EGS - Extended Groth Strip HST/ACS Mosaic -
Primary Web Page

Anton M. Koekemoer (STScI)

First created: 12 Jul 2004
Last updated: 30 Dec 2005

Latest News:

9 December 2005

The full-size HST/ACS mosaics of the EGS have now been completed. This includes re-processing all the original HST exposures, improving the cosmic ray rejection and combining into a single mosaic image. Below is a small picture of the entire mosaic of HST/ACS observations. The field is in fact oriented at about 50 degrees in the sky, so left -> right on the image corresponds to North-East -> South-West.

Mosaic of the full set of 63 HST/ACS pointings (21x3), covering ~70'x12'. Clicking on this image will load a larger version, with a pixel scale of 1"/pixel (4200x720 pixels). The slight curvature is a result of the tangent-plane projection of the image on the sky.

The mosaic FITS files and pictures are available further down in the Mosaic Data Products section.

Summary of Observations

The HST observations of the Extended Groth Strip (EGS) were obtained as HST Cycle 13 GO program 10134 (PI: M. Davis), using the Advanced Camera for Surveys / Wide Field Channel (ACS/WFC), which consists of two 4096x2048 CCDs providing a total of 4096x4096 pixels at a scale of 0.05"/pixel. The observations were obtained between June 21 2004 and March 12 2005, for a total of 126 orbits.

The EGS HST observations were divided into a grid of 21 x 3 pointings, with each pointing observed for a total of 1 orbit each in two filters, F606W (wide V) and F814W (wide I). Each orbit was divided into 4 exposures, each being either 565 seconds in length (V) or 525 seconds (I). Since the nominal orientation of HST changes throughout the year, some visits were obtained at rotations that were offset at increments of 90 degrees from one another. The following shows the exposure times for each filter, together with current ACS/WFC zeropoints published by STScI (as of Dec. 2005).

Filter Exptime/
Other filter properties:
F606W (wide V) 565s 2260s 26.486 26.398 5917.678 7.906457E-20 672.31
F814W (wide I) 525s 2100s 25.937 25.501 8059.761 7.072360E-20 654.64

Each set of 4 exposures was executed in a 4-point dither pattern, designed to move the chips around the sky to mitigate hot pixels and cover the gap between the chips. The primary component of the pattern was a 2-point line dither separated by about 6", with a secondary 2-point line dither at each of the two primary points, separated by 3". This effectively provided 4 exposures, each offset by about 3" along the y-direction and thereby ensuring that the gap was always covered by at least 3 exposures. In addition, the dithers contained an offset of a few pixels along the x-direction, including non-integer pixel offsets in both x and y to ensure good sub-pixel sampling of the HST/ACS point spread function.

Dither Pattern Specifications
Primary_Pattern Secondary_Pattern
Pattern_Type LINE LINE
Pattern_Purpose DITHER DITHER
Number_Of_Points 2 2
Point_Spacing 6.10186 3.06753
Line_Spacing -none- -none-
Coordinate_Frame POS-TARG POS-TARG
Pattern_Orient 93.0546 81.6083

The grid of 21 x 3 pointings was oriented as shown below, aiming to position the pointings as far apart as possible without introducing gaps in the coverage between pointings, thereby yielding uniform coverage of the entire region. Because of the relatively large distortion of the ACS/WCS camera (~7%), some of the observations obtained at 90 degree rotational offsets display a somewhat different amount of overlap with the adjacent pointings.

Layout of the 21x3 grid of 63 HST/ACS pointings on the sky, showing the location of the two 2048x4096 ACS/WFC chips at each pointing (click the image to obtain a larger, more detailed version).

Image Calibration and Initial Processing with MultiDrizzle

All the exposures were first processed through the standard ACS calibration pipeline steps, which include bias subtraction, gain correction, and flatfielding. These steps are carried out automatically by the HST Pipeline when data is requested for retrieval. In addition, manual corrections were applied to create improved background scattered light / flatfield corrections, as well as improving amplifier/quadrant bias offset removal. This processing yielded a total of 504 separate exposures, thus 8 for each of the 63 pointings (4 exposures for each of 2 filters).

Each set of 4 calibrated exposures was then combined with the Pyraf/STSDAS MultiDrizzle software (Koekemoer et al. 2002, HST Calibration Workshop, 337). This step is also carried out automatically by the pipeline, but can be repeated off-line with different parameters. Initial processing of the EGS data was first carried out by Anton M. Koekemoer starting in July 2004, followed by J. Lotz later in 2005. The steps carried out by MultiDrizzle include automatic registration of the 4 exposures, followed by the creation of a clean "median" image which is subsequently used to identify cosmic rays. The resulting cosmic ray masks are then used in the final combination, which also includes removal of the ACS distortion to produce an image with a uniform, rectilinear pixel scale.

The MultiDrizzle images for all the 63 separate pointings were drizzled to a pixel scale of 30 mas/pixel (thus 0.6 of the input pixel scale), in order to take advantage of the sub-pixel sampling provided by the 4-point dither pattern. Each input pixel was shrunk by a factor of 0.8 (specified using the "pixfrac" parameter) before being mapped to the output plane, in order to reduce the degree of correlated noise in the final image.

MultiDrizzle also produced inverse variance map output images, using the input error arrays associated with each exposure. While these images provide an estimate of the degree of noise associated with each pixel, they should be used with care since they do not include any effects of correlated noise in the images. Therefore, any photometry measurements should compare the values from the inverse variance maps with the measurements of correlated noise obtained directly from the images, and scale these appropriately before using them in photometry software. Finally, the drizzled image for each pointing was tied to the DEEP2/SDSS reference frame (J. Lotz), including recent updates from October 2005, yielding an absolute astrometric accuracy of ~0.07" in RA and Dec.


Mosaic Data Products

The corrected astrometric information for each drizzled image as described above was propagated to each of the 504 separate calibrated exposures, which were then used directly as input to subsequent drizzle processing to create the final combined mosaics. The advantage of using the original input exposures is that all the transformations of rotations, shifts, scale changes and distortion removal can be applied to each pixel in a single step, thereby yielding output mosaic images that are of optimal quality since no additional convolutions are introduced. First, an inverse variance map was created for each exposure, containing all the noise terms from the background sky (modulated by the flatfield and geometric distortion), together with the read noise and accumulated dark current for that exposure. The inverse variance maps were then populated with pixel masks representing cosmic rays and bad pixels.

The final combined mosaics are all projected onto a single tangent point on the sky, thus yielding a uniform rectilinear output pixel grid. The definition of this tangent-plane reference frame is as follows:

EGS HST/ACS Mosaic Reference Frame
Right Ascension: 214.825 =  14:19:18.00
Declination: +52.825 = +52:49:30.00
Y-axis Position Angle: -49.7
Number of x-pixels: 140,000
Number of y-pixels: 24,000
Pixel scale: 30 milliarcseconds
Reference x-pixel: 70,000
Reference y-pixel: 12,000

The exposures were drizzled together into a single mosaic for the EGS, at several different pixel scales. Several versions of these mosaics, including the drizzle and weight files (inverse variance) are available through the links below:

Note that the scale=1 (0.05"/pixel) and scale=0.6 (0.03"/pixel) files are too large to fit within the standard 2 Gb file limit, so these are made available as separate sections. These files are available here:

Scale = 2 (0.1"/pixel: 42,000 x 7,200 pixels; 1.2 Gb):

F606W (V):
GIF picture: mos_v_scale2_drz.gif (120Mb)
FITS DRZ file: mos_v_scale2_drz.fits.gz (1.2Gb)
FITS WHT file: mos_v_scale2_wht.fits.gz (1.2Gb)
F814W (I):
GIF picture: mos_i_scale2_drz.gif (120Mb)
FITS DRZ file: mos_i_scale2_drz.fits.gz (1.2Gb)
FITS WHT file: mos_i_scale2_wht.fits.gz (1.2Gb)

Scale = 1 (0.05"/pixel: 84,000 x 14,400 pixels; 4.8 Gb):

To make it easier to transfer and work with this mosaic, it is split into 4 sections along the x-axis, with each section being 21,000 x 14,400 pixels in size. Each section is 1.2 Gb.

F606W (V):
FITS DRZ files: section1 section2 section3 section4(1.2 Gb each!)
FITS WHT files: section1 section2 section3 section4(1.2 Gb each!)
F814W (I):
FITS DRZ files: section1 section2 section3 section4(1.2 Gb each!)
FITS WHT files: section1 section2 section3 section4(1.2 Gb each!)

Scale = 0.6 (0.03"/pixel: 140,000 x 24,000 pixels; 13.4 Gb):

To make it easier to transfer and work with this mosaic, it is split into 8 sections along the x-axis, with each section being 17,500 x 24,000 pixels in size. Each section is 1.6 Gb.

F606W (V):
FITS DRZ files: section1 section2 section3 section4 section5 section6 section7 section8 (1.2 Gb each!)
FITS WHT files: section1 section2 section3 section4 section5 section6 section7 section8 (1.2 Gb each!)
F814W (I):
FITS DRZ files: section1 section2 section3 section4 section5 section6 section7 section8 (1.2 Gb each!)
FITS WHT files: section1 section2 section3 section4 section5 section6 section7 section8 (1.2 Gb each!)

First created: 12 July 2004, Anton M. Koekemoer. Last updated: 30 December 2005, Anton M. Koekemoer.