Home       Data-Access       Data-Processing       NDWFS-Publications       Related-Surveys       Links        Additional-Documentation

Third Data Release (DR3)-- More Complete Description -- Still Under Active Construction

The NDWFS third data release (DR3) consists of the photometrically and astrometrically calibrated FITS images and associated data quality masks and images for all of the Mosaic-1 optical (Bw, R, I) and ONIS K-band imaging obtained for the 27 sub-fields that comprise the Boötes Field of the NOAO Deep Wide-Field Survey. Single-band and matched object catalogues are also available as part of this release. Below we describe the optical imaging, IR imaging, and catalogue construction that provided the data products that form the DR3 release.

Optical (Bw,R,I) Images

All the optical images included in DR3 were obtained with the KPNO Mayall 4m telescope and MOSAIC-1 imager. Some of the R-band images were obtained with an early version of the instrument that was equipped with engineering grade CCDs. The remaining R-band images, all of the Bw-band images, and all of the I-band images were obtained after the original CCDs had been replaced with thinned, more sensitive devices. All of the optical images in this release have been projected to a common scale (0.258" per pixel) and orientation (North up, East left). The same tangent point was used for the projection of all of these images (RA=14:32:05.7120; DEC=+34:16:47.496; J2000). Each of the images covers one of our sub-fields, which are nominally the size of a Mosaic pointing and are offset by approximately 35 arcminutes from cardinal direction neighbors.

While we made every effort to construct a data set homogeneous in its quality, the depth and image quality achieved is variable. Notes on the particulars of the variations of the individual fields are in preparation. Details on the reduction procedures used to process the Mosaic images are available in our guide to the processing of NDWFS Mosaic images.

Astrometric Accuracy

Astrometric solutions were generated as described in our reduction notes using the USNO A2.0 catalogue and USNO astrometric standard fields. A more complete description of these procedures is in preparation and will be linked off of these pages when it is completed and discussed by Jannuzi et al. (2005). The typical RMS of the residuals of these solutions is 0.35 arcseconds.

Photometric Zero-points and Delivered Image Quality of Images

The combined/stacked images (in a given filter) have over-lap regions of approximately one arcminute. This allow us to ensure that their relative calibration is accurate to better than 0.03 of a magnitude. The absolute zero-points have been determined using multiple observations of Landolt standards on photometric nights. The images were either obtained on photometric nights, zero-pointed by separate exposures obtained for that purpose, or tied (using the over-lap regions) to fields observed under photometric conditions. Details will be presented by Jannuzi et al. (2005). The photometry in these images has been referenced to Kron-Cousins B-band, R-band, and I-band. The MAGZERO keyword in each image gives the magnitude of one count in the image (Vega).

The stacked/combined images were made from individual exposures with a range of delivered image quality. We did not try to match or circularize the PSF's of either the final combined images taken with different filters, nor the images that went into the final images. This might limit the usefulness of these images for some studies. Jannuzi et al. (2005) will provide more quantitative information about this issue and will be linked off this web page when it is available.

There is also some variation in image quality across the field, but this is generally quite small. We can roughly characterize the DIQ (delivered image quality) as determined by measurements of the FWHM of stars in the released images (based on measurements made by SExtractor of unsaturated stars).

In the table linked below, we list the photometric zero points for each image derived on the Vega system. Note in particular that while Bw observations of spectrophotometric standards indicate that the AB_zeropoint and Vega_zeropoint for this filter are very similar, our current values have been referenced to B-band Landolt standards. We also list the seeing (delivered image quality) and other selected information. The image headers contain the same information (and more) as the MAGZERO, SEEING, DEPTH, and DEPTHCAT keywords.

Table of selected properties of the optical images.

IR Imaging

All the near-IR observations for the NDWFS were obtained using the 2.1-m telescope of the Kitt Peak National Observatory over a period spanning five years. The time period over which we conducted the survey witnessed many advances in the efficiency and size of near-IR arrays. As a result, observations of the NDWFS fields were obtained using three different instruments. We began the survey with the Ohio State - NOAO Infrared Spectrometer (ONIS) and observed the NDWFS fields primarily in the K-band starting in 1997 September and ending in 2000 April. We then continued the survey using the Simultaneous Quad IR Imager (SQIID) from 2000 May through 2001 May, imaging the field in J, H and K. The Florida Multi-object Imaging Near-IR Grism Observational Spectrometer (FLAMINGOS) was commissioned in late 2001. Its 20 arcmin field of view greatly improved our observing efficiency. We completed the IR portion of the survey with this instrument.

The FLAMINGOS and ONIS data together provide K or Ks-band coverage across the entire Boötes field. The DR3 release includes all of the ONIS K-band data of the Boötes field. The FLAMINGOS images are still be prepared for release. The ONIS instrument uses a 512x1024 InSb array and provides near-IR imaging over a field of view of approximately 2.9'x5.8' sampled at a pixel scale of 0.3407 arcsec/pixel. We observed the Boötes NDWFS fields through a K-band filter using ONIS at the f/8 cassegrain focus of the 2.1m telescope over 50 nights, spread out over a period of three years starting in 1998 March and ending in 2000 April.

On each night, we typically observed in a single strip of roughly constant declination, starting at the west end of the field. Each position was observed for 60 sec (typically split into 8 individual readouts of 7.5 sec each) and the telescope was then offset to the east by 17 arcseconds. As a result of the significant frame-to-frame overlap, each stacked pixel has a total exposure time of roughly 10 min. On a typical clear night, we obtained about 300 images covering an area of approximately 1.4 degree times 5.8 arcminuts, the latter number corresponding to the height of the array. Observations were restricted to times when the target field was at an airmass less than 2.0. On each photometric night, observations were also obtained of photometric standards.

The delivered image quality of the ONIS observations varies significantly across each strip, due not only to seeing variations, but also to focus drifts (which are notoriously common at the f/8 focus of the 2.1m) and an intermittent oscillation in the RA drive which resulted in an east-west elongation of the image of about 1\arcsec in some cases. As discussed below, the final stacked images combine observations from different nights and therefore show significant spatial variations of the point-spread function.

The astrometry and photometry of the ONIS data were verified by comparison to 2MASS and to our optical images (for astrometry). Details will be provided on this page in the next few weeks and by Dey et al. (2005).

The released ONIS images cover portions of the 20 sub-fields discussed previously and were constructed from the original observations to match, approximately, the distribution of the optical sub-fields. As a result, the image quality and other properties of the data can change significantly as a function of position in the released images in DR3. The header keywords for the DR3 IR images are therefore approximations or mean values for the data in the image. In particular, that is why we have used a single DEPTHCAT value for all of the DR3 K-band images. Note that all the ONIS IR images have the same MAGZERO by construction.

NDWFS Catalogues

Access to the NDWFS DR3 catalogues.

NDWFS Source Naming Conventions

NDWFS DR3 Catalogs

All the NDWFS catalogs for the October 2004 Boötes data release were generated with SExtractor 2.3.2 (Bertin & Arnouts 1996, A&A 117, 393). Catalogs were generated in single-band mode for each subfield and band. Merged catalogs were generated from the single-band catalogs. These catalogs are available from this website in ASCII and FITS binary tables.

With the exception of four parameters in the merged catalogs (object name, subfield, and two flags), the catalogs contain standard SExtractor object parameters. These are described in full detail in the SExtractor manual which is available from the SExtractor website (http://terapix.iap.fr/rubrique.php?id_rubrique=91). Please note that in addition to the standard SExtractor header information, some comments have been added to the catalogs.

These catalogs are intended for preliminary science with the NDWFS. We caution that they may not be ideal for all the science possible with the NDWFS imaging dataset. Results derived from the catalogs should, at the very least, be confirmed by inspecting the imaging data.

Single-band catalogs for individual bands and subfields

Single-band catalogs were generated for each band and subfield using SExtractor 2.3.2 in single-band mode. The parameters used when generating the catalogs are listed below. The original image headers are contained in the header of the FITS binary table version of the single-band catalogs.

We have aimed to make the catalogs as deep as possible, while minimizing the number of spurious object detections. Before detecting objects, SExtractor convolves the data with a Moffat profile convolution filter which we have matched to the seeing. For the K-band we assume the seeing (delivered image quality) is 1" (FWHM), which is a valid approximation for much of the data. In the optical the signal-to-noise per pixel and minimum number of connected pixels per object are a function of the measured seeing. This increases the depth of the data, particularly in poor seeing where objects may comprise a large number of pixels with comparatively low signal-to-noise per pixel. The integrated signal-to-noise thresholds for the optical and K-band catalogs are 1.6 and 1.8 respectively.

Merged (Matched) Single-band Catalogues for All of Boötes Field

We have produced a merged BwRIK catalogue for the entire Boötes field. This catalogue is available from this website in ASCII and FITS binary tables. This merged catalogue contains all the object detections and information contained in the original single-band catalogs. This catalogue was produced by matching objects in the individual bands and subfields. The NDWFS subfield and object name are listed for each object in each band.

The catalogue has been split by declination range and band into 16 files. The format of the catalogue is similar to dual-mode SExtractor catalogs, with the nth line in each file corresponding to the same object. When an object is not detected in a given band, valid Extractor values are replaced by -9.9, -99.9, -999.9 etc.

Objects were matched if their centroids were with 1" of each other or if the centroids were within an ellipse defined by two times A_WORLD, two times B_WORLD and THETA_WORLD. Where multiple matches were found, the closest matches were used and the relevant objects were flagged with NDWFS_SPLITMATCH (see object flags).

Catalogue depth (DEPTHCAT)

For each subfield and band, we have determined the 50 percent completeness limit by adding artificial stellar objects to the data and recovering them with SExtractor. The 50 percent completeness limits are recorded by the DEPTHCAT keyword in the released image headers. The stellar objects were given Moffat profiles matched to the observed seeing. Artificial stellar objects of a given magnitude were added to copies of the data and recovered using SExtractor. The process was repeated for a range of magnitudes spanning the expected range of ~100% completeness to ~0% completeness in steps of 0.1 magnitudes. The 50 percent completeness limit for stellar objects was then determined using the completeness measurements and linear interpolation between the data points. We caution that the 50% completeness limit for extended objects with an intrinsic size of ~1" is several tenths of a magnitude brighter than the 50% completeness limit for stellar objects.

Object flags

The October 2004 release catalogs contain 2 SExtractor flags and 2 other flags which provide information about the catalogue data quality.

SExtractor's FLAGS parameter is discussed in the full detail in the SExtractor manual (http://terapix.iap.fr/rubrique.php?id_rubrique=91). Values of three or less for FLAGS are typical for good quality data. SExtractor's IMAFLAGS_ISO has been set to the minimum value of the NCOMBIM mask (optical) or the FLAGIM mask (K-band) within an object isophote. A value of zero indicates there is a region of the object isophote than contains no valid data. High values indicate every pixel within the object isophote contains good data from multiple individual frames.

FLAG_DUPLICATE and FLAG_SPLITMATCH are NDWFS flags which appear in the merged catalogue only. As our pointings/images overlap each other, the same object can be detected in multiple subfields. FLAG_DUPLICATE=1 indicates there is another catalogue entry for this object with better quality data. When the single-band catalogs are being merged, there can be ambiguous matches (e.g., two I-band detections matched to one Bw-band detection). While we use the closest object pairs when multiple matches are found, we flag all the objects with FLAG_SPLITMATCH=1 as errors may be associated with the object match and object photometry.

Spurious objects

While we have attempted to reduce the number of spurious objects within the catalog, some spurious objects remain. The majority are associated with the halos of bright stars, the edges of large resolved galaxies and field edges. There are also residual images in the K-band, which can be found to the east of bright stars. While many spurious objects will have flags in the catalogs (e.g., low IMAFLAGS_ISO values or high FLAGS) some near bright stars and galaxies may have relatively normal flag values. We therefore caution against using the catalog entries near very bright objects.


NOAO Deep Wide-Field Survey -- Last modified: November 26 2007 12:59:58. -- Feedback: jannuzi@noao.edu