Airborne high spectral resolution lidar for measuring
aerosol extinction and backscatter coefficients
Michael Esselborn,
1,
* Martin Wirth,
1
Andreas Fix,
1
Matthias Tesche,
2
and Gerhard Ehret
1
1
Institut für Physik der Atmosphäre, Deutsches Zentrum für Luft- und Raumfahrt (DLR), Oberpfaffenhofen,
82234 Wessling, Germany
2
Leibniz-Institut für Troposphärenforschung (IFT), 04318 Leipzig, Germany
*Corresponding author: Michael.Esselborn@dlr.de
Received 15 August 2007; revised 3 December 2007; accepted 3 December 2007;
posted 4 December 2007 (Doc. ID 86422); published 14 January 2008
An airborne high spectral resolution lidar (HSRL) based on an iodine absorption filter and a high-power
frequency-doubled Nd:YAG laser has been developed to measure backscatter and extinction coefficients
of aerosols and clouds. The instrument was operated aboard the Falcon 20 research aircraft of the
German Aerospace Center (DLR) during the Saharan Mineral Dust Experiment in May–June 2006 to
measure optical properties of Saharan dust. A detailed description of the lidar system, the analysis of its
data products, and measurements of backscatter and extinction coefficients of Saharan dust are pre-
sented. The system errors are discussed and airborne HSRL results are compared to ground-based
Raman lidar and sunphotometer measurements. © 2008 Optical Society of America
OCIS codes: 010.0280, 010.1280, 280.1100, 280.3640, 290.2200, 290.5850.
1. Introduction
Aerosols play a key role in the Earth’s radiative bud-
get because they directly influence the fluxes of solar
and terrestrial radiation within the atmosphere by
absorption and scattering of light [1]. The quantifi-
cation of this effect accounts for accurate and highly
resolved measurements of aerosol extinction and ver-
tical distribution. Using a conventional backscatter
lidar, aerosol extinction cannot be measured directly.
Only with assumption of the aerosol extinction-to-
backscatter ratio, the so-called lidar ratio, aerosol
extinction coefficients can be retrieved by means of
inversion algorithms [2]. However, the aerosol lidar
ratio is a highly variable quantity, so that large errors
in the retrieval must be expected if the lidar ratio is
not known exactly. The only alternate lidar method to
measure aerosol extinction profiles directly is Raman
lidar [3]. The long averaging times associated with
the Raman method, however, have prevented this
technique from being used in airborne operation so
far. Recently, ground-based demonstration measure-
ments of the lidar ratio and other quantities with a
Raman lidar specially developed for airborne opera-
tion have been reported [4]. A high spectral resolution
lidar (HSRL) takes advantage of the different spec-
tral broadening of light, backscattered by molecules
and aerosols. By means of a narrow bandwidth opti-
cal filter the aerosol contribution is separated from
the molecular backscatter. Thus, aerosol backscatter
and extinction coefficients can be measured directly
and no assumption about the lidar ratio is required.
Up to now, only a few HSRL instruments have been
proposed and successfully implemented to measure
atmospheric parameters or particle properties. These
utilize either narrow bandwidth interferometers [5]
to reject aerosol scattering or atomic [6 –8] and mo-
lecular [8–10] vapor filters. The advantages of iodine
vapor filters are the strong rejection of aerosol back-
scatter at low cell temperatures and the marginal
sensitivity to optical alignment and angular diver-
gence of the backscattered light. Furthermore, they
can be designed in compact dimensions and do not
have to be pressure stabilized. Thus, iodine vapor
filters are ideal candidates for airborne HSRL oper-
ation. Airborne HSRL measurements could be dem-
onstrated by our group during the Lindenberg
Aerosol Characterization Experiment (LACE) field
campaign in 1998 using a 10 Hz single-mode Nd:YAG
0003-6935/08/030346-13$15.00/0
© 2008 Optical Society of America
20 January 2008 兾 Vol. 47, No. 3 兾 APPLIED OPTICS 346
laser system aboard the German Aerospace Center
(DLR) Falcon research aircraft [11]. This system was
limited by the relatively low horizontal sampling ca-
pability, low aerosol attenuation within the single-
pass iodine vapor cell and unstable laser frequency
control. Recently, another HSRL based on a 200 Hz
low-power Nd:YAG laser has been deployed during
two field experiments aboard the NASA B200 King
Air [12].
Here, we present a new airborne HSRL, which was
applied during the Saharan Mineral Dust Experi-
ment (SAMUM) field campaign in May–June 2006 to
measure cross sections of extinction and backscatter
coefficients of pure Saharan dust close to its source
regions for the first time to the best of our knowledge.
This system is operated at 100 Hz using a high-power
Nd:YAG laser, which is frequency controlled by
means of a new robust laser frequency stabilization.
For stronger aerosol suppression a new iodine vapor
filter in dual-pass configuration has been designed.
The development of the HSRL system was partially
driven by two forthcoming satellite missions of the
European Space Agency (ESA). The Atmospheric Dy-
namics Mission (ADM [13]) will use a Doppler lidar,
which is expected to also have potential for extinction
profiling [14]. The EarthCARE mission [15] will em-
ploy a HSRL instrument at 355 nm. The airborne
HSRL measurements can be of great help to evaluate
the potential and performance of the spaceborne in-
struments in advance. For validation purposes, air-
borne measurements can be coordinated and allow
nearly real-time spatial overlap along the satellite
footprint.
The purpose of this paper is a detailed description
of our new airborne HSRL system, the retrieval
method of its data products, an error analysis, and a
presentation of measurement examples together with
a comparison to ground-based Raman lidar and
sunphotometer data. A detailed scientific analysis of
the campaign’s results will be the subject of a forth-
coming paper.
2. HSRL Theory and Data Retrieval
The elastic scattering of light by particles small com-
pared to the wavelength is usually described by Ray-
leigh scattering theory [16]. Besides the elastic part,
atmospheric Rayleigh scattering [17] includes the fre-
quency shifted rotational Raman bands associated
with ⌬J ⫽⫾2 transitions, where J is the rotational
quantum number of the scattering molecule. The
dominant central part of the backscatter spectrum,
which arises from elastic scattering together with the
unshifted ⌬J ⫽ 0 rotational Raman branch make up
the Cabannes line [18], which shows pressure and
temperature-dependent Brillouin sidebands [19]. At
atmospheric temperatures close to 300 K the Doppler
broadening of the Cabannes line amounts to 2.6 GHz
for green light with a wavelength of 532 nm. In con-
trast, aerosol backscatter is hardly broadened due to
the relatively slow motion of aerosol particles, so that
it can be characterized by the laser frequency distri-
bution.
Using HSRL, the received atmospheric backscatter
is split into two channels. The narrow bandwidth
optical filter in the molecular channel suppresses the
aerosol backscatter, whereas the combined channel
detects the intensity of both aerosol and molecular
backscatter. Therefore the emitted laser frequency
must be tuned to match the filter absorption line. The
iodine absorption filter eliminates the aerosol back-
scatter and transmits the wings of the Doppler broad-
ened molecular backscatter spectrum. To determine
the amount of molecular backscatter absorbed by the
iodine filter, the HSRL system needs to be calibrated.
This is done by measuring the filter transmission
spectrum and calculating the atmospheric tempera-
ture and pressure-dependent filter transmission with
an appropriate molecular backscatter model. For
measuring the iodine filter transmission spectrum, a
highly attenuated reflection of the pulsed green laser
emission is directed through the receiver assembly
and the laser frequency is scanned. The filter trans-
mission is determined by the product of the iodine
filter transmission and the calculated molecular
backscatter spectrum. In our experiments the trans-
mitted fraction of the molecular backscatter exceeds
30% for atmospheric temperatures higher than 200 K
using iodine absorption line 1109 (line notation fol-
lows Ref. [20]). The partial transmission
m
共r兲 of
the temperature- and pressure-dependent Cabannes
spectrum R共, T, p兲 through the iodine vapor filter
with transmission function 共兲 may be written as
m
共
T, p
兲
⫽
冕
共
兲
冕
R
共
⬘, T, p
兲
l
共
⫺⬘
兲
d⬘d
冕
R
共
⬘, T, p
兲
l
共
⫺⬘
兲
d⬘
, (1)
with l共兲 being the laser spectrum. Figure 1 shows the
measured transmission of the iodine filter cell oper-
ated at a vapor pressure of 53 Pa as a function of
frequency offset from the absorption maximum. The
central absorption line is iodine line 1109. The mo-
Fig. 1. Measured iodine transmission spectrum at 563.244 THz
together with molecular backscatter spectrum calculated with the
S6-Tenti model for an atmospheric temperature of 300 K and
pressure of 1000 hPa.
347 APPLIED OPTICS 兾 Vol. 47, No. 3 兾 20 January 2008
lecular backscatter spectrum was calculated with the
S6 code provided by Tenti et al. [21] using a molecular
mass of 28.8 kg兾kmol for air molecules. Due to the
lack of experimentally verified data for air molecules,
the molecular parameters for the ratio of shear vis-
cosity to bulk viscosity and the ratio of shear viscosity
to thermal conductivity of nitrogen [22] have been
used. The result is given for an atmospheric temper-
ature of 300 K and pressure of 1000 hPa. The trans-
mitted fraction of molecular backscatter is the ratio
of the integrals of the molecular spectrum before
and after filtering. In the case of T ⫽ 300 K, p ⫽
1000 hPa the fraction amounts to
m
⫽ 0.430.
Depending on the bandwidth of the background
filter, spectral components arising from pure rota-
tional Raman scattering have to be considered. Fig-
ure 2 shows the measured transmission function of
the interference filter and the rotational Raman spec-
trum as a function of the frequency offset from the
incident laser wavelength. The redshifted Stokes and
the blueshifted anti-Stokes branch are denoted by S
and O, respectively. The full width at half-maximum
(FWHM) of the interference filter is 1095 GHz. As
can be seen, the first Stokes and anti-Stokes Raman
line of oxygen and the first two lines of nitrogen are
included within the FWHM of the filter. The central
Q branch of the rotational Raman spectrum adds to
the Cabannes line R共, T, p兲 and has the same spec-
tral broadening. Thus, it does not affect the calcula-
tion of
m
. The shifted O and S branches contribute
only 2.5% of the whole Rayleigh cross section. Their
partial transmission through the background inter-
ference filter can be calculated. In our case, 5.7% of
the shifted rotational Raman scattering intensity is
transmitted through the filter. Raman scattering
contributes less than 0.5% to the molecular channel
intensity assuming that at least 1兾3 of the total
Cabannes scattering is detected and the Raman
bands are not attenuated by the iodine absorption
spectrum. Therefore the small impact of Raman scat-
tering on the backscatter calculation is neglected.
The other quantity measured through calibration
is the transmission of aerosol backscatter in the mo-
lecular channel. This quantity decreases with in-
creasing vapor pressure and optical path length
within the filter cell. Assuming no spectral broaden-
ing, the aerosol backscatter spectrum can be de-
scribed by the laser frequency distribution. The
laser’s spectral width of ⬇75 MHz is small compared
to the bandwidth of the iodine filter, which is ⬇2 GHz
at typical filter conditions. Thus, the transmitted
fraction of aerosol backscatter
a
through the filter
can be regarded as the filter transmission at line
center:
a
⫽
共
0
兲
. (2)
The absorption within the iodine filter cell at line
center is usually very strong and the experimental
determination of
a
is restricted by detector noise or
stray light. With a model provided by Forkey et al.
[23,24], iodine absorption spectra at 532 nm can be
calculated for different cell lengths and tempera-
tures. For our HSRL experiments the calculated
aerosol transmission is typically of the order of 10
⫺6
⬍
a
⬍ 10
⫺5
. As will be shown in Subsection 3.B, our
receiver module is laid out for polarization sensitive
detection. At first, the received atmospheric back-
scatter is split into its parallel (superscript 㛳) and
cross-polarized (superscript ⬜) components. The
parallel-polarized channel is split again into the
combined channel (subscript C ) and the molecular
channel (subscript M). The received power from dis-
tance in the three measurement channels can be
written as
P
C
储
共
r
兲
⫽
C
储
E
0
储
c
2
A
r
2
2
共
r
兲
关

m
储
共
r
兲
⫹
a
储
共
r
兲
兴
, (3)
P
M
储
共
r
兲
⫽
M
储
E
0
储
c
2
A
r
2
2
共
r
兲
关
m
共
r
兲

m
储
共
r
兲
⫹
a

a
储
共
r
兲
兴
, (4)
P
⬜
共
r
兲
⫽
⬜
E
0
储
c
2
A
r
2
2
共
r
兲
关

m
⬜
共
r
兲
⫹
a
⬜
共
r
兲
兴
, (5)
where denotes the channel efficiency, c is the speed
of light, A is the telescope aperture; E
0
储
is the pulse
energy,
2
共r兲 is the two-way atmospheric transmis-
sion over range r from the lidar transmitter to the
scattering event, and 
m,a
储,⬜
共r兲 denotes the parallel
and perpendicular backscatter coefficients of mole-
cules and aerosols. Note that
m
共r兲 is a function of
height due to its dependence on atmospheric temper-
ature and pressure. The atmospheric transmission is
composed of a molecular and an aerosol contribution,
2
共
r
兲
⫽
m
2
共
r
兲
a
2
共
r
兲
⫽ exp
冋
⫺2
冕
0
r
共
␣
m
共
r⬘
兲
⫹␣
a
共
r⬘
兲兲
dr⬘
册
, (6)
where ␣
m,a
共r兲 denotes the molecular and aerosol ex-
tinction coefficient. Molecular and aerosol absorption
effects can be neglected at the used lidar wavelength
of 532 nm. Following Rayleigh scattering theory [17],
␣
m
共r兲 is strictly proportional to 
m
共r兲,
Fig. 2. Measured interference filter transmission and calculated
rotational Raman spectrum at 563.244 THz.
20 January 2008 兾 Vol. 47, No. 3 兾 APPLIED OPTICS 348
␣
m
共
r
兲
⫽
8
3
共
45 ⫹ 10⑀
兲
共
45 ⫹ 7⑀
兲

m
共
r
兲
, (7)
where ⑀共532 nm兲 ⫽ 0.222 accounts for molecular an-
isotropy [25]. 
m
共r兲 is given by the product of the
differential backscattering cross section ⭸共兲兾⭸⍀
and the number density N共r兲 of the scatterers [26]

m
共
r
兲
⫽
⭸
共
兲
⭸⍀
N
共
r
兲
, (8)
where N共r兲 can be calculated from measured tem-
perature and pressure data or from standard
atmosphere profiles. The Rayleigh scattering proper-
ties of the atmospheric gas composition are well-
known and tabulated in the literature [25,27]. The
total Rayleigh scattering cross section at 532 nm
amounts to ⫽5.16 ⫻ 10
⫺31
m
2
as interpolated from
data given in Ref. [27]. The differential backscatter-
ing cross section of the Cabannes line only amounts to
⭸共兲兾⭸⍀ ⫽ 5.93 ⫻ 10
⫺32
m
2
sr
⫺1
at 532 nm. How-
ever, the effectively detected cross section depends on
the spectral and polarization properties of the re-
ceiver because of the depolarization of the rotational
Raman lines and their spectral separation. As al-
ready stated the rotational Raman lines partially
transmit through the interference filter. For the
background interference filter and the polarization
beam splitters in our receiver the value of the differ-
ential backscattering cross section reduces to 5.91
⫻ 10
⫺32
m
2
sr
⫺1
.
To determine the amount of total (parallel and
cross-polarized) backscatter, which is required for de-
polarization data processing, the measured powers in
the parallel-polarized combined channel and the
cross-polarized channel has to be calculated:
P
T
共
r
兲
⫽ P
C
储
⫹
C
储
⬜
P
⬜
共
r
兲
. (9)
To determine the relative sensitivity factor
C
储
兾
⬜
a
calibration during the measurement flight is re-
quired. Therefore the receiver module is rotated by
an angle of 45°, such that the plane of polarization of
both the parallel and the cross-polarized channel is
rotated by 45° relative to the transmitter’s plane of
polarization.
C
储
兾
⬜
is then calculated from the ratio
of the signal intensities in the parallel and cross-
polarized channel. To obtain the aerosol extinction
and backscatter coefficients from the measured sig-
nals given in Eqs. (3)–(5) and (9), the signals are
multiplied by r
2
for range correction and divided by
关
m
2
共r兲 ⫻
m
共r兲兴 to account for molecular extinction
and backscatter, respectively. Moreover, the signals
are normalized to the power received from a region at
distance r
0
where the aerosol concentration can be
neglected, i.e., 
a
共r
0
兲⬇0, whereby all constants can-
cel out. Hence, an attenuated backscatter ratio R,
which is the backscatter ratio 关1 ⫹
a
共r兲兾
m
共r兲兴 mul-
tiplied by aerosol transmission is defined as
R
C
储
共
r
兲
⫽
P
C
储
共
r
兲
C
C
储
r
2
m
2
共
r
兲

m
储
共
r
兲
⫽
a
2
共
r
兲
冋
1 ⫹

a
储
共
r
兲

m
储
共
r
兲
册
, (10)
R
M
储
共
r
兲
⫽
P
M
储
共
r
兲
C
M
储
r
2
m
2
共
r
兲

m
储
共
r
兲
⫽
a
2
共
r
兲
冋
m
共
r
兲
⫹
a

a
储
共
r
兲

m
储
共
r
兲
册
,
(11)
R
T
共
r
兲
⫽
P
T
共
r
兲
C
T
r
2
m
2
共
r
兲

m
T
共
r
兲
⫽
a
2
共
r
兲
冋
1 ⫹

a
T
共
r
兲

m
T
共
r
兲
册
, (12)
where 
m,a
T
共r兲 denotes the total (parallel and cross-
polarized) molecular and aerosol backscatter coeffi-
cients. The constant C
C,M
储,T
for each channel is chosen
to set R
C
储
共r
0
兲 and R
T
共r
0
兲 close to 1 and R
M
储
共r
0
兲 to the
value of
m
共r
0
兲. The constant is composed of C
C,M
储,T
⫽
C,M
储,T
E
0
储
A共c兾2兲 and calculated for each profile.
Equations (10) and (11) are used to express the two-
way aerosol transmission by the two measured quan-
tities and the calibration constants:
a
2
共
r
兲
⫽
R
M
储
共
r
兲
⫺
a
R
C
储
共
r
兲
m
共
r
兲
⫺
a
. (13)
The aerosol extinction coefficient ␣
a
共r兲 follows from
aerosol optical thickness (AOT) t
a
共r兲 by differentia-
tion:
t
a
共
r
兲
⫽⫺
1
2
ln
关
a
2
共
r
兲
兴
, (14)
␣
a
共
r
兲
⫽
⭸
⭸r
t
a
共
r
兲
. (15)
To calculate the numerical derivative of the AOT pro-
file, a Savitzky–Golay filter [28] of first order is used.
This is equivalent to a linear regression on a specified
number of data points within a moving window. The
slope of the fitted straight line in each point is taken
as the value of the derivative. This differentiation
method tends to preserve slopes and edges in the
AOT profile, which would be flattened by adjacent
averaging techniques.
To determine the total aerosol backscatter coeffi-
cient 
a
T
共r兲 ⫽
a
储
共r兲 ⫹
a
⬜
共r兲 from the three measured
signals given in Eqs. (3)–(5), the aerosol depolariza-
tion ratio ␦
a
共r兲 has to be analyzed first. Therefore the
equation given in [29] is used:
␦
a
共
r
兲
⫽

a
⬜
共
r
兲

a
储
共
r
兲
⫽
关
1 ⫹␦
m
兴
␦
v
共
r
兲
R
T
共
r
兲
兾
a
2
共
r
兲
⫺
关
1 ⫹␦
v
共
r
兲
兴
␦
m
关
1 ⫹␦
m
兴
R
T
共
r
兲
兾
a
2
共
r
兲
⫺
关
1 ⫹␦
v
共
r
兲
兴
,
(16)
where ␦
m
denotes the volume depolarization ratio of
molecular backscatter and ␦
v
共r兲 the volume depolar-
ization ratio of total backscatter defined as
349 APPLIED OPTICS 兾 Vol. 47, No. 3 兾 20 January 2008
␦
v
共
r
兲
⫽
C
储
⬜
P
⬜
共
r
兲
P
C
储
共
r
兲
. (17)
The value of ␦
m
is highly dependent on the amount of
detected pure rotational Raman scattering and varies
between 3.63 ⫻ 10
⫺3
and 1.43 ⫻ 10
⫺2
depending on
the detection of the Cabannes line alone or the full
inclusion of the pure rotational Raman bands, res-
pectively [30]. As has been discussed before, our in-
terference filters partially transmit the rotational
Raman spectrum and the value of ␦
m
has been calcu-
lated to 6.8 ⫻ 10
⫺3
. Finally, the total aerosol back-
scatter coefficient can be expressed by

a
T
共
r
兲
⫽
冋
R
C
储
共
r
兲
a
2
共
r
兲
⫺ 1
册

m
储
共
r
兲共
1 ⫹␦
a
共
r
兲兲
. (18)
The aerosol lidar ratio S
a
共r兲 is defined as the ratio
of the aerosol extinction coefficient to the backscatter
coefficient:
S
a
共
r
兲
⫽
␣
a
共
r
兲

a
T
共
r
兲
. (19)
Unlike the constant molecular lidar ratio, the aerosol
lidar ratio generally varies with height because it
depends on the aerosol size distribution, its shape,
and chemical composition [31,32].
3. System Description
A. Transmitter
The HSRL described in this paper has been devel-
oped as an extension of the existing airborne DLR
water vapor differential absorption lidar (DIAL) sys-
tem [33]. The laser transmitter used for the HSRL
measurements is a high-power, Q-switched Nd:YAG
laser in a master-oscillator power amplifier configu-
ration. It consists of a low-power master oscillator
and three power amplifiers yielding a pulse energy of
220 mJ at 1064 nm and a repetition rate of 100 Hz. A
schematic of the HSRL transmitter is shown in Fig. 3.
A detailed description of the laser system can be
found in Ehret et al. [34]. To obtain single longitudi-
nal mode operation the master oscillator is injection
seeded by a monolithic Nd:YAG ring laser and stabi-
lized by minimizing the Q-switch build-up time [35].
The seed laser frequency is tuned by changing the
Nd:YAG crystal temperature. The fundamental radi-
ation of the pulsed laser is frequency-doubled using a
temperature stabilized KTP crystal yielding a pulse
energy of 110 mJ. An attenuated reflection of the
green radiation is directed to a frequency stabilizer
that continuously controls the temperature of the
seed laser. The stabilization method is based on
acousto-optic modulation and locks the green laser
radiation to a Doppler-broadened iodine absorption
line. The advantage of this frequency stabilization
technique is the monitoring and stabilization of the
amplified laser output instead of the seed laser fre-
quency. With this new method long-term frequency
fluctuations have been reduced to ⌬ ⫽ 4.8 MHz or in
relative terms ⌬兾⫽8.5 ⫻ 10
⫺9
during airborne
operation. Table 1 summarizes the system proper-
ties.
B. Receiver
The atmospheric backscatter is collected by means of
a 350 mm Cassegrain telescope. A field stop within
the focal plane of the telescope limits the acceptance
angle to 1 mrad. Dichroic beam splitters are used to
spectrally separate the backscatter signals at 1064
and 532 nm. The part of the receiver that detects
atmospheric backscatter at 532 nm is schematically
shown in Fig. 4. The received light is split into its
polarized components with a polarizing beam split-
ter cube. The cross-polarized component is directed
to photomultiplier PMT3 whereas the parallel-
polarized component is transmitted. In both paths a
second polarizing beam splitter is placed to reduce
polarization cross talk. The transmitted parallel com-
ponent of the backscatter signal is split again, such
that half of its energy is directed into the molecular
channel.
Fig. 3. Schematic of the transmitter system.
Table 1. System Parameters of the HSRL System
MOPA Laser system
Pulse energy at 1064 nm 220 mJ
Repetition rate 100 Hz
Pulse width at 1064 nm 18 ns
Linewidth at 1064 nm ⬍75 MHz
Laser divergence at 1064 nm 0.5 mrad
SHG
Pulse energy at 532 nm 100 mJ
Frequency fluctuations ⬍5 MHz
Spectral purity 99.96%
Receiver
Telescope diameter 350 mm
Background filter bandwidth 1 THz
Iodine absorption bandwidth 2 GHz
Optical iodine filter length 380 mm
Detector divergence 1 mrad
PMT quantization 14 bits
PMT sampling rate 10 MHz
20 January 2008 兾 Vol. 47, No. 3 兾 APPLIED OPTICS 350
