Night sky quality monitoring in existing and plann=
ed
dark sky parks by digital cameras
Zolt&a= acute;n Kolláth1,*, Anita Dömény1
[1] 1Eötvös Lorán= d University, Savaria Department of Physics, Szombathely, 9700, Hungary
Received
05 December 2016, Accepted 08 February 2017
Abstract
A crucial part of the qualification of international dark sky places (IDSPs) is the objective measurement of night time sky luminance or radiance. Modern digital cameras provide an alternati= ve way to perform all sky imaging either by a fisheye lens or by a mosaic image taken by a wide angle lens. Here we present a method for processing raw cam= era images to obtain calibrated measurements of sky quality. The comparison of = the night sky quality of different European locations is also presented to demonstrate the use of our technique.
Keywords: Photometry, Ecological light pollution,
Night sky brightness, All-sky imaging
1.
Introduction
The qualification of the night sky is not an
easy task, especially at locations where the level of light pollution is low
[1]. Sky luminance strongly depends on weather conditions and natural pheno=
mena
like airglow. A standard
method for performing night sky brightness measurements is the use of Sky
Quality Meters (SQMs – [2]). However, single measurements with this d=
evice
provide only information about zenith luminance and possible light domes cl=
ose
to the horizon are hidden. A more precise measurement system for this task =
is
the method introduced by the US National Park Service [3], where an
astronomical CCD camera is used to produce a mosaic of the sky (see Figure =
1.).
Although this technique provides the most precise measurements, it is not
available for all organizations or individuals who are interested in the
formation of an IDSP. A cheaper and then a more feasible method for all sky
photometry is the application of DSLR (Digital Single Lens Reflex) or MILC
(Mirrorless Interchangeable-Lens Camera) digital cameras. The sensitivity a=
nd
repeatability of recent cameras give the possibility of scientific quality
measurements when thoroughly calibrated. Such a method was used in the
application of the Hungarian dark sky parks in Zselic and Hortobágy =
[4].
We started sky quality monitoring with a
digital camera in the Zselic Landscape Protection Area and in its neighborh=
ood
in 2008. Now we repeat the measurements with several cameras, included the =
one
which was used during the first survey (the old camera was stored for years
unused). Therefore, it is guaranteed that all the measurements during the 8
year long timespan are comparable. Our first results indicate that there we=
re
no significant changes in sky radiance inside the park.
The long term monitoring is especially
important, since the public lighting in the vicinity of the dark sky park h=
as
undergone major changes during this period of time. For example, the high
pressure sodium lighting was replaced with LED based luminaries in the larg=
est
city (Kaposvár) in the region. The ULR of the lighting has significa=
ntly
decreased as reflected in the shape of the light dome. However, due to the =
rebound effect (an economical phenome=
non
that improving energy efficiency may save less energy than expected), t=
he
effective light pollution is not decreased. Using the RGB information of the
photography, the variation in correlated colour temperature is easily
recognized in the light dome of the city. It indicates that scotopic measur=
es
of the light pollution in the dark sky park originated from Kaposvár=
are
increased. Fortunately, it has no noticeable effects in the Zselic region. =
We
plan further measurements in the Zselic region and the final conclusions wi=
ll
be published in a forthcoming paper.
The result of a similar lighting remodellin=
g of
another Hungarian city, Szekszárd was already published [5]. At that=
location
the sky brightness is decreased after remodelling in all colour bands due to
the improved shielding of the new luminaires. The improvement was significant in=
the
red channel and in a less extent in the blue channel - as expected from the
spectral change from Sodium Lighting to white LED¡¯s. These comparison
measurements have also demonstrated the usefulness of digital camera imager=
y in
light pollution research. Another application of our method was the
quantification the nocturnal sky brightness at a freshwater lake in Germany
[6], to help biomonitoring at the site.
We measured the brightness of the night sky
with calibrated DSLR cameras with a fisheye lens at different dark sky
locations in Europe. After
presenting our method of evaluating digital camera images, we provide some
examples of the use of the method on some selected locations.
Fig.
1. All sky mosaic composed of 10 individual exposures (Samyang 14mm f2.8 le=
ns
with Sony A7S camera). The strong green airglow is clearly visible. Locatio=
n:
Apuseni Nature Park, Padis.
2.
Methods
2.1. The
software DiCaLum
Instead
of developing a standalone software to analyze raw camera data, we decided =
to
use some existing open source computer software to perform some major tasks=
of
processing. The raw camera images are converted to standard PPM image files=
by
dcraw [7]. The platform of our processing program is GNU Octave [8]. GNU Oc=
tave
is a high-level language, developed for numerical computations. It has an
extensive toolkit for manipulating and displaying matrices which are
constructions that can represent images and radiance maps. Its computer language is mostly co=
mpatible
with Matlab. GNU Octave is ea=
sily
extensible via user-defined functions written in its own language. We used =
this
property to develop DiCaLum (the name originates from the words: Digital Ca=
mera
Luminance) a toolkit for night sky radiance and luminance measurements by D=
SLR
and MILC cameras.
DiCaLum is a library of GNU Octave
functions developed to read in digital camera RAW images and convert them to
standard arrays containing the luminance or radiance values of the pixels. =
For
example the following line:
&=
nbsp; [x
, jd] =3D nsu(¡°IMG_1111¡±, ¡°CR2¡±, L_SIG45, C_60D);
reads
in the raw image file ¡®IMG_1111.CR2¡¯. The constants L_SIG45 and C_60D sta=
nd for
the Sigma 4.5mm circular fisheye lens and for Canon EOS 60D camera
respectively. These constants are defined for the lenses and cameras for wh=
ich
calibration data are available. The returned ¡®x¡¯ array contains the radia=
nce
values in natural sky units (NSU – see [10]). Using the matrix arithm=
etics of
GNU Octave, it is easy to convert the NSU to other sky luminance/radiance
units. For example the statement ¡®y =3D 21.6 – 2.5*log10(x)¡¯ conver=
ts the array
to ¡®y¡¯ in mag/arcsec2 units. Additional functions are available to get
different kind of parameters, maxima, minima, mean values, etc. from the
images. The simplest code to plot a false colour image is the following:
&=
nbsp; dicalum;
&=
nbsp; [x
, jd] =3D nsu(¡°IMG_1111¡±, ¡°CR2¡±, L_SIG45, C_60D);
&=
nbsp; imagesc(log10(x),[-1,1]);
colorbar;
This
short code segment reads in the raw image file, then plots it in a logarith=
mic
false color scale in the 0.1-10NSU range.
The
plotting toolkit of Octave can be used to display the luminance map, or any
derivative quantities like the dependence of luminance on zenith angle. We
provide sample scripts to demonstrate the use of these functions. The user =
can
develop her or his own scripts for additional tasks relatively easily by
standard Octave commands.
2.2.
Calibration database in DiCaLum
The
functions provided in DiCaLum are useless without calibration data. Both the
lens and the camera should be calibrated.&=
nbsp;
In the program package we provide calibration values for the devices=
we
used. A major factor is the vignetting effect of the lens: pixel values at =
the
corners (or at the extremes of fisheye images) can be a factor of 4-10 lower
compared to the central pixel values even if the luminance is uniform. [9] tested the vignetting effect o=
f two
identical lenses mounted on two identical cameras and they demonstrated that
vignetting curves determined for one device can be reasonably used to corre=
ct
the vignetting effect of the other device of the same brand.
There
are different methods to fit the vignetting curve of a lens. A traditional
method is to use an integrating sphere which provides uniform luminance in =
the
whole field of view. However, this method depends on a high quality integra=
ting
sphere with a size that is compatible with the size of the lens. In the stu=
dies
presented in [9] 49 white and gray squares were arranged in a semi-circle, =
and
the luminance was measured in parallel with the photographic exposure. Then
based on the measured luminance values of the squares the relative luminanc=
e of
each square and so the vignetting curve can be calculated. The authors prov=
ided
the 6th order polynomial fits of the vignetting measurements.
Instead
of multiple patches of squares and a fixed camera, we used a single stabili=
zed
light source and rotated the camera to obtain the relative luminance values=
at
different locations in the field of view. Our source is a set of six white =
LEDs
with a double diffusor. This provides an almost homogeneous luminance on a =
45
mm diameter disk. Our standard procedure is to take 200-300 images at a
distance of 4-5 meters from the source in a dark room. The lens is focused =
to
infinity, which provides a slightly out of focus images. Figure 2. displays=
a
typical measurement sequence. Here the individual images were negated and p=
lot
on top of each other. For the processing we used DiCaLum itself: all images
were processed separately, the location of the center of the light source a=
nd
its mean instrumental brightness was determined. From the sensor coordinate and rel=
ative
luminance the determination of the vignetting curve is a straightforward ta=
sk.
Fig. 2. The negative images of the vignetting
calibration source plotted on top of each other. The axes represent the pix=
el
distance from the image center (Lensbaby 5.8mm fisheye lens attached to a S=
ONY
A7S camera).
Although
it is possible to fit the measured vignetting curve by a polynomial, the or=
der
of the polynomial needs to be high for a precise fit, at least for fisheye
lenses. In the GNU Octave system, it is more feasible to use a spline fit to
the data and construct a piecewise polynomial structure. We used a cubic sp=
line
with 6 knots – the locations of the inside knots were selected to pro=
vide
optimal fit to the data.
Fig. 3. The relative luminance (Lr) as a functi=
on
of the distance from the image center of the sensor. Left panel: Sigma 4.5mm
lens with f2.8 and f3.2 aperture. Solid line: spline fit to our data, dashed
line: polynomial fit presented in [9]. Right panel: Comparison of the
vignetting curve of two circular fisheye lenses.
Piecewise
polynomial structures are handled in a straightforward way in GNU Octave, t=
hen
it is easy to code the vignetting correction. Figure 3. displays the vignet=
ting
measurements and their fits for two circular fisheye lenses: the Sigma 4.5mm
F2.8 EX DC HSM and the Lensba=
by
5.8mm f/3.5. For the 4.5mm Sigma lens we provide data for the full open (f2=
.8) and
the one stop closed (f3.2) apertures. The 5.8 mm Lensbaby lens is a manual =
one
with now ¡®clicks¡¯ on the aperture ring, so it is recommended to use only =
the
wide open setting – otherwise the repeatability of the measurement is=
not
guaranteed. In general, the l=
enses
have less vignetting when stopped down (higher f-number), but for the low l=
ight
level measurements the widest aperture is needed to reduce exposure time. F=
or
comparison, we present the 6th order polynomial vignetting curve published =
in
[9] by dashed curves in the left panel of Figure 3. At f3.2 the agreement of
the vignetting curves is excellent - at f2.8 it reaches 5% at the center of=
the
curve.
The
calibration of the camera is performed with a lens already corrected for the
vignetting effect. We prefer to use high quality luminance meters to calibr=
ate
the camera system in laboratory conditions. Then the camera is calibrated f=
or a
given set of light-sources in standard photometric (cd/m2) units. The other
possible way to calibrate the camera is performing astronomical photometry =
with
a telephoto lens and processing this data to get the sky background in
mag/arcsec2 units. We prefer the calibration in laboratory and plan to do t=
he
calibration with a sensitive spectroradiometer and different type of source=
s.
In this way the color dependence of the measurements can be estimated more
precisely.
3.
Test results
To
demonstrate the usefulness of digital camera measurements, we selected two
locations with recent observations. These locations are the Montsec
Astronomical Park in Spain and Padis Area in the Apuseni Nature Park in
Romania. Both locations have low light pollution levels and they have a sim=
ilar
elevation above the sea level. The New World Atlas of Artificial Sky Bright=
ness
[11] predict a little bit darker sky in Apuseni (dark blue color) than in
Montsec (light blue color).
The
luminance maps of the sky in false color are presented in Figure 4. The sky
brightness as a function of zenith angle was also calculated. For the
derivation of this function, we selected concentric rings at different zeni=
th
angles of a width of 2 degrees. For each rings the maximum, the minimum and=
the
mean value of the radiance is calculated. On the right panels of Figure 4. e
present all these three curves. This method provides a compact summary on t=
he
luminance distribution in the sky.
During
the beginning of the first night at the Padis observations there was a very
strong green airglow activity. It increased the sky luminance significantly
everywhere on the sky. The upper two panels in Figure 4. clearly show the
effect of increased airglow on sky luminance.
The
zenith brightness at Montsec was excellent (around 22 mag/arcsec2 with no M=
ilky
Way on the sky), but in the luminance map a clear gradient is visible. Clos=
e to
the horizon, it is definitely brighter than at Padis. Around zenith the sky brightness i=
s increased
due to the presence of the Milky Way in the Padis measurements, although to
some extend it can be handled with the minimum radiance values for all zeni=
th
angles, as they represent the portion of the sky which less affected by MW.=
The
effect of the Milky Way is clearly displayed by the deviation of the maximum
and minimum curves.
|
|
|
|
|
|
|
|
|
Fig.
4. Luminance maps calculated from the all-sky (right panels) and the variat=
ion
of luminance vs zenith angle (right panels).
To
compare the three cases presented in Figure 4, we also show the mean zenith
angle - luminance curves in figure 5. The different gradient of sky radianc=
e at
the two locations is clearly demonstrated by comparing the Montsec and Padis
curves. The difference can be originated from different weather and air
conditions, although the available satellite data do not show any clear
difference in the approximate aerosol optical depth for the two locations.
Further radiative transfer modellings of the situations are also planned to
understand the real mechanism behind the observed difference.
Figure
5. also shows the difference between the sky luminance values with different
level of airglow activity. Even with low solar activity, we experienced a v=
ery
strong effect on sky radiance. The variation of sky radiance due to the air=
glow
is around 20% between zenith angles 30 and 70 degrees. The difference incre=
ases
from 0.15 mag/arcsec2 at 30 degrees to 0.25 mag/arcsec2 at 70 degrees. The zenith values are also affected b=
y the
different location of the Milky Way, the airglow must have smaller effect t=
here
than the difference shown in the figure.
Fig. 5. Mean sky brightness vs zenith angle =
211;
comparison of the three observations.
4.
Conclusion
We
developed a computer tool (DiCaLum), based on the GNU Octave programming
language, to evaluate digital camera (DSLR or MILC) images. The basic functions of DiC=
aLum
can be easily extended by standard GNU Octave programming. For example, such
extensions provide publication quality plots of the observations.
With some example measurements, it =
is
demonstrated that:
We
plan to distribute the code as a free software in the near future. The miss=
ing
part is to make the full calibration of a wide range of cameras and lenses.=
Acknowledgements
This
work was partly supported by the EU COST Action ES1204 (Loss of the Night
Network). The Montsec measurements are founded by the Stars4All awareness
platform.
References
[1]
Barentine, J.C. (2016). Going for The Gold: Quantifying and Ranking Visual
Night Sky Quality in International Dark Sky Places. IJSL (in press).
[2]
Cinzano, P. (2005). Night sky photometry with sky quality meter. ISTIL Int.
Rep, 9.
[3]
Duriscoe, D. M., Luginbuhl, C. B., & Moore, C. A. (2007). Measuring
night-sky brightness with a wide-field CCD camera. Publications of the
Astronomical Society of the Pacific, 119(852), 192.
[4]
Kolláth, Z. (2010). Measuring and modelling light pollution at the Z=
selic
Starry Sky Park. In Journal of Physics: Conference Series (Vol. 218, No. 1,=
p.
012001). IOP Publishing.
[5]
Kolláth, Z., Dömény, A., Kolláth, K., & Nagy,=
B. (2016). Qualifying
lighting remodelling in a Hungarian city based on light pollution effects.
Journal of Quantitative Spectroscopy and Radiative Transfer, 181, 46-51.
[6]
Jechow, A., Hölker, F., Kolláth, Z., Gessner, M. O., & Kyba=
, C. C. (2016).
Evaluating the summer night sky brightness at a research field site on Lake
Stechlin in northeastern Germany. Journal of Quantitative Spectroscopy and
Radiative Transfer, 181, 24-32.
[7]
Coffin, D. (2011). Decoding raw digital photos in Linux. Available from:
http://www. cybercom. net/~ dcoffin/dcraw/(date accessed: September the 13.,
2013). from: https://www.cybercom.net/~dcoffin/dcraw/
[8]
Eaton, J. W., Bateman, D., & Hauberg, S. (2007). GNU Octave version 3.0=
. 1
manual: a high-level interactive language for numerical computations. SoHo
Books.
[9] Cauwerts, C., PhD, M. B., & Deneyer, A. (2012). Comparison of the vignetting effects of two identical fisheye lenses. LEUKOS, 8(3), 181-203.<= o:p>
[10]
Kyba, C. C., Tong, K. P. et al. (2015). Worldwide variations in artificial
skyglow. Scientific reports, 5, 8409.
[11]
Falchi, F., Cinzano, P., Duriscoe, D., Kyba, C. C., Elvidge, C. D., Baugh, =
K.,
Furgoni, R. et al. (2016). The new world atlas of artificial night sky
brightness. Science advances, 2(6), e1600377