THE FIRST RELEASE COSMOS OPTICAL AND NEAR-IR DATA AND CATALOG
1
P. Capak,
2
H. Aussel,
3,4
M. Ajiki,
5
H. J. McCracken,
3,6
B. Mobasher,
7
N. Scoville,
2,3
P. Shopbell,
2
Y. Taniguchi,
5,8
D. Thompson,
2,9
S. Tribi ano,
10,11
S. Sasaki,
2,5,8
A. W. Blain,
2
M. Brusa,
12
C. Carilli,
13
A. Comastri,
14
C. M. Carollo,
15
P. Cassata,
16
J. Colbert,
17
R. S. Ellis,
2
M. Elvis,
18
M. Giavalisco,
7
W. Green,
2
L. Guzzo,
16
G. Hasinger,
12
O. Ilbert,
19
C. Impey,
20
K. Jahnke,
21
J. Kartaltepe,
19
J.-P. Kneib,
22
J. Koda,
2
A. Koekemoer,
7
Y. Komiyama,
23
A. Leauthaud,
2,22
O. Lefevre,
22
S. Lilly,
15
C. Liu,
10
R. Massey,
2
S. Miyazaki,
24
T. Murayama,
5
T. Nagao,
23,25
J. A. Peacock,
26
A. Pickles,
27
C. Porci ani,
15
A. Renzini,
28,29
J. Rhodes,
2,30
M. Rich,
31
M. Salvato,
2
D. B. Sanders,
19
C. Scarlata,
15
D. Schiminovich,
32
E. Schinnerer,
21
M. Scodeggio,
33
K. Sheth,
2,17
Y. Shioya,
8
L. A. M. Tasca,
22
J. E. Taylor,
2
L. Yan,
17
and G. Zamorani
34
Received 2006 September 23; accepted 2007 April 6
ABSTRACT
We present imaging data and photometry for the COSMOS survey in 15 photometric bands between 0.3 and 2.4 m.
These include data taken on the Subaru 8.3 m telescope, the KPNO and CTIO 4 m telescopes, and the CFHT 3.6 m
telescope. Special techniques are used to ensure that the relative photometric calibration is better than 1% across the field of
view. The absolute photometric accuracy from standard-s tar measurements is found to be 6%. The absolute calibration is
corrected using galaxy spectra, providing colors accurate to 2% or better . Stellar and galaxy colors and counts agree well
with the expected values. Finally, as the first step in the scientific analysis of these data we construct panchromatic number
counts which confirm that both the geometry of the universe and the galaxy population are evolving.
Subject headinggs: c osmology: o bservations — galaxies: evolution — large-scale structure of universe — surveys
1. INTRODUCTION
Advances in astronomy are often driven by improved accu-
racy and precision along with increases in sensitivity and area of
the available data. The Canada-France Redshift Survey (CFRS;
Lilly et al. 1995), the Hawaii Deep Surveys (Cowie et al. 1999),
and the Hubble Deep Fields (HDFs; Williams et al. 1996;
Casertano et al. 2000) were the first deep imaging and spectros-
copic surveys aimed at understanding galaxy formation and evo-
lution. These discovered the global decline in star formation at
1
Based in part on observations with the NASA/ ESA Hubble Space Telescope,
obtained at the Space Telescope Science Institute, which is operated by AURA,
Inc., under NASA contract NAS5-26555; the Subaru telescope, which is operated
by the National Astronomical Observatory of Japan; the MegaPrime/ MegaCam, a
joint project of CFHTand CEA/ DAPNIA at the Canada-France-Hawaii Telescope,
which is operated by the National Research Council of Canada, the Institute Na-
tional des Science de l’Univers of the Centre National de la Recherche, and the
University of Hawaii; and the Kitt Peak National Observatory, Cerro Tololo Inter-
American Observatory, and the National Optical Astronomy Observatory, which is
operated by the Association of Universities for Research in Astronomy, Inc., under
cooperative agreement with the National Science Foundation.
2
California Institute of Technology, MC 105-24, 1200 East California Bou-
levard, Pasadena, CA 91125.
3
Visiting Astronomer, University of Hawaii, 2680 Woodlawn Drive, Hono-
lulu, HI 96822.
4
AIM, Unite
´
Mixte de Recherche, UMR 7158, CNRS, Universite
´
Paris VII,
CEA /Saclay, F-91191 Gif-sur-Yvette, France.
5
Astronomical Institute, Graduate School of Science, Tohoku University,
Aramaki, Aoba, Sendai 980- 8578, Japan.
6
Institut d’Astrophysique de Paris, UMR 7095, CNRS, Universite
´
Pierre et
Marie Curie, 98 bis Boulevard Arago, F-75014 Paris, France.
7
Space Telescope Science Institute, 3700 San Martin Drive, Baltimore, MD
21218.
8
Physics Department, Graduate School of Science and Engineering, Ehime
University, 2-5 Bunkyo-cho, Matsuyama 790 -8577, Japan.
9
LBT Observatory, University of Arizona, 933 North Cherry Avenue, Tucson,
AZ 85721-0065.
10
American Museum of Natural History, Central Park West at 79th Street,
New York, NY 10024.
11
CUNY Borough of Manhattan Community College, 199 Chambers Street,
New York, NY 10007.
12
Max-Planck-Institut fu
¨
r Extraterrestrische Physik, D-85478 Garching,
Germany.
13
National Radio Astronomy Observatory , P.O. Box O, Socorro, NM 87801-0387.
14
INAFYOsservatorio Astronomico di Bologna, via Ranzani 1, I- 40127
Bologna, Italy.
15
Department of Physics, ETH Zurich, CH-8093 Zurich, Switzerland.
16
INAFY Osservatorio Astronomico di Brera, via Bianchi 46, I-23807 Merate
(LC), Italy.
17
Spitzer Science Center, California Institute of Technology, Pasadena, CA
91125.
18
Harvard-Smithsonian Center for Astrophysics, 60 Garden Street, Cam-
bridge, MA 02138.
19
Institute for Astronomy, 2680 Woodlawn Drive, University of Hawaii,
Honolulu, HI 96822.
20
Steward Observatory, University of Arizona, 933 North Cherry Avenue,
Tucson, AZ 85721.
21
Max-Planck-Institut fu
¨
r Astronomie, Ko¨nigstuhl 17, D-69117, Heidelberg,
Germany.
22
Laboratoire d’Astrophysique de Marseille, BP 8, Traverse du Siphon,
F-13376 Marseille Cedex 12, France.
23
National Astronomical Observatory of Japan, 2-21-1 Osawa, Mitaka, Tokyo
181-8588, Japan.
24
Subaru Telescope, National Astronomical Observatory of Japan, 650
North A’ohoku Place, Hilo, HI 96720.
25
INAFY Osservatorio Astrofisico di Arcetri, Largo Enrico Fermi 5, I-50125
Firenze, Italy.
26
Institute for Astronomy, University of Edinburgh, Royal Observatory,
Blackford Hill, Edinburgh EH9 3HJ, UK.
27
Caltech Optical Observatories, MS 320- 47, California Institute of T echnology,
Pasadena, CA 911 25.
28
European Southern Observatory, Karl-Schwarzschild-Strasse 2, D-85748
Garching, Germany.
29
Dipartimento di Astronomia, Universite
´
di Padova, V icolo dell’Osservatorio 2,
I-35122 Padua, Italy .
30
Jet Propulsion Laboratory, Pasadena, CA 91109.
31
Department of Physics and Astronomy, University of California, Los
Angeles, CA 90095.
32
Department of Astronomy, Columbia University, MC 2457, 550 West 120th
Street, New York, NY 10027.
33
INAFY IASF Milano, via Bassini 15, I-20133 Milano, Italy.
34
INAFY Osservatorio Astronomico di Bologna, via Ranzani 1, I- 40127
Bologna, Italy.
99
The Astrophysical Journal Supplement Series, 172:99Y 116, 2007 September
# 2007. The American Astronomical Society. All rights reserved. Printed in U.S.A.
z < 1 and showed that this was due to star formation occurring in
smaller galaxies at later times (Lilly et al. 1996; Cowie et al.
1999), a phenomenon often referred to as ‘‘cosmic downsizing.’’
At the same time Steidel et al. (1996, 1999, 2003) used the Lyman
break galaxy color selection technique to identify galaxies at high
redshift, dramatically improving the efficiency of spectroscopic
surveys at z > 3. Other selections such as the BzK (Daddi et al.
2004), BX/BM (Adelberger et al. 2004), and distant red galaxy
selections (Franx et al. 2003) have allowed for efficient sorting of
1 < z < 3galaxies.
Photometric redshifts are the logical extension of color se-
lection by estimating redshifts and spectral energy distributions
(SEDs) from many photometric bands. Unlike color selection,
photometric redshifts take advantage of all available informa-
tion, enabling redshift estimates along with the age, star forma-
tion rate, and mass. Unfortunately, photometric redshifts are also
susceptible to systematics in all bands. This increases the calibra-
tion requirements, especially the required photometric accuracy,
for modern cosmological surveys such as the Great Observatories
Origins Deep Survey (GOODS; Giavalisco et al. 2004), the Gal-
axy Evolution from Morphology and SEDs (GEMS) survey (Rix
et al. 2004), and the C osmic Evolution Survey, or COSMOS
(Scoville et al. 2007b).
GOODS and GEMS are designed to study the evolution of
galaxies with look-back time, whereas COSMOS is designed to
probe the evolution of galaxies in the context of their large-scale
structure out to moderate redshift. The desire to study large-scale
structure in COSMOS necessitates a 2 deg
2
area with deep pan-
chromatic data. Such data have been collected at nearly every
observable wavelength from the X-rays to the radio. The study
of large-scale structures places strong calibration requirements
on the COSMOS data; for example, spatial variations in photom-
etry and astrometry must be kept to a minimum, typically less
than 1% for photometry, to ensure high-quality photometric red-
shifts and 0.01
00
positional accuracy for astrometry to enable mea-
surements of weak lensing and correlation functions. Meeting
these calibration requirements is often difficult, as multiple instru-
ment pointings are used to cover the field.
This paper concentrates on the ground-based data reduction,
the multiband optical and near-infrared catalog, and the steps
taken to ensure a high level of photometric consistency. The ob-
serving strategy for the Subaru Suprime-Cam observations, which
form the bulk of our ground-based data, are discussed separately
in Taniguchi et al. (2007). In addition, the absolute photometric
and astrometric system used here is defined in H. Aussel et al.
(2007, in preparation).
An overview of the COSMOS project and its goals are given
in Scoville et al. (2007b). Details of the Hubble Space Telescope
(HST ) observations, including the Advanced Camera for Surveys
(ACS), the Wide Field Planetary Camera 2 ( WFPC2), and the
Near Infrared Camera and Multi-Object Spectrometer (NICMOS)
are found in Scoville et al. (2007a). The ACS data acquisition and
reduction are detailed in Koekemoer et al. (2007), and a mono-
chromatic catalog based only on the HST ACS observations is
presented in Leauthaud et al. (2007). Observations at other wave-
lengths consist of X-ray observations with XMM-Newton ( Hasinger
et al. 2007), ultraviolet observations with GALEX ( Zamojski
et al. 2007), mid-infrared observations with the Spitzer Space
Te lescope (Sanders et al. 2007), submillimeter observations
from the Caltech Submillimeter O bservatory (CSO;Aguirre
et al. 2007) and the I nstitut de Radioastronomie Millime
´
trique
(IRAM) 30 m telescope (Bertoldi et al. 2007), and radio obser-
vations with the Very Large Array (VLA; Schinnerer et al. 2004,
2007).
We begin by presenting an overview of the various data sets
and p hotometric systems, the imaging data products, and the
data reduction in x 2. Point-spread function (PSF) matching is
covered in x3, and the generation of a multicolor catalog is pre-
sented in x 4. Finally, in x 5 we conduct several quality checks,
and suggest several corrections to the absolute photometry.
2. OBSERVATIONS AND DATA REDUCTION
The present COSMOS data were collected on a variety of
telescopes and instruments, as well as from the Sloan Digital Sky
Survey (SDSS) second data release (DR2) archive (Abazajian et al.
2004). This paper covers the processing of the data obtained with
Suprime-Cam ( Komiyama et al. 2003) on the Subaru 8.3 m tele-
scope, Megaprime (Aune et al. 2003; Boulade et al. 2003) on the
3.6 m Canada-France-Hawaii Telescope (CFHT), FLAMINGOS
(Elston 1998) on the Kitt Peak National Observatory (KPNO) 4 m
telescope, and the Infrared Side Port Imager ( ISPI; Probst et al.
2003) on the Cerro Tololo Inter-American Observatory (CTIO)
4 m telescope during the 2004Y2005 observing season.
The telescopes and instruments used for the COSMOS survey
are presented in Table 1. A survey efficiency is given for each
telescope-instrument pair to allow comparisons between the vari-
ous data sets. The survey efficiency is defined as the telescope
collecting area multiplied by the detector imaging area (deg
2
m
2
)
and does not include variations in detector sensitivity, sky back-
ground, or field geometry. This number is most useful for com-
paring observations taken in similar bands. The filter transmission
profiles, including atmospheric transmission, telescope reflectiv-
ity, instrument optical transmission, filter transmission, and de-
tector sensitivity, are plotted in Figures 1 and 2 in units of relative
detector quantum efficiency normalized to 1 at the peak.
The Suprime-Cam, Megaprime, SDSS photometric, and SDSS
survey cameras have filters distinct from each other and from the
Landolt standard star system. Even the SDSS photometric tele-
scope and SDSS survey telescope filter sets differ from one an-
other by 2%Y 4%. To differentiate between these filter systems we
TABLE 1
Telescopes Used for COSMOS Optical/ IR Data in 2005Y2006
Telescope
Telescope Diameter
(m) Instrument
Field of View
(arcmin) Instrument Wavelength Sensitivity Survey Efficiency
a
Filters Used
CFHT ........... 3.6 Mega-Prime 56.4 ; 57.6 3200Y11000 8 9.19 u
, i
CTIO ............ 4 ISPI 10.2 ; 10.2 0.9Y 2.5 m 0.37 K
s
HST .............. 2.5 ACS WFC 3.4 ; 3.4 4000Y11000 8 0.02 F814W
KPNO........... 4 FLAMINGOS 10.8 ; 10.8 0.9Y2.5 m 0.41 K
s
SDSS ............ 2.5 SDSS 25 ; 13.5 ; 13.5 3200Y11000 8 2.49 u, g, r, i, z
Subaru .......... 8.3 Suprime-Cam 34 ; 27 4000Y11000 8 13.8 B
J
, V
J
, g
+
, r
+
, i
+
, z
+
, NB816
a
Defined as the telescope collecting area multiplied by the imaging area in square degrees.
CAPAK ET AL.100 Vol. 172
use a plus sign superscript for the Suprime-Cam Sloan filters and
an asterisk superscript for the Megaprime Sloan filters; no super-
script is used for the SDSS survey filters. The designations U, B, V,
R,andI are used for the Landolt-Johnson-Cousins set, while B
J
and V
J
are used for the Suprime-Cam Johnson set. Conversions
between these systems are discussed in H. Aussel et al. (2007, in
preparation).
The wavelength range, depth, and image quality for all data
presently reduced and included in the versi on 2.0 optical/IR cat-
alog are given in Table 2 and plotted in Figure 3. The depth quoted
is for a 5 measurement in a 3
00
aperture of an isolated point source
at the median seeing given in T able 2. This should be viewed as an
optimistic estimate since most objects are extended and many are
confused with neighboring sources. Taniguchi et al. (2007) present
a discussion of detection sensitivities and completenes s for variou s
Subaru filters. The median photometric depths in the COSMOS i
þ
selected catalog are discusse d in x4. T a ble 2 also gives a first-order
offset to the Vega system; however, a color term must be applied
to get the true Landolt-Vega system magnitudes; these are given
in H. Aussel et al. (2007, in preparation).
2.1. Data Products
We took speci al care in producing data products that simplify
analysis and are tractable on contemporary computers.
35
To do
this we defined a common grid of subimages for all data products.
The starting point for this grid is the COSMOS astrometric cat-
alog, which covers 4 deg
2
(H. Aussel et al. 2007, in preparation)
and is larger than all present or planned COSMOS data sets. The
area is divided into 144 sections of 10
0
; 10
0
, and each section is
covered by an image of size 4096 ; 4096 pixels with a pixel scale
of 0.15
00
. Therefore, adjacent tiles overlap each other by 14.4
00
on
all sides. As a result, the vast majority of objects can be analyzed
on a single image. The layout of the image tiles is shown in Fig-
ure 4. The pixel scale was chosen to be an integer multiple of the
0.05
00
scale used for the HST ACS images. All images and noise
maps are scaled to units of nanojanskys per pixel, which corre-
sponds to a magnitude zero point of 31.4.
For each Subaru and SDSS band an image with the original
PSF and a PSF-homogenized across the field within that band is
Fig. 2.—Transmission profile of the K
s
-band filter from KPNO FLAMINGOS
and CTIO ISPI. This profile is normalized to a maximum throughput of 1 and
includes the transmission of the atmosphere, the telescope, the camera optics, the
filter, and the detector.
Fig. 1.—Filter transmission profiles for the COSMOS optical data set from
CFHT, Subaru, and HST as of 2005 April. These profiles are normalized to a
maximum throughput of 1 and include the transmission of the atmosphere, the
telescope, the camera optics, the filter, and the detector. The HST F475W data
only cover the central 9
0
; 9
0
area; details are given in Scoville et al. (2007a). The
SDSS Abazajian et al. (2004) and Johnson-Cousins systems used by Landolt
(1992) are shown for comparison. Notice the significant differences between the
Johnson-Cousins, SDSS, and other systems. Color conversions are clearly needed
to transform from the COSMOS system to the standard star systems. These are
given in H. Aussel et al. (2007, in preparation).
35
All data products discussed in this paper are publicly available at http://
irsa.ipac.caltech.edu/data /COSMOS/.
COSMOS OBSERVATIONS: OPTICAL/ NEAR-IR DATA 101No. 1, 2007
provided. For the Subaru B
J
, r
þ
, and i
þ
bands, which have ex-
ceptional image quality (0.5
00
Y0.8
00
seeing), a ‘‘best seeing’’ im-
age is also provided. The CFHT images were taken in queue
observing mode, ensuring a consistent PSF for all observations, so
only an original PSF image is provided for these data. Finally, due
to the large variation of the PSF in the CTIO and KPNO data,
only a PSF-homogenized image is provided.
In addition, rms noise maps are provided for each filter. These
are on the same tiling scheme and flux scale as the images. The
rms maps include noise contributions from photon noise, back-
ground subtra ction, flat-fielding, defect masking, saturation, and
cosmic-ray removal. They do not include the photon noise con-
tribution from object flux.
2.2. Subaru Suprime-Cam
The Suprime-Cam instrument ( Komiyama et al. 2003) on the
Subaru 8.3 m telescope has a 34
0
; 27
0
field of view. The camera
has 10 2K ; 4K Lincoln Labs CCD detectors, which have good
sensitivity between 4000 and 10000 8.NineSuprime-Cam
pointings were required to co ver the COSMOS field. Durin g
2004 and 2005, data were obtained in the B
J
, V
J
, g
+
, r
+
, i
+
, and z
+
broadband and the NB816 narrowband filters. These Suprime-
Cam observations, which required special planning, are detailed
in Taniguchi et al. (2007). Further observations in 11 300 8 in-
termediate bands, IA427, IA464, IA484, IA505, IA527, IA624,
IA679, IA709, IA738, IA767, and IA827, and one narrow band,
NB711, were obtained in 2006 and 2007. These new observations
TABLE 2
Data Quality and Depth
Filter Name
Central Wavelength
(8)
Filter Width
(8)
Seeing Range
(arcsec) Depth
a,b
Saturation Magnitude
b
Offset from Vega System
c
u.................................. 3591.3 550 1.2Y 2.0 22.0 12.0 0.921
u
................................ 3797.9 720 0.9 26.4 15.8 0.380
B
J
................................ 4459.7 897 0.4Y 0.9 27.3 18.7 0.131
g.................................. 4723.1 1300 1.2Y1.7 22.2 12.0 0.117
g
+
................................ 4779.6 1265 0.7Y 2.1 27.0 18.2 0.117
V
J
................................ 5483.8 946 0.5Y1.6 26.6 18.7 0.004
r.................................. 6213.0 1200 1.0Y1.7 22.2 12.0 0.142
r
+
................................ 6295.1 1382 0.4 Y1.0 26.8 18.7 0.125
i .................................. 7522.5 1300 0.9Y1.7 21.3 12.0 0.355
i
+
................................ 7640.8 1497 0.4 Y 0.9 26.2 20.0
d
0.379
i
................................ 7683.6 1380 0.94 24.0 16.0 0.380
F814W........................ 8037.2 1862 0.12 24.9
e
18.7 0.414
NB816........................ 8151.0 117 0.4 Y1.7 25.7 16.9 0.458
z.................................. 8855.0 1000 1Y1.7 20.5 12.0 0.538
z
+
................................ 9036.9 856 0.5Y1.1 25.2 18.7 0.547
K
s
................................ 21537.2 3120 1.3 21.6 10.0 1.852
a
This is 5 in a 3
00
aperture for an isolated point source at the native seeing.
b
In AB magnitudes.
c
AB magnitude = Vega magnitude + offset. This offset does not include the color conversions to the Johnsons-Cousins system used by Landolt (1992).
d
Compact objects saturate at i
þ
< 21:8 due to the exceptional seeing.
e
The sensitivity for photometry of an optimally extracted point source is 27.1; for optimal photometry of a 1
00
galaxy it is 26.1.
Fig. 3.—Background-limited depth of COSMOS observations in the ultra-
violet, optical, and infrared. The CFHT, KPNO/CTIO, and Subaru depths are 5
in a 3
00
aperture. The SDSS depths are those quoted in Abazajian et al. (2004). The
depth of the HST ACS observations is given for a 3
00
aperture and a 0.15
00
aperture
with a point source. A 3
00
aperture is optimal for color measurements, while the
0.15
00
aperture is the 5 detection limit for point sources. The Spitzer IRAC
depths are those expected at 5 in a 3
00
aperture for observations taken in 2006.
The GALEX depths are from Zamojski et al. (2007).
Fig. 4.—Layout of image tiles for the COSMOS field.
CAPAK ET AL.102 Vol. 172
have been reduced using the prescription described here, but will
be presented elsewhere.
Objects brighter than 19 mag are saturated in a typical expo-
sure, and under good seeing the saturation level can drop to 22 mag
in long exposures. As a result, it is extremely difficult to astrome-
trically calibrate these data against external astrometric catalogs
such as the SDSS (Abazajian et al. 2004) and USNO-B1.0 (Monet
et al. 2003), which only reach 21 mag. To mitigate this limitation,
a series of short exposures were taken in each band.
2.2.1. Initial Calibration
The first step in our Suprime-Cam data reduction is to measure
a bias level from the overscan region and subtract it from all the
images. Then all bad or saturated pixels are masked. Next, a me-
dian bias frame is constructed from overscan-corrected frames.
Following the overscan correction, this bias frame is subtracted
from data and flat frames in order to remove bias structures. In
particular, the bias level increases near the edges of the CCDs
farthest from the readout register.
A median dome flat for each band is then constructed from
10Y20 bias-subtracted flat-field images. The median dome flats
and median biases are inspected for bad pixels, charge traps, and
other defects that need to be masked. The appropriate median
dome flat, with all defects masked, is then used to normalize all
data frames. Finally, portions of the image vignetted by the guide
probe are masked from all data frames using the position of the
guide probe recorded in the image header.
After the initial calibration catalogs are generated for every
data frame with the IMCAT
36
hfindpeaks and apphot routines
using 5
00
diameter apertures. The large aperture is chosen to min-
imize photometric variations caused by changes in seeing. This
catalog is then used to generate an object mask for each frame
and to calculate the astrometric solution.
The night sky is subtracted in a two-step process to account for
fringing and scattered light. First, a normalized median sky frame
is constructed for each night. The median frame is generated by
masking objects in all frames, normalizing every frame to the
same median flux, and finally median-combining the normalized
images. The median sky is then scaled to the median background
level in each frame and subtracted. This removes both night sky
illumination and fringing.
After subtracting the median sky, residual scattered light is
visible on the images. This residual light affects both the overall
flat field and the background of each individual frame. A correc-
tion to the flat field is described in x 2.2.2, while the light in each
frame is subtracted by masking objects and measuring the me-
dian of the residual background in 128 ; 128 pixel squares. A
background image is then generated by tesselating over the grid
of medians. After subtracting this background, no visible sky
structure is left on the individua l frames. However, this step cre-
ates negative halos around bright stars and very extended galaxies
due to imperfect masking. Fortunately, the amplitude of these ha-
los is similar in all frames and can be accounted for as a residual
background in the combined images.
After sky subtraction, an astrometric solution is calculated sep-
arately for all exposures and CCDs by matching the object cata-
logs to the COSMOS astrometric catalog (H. Aussel et al. 2007, in
preparation) using a fourth-order, two-dimensional polynomial.
The polynomial fits are improved by removing mismatched ob-
jects in an iterative fashion until the solution converges (typically
in two iterations). The resulting scatter between the fit positions
and the final astrometry is always less than 0.2
00
at the 1 level,
independent of position.
Using the astrometric solutions, defects around charge bleeds
from saturated stars are masked using a list of bright stars from
the SDSS and the USNO-B1.0. Cosmic-ray events are removed
by detecting sharp edges in the images. Finally, every frame is vi-
sually inspected to remove internal reflections, satellites, aster-
oids, and other false objects. Once all masking is complete, a new
photometric catalog is generated containing only isolated objects
in unmasked regions.
2.2.2. Scattered-Light Correction
Mechanical and optical constraints make it impossible to baffle
wide-field cameras against all scattered light. The scattered light is
equivalent to an unknown dark current added to each image, and
must be subtracted rather than divided out. As a result, the usual
flat-fielding technique of observing a uniform light source such as
the dome or sky is inaccurate at the 3%Y 5% level.
For Suprime-Cam the scattered-light pattern and strength
change significantly with the lighting conditions and telescope
position. Variations as large as 5% are observed at the edges of
the field between dark, twilight, and dome conditions. Figure 5
shows the difference between two dome flats taken at differ-
ent rotation angles. This effect is similar in amplitude and pattern
to that observe d with the 12K and Meg acam cameras on th e
CFHT.
37
Following the example of CFHT we calculate the true flat by
observing objects at multiple positions on the camera. The true flat
can then be solved for as the flat field, which yields t he same
background-subtracted flux for an object at any position in the
field of view. In practice the flat image is generated by dividing
the focal plane into 128 ; 128 pixel regions r, and calculating a
factor C
r
for each region. The regions are defined so that no re-
gion crosses a detector boundary. As a result, sensitivity varia-
tions due to detectors are also measured. We can also allow for
an additional factor P
e
between exposures to correct for pho-
tometric variations due to atmospheric conditions and seeing.
Fig. 5.—Difference between Suprime-Cam dome flats taken in r
þ
at position
angles of 0
and +90
with chip-to-chip sensitivity variations removed. The scale
is linear with a stretch of 3% to +3% from black to white. Variations in the il-
lumination pattern due to scattered light are clearly visible.
37
See http://www.cfht.hawaii.edu/ Instruments/ Elixir/scattered.html.
36
See http://www.ifa.hawaii.edu/ kaiser/imcat /.
COSMOS OBSERVATIONS: OPTICAL/ NEAR-IR DATA 103No. 1, 2007
