The COSMOS HST/ACS Observations Webpage

[OLD --- COSMOS HST/ACS V1.2 & V1.3 (2005 - 2006)]

DATA ACCESS -- the V1.2 Cycle 12 + 13 drizzled FITS images can all be obtained here, along with their associated weight files (expressed in terms of inverse variance):

Most of the COSMOS team would likely use the images drizzled with North up, which have been drizzled onto exactly the same frame as the Subaru images, except that the pixel scale is either 0.03"/pixel or 0.05"/pixel instead of 0.15"/pixel, thus the Subaru pixel scale can be recovered by binning the images by a factor of 5 (for the 0.03" images) or a factor of 3 (for the 0.05" images).

For those doing lensing work, it is more desirable to have the images in the unrotated frame, ie with the [sci,1] chip at the bottom and the [sci,2] chip at the top. These images have been drizzled to a pixel scale of 30mas/pixel, using a pixfrac of 0.8 and a Gaussian kernel.

The data are also available as a mosaic, currently at 0.05"/pixel; a version drizzled to 0.03"/pixel will also be made available although right now the most common use of the 0.03"/pixel data seems to be for lensing work which requires data to be unrotated and not mosaiced. The 0.05"/pixel mosaic is 100,800 pixels on a side; the pixels were mapped onto this mosaic by applying all transformations in a single step (including rotations, translations, and mapping onto the tangent plane) using MultiDrizzle (Koekemoer et al. 2002). The mosaic is oriented with North to the top and the central coordinate for the tangent plane projection is near the centre of the mosaic (at pixel 50040, 48600), which is the standard COSMOS reference point:

Be warned that this mosaic is *VERY LARGE* (81 Gb!) and may take several days to download. In addition, some filesystems have problems retrieving files > 2Gb in a single step, so to help with this I've also made the mosaic available as a set of 41 pieces, each about 2Gb in size (and gzipped), which can be easily joined back together using the unix "cat" command once they have all been retrieved. The single mosaic as well as the split pieces are all available here:

Some mosaic pictures and other related binned-up files are also available here:

The target coordinates for all the COSMOS pointings in the HST PhaseII file are available here:

July 2005 - COSMOS Complete: All HST Cycle 12 + 13

All 588 orbits of the COSMOS project have now been obtained and are available. Note that unfortunately two of the more recent pointings failed as a result of guidestar problems, leaving a small gap near the top of the mosaic. We are requesting that these be repeated ASAP, preferably with different guidestars.

June 2005 - Full Cycle 12+13 V1.2 Mosaic

The full Cycle 12+13 mosaic is now available, which I've constructed based on all the Cycle 12+13 V1.2 data (588 orbits in I-band obtained during Cycle 12). This mosaic uses tiles that are registered to an accuracy of ~0.2 ACS pixels (10 milliarcseconds), and also has greatly improved cosmic ray rejection along the overlap regions between tiles, so it can be treated properly as a contiguous image. There are still low-level background variations (on the order of 5% of the original sky values), mostly due to scattered light effects. Other remaining low-level effects include faint satellite and asteroid trails, and ghosts from bright stars. These are in the process of being removed for the next version of the data. The mosaic is at the full ACS resolution (50 mas/pix) and is 100,800 pixels on a side; previously only the binned-down versions were available. The pixels were mapped onto this mosaic by applying all transformations in a single step (including rotations, translations, and mapping onto the tangent plane) using MultiDrizzle (Koekemoer et al. 2002). The mosaic is oriented with North to the top and the central coordinate for the tangent plane projection is near the centre of the mosaic (at pixel 50040, 48600), which is the standard COSMOS reference point:

Thursday 31 March 2005 - Full Cycle 12 V1.2 Mosaic

The full Cycle 12 mosaic is now available, which I've constructed based on all the Cycle 12 V1.2 data (261 tiles in I-band obtained during Cycle 12). This mosaic uses tiles that are registered to an accuracy of ~0.2 ACS pixels (10 milliarcseconds), and also has greatly improved cosmic ray rejection along the overlap regions between tiles, so it can be treated properly as a contiguous image. There are still low-level background variations (on the order of 5% of the original sky values), mostly due to scattered light effects. Other remaining low-level effects include faint satellite and asteroid trails, and ghosts from bright stars. These are in the process of being removed for the next version of the data. The mosaic is at the full ACS resolution (50 mas/pix) and is 80,000 pixels on a side; previously only the binned-down versions were available. The pixels were mapped onto this mosaic by applying all transformations in a single step (including rotations, translations, and mapping onto the tangent plane) using MultiDrizzle (Koekemoer et al. 2002). The mosaic is oriented with North to the top and the central coordinate for the tangent plane projection is at the centre of the mosaic (at pixel 40000, 40000).

January 2005 - ACS Cycle 13 V1.2 Data Release

A total of 177 pointings from the Cycle 13 data have so far been obtained. These have been processed through the COSMOS cluster pipeline at STScI to the same level of accuracy as the Cycle 12 data to date (see next item below), in terms of data calibration and astrometric alignment. Additional data will be processed to the same level of accuracy, as new images arrive.

Monday 22 November 2004 - ACS Cycle 12 V1.2 Data Release

The V1.2 images for the full 270 orbits from Cycle 12 have been produced by completely re-drizzling using improved absolute astrometry, together with the latest improved geometric distortion corrections for the ACS camera. The images have been registered onto the fundamental astrometric frame for COSMOS, which makes use of the CFHT Megacam i-band image. The CFHT image (courtesy of Olivier LeFevre, Herve Aussel, Henry McCracken, et al.) is tied to the USNO-B1.0 system, which in turn has been tied to the VLA 1.4 GHz image (courtesy Eva Schinnerer and Chris Carilli). These V1.2 ACS images therefore supercede the V1.0 and V1.1 images, and should provide an absolute accuracy of order 0.05-0.1" or better.

All images have the same pixel scale = 1 ACS pixel = 0.05"/pixel. Images with North up are 5600x5600 pixels (thus 120 Mb each) while the unrotated images are smaller at 4600x4600 pixels (80 Mb each). The single-exposure images are drizzled onto the same scale and orientation as the combined images.

Thu 17 Jun 2004 - ACS Cycle 12 Data Release V1.1: Improved Astrometry

The full 270 orbits from Cycle 12 have now been registered onto the fundamental astrometric frame for COSMOS, which makes use of the CFHT Megacam i-band image. The CFHT image (courtesy of Olivier LeFevre, Herve Aussel, Henry McCracken, et al.) is tied to the USNO-B1.0 system, which in turn has been tied to the VLA 1.4 GHz image (courtesy Eva Schinnerer and Chris Carilli).

NOTE: The CFHT catalog that was used here was a preliminary catalog that Herve Aussel had shifted onto the VLA frame. In the meantime, Henry McCracken is using this CFHT catalog to completely regrid the CFHT image and then generate a new catalog from that. Once this new, final, master catalog has been generated, it will be used to further refine the astrometry of the HST images. However, the current V1.1 release should already be a substantial improvement over the V1.0 images.

The initial astrometry in the earlier V1.0 data release was based purely on the astrometric information from the HST pipeline, which in turn was based on an assumed knowledge about the guidestar positions in GSC1. However, the GSC1 positions are typically off by ~1-2", and in some cases in the COSMOS data up to 3-4", therefore the absolute astrometry in the original V1.0 data release was also inaccurate to this level. These guidestar positional uncertainties also introduced slight errors in the image orientation, typically to the level of 0.02 degrees or less.

The V1.1 ACS images, which have been registered onto the VLA/USNO-B1.0/CFHT frame, now have absolute astrometry as well as relative alignment between each ACS tile to better than ~0.1", and have also corrected the orientation uncertainties. At levels below ~0.1", small differences are still apparent (of the order 1 ACS pixel = 50 mas or less). These can be addressed by incorporating improvements to the ACS distortion model that have been developed at STScI in recent months, and these improvements will be included in a subsequent COSMOS data release. However, the astrometric changes between V1.1 and any subsequent release will likely be below the level of 0.1" and will be primarily aimed at improving the registration between adjacent ACS tiles to an accuracy of 2-5 milliarcsec.

Therefore, the present V1.1 release can be considered accurate to an absolute astrometric level of approximately 0.1", which is an improvement of 10-50x over the original V1.0 astrometry.

Tue 18 May 2004 - Total now 270 orbits: CYCLE 12 COMPLETED!!

The final 7 pointings from Cycle 12 have been obtained and processed:
31-19, 31-20, 31-21, 31-22, 31-23, 31-24, 31-25

Fri 14 May 2004 - Total now 263 orbits

An additional 11 pointings have been obtained and processed:
30-15, 30-25, 31-01, 31-02, 31-03, 31-04, 31-06, 31-08, 31-16, 31-17, 31-18

Thu 6 May 2004 - Total now 252 orbits

An additional 10 pointings have been obtained and processed:
30-20, 31-05, 31-07, 31-09, 31-10, 31-11, 31-12, 31-13, 31-14, 31-15

Tue 4 May 2004 - Total now 242 orbits

An additional 15 pointings have been obtained and processed:
26-11, 26-13, 26-14, 26-15, 26-16, 26-17, 26-18, 26-19, 26-20, 26-21, 26-22, 26-23, 26-24, 30-05, 30-10

Sun 2 May 2004 - Total now 227 orbits

An additional 10 pointings have been obtained and processed:
24-10, 24-17, 26-02, 26-03, 26-04, 26-05, 26-06, 26-08, 26-09, 26-10

Thu 29 April 2004 - Total now 217 orbits

An additional 2 pointings have been obtained and processed:
24-01, 24-02

Wed 28 April 2004 - Total now 215 orbits

An additional 8 pointings have been obtained and processed:
25-25, 26-01, 32-01, 32-02, 32-09, 32-10, 32-19, 32-20

Fri 23 April 2004 - Total now 207 orbits

An additional 4 pointings have been obtained and processed:
25-15, 25-22, 25-23, 32-11

Tue 20 April 2004 - Total now 203 orbits

All additional pointings obtained before 20 April have now been processed, bringing the total number of orbits to 203 (9 of which are g-band images overlapping with their i-band counterparts). The new pointings obtained after 31 March are:

DATA ACCESS: Note that the newest data are accessible via ftp, but due to security issues here our computing division needs to know the IP address numbers of the specific machines that COSMOS team members will be using to ftp the data. So, could everyone please let me know these IP address numbers and I will ask our computer people to add your machines to the access lists. You'll be able to download the newest data only after this is done.

Thu 15 April 2004 - Single-Exposure Images Released

In order to help with examining interesting features on the images, eg faint, long, curved "arcs" (most of which may be due to slow-moving solar system objects), the single-exposure images are being made available, all drizzled onto the same scale and orientation as the final images. They are available here: drz_single/

Wed 31 March 2004 - Total now 131 orbits

An additional 3 pointings have been obtained and processed: 19-24, 19-25, 23-05

Tue 30 March 2004 - Total now 128 orbits

An additional 3 pointings have been obtained and processed: 19-20, 19-21, 19-22

An issue has been identified with the data taken in October and November 2003, (mostly near the centre of the field, with a few other pointings across the field) which seem to have a scattered light problem, evident as a large-scale imprint on the final output images. This is possibly due to the fact that those observations were obtained within 50 - 60 degrees of the sun. Efforts are underway to model and subtract this from the images. However, the effect on the scientific quality of object measurements (photometry, SExtractor results etc) is likely to be insignificant.

Thu 26 March 2004 - ACS Data V1.0 Release: 125 pointings

All pointings up to March 25, 2004 (125 orbits total, with 116 pointings in i-band, and an overlapping 9 pointings in g-band) have been reprocessed and improved significantly since last year's V0.5 release, as follows:

• Flat fielding: The flatfield reference files have been improved significantly, so that the images are now flat to below about 1% of the sky level.
• Cosmic ray rejection: The CR rejection has now finally been optimized for the COSMOS datasets (4-point dithers with 500s exposures), which has corrected some over-rejection in the bright cores of some objects that was apparent in the V0.5 data, as well as eliminating almost all the spurious residual CRs in the field. The only CRs remaining now are simply the result of 4 cosmic rays overlapping at the same place in all 4 exposures, but these are very infrequent. Some care should be taken along the edges and in the gap along the centre of the images, where the depth is only 3 exposures deep, so I recommend examining the weight files to quantify the exposure depth across the images.
• Astrometric improvements: Some attempt has been made to improve the absolute astrometry of the images, although this work is still on-going, so an update will be posted here as soon as that is complete.