALIPSO
Saer = 70 ± 25 sr considering the variability of
this subtype. It is most likely that the presence of marine particles
over the Mediterranean area influences the mean lidar ratio value for
this category. This effect was described by and
, where the marine particles can act as an external
mixture and reduce linearly the lidar ratio values.
The EARLINET lidar ratio value for dust is 51 ± 10 sr and
is higher than the CALIPSO Saer = 40 ± 20 sr,
however comparable considering the variability of the parameter, even
in the lower limits of the standard deviation. The measured lidar
ratio is in accordance with other studies e.g.,
and field experiments on dust sources e.g.,. Moreover, the mean EARLINET lidar ratio
exceeded the CALIPSO modeled value for all the examined
sites. Typically, the source region of the dust outbreaks is the
Western Saharan region where according to numerous studies
e.g., lidar ratio
at 532 nm is around 55–58 sr.
The mean polluted dust lidar ratio is 53 ± 14 sr and is in
good agreement with the Saer = 55 ± 22 sr used in
the CALIPSO retrievals, however the lidar ratio varies significantly
with location. The lidar ratio value assumed by CALIPSO for polluted
dust seems to be appropriate for continental sites as Leipzig,
Saer = 52 ± 8 sr. A fair agreement is observed
also for a southern Europe continental site such as Potenza, even if
the mean value is greater than the CALIPSO lidar ratio,
Saer = 64 ± 15 sr. For all the other sites, the
mean lidar ratio values stay below the CALIPSO assumed value of
55 sr – for Granada Saer = 45 ± 11 sr, for
Évora Saer = 42 ± 9 sr, and for Naples
Saer = 38 ± 15 sr. The main reason of this
divergence is the presence of marine particles in the mixture, which
are not taken into account for the CALIPSO polluted dust category
. These results underline the large variability of the
polluted dust lidar ratio and its dependence on the mixture of
particles.
Assessing the impact of lidar ratio
In the light of the disparity observed in the lidar ratios of clean
continental, dust, and polluted dust subtypes, we assessed the impact of
introducing the calculated EARLINET values into the CALIPSO extinction
retrieval. Hence, the lidar ratio values of the subtypes of dust, polluted
dust, and clean continental are set to Saer = 51 sr,
Saer = 53 sr, and Saer = 47 sr, respectively. The
CALIPSO typing data coming from the Vertical Feature Mask are weighted
according to the alternative lidar ratio values and they are multiplied by
the respective backscatter coefficient to estimate the extinction profiles.
Figure summarizes the columnar mean relative differences
between the CL3* extinction profiles and the lidar ratio corrected CL3*
profiles for each aerosol subtype (i.e., clean continental, polluted dust,
dust) and the combination of them.
Mean relative differences between CL3* and the corrected CL3*
extinction coefficient. The corrected CL3* extinction coefficient is
retrieved when introducing the EARLINET-estimated lidar ratio for clean
continental (CC), dust (D), and polluted dust (PD) subtypes as well as for
the category Combined (CC + D + PD).
The rate of the change caused by the adjustment of the lidar ratio
depends on the observations frequency of the aerosol subtype and on
the backscattering intensity of each feature. By this, we highlight
that the almost 10 sr increase of the clean continental lidar
ratio produces an extinction increase of less than 1 %, whilst the
use of 53 sr instead of 55 sr for the polluted dust
creates a decrease of about 3 %. Consequently, the clean
continental lidar ratio inference produces an almost insignificant
change in the extinction profile, whereas for the polluted dust, small
difference in lidar ratio value leads to small underestimation of the
extinction retrieval. Moreover, we should consider that this subtype
is systematically overused by CALIPSO and,
therefore, the impending re-typing of the wrongly flagged polluted
dust features will lead to an increase of the dust, polluted
continental fraction, which will affect the lidar ratio. The potential
improvement of the CALIPSO dust retrievals by using a dust lidar ratio
of 51 sr produced a 5 % increase, confirming that
a regional correction and spatial constant value can enhance the
extinction retrievals .
In synthesis, we observed that, even if the aerosol layer is perfectly
identified, the retrieved extinction is affected by the input value of
lidar ratio as, in many cases, it might not represent the local
aerosol situation. The latter is also the outcome of previous
studies e.g.,,
concluding that the usage of incorrect lidar ratio would lead to
errors in the AOD . Here, we suggest regional-corrected values of lidar ratio to improve the CALIPSO extinction
retrieval based on independent, range-resolved lidar ratio profiles
measured on a continental scale.
Conclusions
The comparison of CALIPSO to advanced ground-based lidar systems is
essential to understand whether CALIPSO measurements are representative of
the corresponding station-surrounding area in a climatological sense
and if there are systematic deviations due to assumptions in the
CALIPSO retrievals. CL3* data were compared against EARLINET monthly
averages obtained by profiles measured during satellite
overflights. CALIPSO monthly profiles yielded lower extinction values
comparing to EARLINET ones. A total mean AOD difference of -0.05 was
found. There are many possible reasons for the observed differences,
of which the most important are: difference in sampling volumes and
the spatial variability of the aerosol fields, problems/limitations
with the CALIPSO measurements, and uncertainty of the CALIPSO
assumptions. A mean relative difference of 18 % was found for
the aerosol backscatter coefficient, while a considerably larger
difference – 25 % – was obtained for the extinction
coefficient. The better agreement on backscatter has to be ascribed to
the higher impact of lidar ratio assumption on extinction rather than
on backscatter. Observe that the improvement in the
backscatter comparison is mainly associated to the low troposphere
where both the CALIPSO typing and the lidar ratio inference are more
complex.
The comparison on aerosol typing showed a robust identification of
dust subtype demonstrating the good performance of the CALIPSO
polarization-sensitive observations that facilitate the correct
identification of irregular shaped particles. A CALIPSO overestimation
of the polluted dust subtype was identified and it was found to be
most prominent in the lowest height ranges. This reflects the effects
of a known bug suggesting that a part of the aerosol loading will be
reclassified as polluted continental or smoke, and hence will enhance
the corresponding extinction estimates. The polluted and clean
continental subtypes produced the poorest agreement. The polluted
continental disparity of the data sets, typically in the regions
adjoining the PBL, affects the extinction retrievals and can be
attributed to the CALIPSO polluted dust overuse as well as to the
local aerosol content. The clean continental subtype is the least
encountered aerosol type observed and it characterizes the typical
aerosol background conditions over the stations. In most of the cases,
the minimum levels of the signal-to-noise ratio needed to retrieve the
extinction coefficient for this aerosol subtype is not met by the
EARLINET systems. The marine particles by the CALIPSO classification
scheme are surface-dependent, and furthermore no mixing with other
aerosol types is considered. On the other side, according to the
EARLINET observations, the presence of marine particles mixed to other
types (i.e., smoke, polluted continental) is a common situation over
the Mediterranean Sea.
A type-by-type comparison of CALIPSO modeled against EARLINET measured
lidar ratio was carried out. The most notable differences were found
for the clean continental, dust, and polluted dust subtypes. The mean
clean continental EARLINET lidar ratio was 47±4 sr and
diverges about 10 sr from the modeled value. In the CALIPSO
scheme, this aerosol subtype is intended as the background aerosol and
deemed not to be influenced by continental pollution, whereas these
conditions are unlikely in a highly populated region as Europe. The
dust EARLINET lidar ratio value is 51±10 sr and is
greater than the CALIPSO 40 sr, highlighting the low CALIPSO
lidar ratio inference. The mean polluted dust lidar ratio was 53±14 sr and is in good agreement with the 55 sr used in
the CALIPSO retrieval codes. However, the EARLINET sites in the
Mediterranean area indicate the existence of mixtures with marine
particles that are not accounted for in the CALIPSO polluted dust
subtype.
In accordance to previous studies, we have quantitatively shown the
improvement of CALIPSO product by adjusting the assumed lidar ratio
values taking as reference the corresponding EARLINET
measurements. Based on our findings, we suggest the regional tuning of
the dust lidar ratio. Marine particles should be taken into account in
the polluted dust subtype, at least in areas like the Mediterranean,
where the flow of these particles inland change the composition
affecting the CALIPSO optical properties retrieval. The correction of
the space-based extinction retrieval enhanced the climatic relevant
AOD about 3 % regionally. Generally, the backscatter comparison
showed a better agreement with respect to the extinction comparison;
hence backscatter could be coupled in the CL3 files offering more
robust data – for instance, for model validation and climatological
studies.