4.5 Absolute Astrometry

HST Pointing Accuracy

Historically, the accuracy of HST absolute astrometry has been limited primarily by uncertainties in the celestial coordinates of the guide stars as specified in the Guide Star Catalog. GSC 1.1 had nominal rms errors of ~0.5 arcsec per coordinate, with errors as large as ~1‐3 arcsec reported near the plate edges. This accuracy improved substantially in October 2005 (during HST Cycle 15) with the introduction of GSC 2.3.2, where rms errors per coordinate were reduced to ~0.3 arcsec over the whole sky.  An updated version of the catalog (GSC 2.4.0) was released in October 2017, improving the celestial coordinates with the positions from Gaia DR1 and reducing errors to < 30mas over the entire sky. Thus, after including uncertainties in the positions of the science Instruments (SIs) in the alignment of the focal plane to the Fine Guidance Sensors (FGS), the total error in HST absolute astrometry is ~1 arcsec for observations made with GSC1.1, ~0.3 arcsec for those made with GSC 2.3.2, and ~0.1 arcsec when using the newer GSC 2.4.0.  A summary of the GSC catalogs and associated errors over the HST lifetime is provided in Table 4.4. 

Information on the guide stars used for a particular observation can be found in the FITS header of the associated jitter file, which can be retrieved from MAST and end with the suffix jif.fits. Additional information such as the positional uncertainty and parallax of these guide stars can be found by using the web form here.

Table 4.4: Key Guide Star Catalog releases and associated errors

Absolute Astrometry Improvements in MAST

The coordinates populated in the FITS headers of HST observations retrieved from DADS (the HST Data Archiving and Distribution Service) were derived based on the guide star coordinates in use at the time of the observation. As the accuracy in these catalogs were refined over time, the pointing accuracy of HST has also improved. Table 4.4 lists the catalog in use at the time of installation of the three main imaging cameras (WFPC2, ACS, and WFC3) and the typical errors at each epoch.

The goal of the HST Astrometry Project is to correct these inconsistencies in the archival data products as much as possible.  As observations are processed or reprocessed in the HST pipeline, their World Coordinate System (WCS) will be updated to use the most accurate solution available. There are two types of corrections that can be performed:

• a priori: correct the coordinates of the guide stars in use at the time of observation to the coordinates of those guide stars as determined by Gaia by applying a global offset to the WCS
• a posteriori: identify sources in the HST image and cross-match with positions from an external reference catalog (such as Gaia) to derive an improved WCS based on fitting x/y to RA/Dec

Note that a priori corrections are only relevant for observations which executed prior October 2017 (e.g. prior to the release of GSC 2.4.0), and these will still include small errors in the alignment of the science instruments to the HST focal plane. The a posteriori corrections are limited to imaging instruments for which there are an adequate number sources to define a reference catalog for matching. These solutions remove uncertainties in the focal plane and are expected to have the smallest absolute astrometric error.

New Data Products for WFC3, ACS, and WFPC2

Enhanced data products were delivered to MAST in several stages and are summarized below. 

• December 2019 - Improved astrometry for WFC3 and ACS imaging data includes two corrections to the header world coordinate system (WCS). The first includes a new version of HST's Guide Star Catalog (GSC, version 2.4.0) which updates the coordinates of the guide stars with the positions from Gaia DR1. This reduces the typical uncertainties in the positions of the guide stars to <~100 mas over the entire sky. Combining this new information with the knowledge of the instrument distortions, an a priori correction has been made for all WFC3 and ACS observations in order to lock all HST observations onto a common absolute reference frame. When possible, an additional a posteriori correction has been applied by aligning sources in each HST image to an external reference catalog (e.g. Gaia). While some observing modes cannot be aligned to Gaia (e.g. grism and moving target observations) or the alignment may fail due to a lack of sources in either the HST image or the Gaia catalog, approximately 70% of the WFC3 and ACS frames have been aligned successfully. For these data products, the typical pointing uncertainty is reduced to ~10 mas, although the uncertainties increase for observations further in time from the Gaia reference epoch (2015.0 for DR1, 2015.5 for DR2, and 2016.0 for eDR3). The software used to create the new data products is described on the Pipeline Astrometric Calibration page.
• December 2020 - Production of new advanced data products for ACS and WFC3 begins in the MAST pipeline. These Hubble Legacy Archive (HLA)-style mosaics comprise the data from a single HST visit which are aligned to a common astrometric reference frame. These new 'Hubble Advanced Products' (HAP) are referred to as 'Single Visit Mosaics' (SVMs) and are described in a MAST Newsletter article. The data products are all drizzled onto the same north-up pixel grid and may include improved relative alignment across filters for datasets acquired within the same visit, enabling easy comparison of the images through multiple filters. When possible, sources in the images have been aligned to Gaia to improve the WCS. SVM data products with both relative alignment (by filter) and absolute alignment to Gaia will contain the string 'FIT_SVM_GAIA' in the 'WCSNAME' keyword in the science extension of the image header. The software used to compute these new data products is described in the DrizzlePac documentation for Single Visit Mosaic Processing.
• November 2021 - Source catalogs are now included as part of the SVM data products. Because the SVM drizzled images include an additional relative alignment across filters in a visit, these are used to generate point source and segment catalogs during pipeline processing. These catalogs supersede those produced by the Hubble Legacy Archive and will be the basis of the next version of the Hubble Source Catalog.
• April 2022 - A new type of Hubble Advanced Product is distributed through MAST. These are cross-visit, cross-proposal mosaics called 'Multi-Visit Mosaics' (MVMs) which combine all ACS/WFC, WFC3/UVIS, or WFC3/IR images falling within a pre-defined 0.2° x 0.2° ‘sky cell’ for each detector+filter and drizzled to a common all-sky pixel grid.  MVM products are described in the following MAST Newsletter article and complement the Single Visit Mosaics (SVMs). Because they may include observations spanning a large date range, MVMs may have photometric errors of several percent or systematic alignment errors when combining visits which are aligned to different reference catalogs. These products are therefore recommended as ‘discovery images’ for comparing observations in different detectors and passbands and not for precise photometry.
  
• December 2023 - MAST begins producing new WFPC2 data products with improved absolute astrometry. The calibrated images combine the previously separate science array 'c0m.fits' and data quality array 'c1m.fits' files into a new 'flt.fits' data product, similar to ACS and WFC3. These are used to produce new 'drw.fits' drizzled data products, where the 'drw' suffix reflects the improved WCS. These are drizzled to the scale of the PC chip (0.0455"/pix) and replace the old 'drz.fits' products, which were drizzled to the scale of the WF chips (0.0996"/pix). For more details, see the notebook Drizzling new WFPC2 flt data products.
  
  

WCS Naming Conventions

Successfully aligning an observation to Gaia using the a posteriori processing will result in an update of the 'active' WCS of the image with the new solution and a new 'headerlet' extension. This headerlet includes the WCS keywords which define the transformation from pixels to Gaia-aligned positions on the sky, as well as information about how this solution was derived and the uncertainty of the fit.

The various WCS solutions are identified by the WCSNAME keyword found in each FITS headerlet and use the following naming convention: 

wcsName = OriginalSolution - CorrectionType

 where OriginalSolution may be either

•    OPUS : initial ground system wcs, no distortion correction
•    IDC_xxxxxxxxx : initial distortion corrected wcs  (where xxxxxxxxx = geometric distortion model used, eg. the rootname of the IDCTAB reference file)

 and CorrectionType may have several forms

•    GSC240 : 'a priori' WCS where guide star coordinates are corrected from the original reference frame (e.g. GSC1.1 or GSC2.3) to the Gaia DR1-based GSC2.4.0
•    HSC30 :   'a priori' WCS corrected from the original reference frame to the Hubble Source Catalog (HSC v3.0) frame, which is based on Gaia DR1
•    FIT-IMG-RefCat  : 'a posteriori' WCS matched to a reference catalog, where 'IMG' implies each FLT is separately aligned to the reference catalog
•    FIT-REL-RefCat   : 'a posteriori' WCS matched to a reference catalog, where 'REL' implies that FLTs within the same filter within the same visit are aligned before a global catalog alignment
•    FIT-SVM-RefCat : 'a posteriori' WCS matched to a reference catalog, where 'SVM' implies that FLTs in multiple filters within the same visit are aligned before a global catalog alignment

and REFcat may be one of the following when an adequate number of matches are found in the HST frame to compute the linear transformations (shift, rotation, scale) to sky coordinates:

•    Gaia eDR3
•    GSC 2.4.2
•    2MASS

A list of possible WCSNAME values in the FITS image headers is provided in Table 4.5. More details on interpreting the WCS names may be found on the Astrometry in Drizzled Products page.

Table 4.5: Sample WCSNAME keyword values and the corresponding WCSTYPE description. 

Implementation

The key to implementing improved astrometry is the use of headerlets, self-contained FITS extensions containing a WCS transformation which can be attached to a FITS file and applied to the primary WCS. MAST data processed with the Enhanced Pipeline Products code will have headerlets added as extra extensions to the FITS file. An observation can have multiple headerlets, each of which may have astrometry derived by differing methods. As HST data is processed/reprocessed, all available headerlets will be present as FITS extensions in the archived image with the best solution applied to the primary WCS.  More details on how the WCS information is stored in headerlets may be found on the page Astrometry in Drizzled Products.

Once the observations are downloaded from MAST, the python notebook 'Improving Astrometry Using Alternate WCS Solutions' will familiarize users with how to check the active WCS or switch to an alternate WCS, if desired.  Alternatively, the WCS solutions may be downloaded directly from MAST as separate 'hlet.fits' headerlet files and applied to existing FITS data. Python functions for creating, updating, and applying headerlets to FITS images are described via the Headerlet User Interface.

Caveats

While the majority of calibrated HST data products are now aligned to a common absolute reference frame, further improvements may be possible via manual realignment using the drizzlepac tools.  This is particularly true for exposures acquired in the same visit in which the WCSNAME keyword values do not match. Additionally, the quality of the WCS may be compromised when the number of sources used for the alignment is small, e.g. the NMATCHES keyword has a value less than ~10. 

For standard drizzled data products 'ipppssoot_drz.fits', alignment errors may be present in the following situations:

• Short and long exposures obtained in the same visit may no longer be aligned due to potentially different number of Gaia matches.
• Exposures in different filters (eg. narrowband vs broadband) acquired in the same visit may no longer be aligned to one another, e.g. if each filter had a different number of Gaia matches.
• Grism images may now be offset from their direct image counterparts, where only the later may be aligned to an external reference catalog. In order to preserve relative alignment between grism and direct images, users may wish to back out the updated WCS solutions entirely, as described in Section 4 of the notebook, 'Improving Astrometry Using Alternate WCS Solutions'.

The Single Visit Mosaic drizzled data products 'hst_ipppssoo_*_drz.fits' are recommended for the best alignment across all exposures acquired in a single visit, and the WCSNAME keywords will contain the string 'FIT_SVM', as shown in Table 4.5.  If desired, these WCS solutions may be downloaded as separate headerlet files, applied to standard calibrated data products ('ipppssoot_flt.fits') and redrizzled.