(Open Access) Aerosol lidar intercomparison in the framework of the EARLINET project. 3. Raman lidar algorithm for aerosol extinction, backscatter, and lidar ratio (2004) | Gelsomina Pappalardo

Aerosol lidar intercomparison in the framework of

the EARLINET project. 3. Raman lidar algorithm

for aerosol extinction, backscatter, and lidar ratio

Gelsomina Pappalardo, Aldo Amodeo, Marco Pandolﬁ, Ulla Wandinger,

Albert Ansmann, Jens Bo¨ senberg, Volker Matthias, Vassilis Amiridis,

Ferdinando De Tomasi, Max Frioud, Marco Iarlori, Leonce Komguem,

Alexandros Papayannis, Francesc Rocadenbosch, and Xuan Wang

An intercomparison of the algorithms used to retrieve aerosol extinction and backscatter starting from

Raman lidar signals has been performed by 11 groups of lidar scientists involved in the European Aerosol

Research Lidar Network 共EARLINET兲. This intercomparison is part of an extended quality assurance

program performed on aerosol lidars in the EARLINET. Lidar instruments and aerosol backscatter

algorithms were tested separately. The Raman lidar algorithms were tested by use of synthetic lidar

data, simulated at 355, 532, 386, and 607 nm, with realistic experimental and atmospheric conditions

taken into account. The intercomparison demonstrates that the data-handling procedures used by all

the lidar groups provide satisfactory results. Extinction proﬁles show mean deviations from the correct

solution within 10% in the planetary boundary layer 共PBL兲, and backscatter proﬁles, retrieved by use of

algorithms based on the combined Raman elastic-backscatter lidar technique, show mean deviations from

solutions within 20% up to 2 km. The intercomparison was also carried out for the lidar ratio and

produced proﬁles that show a mean deviation from the solution within 20% in the PBL. The mean value

of this parameter was also calculated within a lofted aerosol layer at higher altitudes that is represen-

tative of typical layers related to special events such as Saharan dust outbreaks, forest ﬁres, and volcanic

OCIS codes: 010.3640, 280.1100, 290.1090, 290.5860, 290.2200.

1. Introduction

Atmospheric aerosols, which originate both from

natural sources and from human intervention, con-

siderably affect the Earth’s radiation balance,

1,2

and they are considered to be one of the major

sources of uncertainty in climate forcing predic-

tions. For this reason, information on their global

distribution, their vertical and horizontal extent,

and their time of residence in the atmosphere are

necessary.

G. Pappalardo 共pappalardo@imaa.cnr.it兲, A. Amodeo, and M.

Pandolﬁ are with the Istituto di Metodologie per l’Analisi Ambi-

entale, Consiglio Nazionale delle Richerche, Contrada S. Loja,

85050 Tito Scalo 共Potenza兲, Italy. U. Wandinger and A. Ansmann

are with Institute for Tropospheric Research, Permoserstrasse 15,

04303 Leipzig, Germany. When this research was conducted, J.

Bo¨senberg and V. Matthias were with the Max-Planck-Institut fu¨r

Meteorologie, Bundesstrasse 55, 20146 Hamburg, Germany; V.

Matthias is now with Institute for Coastal Research, GKSS Re-

search Center, Max-Planck-Strasse 1, 21502 Geesthacht, Ger-

many. V. Amiridis is with the Laboratory of Atmospheric

Physics, Aristotle University of Thessaloniki, Box 149, 54124 Thes-

saloniki, Greece. F. De Tomasi is with the Istituto Nazionale per

la Fisica della Materia, Dipartimento di Fisica, Universita`di

Lecce, via Arnesano, 73100 Lecce, Italy. M. Frioud is with the

Observatory of Neuchaˆtel, Rue de l’Observatoire 58, 2000

Neuchaˆtel, Switzerland. M. Iarlori is with the Dipartimento di

Fisica, Universita` degli Studi–L’Aquila, via Vetoio Loc. Coppito,

67010 L’Aquila, Italy. L. Komguem is with the Department of

Physics, University of Wales, Aberystwyth Ceredigion SY23 3BZ,

United Kingdom. A. Papayannis is with the Department of Phys-

ics, National Technical University of Athens, Heroon Polytechniou

9, 5780 Zografou, Greece. F. Rocadenbosch is with the Depart-

ment of Signal Theory and Communication, Universitat Polite`c-

nica de Catalunya, Jordi Girona 1-3, Edif. D4-016, 08034

Barcelona, Spain. X. Wang is with the Istituto Nazionale per la

Fisica della Materia, Complesso Universitario Monte S. Angelo,

via Cintia, 80126 Napoli, Italy.

Received 30 November 2003; revised manuscript received 20

May 2004; accepted 27 May 2004.

0003-6935兾04兾285370-16$15.00兾0

5370 APPLIED OPTICS 兾 Vol. 43, No. 28 兾 1 October 2004

The European Aerosol Research Lidar Network

共EARLINET兲 is the ﬁrst European network of 22 ad-

vanced lidar stations operating to provide a quanti-

tative climatological database of the horizontal,

vertical, and temporal distribution of aerosols over

Europe.

3–6

Lidars are powerful tools for providing

quantitative measurements of the optical properties

of aerosols with high spatial and temporal resolution

and with a high level of accuracy. In particular, the

use of combined Raman elastic-backscatter lidar per-

mits the independent measurement of aerosol extinc-

tion and backscatter as a function of height.

7,8

has long been known that the elastic-backscatter li-

dar system is limited by the fact that only one signal

is measured, whereas two parameters, backscatter

and extinction, determine the lidar signal.

Three

methods have been demonstrated to provide indepen-

dent aerosol extinction measurements: high-

spectral-resolution lidar,

Raman lidar

7,8

and

multiple zenith-angle measurements.

11–13

The Ra-

man and the high-spectral-resolution techniques

both rely on the detection of a pure molecular back-

scatter signal, but the latter requires a much higher

technical effort to suppress aerosol backscattering.

For reasons of technical practicability, the preferred

method within the EARLINET is a combination of

Raman and elastic scattering at one emission wave-

length near 355 nm. A large effort on the part of the

EARLINET community has been devoted to upgrad-

ing the Raman capability. At present, nine lidar

stations are able to perform measurements of nitro-

gen Raman scattering in the UV simultaneously with

elastic backscatter. One of these lidar stations 共at

Leipzig兲 is also able to measure nitrogen Raman scat-

tering in the visible domain.

The Raman lidar technique has been widely and

successfully used for measurements of aerosol extinc-

tion. The Raman lidar technique is also used for

operational lidar systems with automated data anal-

ysis.

The most critical part of this method arises

from the need to calculate the derivative of the loga-

rithm of the ratio between the atmospheric number

density and the range-corrected lidar-received power

in conjunction with data averaging and handling op-

erations. Much care is needed in data-averaging

and -handling operations to prevent miscalculation in

the estimation of both the aerosol extinction coefﬁ-

cient and the statistical error.

Because of the importance of the Raman technique,

data simulations to test and to improve Raman algo-

rithms used by each group of scientists within the

EARLINET network have been prepared.

4,15

Fur-

thermore, the simulations serve to draw attention to

special problems, such as appropriate averaging and

error determination, in the analysis of Raman lidar

data.

In this paper we present the results of an inter-

comparison of the algorithms used for retrieving

aerosol extinction and backscatter by use of a Raman

lidar, performed by the 11 groups of scientists within

EARLINET whose locations are listed in Table 1.

All these groups, except for the Neuchaˆtel and Bar-

celona groups that participated in this intercompari-

son with the intention to upgrade their systems with

a Raman channel in near future, have a lidar system

with a Raman channel. Algorithms used for the Ra-

man extinction retrieval were tested with synthetic

lidar data, with realistic experimental conditions

that are common for most of the groups and with

realistic aerosol load and properties taken into ac-

count.

This paper is the third in a sequence of papers

about aerosol lidar intercomparisons in the frame-

work of the EARLINET. These intercomparison ex-

periments were performed to produce a high-quality

standard of data originating from different systems

within the network. The results of the instrument

intercomparison were published in part one of the

series,

and those related to the aerosol backscatter

algorithm intercomparison, starting from elastic li-

dar signals, were published in part two.

This paper completes the algorithm intercompari-

son experiment, with an intercomparison of the algo-

rithms for the independent retrieval of both aerosol

extinction and backscatter starting from Raman and

elastic lidar signals. The paper is organized as fol-

lows: In Section 2 a brief description of the Raman

method is presented. Section 3 is devoted to a dis-

cussion of the data analysis applied for retrieving the

aerosol extinction coefﬁcient, including data han-

Table 1. Locations of Participating Groups and Raman Algorithms Used

Location of

Lidar Station Code Raman Algorithm

Aberystwyth ab Linear and quadratic ﬁt

Athens at Sliding average ﬁlter and polynomial ﬁt

Barcelona ba Weighted gliding window for spatial averaging, least-squares linear ﬁt

Hamburg hh Sliding average

L’Aquila la Second-order digital Savitzky–Golay ﬁlter

Lecce lc Sliding linear least-squares ﬁt

Leipzig le Sliding linear least-squares ﬁt

Napoli na Sliding linear ﬁt

Neuchaˆtel ne Sliding average

Potenza po Sliding linear least-squares ﬁt

Thessaloniki th Least-squares ﬁt

1 October 2004 兾 Vol. 43, No. 28 兾 APPLIED OPTICS 5371

dling, sources of error, and calculation of the effective

spatial resolution. In Section 4 the simulation used

for the intercomparison is described. In Section 5

the results from the various participating teams of

scientists are presented and discussed: Besides the

deviations between the retrieved aerosol extinction

proﬁles and the initial proﬁle used in the simulation,

the results of the intercomparison of the retrieved

aerosol backscatter proﬁles obtained by use of the

combined Raman elastic-backscatter lidar method

and of the lidar ratio proﬁles obtained from the inde-

pendent retrieval of the aerosol extinction and back-

scatter proﬁles are reported, as well.

2. Methodology

Raman scattering is an inelastic pure molecular scat-

tering

that has been successfully used in lidar

remote-sensing techniques since the late 1960s.

17–19

In a Raman lidar, wavelength ␭

of the scattered

light is shifted with respect to emitted laser wave-

length ␭

, and such a shift depends on the scattering

molecule.

For detection of the Raman scattering of

a gas with known atmospheric density, such as ni-

trogen or oxygen, the backscatter coefﬁcient in the

Raman lidar equation is known, and only the aerosol

extinction and its wavelength dependence remain as

unknowns.

The Raman lidar equation can be written as

P共␭

, z兲 ⫽ P共␭

兲C

共 z兲

␤

共␭

, z兲

exp

再

⫺

兰

关␣共␭

, ␨兲

⫹ ␣共␭

, ␨兲兴d␨

冎

, (1)

where P共␭

, z兲 is the power received from distance z

at Raman wavelength ␭

, P共␭

兲 is the emitted power

at wavelength ␭

, C

共z兲 is a function that depends on

the overlap function and on all the range-

independent system parameters, ␤

共␭, z兲⫽N共z兲␴

共␭兲

is the Raman backscatter coefﬁcient, where N共z兲 is

the atmospheric number density of the Raman scat-

terer and ␴

共␭兲 is the Raman backscatter cross sec-

tion, ␣ is the range-dependent total volumetric

extinction coefﬁcient at wavelengths ␭

and ␭

, and ␨

is the range integration variable.

Assuming a wavelength dependence of the aerosol

extinction ␣

aer

⬀␭

⫺k

, the Raman lidar equation can

be solved for the aerosol extinction at the emitted

laser wavelength

␣

aer

共␭

, z兲 ⫽

冋

N共 z兲

P共␭

, z兲 z

册

⫺ ␣

mol

共␭

, z兲 ⫺ ␣

mol

共␭

, z兲

1 ⫹ 共␭

兾␭

兲

, (2)

where d␴

共␭兲兾dz ⫽ 0 has been used. The molecular

extinction can be calculated from Rayleigh scattering

coefﬁcients and atmospheric number density proﬁles

retrieved from models or radiosonde measurements.

With the detection of the Raman scattered light,

independent aerosol extinction proﬁles can be deter-

mined. One can also use this information to derive

the aerosol backscatter without any assumption

about the extinction-to-backscatter ratio 共lidar ra-

tio兲,

which is an important parameter because it is

directly related to the microphysical properties of the

particles. One then calculates the backscatter pro-

ﬁle by forming the ratio of the elastic and the Raman

backscattered signals at height z and at a calibration

height z

, with an assumption about the reference

value of the aerosol backscatter at height z

␤

aer

共z

兲.

21–23

3. Data Analysis

In the analysis of Raman lidar measurements of aero-

sol extinction 关Eq. 共2兲兴 it is necessary to calculate the

derivative of the logarithm of the ratio between the

atmospheric number density and the range-corrected

lidar-received power. The statistical ﬂuctuations of

the Raman lidar signal can produce large ﬂuctua-

tions in the derivative and thus in the aerosol extinc-

tion proﬁle. Therefore the employment of data-

smoothing techniques is generally necessary.

Several methods such as data binning, sliding av-

erages, Kaiser ﬁlters,

and Savitsky–Golay ﬁlters

are usually employed for handling lidar signals. A

further method is the calculation of the derivative by

means of the least-squares technique, after the loga-

rithmic function in Eq. 共2兲 has been approximated

with a ﬁrst- or -second order polynomial within a

height range.

It is important to note, however,

that the application of a particular method of data

handling inﬂuences not only the ﬁnal result but also

the spatial resolution and the error.

A. Errors

For calculation of the aerosol extinction coefﬁcient it

is important to consider the main sources of uncer-

tainties, which are as follows:

共a兲 The statistical error that is due to signal de-

tection,

共b兲 The systematic error associated with the esti-

mate of temperature and pressure proﬁles,

8,27,28

共c兲 The systematic error associated with the esti-

mate of the ozone proﬁles in the UV,

共d兲 The systematic error associated with

wavelength-dependence parameter k,

8,26,29,30,31

共e兲 The systematic error associated with multiple

scattering,

8,26,32

共f 兲 The error introduced by data-handling proce-

dures such as signal averaging during varying atmo-

spheric extinction and scattering conditions.

8,33

In this paper we present an intercomparison of

several algorithms for the retrieval of aerosol extinc-

tion starting from synthetic data, and for this reason

only statistical errors that are due to signal detection

are taken into account; all the systematic effects are

neglected. For this intercomparison, each lidar

group provided proﬁles with their corresponding er-

5372 APPLIED OPTICS 兾 Vol. 43, No. 28 兾 1 October 2004

rors, which were calculated by analytical or numeri-

cal techniques. Analytical calculation is performed

by application of the error propagation rules in Eq.

共2兲. When complicated or nonlinear data-handling

procedures are applied, a numerical procedure based

on a Monte Carlo technique could be useful. This

procedure is based on the random extraction of new

lidar signals, each bin of which is considered a sample

element of a given probability distribution with the

experimentally observed mean value and standard

deviation. The extracted lidar signals are then pro-

cessed with the same algorithm to produce a set of

solutions from which the standard deviation is calcu-

lated as a function of height. Using both analytical

and numerical procedures will produce an error that

will depend on the noise of the signal and on the

particular algorithm used.

B. Effective Spatial Resolution

As was mentioned at the beginning of this section, the

application of a particular procedure of data-handling

inﬂuences the effective spatial resolution. In fact,

even if the retrieved extinction proﬁle presents the

same spatial resolution as do the raw Raman lidar

data, the data-handling procedures that are included

in the algorithm used to retrieve aerosol extinction

produce a loss of information that results in lower

spatial resolution, which we call the effective spatial

resolution. To test the effective spatial resolution

for the algorithm used, we use a step function

method. This method checks the ability to resolve

two narrow and well-separated structures in the

aerosol extinction proﬁle. In this method an ideal

synthetic Raman lidar signal is generated; an aerosol

extinction proﬁle that equals zero everywhere except

at two heights is used as input data. The synthetic

Raman lidar signal shows a step structure, with two

jumps that coincide with the two heights where the

aerosol extinction is different from zero. By select-

ing the type of data handling and changing the dis-

tance between the two peaked structures in the

aerosol extinction proﬁle it is possible to check the

minimum distance for which the two retrieved peaks

are resolved according to the Rayleigh criterion

that

is commonly used in spectroscopy to decide when two

neighboring spectral lines can be considered resolved.

This minimum distance is the effective spatial reso-

lution of the retrieved aerosol extinction proﬁle.

Figure 1 shows the application of the step function

method. Figure 1共a兲 shows an aerosol extinction

proﬁle that is equal to zero everywhere except at a

number of heights where peaks are present. The

aerosol extinction proﬁle in the ﬁgure has a spatial

resolution of 15 m and consists of 5 pairs of peaks that

are separated by 5, 6, 7, 8, and 9 points, which cor-

respond to 75, 90, 105, 120, and 135 m, respectively.

Fig. 1. 共a兲 Synthetic Raman lidar signal at 355 nm 共on a log scale兲 with a step structure and the corresponding peaked aerosol extinction

proﬁle 共on a linear scale兲 that are zero everywhere except for a series of couples of heights separated by 75, 90, 105, 120, and 135 m, starting

from the low heights. 共b兲 Aerosol extinction proﬁles retrieved by application of the 5-, 7-, and 9-point sliding linear ﬁts to the synthetic

Raman lidar stepping signal.

1 October 2004 兾 Vol. 43, No. 28 兾 APPLIED OPTICS 5373

The same ﬁgure also shows the corresponding nitro-

gen Raman lidar signal simulated at 355 nm and

characterized by steps that coincide with the peaks of

the aerosol extinction proﬁle. In Fig. 1共b兲 the results

of the application of the sliding linear ﬁt algorithm to

the synthetic Raman lidar signal are illustrated: 5-,

7-, and 9-point sliding linear ﬁts have been applied.

For example, the 9-point sliding linear ﬁt is able to

resolve perfectly the two peaks separated by 135 m

共from 1.4 to 1.6 km of height兲, and it is also able to

resolve the two peaks separated by 120 m 共1 to 1.2 km

of height兲 according to the Rayleigh criterion, but it is

not able to resolve the two peaks separated by 105 m

共0.7 to 0.9 km of height兲; therefore the effective res-

olution of the 9-point sliding linear ﬁt is 120 m. In

the same way, Fig. 1共b兲 shows that the 7-point sliding

linear ﬁt is able to resolve all the peaks up to the

minimum separation of 90 m 共from 0.4 to 0.6 km of

height兲; therefore, in this case, the effective spatial

resolution is 90 m. The 5-point sliding linear ﬁtis

able to resolve all the peaks of the aerosol extinction

because the minimum separation of the peaks in this

example is just 75 m, corresponding to 5 points.

It is important to note that the effective spatial

resolution is always better than the expected resolu-

tion, based just on the number of points used in the

data-smoothing procedure: In fact, in the 9-point

sliding linear ﬁt the effective resolution is 120 instead

of 135 m, and for the 7-point sliding linear ﬁt the

effective resolution is 90 instead of 105 m. This

technique was used by all the lidar groups to retrieve

the effective spatial resolution of the solutions in the

intercomparison.

4. Intercomparison of the Raman Lidar Algorithms

The main goal of the Raman algorithm intercompari-

son experiment is to test the algorithms used by each

group within the EARLINET network for the re-

trieval of the aerosol extinction proﬁle starting from

nitrogen Raman lidar signals. This intercompari-

son experiment has the further purpose of testing the

algorithm used for the independent retrieval of the

aerosol backscatter and of the lidar ratio. For this

purpose, synthetic Raman and elastic lidar signals

were generated, with typical experimental conditions

and realistic aerosol properties and load taken into

account. The intercomparison was blind, to repro-

duce real conditions in which the solution is not

known but also to prevent any possible inﬂuence on

the groups in their retrieval of the results.

Synthetic elastic and Raman lidar signals at 355

and 532 nm were calculated with a lidar simulation

model designed by the Institute for Tropospheric Re-

search, Leipzig, Germany. Temperature, pressure,

extinction, backscatter, and lidar ratio proﬁles, as

shown in Fig. 2, were used as input data. In the

Fig. 2. Input data for the simulation at 355 nm 共thinner curves兲 and at 532 nm 共thicker curves兲.

5374 APPLIED OPTICS 兾 Vol. 43, No. 28 兾 1 October 2004

Aerosol lidar intercomparison in the framework of the EARLINET project. 3. Raman lidar algorithm for aerosol extinction, backscatter, and lidar ratio

Figures

Citations

EARLINET: towards an advanced sustainable European aerosol lidar network

The Retrieval of Profiles of Particulate Extinction from Cloud-Aerosol Lidar Infrared Pathfinder Satellite Observations (CALIPSO) Data: Algorithm Description

Systematic lidar observations of Saharan dust over Europe in the frame of EARLINET (2000-2002)

EARLINET correlative measurements for CALIPSO: First intercomparison results

The automated multiwavelength Raman polarization and water-vapor lidar Polly XT : the neXT generation

References

Numerical recipes in Pascal : the art of scientific computing

Climate change 2001: The scientific basis. Contribution of Working Group I to the Third Assessment Report of the Intergovernmental Panel on Climate Change

Independent measurement of extinction and backscatter profiles in cirrus clouds by using a combined Raman elastic-backscatter lidar

Measurement of atmospheric aerosol extinction profiles with a Raman lidar

Raman lidar system for the measurement of water vapor and aerosols in the Earth’s atmosphere

Related Papers (5)

Stable analytical inversion solution for processing lidar returns

Analysis of atmospheric lidar observations: some comments

Independent measurement of extinction and backscatter profiles in cirrus clouds by using a combined Raman elastic-backscatter lidar

AERONET-a federated instrument network and data archive for aerosol Characterization

Aerosol-type-dependent lidar ratios observed with Raman lidar

Frequently Asked Questions (1)

Q1. What contributions have the authors mentioned in the paper "Aerosol lidar intercomparison in the framework of the earlinet project. 3. raman lidar algorithm for aerosol extinction, backscatter, and lidar ratio" ?