Nigel Hambly, Tim Hawarden, Andy Adamson, Mark Casali, Steve Warren, Sandy Leggett
email: nch roe.ac.uk, tgh roe.ac.uk, a.adamson jach.hawaii.edu, mmc roe.ac.uk, s.j.warren ic.ac.uk, s.leggett jach.hawaii.edu
This document is intended as a 'top-level' analysis of the requirements for photometric calibration of WFCAM data (for technical details and references for further reading concerning WFCAM, see Appendix A). It is our intention that this document be available to all interested parties, and that such parties should feel free to contribute ideas at any time as they see fit. All comments/suggestions etc. should be emailed to nch roe.ac.uk, with CCs to the above list. The ultimate intention of this document is to specify the calibration procedure. The initial aim, however, is to promote an open debate amongst all interested individuals concerning WFCAM photometric calibration.
The task of photometric calibration for WFCAM data is broken down into the following basic scheme:
The final category is intended to be the result of the top-level analysis. It should indicate the method of choice, the reasons for that choice and an outline of what should happen next as this recommendation is taken forward. The basic steps are described in more detail as follows:
Taking the UKIDSS consortium proposal (see Appendix A for reference) as indicative of the science that will be done using WFCAM, Table 1 gives details of the wavebands that will require routine calibration. Note the inclusion of non-standard passbands in addition to the usual JHK; filters other than those detailed in Table 1 will have to be calibrated on an ad hoc basis by those observers using them.
|Y||0.97||1.02||1.07||to distinguish L/T dwarfs from QSOs|
Table 1. Filters requiring routine photometric calibration. Notes: 1. wavmin and wavmax are 50% cut-on and cut-off values. 2. JHK filters will be Mauna Kea Observatory (MKO) system filters. 3. Y is similar to MKO Z but has a bluer short wavelength cutoff.
For present purposes, we define calibration as meaning the definition of data acquisition procedures (i.e. definition of targets and observing procedures) that will enable the derivation of flux levels (be they in units of magnitudes or flux density) for image and object catalogue products from the camera. This includes specification of filter systems (previous Section) and specification of the level of calibration (e.g. zeropoints, extinction, first/second order colour terms) but we will not discuss, for example, specification of photometric measures (e.g. isophotal/total/aperture magnitudes) or flux measures (e.g. Janskys). 'Routine' calibration means those procedures that will result in a stable photometric system over timescales of whole nights to that of the scientifically productive life of WFCAM.
A reasonable goal is that the scientific 'usefulness' of WFCAM data should not be limited by the accuracy of the external calibration - i.e. as far as possible the errors on photometric measures should be sky-noise limited as opposed to limited by uncertainties in the calibration zeropoint.
One possible approach to specifying the required accuracy is to look at current science drivers. From the UKIDSS consortium and VISTA survey specifications, calibration is required at the level of ~2% with a goal of ~1% for all wavebands.
However, it is worth noting that there is a legacy aspect to survey instruments such as WFCAM. The final WFCAM archive is likely to be used long after the observations are complete and initial science results published. Datamining will be done on the entire archive for some time. Comparison of observed fluxes in many wavebands and with model computations is becoming increasingly common in astronomy, so it would seem a reasonable goal to establish absolute photometric zeropoints at the level of a few percent in all wavebands to maximise the potential return from the investment of time and money in building and operating WFCAM.
For the present purposes, we consider the following calibration methods:
In practice, some of the above methods are likely to be used (in isolation or in some combination) as survey observations progress. Here, we wish to establish a recommended calibration procedure that can be generally applied to yield requisite photometric accuracy for science exploitation.
Discussion: It seems safe to assume that photometric precision from WFCAM observations will be limited by positional dependent systematic errors. Some of these may not change significantly over the lifetime of the instrument - e.g. relative device gain, device sensitivity. Vignetting may change subtly with pointing position due to flexure and will certainly change with instrument changes. Of course, vignetting will be removed by standard flat-fielding. One point worth noting is the effect of scattered light which introduces errors into the flatfield (e.g. Manfroid, Selman & Jones 2001, and references therein). These can be analysed and corrected using the procedure described in Manfroid et al. in any suitably dense star field; the star field does not require absolute calibration since flatfield errors can be determined from measurements of differences in stellar magnitudes when the same stars are measured in different positions within the FOV. However, it would seem safe to assume that some time-varying positional systematics other than the above may be present; control of those systematics demands frequent observation of sources (with known fluxes) scattered over the device fields-of-view. The only options under consideration above that are capable of controlling time-variable position-dependent systematic errors are options 2 & 3 which use sources of known flux scattered over the device FOV. However, there are compelling arguments against using 2MASS fluxes (option 2) to derive WFCAM zeropoints (see Appendix D). This in turn indicates that the only useable option is 3.
Based on the above considerations, the recommended procedure is to use option 3: calibration via standard fields established before survey observations commence. Appendix E considers some possibilities. WFCAM first light is currently scheduled for the final quarter of 2003. We therefore have 4 UKIRT observing Semesters (02A/02B/03A/03B) in which to undertake an observing programme for the establishment of standard fields. Given the proximity of the deadline for 02A (September 30th 2001) and the need to carefully design the observing programme it would seem reasonable to aim to finalise plans over the next 6 months with the aim of a PATT proposal to be submitted for 02B before March 31st 2002.
The UKIRT infrared wide-field imager WFCAM is now proceeding apace; for many more details see the project home page. Currently scheduled for first light in the final quarter of 2003, WFCAM will be used for UKIDSS surveys, as well as classical PATT observing. The data rate and characteristics of this large camera (based on 2×2 2048-pixel devices) make it important that standardised pipeline reductions be carried out for both the survey and non-survey data; furthermore, it is only through the implementation of a general science archive that full exploitation of the data will be possible. JAC/ATC are working closely with the UK wide-field astronomy units in Edinburgh and Cambridge to make arrangements for pipeline processing and archiving of all WFCAM data; photometric calibration is of course fundamental to the scientific usefulness of WFCAM data and the overall success of the project.
Currently, the UKIRT infrared imager is UFTI. This is based on a 1024×1024 'Hawaii' device with 0.092 arcsec/pix and a corresponding FOV of 2.5 arcmin2. This is soon to be replaced by UIST (latest projection is end of 2001) which uses a 1024×1024 'Aladdin' device, but which will have a 0.12 arcsec/pix mode yielding a FOV of 4 arcmin2. These contrast with WFCAM, which will have 4 2048×2048 devices (spaced at 90% detector width in the focal plane) and 0.4 arcsec/pix. Hence, the WFCAM FOV is 186 arcmin2 per detector. INGRID (WHT) has a FOV of 16 arcmin2.
The JHK filters currently under consideration for WFCAM will be Mauna Kea Observatory (MKO) standard system filters. Note that, for example, 2MASS (Appendix C) uses significantly different J and Ks filters.
The fundamental bright magnitude limit of observation with WFCAM will be dictated by the finite readout time (currently expected to be 0.6s for a 128-channel readout). For a standard reset-read-integration-read sequence, the device arrays must not saturate within 2×0.6s which yields an absolute maximum brightness of m~10. In practice, latency/remanence problems will possibly push this limit fainter.
Existing lists of primary photometric standards include Casali & Hawarden (1992); Persson et al. (1998); Hawarden et al. (2001).
Work in progress includes UFTI observations of the Persson standards and Landolt optical standards and also the establishment of Y (i.e. similar to Z in the MK list of Tokunaga) standards. Work to date has concentrated on JHKLM; calibration of non-standard (e.g. H2) bands could be achieved synthetically by multiplying instrument and filter responses and assuming, for example, a K zeropoint for H2. Ideally, primary photometric standards should be established for non-standard passbands.
The 2µm All Sky Survey aims to survey the infrared sky to 10-sigma limits of J=15.8, H=15.1 and Kshort=14.3 (for point sources). Some references where useful info can be found include: 2MASS overview, the 2MASS data-release explanatory suplement, and Nikolaev et al. (2000) & Carpenter (2001). The global properties can be summarised as follows: Minimum photometric error (at bright mags for point sources) is ~2% (JHK), maximum depends on mag. of course, and is of course ~10% at the 10-sigma limits above.
For the "secondary" photometric calibrators;
sigma-J ~ 3% }
sigma-H ~ 5% } at m=14
sigma-K ~ 8% }
Astrometry: from the Second Incremental Release (SIR):
SIR v Tycho-AC : RMS ~0.11 arcsec (n=106, Vlim = 11.0)
SIR v Tycho-2 : RMS ~0.18 arcsec (n=1.5x106 Vlim = 12.5)
SIR v UCAC1 : RMS ~0.12 arcsec (n=107 Vlim = 16.0)
(Note: Tycho-AC: Urban et al. 1998, AJ, 115, 2161; Tycho-2: Hog et al. 2000, A&A, 355, L27; UCAC1: Zacharias et al. 2000, AJ, 120, 2131)
As at 10/10/2000, the full 2MASS data products were "expected in late 2002".
Assuming single sky-limited exposures have half full-well at sky, then the dynamic range would be around 5mags. Assuming the sky-limited 3-sigma point source detection limits are 20,19,18 at J,H,K (e.g. the original WFCAM Technical Report submitted to GBFC), then stars as bright as J,H,K=15,14,13 would be just unsaturated in these ~10s exposures. Since the 2MASS point source counts take a dive at around J,H,K=16,15,14ish this means there's potentially about a 1mag range where 2MASS sources are not over-exposed in single-shot sky-limited WFCAM exposures; 2MASS point sources have the advantage of being well above the faint confusion limit and hence immune to photometric errors due to crowding. Point source counts are detailed in a 2MASS document.
(the normalisation is a bit strange here - these plots seem to be in no. of sources per pi sq. deg per 0.25 mag, which is very approximately per sq. deg. per mag; this checks out correctly with Hammersley et al. 1999). At m=15:
No. of pt. srcs.
per sq. deg. per mag.
|No. per WFCAM chip|
Of order ten stars per chip is sufficient to tie the photometric zeropoint of the data to 10%/sqrt(10), or ~3% assuming all errors are normally distributed (which of course won't be true - e.g. variables, confusion/blending etc etc). However the big problem with using observations employing quite different filters and made from a different site is likely to be the absolute zero-point (see the next Section).
The models in the above references can be used to predict point source number counts at magnitudes fainter than the 2MASS limit to examine the number of stars we might expect to have for the purposes of statistical calibration. For example, at l=301°, b=+78° there are ~300 sources per square degree per mag. at JHK=17; the total number of unsaturated point sources is likely to be ~500 per square degree or 25 per device .... (sufficient for statistical calibration?).
We remind everyone that there are two (at least!) kinds of accuracy: the intra-atmospheric accuracy, which can be quite high, and the extra-atmospheric accuracy (effectively absolute calibration) which is another kettle of fish.
The accuracy with which one can relate intra-atmospheric observations to one another is much the higher and enables the exploitation of precise colours/flux differences even though absolute fluxes are not well known. With well defined photometric systems (see below) the internal/intra-atmospheric accuracies could be well below 1%, i.e. well below 1/100 mag.
The caveat is the transformability of observations from site to site and from night to night, given variations in atmospheric water etc.
The 2MASS filters are not well designed (especially the J band which includes most of the J/H water absorption) and transformations between sites are likely never to be reliable, the more so as they observed at relatively low, wet sites and we plan to work at a very high, dry one. See the profiles in Fig. 1 of Carpenter (2001). The effect of the overlap of the filter bandwidth into the atmospheric water wavelengths is twofold.
First: The Forbes effect - see Hawarden et al. (2001) and references therein. The non-linearity which arises from extinction being caused by a mixture of saturated and unsaturated spectral lines, means that transformations between observations made though different amounts of water vapour are always poor, at least unless quite detailed modelling using H2O meter values is undertaken.
Second, the Forbes effect especially means that absolute calibrations are even more difficult because of the strong curvature of the extinction curve between 1 airmass and 0 airmass, caused by the lines which are saturated at all terrestrial altitudes, and therefore relatively unchanging with height of H2O, which become unsaturated. As you got to 0 airmass, all of a sudden you are getting back bits of spectrum which were wholly opaque before.
Much depends on what is required in the way of absolute accuracy. If 25% is good enough, we can go with 2MASS straight off. However, we have already argued that for the purposes of legacy science and general exploitability of the WFCAM science archive, it would be a great pity if absolute fluxes were only known to ~25%.
It is clear that establishing a network of standard fields around the sky before we have WFCAM on the telescope, i.e. employing UFTI or UIST, would be impractically slow. We should start by aiming to establish just a few 'master' calibration areas pre-WFCAM, and then use these to characterise the instrument once it is on the telescope, subsequently bootstrapping more fields around the sky onto these master fields as part of the WFCAM programme prior to any survey observations proper.
The requirements of such a master field are: i) extend over ~14×14 arcmin2 to cover a WFCAM device FOV; ii) the density of sources should be adequate to begin to characterise the position-dependent systematics present when observing with WFCAM; iii) the distribution in colour should be adequate to characterise any colour terms. Any network of fields should not all be equatorial - some should be available at airmass ~1.0 since survey observations will be made here.
One possibility is to establish master fields near or around existing UKIRT Faint Standards. The procedure in outline would be to measure up ~10 or so m=16 stars over a 14×14 arcmin2 field. If we assume 10 stars, 5 filters, 10 observations per star and 1min integrations including all overheads, then such a master field could be established in ~10hr or ~1 night (of course, observations should be split over a few nights to ensure realistic error estimates and limit observations to reasonable HAs). Given a few nights of UFTI/UIST observing per Semester in 2002/2003, several master calibration fields could be established in preparation for the commissioning phase of WFCAM.
One possible flaw in the above is the colour distribution of the stars which would essentially be random. Another possibility would be to use clusters where we can ensure a certain (wide) colour distribution. There are 22 Globular Clusters above |b|=20° and accessible to UKIRT. Most have horizontal branches at 16 to 18 at V and BHB populations, so one is assured of some stars at around these mags at K over a wide range in colour. The density of stars can be 'tuned' to requirements by observing at a chosen radius from the cluster centre.