New water vapor line parameters in the 26000-13000 cm-1 region

TLDR: In this article, the authors presented new measurements of water vapor line parameters in the 26,000 − 13, 000 cm − 1 spectral region, combining a high-resolution Fourier transform spectrometer with a long-path absorption cell, thus allowing the observation of very weak, previously unobserved, lines.

Abstract: The radiative properties of water vapor play an important role in the physical and chemical processes occurring in the atmosphere. Accurate knowledge of the line parameters for this species is therefore needed. This work presents new measurements of water vapor line parameters in the 26 000– 13 000 cm −1 spectral region. The measurements were obtained by combining a high-resolution Fourier transform spectrometer with a long-path absorption cell, thus allowing the observation of very weak, previously unobserved, lines. A total of more than 9000 lines have been identified and their position, integrated cross section and self-broadening parameter have been determined. The dependence of the line parameters on nitrogen buffer gas pressure (0– 800 hPa ) has also been studied. The complete line list presented here is primarily compared to the HITRAN spectroscopic database, most frequently used in atmospheric calculations.

Figures

Fig. 1. Raw pure water vapor spectrum (18:5 hPa) in the visible and near ultraviolet region. and are the stretching and bending modes, respectively.

Fig. 4. Comparison with the ESA experimental and calculated values of line cross section measurements in this work (14 230–14 215 cm−1 interval). The line at 10−25 cm molecule−1 represents the lower limit of the ESA experimental values in this spectral range.

Fig. 8. Mean wavenumber shift parameter N2 (cm −1 atm−1), calculated for each polyad. The mean value obtained from 88 lines given in [17] and belonging to the 4 polyad is also shown. The value of N given for each polyad is the number of lines taken into account in the average.

Table 1 Summary of the results and comparison with the most recent literature data. S min and S max refer to the minimum and maximum absorption cross sections

Fig. 3. Comparison of the line integrated cross sections with respect to the HITRAN values (S , given in cm molecule−1 at 296 K). The inset gives an expanded portion of the plot for cross sections smaller than 1:25 × 10−24 cm molecule−1. Two distinct regions are obvious: Values of 1.03 and 0.90 are measured for the ratio S (this work)=S (HITRAN) above (crosses) and below (circles) the value S = 7× 10−25 cm molecule−1, respectively.

Fig. 7. Pressure dependence of the line pro;le for two well-de;ned lines of medium intensity (2× 10−25 cm molecule−1) belonging to the 6 polyad.

Full text

Journal of Quantitative Spectroscopy &
Radiative Transfer 74 (2002) 493–510
www.elsevier.com/locate/jqsrt
New water vapor line parameters in the 26000–13000 cm
−1
region
Pierre-Francois Coheur
a;1
, Sophie Fally
a
, Michel Carleer
a
, Cathy Clerbaux
a;2
,
Reginald Colin
a; ∗
, Alain Jenouvrier
b
, Marie-France Merienne
b
, Christian Hermans
c
,
Ann Carine Vandaele
c
a
Laboratoire de Chimie Physique Mol

eculaire, Universit

e Libre de Bruxelles, 50 Av. F.D. Roosevelt,
B-1050 Brussels, Belgium
b
Groupe de Spectrom

etrie Mol

eculaire et Atmosph

erique, UFR Sciences, Moulin de la Housse B.P. 1039,
51687 Reims Cedex 2, France
c
Institut d’A

eronomie Spatiale de Belgique, Av. Circulaire 3, B-1180 Brussels, Belgium
Received 1 August 2001; accepted 23 November 2001
Abstract
The radiative properties of water vapor play an important role in the physical and chemical processes oc-
curring in the atmosphere. Accurate knowledge of the line parameters for this species is therefore needed.
This work presents new measurements of water vapor line parameters in the 26 000–13 000 cm
−1
spectral
region. The measurements were obtained by combining a high-resolution Fourier transform spectrometer with
a long-path absorption cell, thus allowing the observation of very weak, previously unobserved, lines. A total
of more than 9000 lines have been identied and their position, integrated cross section and self-broadening
parameter have been determined. The dependence of the line parameters on nitrogen buer gas pressure
(0–800 hPa) has also been studied. The complete line list presented here is primarily compared to the
HITRAN spectroscopic database, most frequently used in atmospheric calculations. ? 2002 Elsevier Science
Ltd. All rights reserved.
Keywords: Water vapor; Line parameters; Atmospheric radiation
∗
Corresponding author. Fax: +32-2-650-4232.
E-mail address: rcolin@ulb.ac.be (R. Colin).
1
P.F. Coheur and A.C. Vandaele are, respectively, Scientic Research Worker and Postdoctoral Researcher with the
Fonds National de la Recherche Scientique.
2
Also at Service d’Aeronomie, Universite de Paris 6, Paris, France.
0022-4073/02/$ - see front matter ? 2002 Elsevier Science Ltd. All rights reserved.
PII: S 0022-4073(01)00269-2

494 P.-F. Coheur et al. / Journal of Quantitative Spectroscopy & Radiative Transfer 74 (2002) 493 – 510
1. Introduction
Water in all its three phases has a key role in the climate system of the Earth. Water vapor,
more particularly, is the most abundant greenhouse gas, an important chemical reactant and a carrier
of latent heat. Being also the most variable atmospheric species, it intervenes in the climate sys-
tem through dierent feedback mechanisms that are still not completely understood, due to the lack
of observations. These are made dicult as a result of the inhomogeneous horizontal and vertical
distribution of the species in the atmosphere. At present, the measurements of atmospheric water
vapor are best performed from space-borne instruments (balloon-borne radiosondes or satellite-borne
infrared or microwave sensors), which give either good horizontal or vertical resolution. The satellite
measurements require high-quality laboratory data, which are not always available. Accurate spec-
troscopic measurements are in particular needed to obtain information on the absorption properties
of atmospheric water vapor, which are important, among other things, to understand the Earth’s
energy balance. This problem has gained interest in recent years, since water vapor was proposed
as a possible candidate for the “missing absorber” [1], that is the species, which is responsible for
the excess of absorption when observations are compared to models (see, for example, [2]). The
rational for this suggestion stems from the fact that water vapor has a complex and dense vibrational
absorption spectrum, extending from the infrared to the near ultraviolet and characterized by a large
number of weak lines, which have not all been identied. This suggestion that water vapor is the
“missing absorber” has, however, been contradicted by recent calculations [3], which were shown
to agree satisfactorily with the observations.
In the infrared below 4500 cm
−1
, the spectroscopy of water vapor has been studied thoroughly and
line parameters, generally of good quality, are readily available through the HITRAN [4,5] or GEISA
[6] databases. Recent eorts in this spectral region were especially aimed at a better understanding
of the pressure dependence of the line prole [7–9]. In the near infrared, above 4500 cm
−1
,itis
believed that the line lists implemented in the databases are quite extensive but, for some parameters,
not suciently accurate [10]. It is in the visible and near ultraviolet, where the absorption is weak
as compared to the infrared, that the data are the most incomplete [10]. A precise quantitative
analysis of the absorption of water vapor in the visible spectral range is, however, particularly
important for the atmospheric radiative budget, as it produces an attenuation of the solar radiation
at its maximum of emission. A good knowledge of the weak lines in this visible region is also
required for the retrieval of other atmospheric gases absorbing in the same spectral range, as for
instance NO
2
[11]. At present, the most complete sets of water vapor line parameters available for
the visible region are those of Camy-Peyret et al. [12] and of Mandin et al. [13], between 25 250
and 16 500 and 16 500 and 13 200 cm
−1
, respectively. The pressure and temperature eects have,
however, not been studied in these earlier works. The line parameters measured by these groups
have been implemented in the HITRAN database, where additional calculated values for water have
also been added. Recently, in response to an announcement of opportunity of the European Space
Agency [14], a new water vapor line list covering the spectral range from 20 000 to 8000 cm
−1
has
been constructed [15] and partly published [16]. Still in a preliminary form, this database (referred
to as the ESA database in the following) is very extensive but includes a mixture of HITRAN lines,
experimental measurements and calculations. Hence, none of the HITRAN and ESA database appears
to be homogeneous over the entire spectral range. Aside the databases, some recent measurements
have improved the line parameters in restricted spectral regions. In particular, the red portion of

P.-F. Coheur et al. / Journal of Quantitative Spectroscopy & Radiative Transfer 74 (2002) 493 – 510 495
the water vapor absorption spectrum has been analyzed by the use of laser techniques [17–20],
which provide both high-resolution and long absorption path-length. However, these techniques are
generally limited to a narrow spectral interval and are not fully adequate for determining absolute
line intensities, because of the diculty encountered in measuring the total absorption path with
good accuracy. In the 13 200 cm
−1
spectral region, Grossmann and Browell [17,18] have provided
an extensive set of intensities and broadening parameters for water vapor in dierent buer gases.
In the blue region, Harder and Brault [11] have measured line intensities for water vapor lines
lying between 22 721 and 22 230 cm
−1
(atmospheric window for the NO
2
retrieval), which they
subsequently used to calibrate the water vapor lines of an atmospheric solar spectrum recorded at
the Kitt Peak National Observatory. This procedure allowed the authors to measure parameters for
about 40% more lines than in the work of Camy-Peyret et al. [12], suggesting, as expected, that the
line lists implemented in the databases could still be improved by new laboratory measurements.
It is the aim of this work to give a homogeneous and extensive set of water vapor line parameters
in the region 26 000–13 000 cm
−1
. To do this, we have performed absorption measurements of pure
and nitrogen diluted water vapor by using high-resolution Fourier transform spectroscopy and a long
multiple reection absorption cell. This paper is a development of previous work undertaken by
our group [21,22], in which the spectroscopic assignments of water vapor in the visible and near
ultraviolet are given. In the present contribution, we focus on the measurement of the line integrated
absorption cross section and width and on the dependence of the line parameters with the nitrogen
buer gas pressure. A special emphasis is put on the comparison of the line parameters measured
is this work with those implemented in the HITRAN and ESA spectroscopic databases.
2. Experimental
A detailed description of the experimental setup and procedure has been given previously [21].
Briey, the absorption spectra of water vapor were recorded using a high-resolution Fourier Trans-
form Spectrometer (Bruker 120M) coupled to a White multiple-reection absorption cell of 50 m
base path. With a 450 W high-pressure Xenon arc lamp taken as the light source, a total absorption
path of 602:32 m was selected in order to maximize the signal-to-noise ratio (S=N). All spectra
were recorded at room temperature (291:3 K) and at a resolution of 0:06 cm
−1
(15 cm maximum
optical path dierence). A silicon diode was used to cover the visible part of the spectrum (23 000–
13 000 cm
−1
), where the co-addition of 2048 interferograms (total recording time of 12 h) proved to
be adequate, leading to a S=N ratio of 2500 (expressed as the maximum signal amplitude divided by
twice the root-mean-square noise amplitude). A GaP detector was used above 23 000 cm
−1
, where
twice the number of interferograms was accumulated to produce a similar S=N ratio to that obtained
in the lower wavenumber region. Dierent optical lters were used in combination with the silicon
and GaP detectors.
The pure water vapor spectra were measured by introducing 18:5 hPa into the absorption cell, in
order to allow the observation of weak lines while avoiding condensation on the mirrors and on the
cell windows. The N
2
diluted water vapor mixtures were prepared by progressively adding nitrogen
into the cell lled with water vapor. Four dilutions, characterized by 170, 337, 569 and 802 hPa
total pressure were analyzed in this work. The temperature within the cell was measured with three
platinum wire thermometers and the pressure with an MKS Baratron capacitance manometer.

496 P.-F. Coheur et al. / Journal of Quantitative Spectroscopy & Radiative Transfer 74 (2002) 493 – 510
3. Data analysis
A Fourier transform interpolation factor of 16 was rst applied to all the recorded spectra to
ensure that the lines are well represented. From all of these, the atmospheric water absorption,
occurring in the 6 m long absorption path between the cell and the entrance of the spectrometer,
was carefully removed, as described earlier [23]. This external contribution to the absorption was
typically of the order of 1% of the total water vapor absorption, depending on the humidity within the
room.
The wavenumber scale was calibrated using the iodine lines. All spectra were rst calibrated with
respect to one another, in air, using the molecular atmospheric oxygen lines around 13 000 cm
−1
.
The corresponding wavenumber scale was then adjusted by comparing a set of iodine line positions
between 16 000 and 19 000 cm
−1
, recorded simultaneously with one of the water spectrum, with
those listed in [24,25], but transformed to air using Edlen’s formula [26]. The wavenumber scale was
transformed into vacuum wavenumbers by applying to the line positions a fth degree polynomial
based on Edlen’s formula.
The water vapor line parameters were extracted from the spectra using the WSPECTRA program
[27], which ts one by one the lines to a theoretical prole, convolved by the instrumental function.
The baseline is also tted by WSPECTRA. The program operates in two steps, each relying on a
non-linear least-squares procedure. In the rst step, both a Gaussian and a Lorentzian contribution to
the prole are tted, with the amount of Lorentzian width being characterized by a so-called damping
factor. Based on these rst values of the line parameters, the program rets the line to a chosen
prole (Rautian–Sobel’man [28] or Galatry [29]), with the Gaussian width xed to its calculated
value, depending on temperature and wavenumber. The results of the t are expressed in terms of
the line position , area A and Lorentzian full-width at half-maximum , all in wavenumber units.
In the case of the Rautian–Sobel’man and Galatry line proles, an additional narrowing parameter is
tted. The line area and Lorentzian width are then converted into more conventional line integrated
cross sections S

(cm molecule
−1
) and pressure-broadening parameters  (cm
−1
atm
−1
).
The line integrated cross sections (S

) and self-broadening parameters (
self
) were determined
from the pure water vapor spectra, except when the lines were saturated. In this case, the lowest
dilution providing non-saturated lines was used for the measurement of the cross section. In regions
where several pure water vapor spectra were recorded using dierent combinations of detectors and
lters, S

and 
self
were obtained by calculating the averaged value of the individual parameters. In
the averaging procedure, the root mean square (RMS) value of the t was used as the weighting
factor. The relationships used to transform line areas and Lorentzian line widths into line integrated
absorption cross sections and self-broadening parameters are as follows:
S

= A
P
0
T
n
L
T
0
Pl
; (1)

self
=

2P
; (2)
where T and P are the temperature and the pressure in the cell (T being 291:3K), T
0
= 273:15 K,
P
0
is one atmosphere, l is the absorption path length in cm and n
L
is the Loschmidt number
(n
L
=2:68676 × 10
19
molecule cm
−3
).

P.-F. Coheur et al. / Journal of Quantitative Spectroscopy & Radiative Transfer 74 (2002) 493 – 510 497
The N
2
-broadening parameters (
N
2
) and frequency shifts (
N
2
,incm
−1
atm
−1
) are considered to
be related to the partial H
2
O and N
2
pressures in the cell by the usual linear expressions
=2=(
self
× P
H
2
O
)+(
N
2
× P
N
2
); (3)

P
N
2
=  +(
N
2
× P
N
2
); (4)
where  is the wavenumber measured from the pure water vapor spectrum at 18:5 hPa. 
N
2
and 
N
2
were accordingly calculated by solving the systems of equations, each dened by the total pressure,
using general linear least-squares ts [30]. In these calculations, the statistical weights were taken
as the RMS values of the tted line parameters. The RMS values resulting from the least-squares
procedure was adopted as the nal statistical uncertainties on the line parameters.
For the comparison with literature and the databases, we have converted the line parameters mea-
sured here at 291:3K (T
1
) to the temperature T
2
, usually 296 K, by using the following approximate
relations:
S

(T
2
)=S

(T
1
)

T
1
T
2

3=2
e
−
hc
k
E

(T
1
−T
2
)
T
1
T
2
; (5)
(T
2
)=

T
1
T
2

n
(T
1
); (6)
(T
2
)=

T
1
T
2

n
(T
1
): (7)
A value of 0.6 is used for n. As discussed in [9], although this value of n may not be correct for all
transitions, it has only a weak inuence on the line parameters: an error of 50% on n would only
produce a dierence of 0.5% on the retrieved parameters.
The uncertainties in wavenumbers, cross sections, half-widths and pressure shifts have been es-
timated. First, a statistical uncertainty resulting from the tting procedure was obtained for each
parameter, as detailed above (this uncertainty on the parameters is dierent from line to line and
generally substantially larger for the weak or blended lines than for the strong and well-resolved
lines). Then, in addition to the statistical error, a systematic error arising from the uncertainties
on the pressure, the temperature, the path length and the calibration procedure was taken into ac-
count. Assuming uncertainties for the temperature and pressure measurements of 1% and on the
path length of 0.5%, we estimate the systematic uncertainty on the position and the cross sections
to be 2 × 10
−5
% and 2.5%, respectively. For the pressure-dependent parameters (
self
, 
N
2
and 
N
2
),
the systematic uncertainty is 1%. The absolute uncertainty on the parameters can be recalculated
considering both types of uncertainties.
4. Results and discussion
An overview of the water vapor absorption spectrum from 26 000 to 13 000 cm
−1
is presented in
Fig. 1, with the bands identied by their associated polyad number ( and  refer to the stretch-
ing and bending modes, respectively). The two overlapping regions, corresponding to the two main

Frequently asked questions

Q1. What are the contributions in this paper?

This work presents new measurements of water vapor line parameters in the 26 000–13 000 cm−1 spectral region. The dependence of the line parameters on nitrogen bu < er gas pressure ( 0–800 hPa ) has also been studied. 

Q2. How many measurements are used to obtain a mean value of self?

In the spectral region examined, a mean value of self =0:44 cm−1 atm−1 is obtained by taking all experimental measurements weighted with their associated statistical errors. 

Q3. What is the advantage of the line list presented in this work?

The authors believe that the line list presented in this work, which has the advantage of being, on the one hand, much more extensive than actual databases, with a cross section dynamic range of 105, and on the other hand, fully homogeneous for the entire visible region investigated, will improve the accuracy of upcoming atmospheric studies. 

Q4. How was the water vapor spectra measured?

The pure water vapor spectra were measured by introducing 18:5 hPa into the absorption cell, in order to allow the observation of weak lines while avoiding condensation on the mirrors and on the cell windows. 

Q5. How much of the water vapor absorption was the external contribution to the spectra?

This external contribution to the absorption was typically of the order of 1% of the total water vapor absorption, depending on the humidity within the room. 

Q6. Why is the uncertainty of the line parameters less than what was given in previous publications?

This is slightly less than what was given in their previous publications [21,22], essentially because the analysis of several spectra, recorded under di<erent pressure conditions, allows a better discrimination of the weak features from the noise. 

Q7. What is the significance of the line list?

It is to note that the ESA line list includes a very extensive set of calculated values for weak lines, which do not show a good agreement with their measurements, but which adds another 3% to the total absorption. 

Q8. How is the spectral distribution of water vapor studied?

In the infrared below 4500 cm−1, the spectroscopy of water vapor has been studied thoroughly and line parameters, generally of good quality, are readily available through the HITRAN [4,5] or GEISA [6] databases. 

Q9. What is the relationship between the line integrated cross sections and the self-broadening parameters?

The relationships used to transform line areas and Lorentzian line widths into line integrated absorption cross sections and self-broadening parameters are as follows:S = A P0T nLT0Pl ; (1)self = 2P ; (2)where T and P are the temperature and the pressure in the cell (T being 291:3 K), T0 = 273:15 K, P0 is one atmosphere, l is the absorption path length in cm and nL is the Loschmidt number (nL = 2:68676× 1019 molecule cm−3). 

Q10. Why is the self-broadening parameter larger than the average?

This could be due in part to the fact that the lines observed at higher energy involve more rotational levels of low J ′′, which are generally characterized by larger values of the broadening parameter [18]. 

Q11. What is the reason for the suggestion that water vapor is the “missing absorber”?

This problem has gained interest in recent years, since water vapor was proposed as a possible candidate for the “missing absorber” [1], that is the species, which is responsible for the excess of absorption when observations are compared to models (see, for example, [2]). 

Q12. What was the statistical uncertainty for the line integrated cross sections and self-broadening parameters?

a statistical uncertainty resulting from the ;tting procedure was obtained for each parameter, as detailed above (this uncertainty on the parameters is di<erent from line to line and generally substantially larger for the weak or blended lines than for the strong and well-resolved lines).