ARTICLE IN PRESS

Planetary and Space Science 56 (2008) 14671479

www.elsevier.com/locate/pss

Stereoscopic imaging of the hydroxyl emissive layer at low latitudes

G. Moreelsa,, J. Clairemidia, M. Faivreb, D. Mougin-Sisinia, M.N. Kouahlaa,
J.W. Meriwetherb, G.A. Lehmacherb, E. Vidalc, O. Velizc
a

Observatoire de Besanc- on, Institut Utinam, CNRS-UMR 6213, Universite de Franche-Comte, BP 1615, 25010 Besanc- on, France
b
Astronomy and Physics Department, Clemson University, SC 29634, USA
c
Instituto Geofisico del Peru`, Calle Badajoz 169, Mayorazgo IV Etapa, Lima, Peru
Received 30 December 2007; received in revised form 11 April 2008; accepted 14 April 2008
Available online 4 May 2008

Abstract
The hydroxyl nightglow layer is an excellent tracer of the dynamical processes occurring within the mesosphere. A new stereo-imaging
method is applied that not only measures the altitude of the airglow layer but also provides a three-dimensional map of the OH-layer
centroid heights. A campaign was conducted in July 2006 in Peru to obtain NIR images of the OH nightglow layer which were
simultaneously taken for two sites separated by 645 km: Cerro Cosmos (121090 08.200 S, 751330 49.300 W, altitude 4630 m) and Cerro Verde
Tellolo (161330 17.600 S, 711390 59.400 W, altitude 2330 m). Data represented by pairs of images obtained during the nights of July 2627 and
2829 are analyzed to yield satellite-type views of the wave eld. These are obtained by application of an inversion algorithm. In
calculating the normalized cross-correlation parameter for the intensity, three-dimensional maps of the OH nightglow layer surface are
retrieved. The mean altitude of the emission prole barycenter is found to be at 87.1 km on July 26 and 89.5 km on July 28. In these two
cases the horizontal wavelengths determined are 21.1 and 24.6 km with periods of 18 and 34 min, respectively. A panoramic view of the
OH nightglow emission obtained on July 29 at 8 h519 h26 UT is presented, in which the overall direction of the waves is found to be
NNW to SSE, azimuth 15013301 (counted from South). The wave kinetic energy density at the OH nightglow layer altitude is
3.9  104 W/kg, which is comparable to the values derived from partial reection radiowave data.
r 2008 Elsevier Ltd. All rights reserved.
Keywords: Airglow layer; Gravity waves; Stereoscopy; 3D imaging; Sky background

1. Introduction
The OH nightglow layer is an excellent tracer of the
chemical and dynamical processes occurring in the mesosphere. Chemical processes between minor species produce
excited molecules such as OH* or O2 (b1S, a1D) which in
their turn give rise to spontaneous emissions of photons.
Quenching or collisional deactivation resulting from the
exponential increase of pressure with decreasing altitude
produces a de-excitation of the excited species. In the case
of the OH* nightglow emission, the production mechanism
is the reaction of atomic hydrogen with ozone (Bates and
Nicolet, 1950). The prole of hydrogen increases with
Corresponding author. Tel.: +33 3 81 66 69 18;
mobile: +33 6 74 05 40 05; fax: +33 3 81 66 69 44.
E-mail address: guy.moreels@obs-besancon.fr (G. Moreels).

0032-0633/$ - see front matter r 2008 Elsevier Ltd. All rights reserved.
doi:10.1016/j.pss.2008.04.012

altitude, whereas the prole of ozone presents a secondary

maximum between 80 and 90 km and an associated
minimum at 76 km. The OH nightglow emission is mainly
a function of the density product [H] [O3] which has a
maximum located near 87 km. Chemical models predict
that the OH* nightglow layer is 7 km thick within the
8292 km altitude range. Early studies of the evolution of
the OH nightglow layer intensity throughout twilight
periods and successive nights revealed that nighttime
behavior of the OH intensity did not follow a regular
pattern as predicted by chemical models but exhibited
variations connected to dynamical processes. Models of the
mesosphere chemistry (Yee et al., 1997; Swenson and
Gardner, 1998) show that, as soon as the mesosphere is no
longer illuminated by the Sun, the ozone concentration
should rapidly increase, while the production of atomic
hydrogen produced during the day by water vapor

ARTICLE IN PRESS
1468

G. Moreels et al. / Planetary and Space Science 56 (2008) 14671479

dissociation should decrease. The net result of the atmospheric composition for the product [H] [O3], if only
chemical processes were considered, should be a steady
decrease in intensity through the nighttime hours.
The inuence of dynamical processes was initially
identied by Krassovsky (1972), and the importance of
atmospheric tides and gravity waves was pointed out by
Takahashi et al. (1984) and Hines and Tarasick (1987).
Tides have periods that are harmonics of 24 h (Teitelbaum
et al., 1989; Meriwether et al., 1994; Sijvee and Walterscheid, 1994). Gravity waves induce variations in the O(1S),
O2(1S) and OH* nightglow intensities originating in the
mesosphere and lower thermosphere (Battaner and Molina, 1980; Swenson et al., 1995; Taylor et al., 1997; Zhang
and Shepherd, 1999). Their existence was demonstrated by
analysis of data obtained by the Rayleigh lidar (Clemesha
et al., 1966; Hauchecorne et al., 1987; Chanin and
Hauchecorne, 1981). An important advance in the understanding of dynamical processes was achieved when IR
high sensitivity photographic lms became available for
nightglow studies (Peterson and Kieffaber, 1973; Moreels
and Herse, 1977; Herse et al., 1980). This approach to twodimensional (2D) imaging of the mesospheric nightglow
layers was rapidly superseded by the application of digital
cameras when CCD low noise imaging systems started to
be used in astronomy (Taylor et al., 1995; Pautet et al.,
2001).
Images of the mesospheric near-infrared (NIR) nightglow layer frequently show the presence of stripes or
relatively well-organized bands. These structures provide a
clear indication of the propagation of gravity waves in the
upper atmosphere. Photometric measurements of the OH
nightglow emission intensity and of its nocturnal evolution
show that the chemical and dynamical mechanisms that
govern the steady state of the OH* responsible for the
emission have comparable amplitudes (Moreels et al.,
2007). To obtain a detailed view of the inuence of gravity
wave propagation upon the layer, we developed a new
approach that is based upon the calculation of a stereoscopic relief map of the layer using data from two
overlapping views of the same portion of the OH nightglow
layer.
In this paper, preliminary results obtained in Peru in July
2006 are presented. The instrumentation consisting of two
CCD cameras taking simultaneous NIR pictures of the sky
is described in Section 2. In the following section, the basic
principles of stereoscopic imaging are presented. Its
application for calculating the geometry and shape of the
surface of the photometric barycenter of the OH nightglow
layer is explained. The results of observations performed
on the two nights of July 26 and 28, 2006 are presented in
Section 4. Two three-dimensional (3D) maps of the altitude
of the emission centroid are compared with 2D graphs of
the emission intensity recorded at the two sites with a
baseline separation of 645 km. In Section 5, our results are
compared with other observational studies of the mesosphere-lower thermosphere dynamics such as the series of

sodium temperature lidar observations conducted in years

20022003 during the Maui MALT program.
2. Instrumentation and geographic observation sites
2.1. Instrumentation
Two identical CCD cameras are used to take simultaneous NIR images of the sky. The elds of view from the
two cameras overlap with a common volume region for
which the altitude relief of the OH nightglow layer will be
retrieved. Each camera is mounted on a computer-driven
alt-azimuthal platform. The detector of each camera is a
512  512 pixel (19 mm  19 mm) silicon front-illuminated
single-frame CCD array, that is thermoelectrically cooled
to 40 1C to reduce the noise level to 3 electrons/pixel/s.
The exposure time of 90 s is set by an electro-mechanical
shutter. A fast f/1.3 Angenieux lens forms an image of the
sky on the sensitive surface of the CCD. The focal length of
the lens in the NIR range was f 15.9 mm dening a
361  361 eld of view. A Schott RG780/2 mm lter was
mounted in front of the lens in order to transmit the NIR
radiation above the wavelength of 780 nm. The spectral
bandpass is limited in the infrared by the silicon cut-off
transmission of the CCD at 1050 nm. This bandpass region
includes the ve bands of the Dv 4 sequence of OH
represented by the (4,0), (5,1), (6,2), (7,3), and (8,4)
vibrational transitions.
2.2. Observation sites
The conguration chosen for the two observing sites in
Peru was two facing elds of view delimiting a common
volume at the midway point between the two sites (Fig. 1).
Given the nominal altitude of the OH nightglow layer,
87 km, the sites were chosen at a distance of 645 km from
each other. The rst site was located at Cerro Cosmos
(121090 08.200 S, 751330 49.300 W, altitude 4630 m) near the city
of Huancayo. In this place, a solar observatory had been
active until 1988. The second site was located at Cerro
Verde Tellolo (161330 17.600 S, 711390 59.400 W, altitude
2330 m) near Arequipa. Electric power for each site was
provided by a generator. The experimental procedure
consisted of observing a common volume nightglow region
from reciprocal directions. The azimuths for the optical
axes were 3181 (counted clockwise from South) for Cerro
Cosmos and 1381 for Cerro Verde. For both sites the
elevation angle was the same: 181400 . Communication was
secured by the satellite Iridium global communication
array. The UT time reference was provided by a GPS
system.
3. Stereoscopic imaging
Stereoscopic or triangulation methods have been used to
determine the altitude of meteors, of aurorae (Stormer,
1910, 1911) or noctilucent clouds (Witt, 1962) ever since the

ARTICLE IN PRESS
G. Moreels et al. / Planetary and Space Science 56 (2008) 14671479

Fig. 1. Map of Peru showing the locations of the two sites for stereo
observations: Cerro Cosmos (121090 08.200 S, 751330 49.300 W, altitude
4630 m) near Huancayo and Cerro Verde Tellolo (161330 17.600 S,
711390 59.400 W, altitude 2330 m) near Arequipa. The distance between the
two sites is 645 km. The elds of view of the cameras are aligned along
reciprocal directions, in a facing mode.

sensitivity of photographic plates became sufcient to

record low-contrasted or nighttime low-level atmospheric
emissions. Here, our purpose is to obtain a representation
of the surface map dened by the photometric barycenter
of the OH mesospheric nightglow layer. The major
assumption is as follows. If, at a given geometric altitude,
the atmospheric density is lower than in the neighbor
points at the same altitude, the quenching will be less
effective below the emissive layer and the OH emission
more intense. The altitude of the photometric barycenter
will be slightly lower. Conversely, if the density is higher,
the quenching will be more efcient on the layer lower side
and the centroid altitude of the OH nightglow layer slightly
higher. As a consequence, the propagation of a density
wave through the emissive layer should result in a
modulation of the geometric altitude of the emission
barycenter. In other words, the direct measurement of the
altitude of the emissive layer should provide experimental
evidence regarding the location of gravity waves allowing
the study of the propagation of gravity waves in regard to
direction and speed of wave events.

3.1. Overview of stereoscopic vision

A brief overview of the stereoscopic principle of methods
currently used in 3D-vision is presented here. Let us
consider an object such as a ruler, a box or a grid.

1469

Information about depth can be retrieved if two cameras are

used to take photographs of the same part of the object.
Following Faugeras (1993, p. 35), we dene as the retinal
plane the plane in which the image is formed by a camera
through an operation called perspective projection. Using
pinhole models for the cameras, let us call C1 and P1, C2
and P2 their respective optical centers and retinal planes as
shown in Fig. 2. The C1C2 line joining the camera centers is
called the baseline. Let us consider a physical point M of the
observed object. The image of M produced by camera (1) is
called m1. Its coordinates in the retinal plane are
(u, v). In the same way, the image of M produced by
camera (2) is called m2 (u0 , v0 ). Cartesian coordinates are
used in accordance with the orientation of the pixels of the
retinal planes. The ve points M, C1, C2, m1 and m2 dene a
plane called epipolar plane related to the physical point
M. This is a rst constraint called the epipolar constraint.
For the experimenter who considers a specic point m1
in the image plane P1, the problem is to nd the matching
point m2 in the image plane P2. The important step
consists in establishing biunivocal relations between pairs
of points, or point matching. If the object has recognizable
elements such as lines, corners or apexes that can be
identied on both pictures, the equations of the lines
(m1C1) and (m2C2) provide the coordinates of their
intersection M.
In other cases, other methods can be used. First, we note
that the intersection of the epipolar plane and the retinal
P1 planes is a line (D1) containing the point m1. As a result,
the epipolar plane also cuts the retinal P2 plane along a line
(D2) which contains the matching point m2 (Hartley and
Zisserman, 2003, p. 239). The line (D2) is called the
epipolar line related to point m1. It contains the epipole E2
that is the intersection of the baseline (C1C2) with the
retinal plane P2. In order to locate the point m2 matched
with point m1, another constraint needs to be dened.
Epipole

Epipole

E2

C2
m3
Epipolar
segment

C1

E1
m1

m2
m4

(D1)

(D2)

2
M

Fig. 2. Epipolar geometry. Two pinhole cameras are used. Their

respective centers and retinal planes are called C1 and P1 for camera
(1), C2 and P2 for camera (2). The C1C2 line is called the baseline. The
images of a physical point M are called m1 for camera (1) and m2 for
camera (2). The point m2 is located on a line D2 called epipolar line that
is the intersection of the epipolar and retinal planes. The so-called
epipolar constraint states that the ve points M, C1, C2, m1 and m2
dene a plane called epipolar plane.

ARTICLE IN PRESS
1470

G. Moreels et al. / Planetary and Space Science 56 (2008) 14671479

First, in our problem, the retinal planes are parallel to the

mesospheric emissive layer and the optical axis of each
camera, called in the following text virtual camera is
perpendicular to the retinal plane. Since the OH nightglow
layer is 7 km thick and is distributed in altitudes between
80 and 100 km, the position of the m2 point to be matched
with m1 is restricted to a line segment m3m4 called the
epipolar segment (Fig. 2).
3.2. Stereoscopic imaging of the mesospheric emissive layer

Verde cameras whose optical centers are, respectively,

called C1 and C2. Let us consider a point m of coordinates
(u, v) in the rst image (Cerro Cosmos). The matrix of
perspective projection is called P. It is invertible. The
equation of the optical line (C1m) is
C 1 m fM C 1 lP1 m; l 2 R g

Now, as shown in Fig. 4, let us call m1 a point of the

retinal plane of the virtual camera (1) and M the projected
point of m1 on the emissive layer plane. M is a point of the
optical line (C1m1). The projection of that line in the frame
of the retinal plane of the second virtual camera (Cerro
Verde) denes the epipolar segment as shown in Fig. 2. The
limiting points of the epipolar segment correspond to
altitude limits for the emissive layer supposed to be 80 and
100 km. Given a point m1 (u, v) in the frame of the rst
image, and an associated point m2 of coordinates (u0 +du0 ,
v0 +dv0 ) in the second image, we can dene a disparity
(Faugeras, 1993, p. 176) that is the vector (du0 , dv0 )
associated with the point m1. The disparity domain is
constrained by the epipolar geometry.
An additional constraint must be dened to precisely
locate the matched point m2 on the epipolar segment.
In the present case, the object is diffuse and does not
exhibit discrete features such as corners, peaks or lines. We
use the normalized cross-correlation (NCC) criterion
dened by Eq. (2) (Ma et al., 2004, p. 89). Two small
windows are dened around the two matching pixels that
we try to locate in the retinal planes P1 and P2. The
operator compares the intensities in the small windows

The observational procedure consists in taking simultaneous near-IR photographs of the sky at Cerro Cosmos
and Cerro Verde Tellolo. The optical axes are aligned, but
in reciprocal directions in order to delimit a common
volume region of the OH nightglow layer observed by both
ground-based real cameras. As shown in Fig. 3, the
image recorded by each real camera is a projective view of
the area of the emissive layer delimited by the eld of view
of the camera, whose axis is pointed at a zenith angle y.
Here, y 711200 . In order to obtain a realistic image of the
emissive layer, we introduce the concept of a virtual
camera. This ctitious camera is located on the vertical line
having its origin at the real camera. A point A of the
emissive layer has an image called A0 in the focal plane of
the real camera. The matrix transformations used to obtain
the coordinates of the corresponding ctitious Av in the
focal plane of the virtual camera are explained by Pautet
and Moreels (2002).
Two important optical and photometric effects need to
be taken into account in these transformations. First, a
Pp
Pp

kp
lp I 1 x k; y l  I 1 I 2 x dx k; y dy l  I 2

q
q

NCCx; y; dx; dy P
P
P
P
p
p
p
p
2
2

I
x

k;
y

l

I

I
x

dx

k;
y

dy

l

I

1
2
kp
lp 1
kp
lp 2

correction for refraction must be applied, particularly if the

observations are performed at low elevation angles.
Second, a correction for the Van Rhijn effect, i.e., the
intensication of the observed emission at slant angles,
must also be calculated. The result of the linear transformations is an image having the shape of a pie chart angular
sector. This image shows the view that a cosmonaut on
board a satellite could have when using a near-IR camera
and looking downward (Pautet et al., 2002).
Fig. 4 shows the two virtual cameras associated with the
real cameras located in the two stereoscopic observation
sites. The satellite-type images provided by these cameras
are now used to retrieve the surface map describing the
altitude of the photometric barycenter of the emissive
layer. In Fig. 4, for sake of clarity, the image plane is not
placed behind the lens as in normal cameras, but in front,
between the lens and the scene. However, the camera is
assumed to work according to the pinhole model. We can
now apply the epipolar geometry to the satellite-type
images produced by the virtual Cerro Cosmos and Cerro

(1)

(2)

The terms I1(x, y) and I2(x+dx, y+dy) are the

intensities of the matched points m1 and m2. The projection
frames noted by the subscripts 1 and 2 refer, respectively,
to the Cerro Cosmos and Cerro Verde sites. They are
bound to the virtual retinal planes dened in Figs. 3 and 4.
Square windows of dimensions (2p+1,2p+1) are dened
around the points m1 and m2. The mean intensities I 1 and
I 2 are the averages of the values over the (2p+1)2 pixels of
the windows. The position of the matched point m2 is the
result of an iteration. The small variations k and l represent
increments related to the (x, y) and (x+dx, y+dy)
coordinates of both points. In the present case, the size
of the windows for the correlation around the matched
points was 13  13 pixels which is comparatively small in
comparison with the size of the periodic features appearing
in the pictures. The NCC criterion offers the advantage of
being insensitive to the difference of contrast between the
images.
To calculate the altitude of the centroid of the emissive
layer, the height of the peak intensity of the OH nightglow

ARTICLE IN PRESS
G. Moreels et al. / Planetary and Space Science 56 (2008) 14671479

x,y,z frame associated

with the virtual camera

x
f
Focal
length

z
Virtual camera
retinal plane

Bv

1471

Av D v

L
Virtual camera
altitude

Observation
site altitude

B
c
H

RH =
R0 + H
H: emissive
layer altitude

real
camera

z1
x1,y1,z1
frame
associated
with the real
camera

x1 A
D
y1

Real camera
retinal plane

R0
Earth
radius

Earthcenter

Fig. 3. Inversion of the perspective effect to obtain a satellite-type view of the OH nightglow layer. The initial image is taken with the real camera
pointed at a zenith angle y. Here, y is chosen in the 66741 range. A virtual camera located above the real camera is introduced. The image in the retinal
plane of the virtual camera is homothetic to the area of the emissive layer within the eld of view of the real camera. The altitude L of the virtual camera is
a parameter that may be adjusted. Here L 500 km.

layer is expected to lie between 82 and 92 km (Baker

and Stair, 1988). Fig. 5 shows the geometry of the views.
The two virtual cameras C1 and C2 look down upon the
emissive layer whose photometric barycenter is initially
supposed to be located at 87 km. Given a pixel m1 in the
virtual image plane, let us call N the point along the optical
line C1m1 whose altitude is 87 km. Let us use small letters
to designate the optical centers c1 and c2 of the real
cameras. The optical ray from the real c1 camera that goes
through m1 is c1M1NM2. As seen from the real camera c2,

the projected segment at the considered altitude is M3M4

and the epipolar segment in the retinal plane is m3m4. This
scheme provides a geometrical constraint used in conjunction with the cross correlation criterion to retrieve the
surface map of the layer over areas of 50  50 km2.
4. First results of stereoscopic observations in Peru
In this section we present the initial stereoscopic results
obtained during the observing campaign conducted in Peru

ARTICLE IN PRESS
G. Moreels et al. / Planetary and Space Science 56 (2008) 14671479

1472

C2

C2
C1
z

Virtual C1
camera

Epipolar
segment

m2

Virtual
camera

m3
m4 m2

m1

Imageplane
2048x2048

m1

L2 = 500km
Mesospheric
layer

M
L1 = 500km
z1

92
87
82 km

x1

c1
y1

M4

h
1
h

CERRO
VERDE

X
Z

Fig. 4. Stereoscopic system using the two virtual cameras. O is the center
of Earth, supposed spherical. Only one real camera is drawn. For sake of
clarity, the retinal planes of the virtual cameras are not drawn behind the
lens, as in normal cameras, but in front, between the lens and the scene.

in July 2006. The rst step in the image reduction

procedure was to nd pairs of images taken simultaneously
at both sites, Cerro Cosmos and Cerro Verde. The limiting
parameter was frequently the nighttime cloudiness at Cerro
Cosmos. Images showing well-dened emissive structures
are selected. The standard image reduction procedure is
applied to the raw les called Imraw. It consists in
calculating the ratio
Imraw  Imdark =Imff  Imbias

M1

Frame associated
with the real camera

Frame associated
with the center of Earth

M3

c2

Mesospheric
layer

M2

(3)

In this expression Imdark is the image of the dark current

taken with the same exposure time as the eld of view
image. Imff is the image of the at eld obtained in taking
2030 pictures of the sky at dawn before sunrise. Imbias is
an image of the offset, taken with a very short exposure
time, equal to the average time for the at-elds.
The next image processing step consists in removing the
stars in the images by digital ltering. Using the histogram
of the image pixel intensities, a binary image of the stars in
the picture is created. A star image is replaced by the
median value of the pixels inside a square area of size 9  9
around the star whose image is removed. A second median
lter of larger size suppresses the residual traces of stars.
The specic structures in the emissive layer clearly appear
and show well-organized bands with separations that
diminish at low elevation angle, as a consequence of the
perspective projection. The ltering process is quite

Earth
Ground-basedcameras

c1
CERRO
COSMOS

Fig. 5. Determination of the altitude of the centroid of the emissive layer.

Given a point m1, its matched point is located on the epipolar segment
m3m4, because the centroid altitude of the OH layer intensity prole is
known to be between 82 and 92 km (Baker and Stair, 1988). In using the
constraint related to the NCC criterion, the altitude of the emissive point
corresponding to point m1 can be determined.

important because residual star images will spoil the

images used in the stereo-correlation calculations. In
addition, images of bright stars or planets, mainly if
saturated, sometimes cannot entirely be removed.
The second step is to generate the perspective inversion
of the image taken with the ground-based real camera.
Given a point of the emissive layer supposed to be spherical
of radius R+h, the problem is to calculate the coordinates
of the corresponding point of the virtual retinal plane. This
implies the application of ve matrix transformations as
explained in Pautet and Moreels (2002).
Two stereoscopic pairs of images are now presented and
analyzed. The rst one was taken on July 26, 2006 at 22 h15
(UT-5). Fig. 6a and b show the images taken by the real
cameras. On both images, two series of 9 (Cerro Cosmos)
and 11 (Cerro Verde) stripes extend from the horizon to
higher elevation angles along the general longitudinal
direction of the elds of view. The following panels, 6c
and d, show the satellite-type images of the emissive layer
within the limits of the opposite elds of view. The stripes
are regularly aligned along a NNW, SSE direction at
azimuth 1501. The horizontal wavelength is lh 21.1 km.
The wave temporal period is measured in using a series of
40 images taken at regular intervals between 22 h00 and
23 h01. Its value T 18 min. The two satellite-type images
are superposed in Fig. 7 to show the common area in both
elds of view. The numerically coded intensities in a

ARTICLE IN PRESS
G. Moreels et al. / Planetary and Space Science 56 (2008) 14671479

1473

Fig. 6. Pair of simultaneous images taken on July 26, 2006, 22 h15 (UT-5) at Cerro Verde (a, c) and Cerro Cosmos (b, d). The images in a and b are taken
at horizontal coordinates 1381, 201 (Cerro Verde) and 3181, 241 (Cerro Cosmos). Azimuths are counted clockwise from South. The images in c and d
correspond to a and b. They are satellite type views of the emissive layer.

common area of 50  50 km2 retrieved from the observations at Cerro Cosmos and Cerro Verde are shown,
respectively, in Fig. 8a and b. The reconstructed intensities
are given in the same arbitrary units in both panels 8a and
b. Using the stereo-imaging software called Triangul9
(Faivre, 2004), a 3D map of the photometric barycenter
of the emissive layer over the same area is constructed
(Fig. 8c). This relief map may be compared with the coded
representations of the intensities.
The same wavy structure and the same number of
undulations: 2.7 are present in the three panels 8ac. The
3D surface map shows ve maxima at 88.6, 89.0, 89.2, 89.4
and 90.4 km and three minima at 83.4, 84.0 and 85.4 km.
The measured wave amplitude is 2.6 km. The mean altitude
is 87.1 km. A cut along a vertical plane perpendicular to the
general direction of the wave fronts is shown in panel (8d).
The horizontal wavelength is lh 21.1 km. The uncertainty level on the retrieved altitudes is given by the relative
measure of the precision of point matching. An uncertainty
parameter is calculated using the NCC parameter dened
in Eq. (2). For each NCC value, comprised between 1
and +1, we assign, in an afne way, an uncertainty
parameter whose value is equal to 0 when the NCC is 1 and
10 km (the authorized interval of layer thickness) when the
correlation is poor and the NCC is 1. In fact, the vertical

Fig. 7. Superposition of the elds of view corresponding to the images

simultaneously taken on July 26, 22 h15. A wavy sector is apparent. The
azimuthal direction of the wave fronts is 1501 (counted clockwise from
South). The sign + shows the location of the Cerro Verde observation
site.

ARTICLE IN PRESS
1474

G. Moreels et al. / Planetary and Space Science 56 (2008) 14671479

Fig. 8. Coded representations of the intensities of the simultaneous images recorded at Cerro Verde (a) and Cerro Cosmos (b) on July 26, 2006, 22 h15
(UT-5 h). The vertical scales in panels (a) and (b) are identical. They give the local intensity in arbitrary, but identical units. In panel c, a stereoscopic 3-D
reconstitution of the layer barycenter altitude is presented. In panel d, a cut of the relief map c in a direction perpendicular to the wave crests shows the
periodicity of the waves.

bars drawn in panel 8d are not true error bars, but they
measure the probability for the altitude to be comprised
between the bar limits.
The second pair of stereoscopic images was obtained
during the night of July 28, 2007 at 20 h15 (UT-5). It is
presented in Fig. 9 where panels a and b show the images
taken simultaneously in both sites and panels c and d the
satellite-type views after inversion of the perspective effect.
The two images in panels c and d are superposed in Fig. 10.
Two wavy areas can be identied in this view: a northern
part where the direction of the stripes is transverse,
perpendicular to the direction of the elds of view and a
southern part where the stripe direction is longitudinal, as
in Fig. 7. The measured horizontal wavelengths are,
respectively, lh 24.6 and 22.8 km for the transverse and
longitudinal waves. The 3D representations of the coded
intensities are presented in Fig. 11a and b for a 50  50 km2
area inside the superposition zone where the waves are
transverse. The reconstructed intensities are given in the
same arbitrary units in both panels 11a and b. The altitude

of the centroid is shown in panel 11c. Two and half

transverse waves can be identied. A cut along a vertical
plane perpendicular to the general direction of the wave
fronts is shown in panel 8d. The mean altitude is 89.5 km
and the wave amplitude is 2 km. The temporal period is
34 min. The horizontal wavelength is lh 24.6 km. The
wavelengths measured here at a low latitude site are
comparable to those reported by Faivre et al. (2003) at
mid-latitude (Chateaurenard, 441410 N, 61540 E). Using
Fourier analysis, they determined horizontal wavelengths
values comprised between 22 and 55 km which are
comparable to the values measured here.
In Figs. 12 and 13 we present an overall view of the OH
nightglow layer over Cerro Cosmos obtained in making a
3601 panorama at an elevation angle of 161 on July 29 from
03 h51 till 04 h26 (UT-5). 12 photographs are rst assembled
in Fig. 12, then inverted to show the emissive layer as a disk
having a radius 1100 km. Two wide arches are seen in
Fig. 12 that converge at two opposite vanishing points.
The azimuths of both points provide the general direction of

ARTICLE IN PRESS
G. Moreels et al. / Planetary and Space Science 56 (2008) 14671479

1475

Fig. 9. Pair of simultaneous images taken on July 28, 2006, 20 h15 (UT-5) at Cerro Verde (a,c) and Cerro Cosmos (b,d). The images in (c) and (d)
correspond to (a) and (b). They are satellite type views of the OH nightglow layer.

the wave fronts present in the satellite-type view of

Fig. 13, 15013301. A corrugated region extending over
150  150 km2 can also be seen in the NW part of the image.
In a previous paper, Faivre et al. (2007) showed that the
intensity variation of the OH airglow may be explained by
a quenching efciency that varies with variable air density.
When the air density increases, due to the propagation of a
gravity wave, quenching erases the lower side of the
emissive layer. As a result, the intensity decreases and
the barycenter altitude of the layer is higher. Using the
expression of the density of air in a gravity wave given by
Swenson and Liu (1998), Faivre et al. (2007) showed that
the altitude of the airglow layer can vary by several
kilometers. Knowing the density of air at 87 km,
r 1.4  1020 mol/m3, the period T and amplitude a for
the wave oscillation, the local kinetic energy density for the
vertical motion can be calculated using
W 2rP2 a2 =T 2

(4)
4

Fig. 10. Superposition of the elds of view corresponding to the images

simultaneously taken on July 28, 20 h15. A wavy sector is apparent. Its
overall direction is perpendicular to the line of sight.

A value of 3.9  10 W/kg is obtained. This number

may be compared with the measurements of gravity wave
energy densities using partial reection radiowave observations (Manson et al., 1981, 2002). These authors reported
values=1  103 W/kg at altitude 90 km for waves having

ARTICLE IN PRESS
1476

G. Moreels et al. / Planetary and Space Science 56 (2008) 14671479

Fig. 11. Coded representations of the intensities of the simultaneous images recorded at Cerro Verde (a) and Cerro Cosmos (b) on July 28, 2006, 20 h15
(UT-5). The vertical scales in panels (a) and (b) are identical. They give the local intensity in arbitrary, but identical units. In panel (c), a stereoscopic 3-D
reconstitution of the layer barycenter altitude is presented. In panel (d), a cut of the relief map 11c in a direction perpendicular to the wave crests shows the
periodicity of the waves.

Fig. 12. Panorama display of 12 consecutive images taken at Cerro Cosmos from 03 h51 to 04 h26 (UT-5) on July 29, 2006. Elevation angle: 161 above
horizon. The atmospheric emission shows two series of arches that converge at two vanishing points in opposite directions at azimuths 1501 and 3301. This
shows that wave fronts have an overall azimuth direction 1503301.

periods To1 h. As their observations were conducted at

higher latitude in Saskatoon (521N), we note that the
energy density of the waves presently measured in the
equatorial region was two times smaller. However, this
statement cannot be ruled as general, because gravity
waves are also highly variable as a function of seasons
during the year.
5. Discussion
The stereo-imaging method developed here is an original
procedure to investigate the dynamic structure of the

mesosphere in the 8292 km range. Compared to rocket

and satellite observations, it has three specic characteristics. First, it is a purely geometric method to retrieve the
altitude of the OH emissive layer. Second it provides a
detailed description of the signature of gravity waves in
areas of dimensions 50  50 km2 presented as surface
brightness maps of the emissive layer barycenter. Third,
it provides a direct measurement of the kinetic energy
density of the gravity waves at that levels.
Many measurements of the altitude and thickness of the
OH layer have been performed since its identication in
1950. Rockets carrying photometers and/or spectrometers

ARTICLE IN PRESS
G. Moreels et al. / Planetary and Space Science 56 (2008) 14671479

Fig. 13. Satellite type view of the NIR mesospheric emissive layer centered
on Cerro Cosmos, near Huancayo, Peru. North is up, East is on the right.
Date: July 29, 2006, 03 h5104 h26 (UT-5). The view corresponds to the
panorama displayed in Fig. 12. The wave fronts have an overall NNW to
SSE direction at azimuth 1503301. The radius of the displayed eld of
view is 1100 km at the altitude of the emissive layer. A corrugated region
extending over 150  150 km2 can be seen in the NW part of the image.

were frequently used (Packer, 1961; Tarasova, 1963; Baker

and Waddoups, 1967; Evans et al., 1973; McDade et al.,
1987). In their review of the rocket measurements of the
altitude of the OH airglow layer, Baker and Stair (1988)
mention that the value of the mean altitude of the peak OH
volume emission rate is 87.4 km. They recommend as
working numbers for the altitude and thickness 8672.6
and 8.673.1 km.
Limb-sounding observations of the airglow layer by
WINDII on board UARS were performed on a global
scale basis (Zhang and Shepherd, 1999; Melo et al., 2000).
The reported peak altitudes for the 40S40N latitude range
were comprised between 86.4 and 89.4 km during spring
1992 and winter 19921993. For the ve past years, two
instruments on board ENVISAT have performed limb
observations of the mesospheric emissions. SCIAMACHY
produces near-global retrievals of mesopause OH rotational temperatures (von Savigny et al., 2004). GOMOS
measures the column density of ozone and water vapor in
observing the disappearance of some stars, in the stellar
channel, on the Earth horizon. Before the occultation
sequence, the water vapor spectroscopic channel provides
spectra of the OH (84) band in the 926954 nm range with
a resolution Dl 0.2 nm. Preliminary results obtained in

1477

the tropical region in January 2004 showed that the

rotational temperature was 181 K and the altitude of the
OH peak was 89 km (Hauchecorne et al., 2007).
Measuring the altitude of the airglow layer from the
ground is possible either in using a geometric method as
described in Section 3 or in using correlated airglow and
sodium lidar observations. Zhao et al. (2005) performed
simultaneous measurements of the 80105 km temperature
prole using a Na lidar and of the rotational temperature
of OH with the CEDAR mesospheric temperature mapper.
This conguration allowed them to retrieve the altitude of
the OH layer and record its variation, together with the
temperature throughout the night. The general trend that
they report in their Figs. 8 and 9 for January 14 and April
11, 2002 is an altitude decrease from 92 to 84 km during the
9 h course of the nights.
A comparison between the different methods is hardly
possible as the locations and areas of the regions of interest
are very different for the various measurements. Rocket
soundings provide measurements that are partly recorded
in situ. Satellite observations generally encompass large
areas. The optical path in the case of limb observations can
be as long as several hundreds kilometer. In all cases, it is
necessary to take into account the satellite motion during
the exposure time, which limits the spatial resolution of the
map. The data obtained with the stereo-imaging system
provides the OH airglow altitude and its variation over the
area of interest, i.e. a rectangular area of 50  50 km2.
The altitudes measured at 22 h15 on July 26 and 20 h15 on
July 28, 2006 are, respectively, 87.1 and 89.5 km. These
numbers, recorded in the rst part of both nights are
coherent with the data reported by Zhao et al. (2005) for
Haleakala, Maui, a site of comparable latitude, but in the
northern hemisphere.
6. Summary and conclusion
A new stereoscopic observational program aimed at
obtaining an original description of the dynamical regime
of the mesosphere at the level of the NIR nightglow
emissive layer was recently initiated. Two CCD cameras
operated in a vis-a`-vis mode obtained simultaneous images
of the OH vibrationalrotational emissions of the Dv 4
sequence. An observing campaign conducted in Peru in
July 2006 gave evidence of the presence of regularly aligned
bands with horizontal wavelengths of 21.1 km on July 26
and 24.6 km on July 28. The ground-based images are
transformed into satellite-type images equivalent to the
views that would be taken by a downward looking virtual
camera located at an altitude of 500 km. Using a
stereocorrelation method, a 3D representation of the
altitude of the photometric barycenter is presented for an
area of 50  50 km2. The surface is corrugated and shows
on July 26 a wave pattern with horizontal wavelength
l1 21.1 km, an average altitude and amplitude of
87.1 km, and 2.6 km, respectively. For the observations
of July 28 these values are, respectively, l2 24.6 km,

ARTICLE IN PRESS
1478

G. Moreels et al. / Planetary and Space Science 56 (2008) 14671479

mean altitude 89.5 km and amplitude 2 km. The kinetic

energy density of the gravity wave is 3.9  104 W/kg. This
value is comparable to energy densities retrieved using
partial reection radiowave data.
As mentioned by von Savigny et al. (2004), the synergy
between satellite instruments and ground-based observations frequently provides important complementary information. In the present case, the stereo-imaging method is
based on purely geometric calculations. It is a method for
determining the altitude of airglow layers that is independent and different from rocket and satellite methods. We
plan to conduct further measurements at future sites,
among which is probably lAsekrem (Hoggar, Africa), in
order to extend the geographical coverage of gravity wave
structures. Another application will be to conduct a stereoimaging program related to cirruses in conjunction with
lidar measurements made from the ground at Observatoire
de Haute-Provence and on board the CALIPSO satellite.
Acknowledgment
Sweetest thanks are due to Marie-Jeanne for invaluable
help during this scientic program.
References
Baker, D.J., Stair, A.T., 1988. Rocket measurements of the altitude
distributions of the hydroxyl airglow. Physica Scripta 37, 611622.
Baker, D.J., Waddoups, R.O., 1967. Rockets measurements of midlatitude night airglow emissions. Journal of Geophysical Research 72
(19), 48814883 (Corrections for this paper: 1968. Journal of
Geophysical Research 73(7), 25462547).
Bates, D.R., Nicolet, M., 1950. The photochemistry of atmospheric water
vapor. Journal of Geophysical Research 55, 301327.
Battaner, E., Molina, A., 1980. Turbopause internal gravity waves,
557.7 nm airglow, and eddy diffusion coefcient. Journal of Geophysical Research 85, 68036810.
Chanin, M.L., Hauchecorne, A., 1981. Lidar observation of gravity and
tidal waves in the middle atmosphere. Journal of Geophysical
Research 86, 97159721.
Clemesha, B.R., Kent, G.S., Wright, R.W., 1966. Laser probing the lower
atmosphere. Nature 209, 184.
Evans, W.F.J., Llewellyn, E.J., Vallance Jones, A., 1973. Altitude
distribution of hydroxyl bands of the Dv 2 sequence in the
nightglow. Canadian Journal of Physics 51, 12881291.
Faivre, M., 2004. Etude par imagerie stereoscopique et par interferometrie
des structures dynamiques de la haute atmosphe`re. The`se de
lUniversite de Franche-Comte.
Faivre, M., Moreels, G., Pautet, D., Clairemidi, J., 2003. Fourier analysis
of a virtual satellite view of the NIR atmospheric emission over
Europe. Advances in Space Research 32 (5), 843848.
Faivre, M., Moreels, G., Pautet, D., Clairemidi, J., Dumont, F., Lorin, O.,
Colas, F., 2007. Chemical and dynamical processes in the mesospheric
layer: rst results of stereoscopic observations. Journal of Geophysical
Research 112, DOI306.
Faugeras, O., 1993. Three dimensional Compact VisionA Geometric
Viewpoint. The MIT Press, Cambridge, MA, p. 695.
Hartley, R., Zisserman, A., 2003. Multiple View Geometry in Computer
Vision, second ed. Cambridge University Press, Cambridge, p. 672.
Hauchecorne, A., Chanin, M.-L., Wilson, R., 1987. Mesospheric
temperature inversions and gravity wave dynamics. Geophysical
Research Letters 14, 935939.

Hauchecorne, A., Bertaux, J.-L., Moreels, G., 2007. Determination of

mesopause temperature from GOMOS OH emission spectra. Paper
3P12.3, ENVISAT Symposium, Montreux, Switzerland, April 2007,
pp. 2327.
Herse, M., Moreels, G., Clairemidi, J., 1980. Waves in the OH emissive
layer: photogrammetry and topography. Applied Optics 19, 355362.
Hines, C.O., Tarasick, D.W., 1987. On the detection and utilisation of
gravity waves in airglow studies. Planetary and Space Science 35,
851856.
Krassovsky, V.I., 1972. Infrasonic variations of OH emission in the upper
atmosphere. Annales de Geophysique 28, 739746.
Ma, Y., Soatto, S., Kos ecka, J., Shankar Sastry, S., 2004. An Invitation
to 3-D Vision, From Images to Geometric Models. Springer, Berlin,
p. 526.
Manson, A.H., Gregory, J.B., Meek, C.E., 1981. Atmospheric waves
( 10 min30 days) in the mesosphere and thermosphere at
Saskatoon (521N, 1071W), October 1978September 1979. Planetary
and Space Sciences 29 (6), 615625.
Manson, A.H., et al., 2002. Seasonal variations of the semi-diurnal and
diurnal tides in the MLT: multi-year MF radar observations from
2701N, modeled tides (GSWM, CMAM). Annales de Geophysique
20, 661677.
McDade, I.C., Llewellyn, E.J., Murtagh, D.P., Greer, R.G.H., 1987.
Eton5. Simultaneous rocket measurements of the OH Meinel Du 2
sequence and (8,3) band emission proles in the nightglow. Planetary
and Space Science 35 (9), 11371147.
Melo, S.M.L., Lowe, R.P., Russell, J.P., 2000. Double-peaked hydroxyl
airglow proles observed from WINDII/UARS. Journal of Geophysical Research 105 (D10), 1239712403.
Meriwether, J.W., Dao, P.D., McNutt, R.T., Klemetti, W.,
Moskowitz, W., Davidson, G., 1994. Rayleigh lidar observations of
mesosphere temperature structure. Journal of Geophysical Research
99, 1697316987.
Moreels, G., Herse, M., 1977. Photographic evidence of waves around the
85 km level. Planetary and Space Science 25, 265273.
Moreels, G., Clairemidi, J., Faivre, M., Pautet, D., Rubio da Costa, F.,
Rousselot, P., Meriwether, J.W., Lehmacher, G.A., Vidal, E., Chau,
J.L., Monnet, G., 2007. Near-infrared sky background uctuations
at mid- and low latitudes. Experimental Astronomy, doi:10.1007/
s10686-008-9089-6.
Packer, D.M., 1961. Altitudes of the night airglow radiations. Annales de
Geophysique 17, 6775.
Pautet, D., Moreels, G., 2002. Ground-based satellite-type images of the
upper atmosphere emissive layer. Applied Optics 41, 823831.
Pautet, D., Moreels, G., Rousselot, P., Reyle, C., Clairemidi, J., Boice,
D.C., 2001. Evidence of a 2200 km extended wave system at the
mesopause level. Advances in Space Research 27, 11951200.
Pautet, D., Moreels, G., Rousselot, P., Reyle, C., Clairemidi, J., 2002.
Observing at Ouka meden through an extended wave system of the
upper atmosphere emissive layer. ASP Conference Series 266, 2835.
Peterson, A.W., Kieffaber, L.M., 1973. Infrared photography of OH
airglow structures. Nature 242, 321322.
Sijvee, G.G., Walterscheid, R.L., 1994. Six-hour zonally symmetric tidal
oscillations of the winter mesopause over the South Pole Station.
Planetary and Space Science 42 (6), 441453.
Stormer, C., 1910. Photographies des aurores boreales et nouvelle
methode pour mesurer leur altitude. Comptes Rendus de lAcademie
des Sciences 150, 16311634.
Stormer, C., 1911. Resultats des mesures photogrammetriques de
laltitude de laurore boreale a` Bosekop aux mois de fevrier et de
mars 1910. Comptes Rendus de lAcademie des Sciences 152,
11941196.
Swenson, G.R., Gardner, C.S., 1998. Analytical models for the responses
of the mesospheric OH* and Na layers to atmospheric gravity waves.
Journal of Geophysical Research 103 (D6), 62716294.
Swenson, G.R., Liu, A.Z., 1998. A model for calculating acoustic gravity
wave energy ux in the mesosphere from OH airglow. Geophysical
Research Letters 25, 477480.

ARTICLE IN PRESS
G. Moreels et al. / Planetary and Space Science 56 (2008) 14671479
Swenson, G.R., Espy, P.J., Taylor, M.J., 1995. ALOHA-93 measurements
of intrinsic AGW characteristics using airbone airglow imager and
ground based Na wind temperature lidar. Geophysical Research
Letters 22, 28412844.
Takahashi, H., Sahai, Y., Batista, P.P., 1984. Tidal and solar cycle effects
on the OI 5577 A Na D and OH (8,3) airglow emissions observed at
231S. Planetary and Space Science 32 (7), 897902.
Tarasova, T.M., 1963. Night Sky Emission-Line Intensity Distribution
With Respect to Height. Space Research, vol. III. North-Holland
Publ. Co., Amsterdam (pp. 162172).
Taylor, M.J., Bishop, M.B., Taylor, V., 1995. All-sky measurements of
short-period waves imaged in the OI (557.7 nm) and Na (589.2 nm) and
near infrared OH and O2 (0,1) nightglow emission during the
ALOHA-93 campaign. Geophysical Research Letters 22, 28332836.
Taylor, M.J., Pendleton Jr., W.R., Clark, S., Yakahashi, H., Gobi, D.,
Goldberg, R.A., 1997. Images measurements of short period gravity
waves at equatorial latitudes. Journal of Geophysical Research 102,
2628326299.
Teitelbaum, H., Vial, F., Manson, A., Giraldez, R., Massebeuf, M., 1989.
Nonlinear interaction between the diurnal and semidiurnal tides:

1479

terdiurnal and diurnal secondary waves. Journal of Atmospheric and

Terrestrial Physics 51, 627634.
von Savigny, C., Eichmann, K.-U., Llewellyn, E.J., Bovensmann, H.,
Burrows, J.P., Bittner, M., Hoppner, K., Offermann, D., Taylor, M.J.,
Zhao, Y., Steinbrecht, W., Winkler, P., 2004. First near-global retrievals
of OH rotational temperatures from satellite-based Meinel band
emission measurements. Geophysical Research Letters 31, L15111.
Witt, G., 1962. Height, structure and displacements of noctilucent clouds.
Tellus 14 (1), 118.
Yee, J.-H., Crowley, G., Roble, R.G., Skinner, W.R., Burrage, M.D.,
Hays, P.B., 1997. Global simulations and observations of O(1S),
O2(1S) and OH mesospheric nightglow emissions. Journal of
Geophysical Research 102 (A9), 1994919968.
Zhang, S.P., Shepherd, G.G., 1999. The inuence of the diurnal tide on
the O(1S) and OH emission rates observed by WINDII on UARS.
Geophysical Research Letters 26, 529532.
Zhao, Y., Taylor, M.J., Chu, X., 2005. Comparison of simultaneous Na
lidar and mesospheric nightglow temperature measurements and the
effects of tides on the emission layer heights. Journal of Geophysical
Research 110, D09S07.