[Skip to Content]


General Links
Professional Links
Australian Human Resources Institute Diversity Awards 2013 Finalist

Imaging Observations with IRIS2


Imaging Strategies

Compared with optical imaging, infrared imaging is a lot more challenging. Not only is the infrared sky background typically hundreds of times brighter than your target, but the intensity can also vary by 10% or more in just a few minutes, and by factors of two from night-to-night and from season-to-season. The sky emission in the J- and H-bands is dominated by families of emission lines arising from vibrational state transitions of OH radicals. Pressure waves in the atmosphere cause the OH sky emisison to fluctuate in time and space, as dramatically illustrated by the 2MASS Wide-field Airglow Experiment.

In the K-band, the short-wavelength tail of the thermal emission from the earth's atmosphere (a ~300 K blackbody whose flux peaks near 10 µm) dominates, and although brighter than the OH emission in the J- and H-bands, varies much more slowly. As the night-time temperature at Siding Spring Observatory is typically 25 degrees warmer in summer than in winter, the K-band sky brightness is correspondingly greater, resulting in a 0.5 mag decrease in sensitivity. For this reason, observers are recommended to use the Ks filter instead, which has a slightly "bluer" passband than the K filter which makes it less susceptible to thermal effects. Results in the Ks filter can then be transformed to the corresponding magnitude in the MKO K filter, as described below.

To further complicate matters, differences between the illumination pattern of the dome flatfield screen and the sky itself, coupled with slow variations in the background pattern internal and external to IRIS2, mean that the images must be "self-flattened" using other images as close in time and position as possible. This leads to the standard observing strategy of dithering (also called "jittering") successive observations by at least a few arcseconds, so that (in uncrowded fields), simple medianing of at least 3 images will result in an image of the sky which is free of stars, and which simultaneously represents the broad-scale illumination of the array, and the pixel-to-pixel response to this illumination. As a bonus, this same strategy also allows bad pixels in the array to be "filled in" by shifting the images back to a common reference frame, and using data from only the good pixels to reconstruct the image pixel-by-pixel.

An excellent illustrated discussion of imaging strategies in the near-infrared, addressing questions such as "How often should I dither?" and "Do I really need to spend as much time on blank sky as on my object when observing extended sources?" (Answer: Yes you do!! ) can be found in the paper by Vaduvescu & McCall (2004, PASP, 116, 640; astro-ph/0404337). Here we discuss a couple of the major issues observers have to grapple with when planning their observations.

How big?

The first question is therefore how much to dither by? The answer depends to some degree on how compact or extended your targets are, which will usually fall into one of 5 regimes:

  1. Compact (<1 arcminute) or unresolved sources
    For sources which are compact (smaller than one arcminute or so), or unresolved on the scale of the one arcsecond or so seeing at SSO, the solution is simple. Edit a copy of the AAO_Jitter.tcl sequence and set the jitter parameter to be larger than the expected size of your targets or anything else in the field. In the case of stars or high-redshift objects in relatively uncrowded fields, set jitter to at least 10 arcseconds, to allow for the presence of field galaxies or haloes around bright foreground stars. ORAC-DR can identify and mask out such structures, but it is best to make sure jitters are large enough to avoid any overlap in adjoining images. Jitters of up to 60 arcseconds or so are possible, but beyond this, astrometric distortion (even after correction) will make reconstructing a mosaic harder, coupled with the fact that the region common to all dithered images (giving the best signal-to-noise) shrinks as jitter is increased.

    The choice of a FIXED or RANDOM dither pattern is not so important; random dithers make it less likely that artefacts due to (say) array or quadrant boundaries will be apparent in the final mosaic, but these are seldom noticeable even with a fixed dither pattern. In any event, each sub-mosaic formed from a set of fixed dithers will have a small random displacement applied of up jitter arcseconds from the previous one.

  2. Source size between 1 and 3 arcminutes
    As sources occupy a larger fraction of the array area, there comes a point where offsetting to blank sky becomes necessary. This inevitably leads to a reduced efficiency as not all the observing time will include the object of interest. There is a narrow "window of opportunity" in source size of 1-3 arcminutes in which it is possible to keep the object within the field of view at all times, but cycling between the four readout quadrants of the array. Simply slew to your target object, place it on or near the reference pixel, then execute a copy of the AAO_Quadrant_Jitter.tcl sequence after setting the object name and exposure parameters.

    There is a small price to pay for this efficiency however. As the flatfield is created by masking out the entire quadrant containing the object and using only the data from the other 3 quadrants, the cosmetic quality of the flatfield and the reduced mosaic will inevitably be poorer than one for which more dither positions are available. In addition, the overlap region between the 4 large dithers having the highest and most uniform signal-to-noise is just 1/4 of the IRIS2 field of view (3.8' x 3.8'). To do surface photometry of an extended object, you will need at least the outer 1/4 of this to determine the sky level, so objects much bigger than 2 arcminutes or so may still not be suitable for this approach.

  3. Source size between 3 and 7 arcminutes
    When your object of interest all but fills IRIS2's field-of-view (FoV), your only option is to interleave regular dithered images of your source with dithered images of adjacent blank sky. By "adjacent", we mean displaced by at least one FoV, but much further than this and the sky illumination may no longer be representative of that across the object (see the 2MASS Wide-field Airglow Experiment for an illustration of this), and overheads due to telescope offsetting will become significant. Offsets due north or south are fastest, as they do not involve polar axis rotation. However, be sure to examine the sky frames as they are taken, lest a bright star or other extended source ruin the flatfield. If so, you may need to change the sign or magnitude of the declination offset parameter ddec3 in the AAO_Chop_to_Sky.tcl sequence. Note that the CHOP_SKY_JITTER recipe in ORAC-DR will expect ndith images of your object to be bracketed by (ndith + 1) dithered images of adjacent sky (so the sky level in each object image can always be interpolated), which is why AAO_Chop_to_Sky.tcl always starts and ends on a sky position.

  4. Source size between 7 and 10 arcminutes
    If your object slightly overfills IRIS2's FoV, it should still be possible to obtain a reasonably good mosaic, but getting a uniform sky background over the final image will be challenging. The sequence AAO_Extended_3x3.tcl makes a 3 x 3 mosaic of your source with a generous (but recommended) 50% overlap between pointings, interleaving each of the 9 images with 10 bracketing dithered sky frames. To reduce the risk of introducing large scale sky gradients across the final mosaic, the 9 object pointings are done in a zig-zag pattern rather than sequentially in any axis.

    Extension to larger mosaics (e.g. 5 x 5) is possible, but you may want to consider applying for WFCAM or VISTA instead...

  5. Mapping large uncrowded areas
    If you wish to image a contiguous field on the sky larger than the IRIS2 FoV, and it does not contain any significant extended structures, then offset sky frames will not be needed. You can use the AAO_LargeAreaNew.tcl sequence to map a region of N (RA) x M (Dec) IRIS2 fields, with an overlap of 10-20% between each. You can specify the number and size of dithers on each field. Note that ORAC-DR will make a mosaic of each field, but will not attempt to make one "super-mosaic" of all N x M fields, as this would be somewhat unwieldy and difficult to maintain a uniform sky level. The other nice feature of this sequence is that if it is terminated early, it can be restarted at any particular field and proceed from there, rather than having to repeat fields already covered.

How long?

The other big questions are how long to spend integrating between dithers, and how often is a new sky flatfield required? Experience has shown that about one minute per dither is fine with IRIS2. Much shorter than this, and the overheads involved in offsetting the telescope more frequently and writing more images to disk start to become a significant fraction (30% or more) of the total elapsed time. Spending longer on every position (e.g. 2 minutes) may make observing slightly more efficient, but does mean that the time required to collect a minimal set of dithered images (3 in theory, but at least 5 in practice) becomes comparable to the timescale on which changes in the sky background structure (not just the level) begin to become noticeable.

This then leads us on to the second issue - combining images taken over a timespan of 10 minutes or more to make one flatfield will often result in residual structure in the flatfielded images, and resultant poor-quality mosaics. This is especially true in H- and (to a lesser extent) in J-band than in K-band, since the OH emission varies much faster than the thermal background. However, every night is different and even the most brilliantly clear, photometric nights can show wild and rapid fluctuations in sky brightness which may be almost impossible to overcome, no matter how frequently one dithers. The bias image subtraction option in Skycat, together with the reduced mosaics produced by ORAC-DR, will quickly indicate to you just how stable the night is, and help you decide if more frequent dithers, or fewer dithers per mosaic may be required. If, after the event, you feel that (say) 5 dithers per mosaic may have been more appropriate than 9, then all is not lost. It is possible to modify the number of dithers per mosaic ORAC-DR expects, and have it re-reduce the data to see if this helps.

Observations of extended sources for which offsets to blank sky are required (regimes 3 & 4 above) are even more susceptible to such sky instabilities. To obtain the 5 object and 6 sky frames expected by the CHOP_SKY_JITTER recipe within this 5-10 minute window means no more than 30 seconds or so should be spent integrating at any one position.

Photometric Standards

Photometric calibration of infrared data has become a lot simpler in recent years with the availability of the 2MASS Point Source Catalog data. Nevertheless, it is important that you keep in mind the importance of transforming between the "natural" system of IRIS2, which most closely resembles the Mauna Kea Observatory (MKO) system, and other systems in common use such as 2MASS, Las Campanas Observatory (LCO), etc.

The Carter & Meadows standards and CIT/CTIO standards are almost too bright to be observed in the shortest integration time (0.6s with the science-grade array). The UKIRT Faint Standards are well-suited to 3-10 sec exposures, and their transformations from the MKO system used in IRIS2 to other systems such as 2MASS and the LCO system are now well-determined. Other standards in common use include the LCO Red stars, the HST/NICMOS Faint Standards, and the Hunt standards. As usual, it is recommended that standard stars be observed over a range of airmasses similar to that of your targets.

In practice, the all-sky coverage of 2MASS and the wide field of view of IRIS2 means that virtually any IRIS2 image can be calibrated photometrically, even if the night itself was not perfectly photometric, and without even needing to know the extinction. Once the ORAC-DR reduced mosaic appears in gaia, select Data-Servers -> Catalogs -> 2MASS Catalog at CDS. Click on Set from Image, then Search. This will bring up a list of 2MASS stars within the image and their J, H, and Ks magnitudes, then overplot them on your image. Although the alignment may not be perfect, you should easily be able to associate a particular star in your image with its catalog data, by clicking on its red marker. You can then use Image-Analysis -> Aperture photometry -> Results in magnitudes... to derive a zero-point for the image by comparing the measured aperture magnitudes with the known 2MASS magnitudes for several stars in the image. You should use the transforms below to correct for differences between the MKO system of IRIS2, and 2MASS. You can also use the 2MASS catalog data to improve the astrometric calibration using Image-Analysis -> Astrometry calibration -> Fit to star positions... Select Help -> Astrometry Overview... from the Astrometry popup window for advice on how to proceed.

Transforms to Other Photometric Systems

The J,H,K,Ks filters are the same MKO infrared filters as those discussed in Alan Tokunaga's 2001 paper. This paper contains links to other work using NSFcam on the IRTF and IRCAM3 on UKIRT to define colour terms for these filters. The colour terms for IRIS2 will be different due to those instruments using InSb (not HgCdTe) detectors, and having different optical trains.

Deriving transformations from the IRIS2 system to other systems require that enough stars spanning a wide range in colour be observed. Only a preliminary transformation to the MKO system is available for the "Mark 1" science-grade array in use between March 2002 and April 2005, based on observations of only a handful of UKIRT Faint Standard stars over 3 separate nights in July 2002:

(J - H)_MK = 0.90 +/- 0.05 (J - H)_IRIS2
(H - K)_MK = 1.05 +/- 0.05 (H - K)_IRIS2
(J - K)_MK = 0.95 +/- 0.05 (J - K)_IRIS2

Perhaps most useful is the transformation between the MKO K band, and the Ks band more often used with IRIS2 for reduced background:

K_MK = Ks_IRIS2 - 0.13 (H - K)_IRIS2
= Ks_IRIS2 - 0.12 (H - K)_MK

i.e., redder stars will appear brighter in K than in Ks.

For the "Mark 2" science-grade array in use since May 2006, a more complete set of transformations to other systems has been derived from observations on the night of 2 Sep 2006 of UKIRT Faint Standards FS 140 and 144; NICMOS standards S279-F, S024-D, S071-D, S808-C, and S889-E; and LCO red stars L547 and BRI2202. The zero-points (magnitude of a star in the MKO system giving 1 ADU/s in a 10 arcsecond diameter aperture at 1 airmass), extinction (mag/airmass), and colour terms for this night are given by the following relations:

JMKO = 22.58 - 2.5 log (I / t) - 0.076 (X - 1) - 0.002 (J - K)MKO
HMKO = 22.79 - 2.5 log (I / t) - 0.048 (X - 1) + 0.038 (H - K)MKO
KMKO = 22.35 - 2.5 log (I / t) - 0.090 (X - 1) + 0.003 (J - K)MKO
Ks2MASS = 22.37 - 2.5 log (I / t) - 0.097 (X - 1) - 0.009 (J - Ks)2MASS

where   I = total integrated ADU in 10" aperture;
             t = integration time in seconds;
and     X = airmass of observation.

Note that we do not attempt to extrapolate to zero airmass, due to the Forbes effect (i.e. the extinction curve is non-linear between 0 and 1 airmass) - see Section 2.4 of Tokunaga & Vacca (2007; astro-ph/0702285).

Although included in the MKO filter set, the Ks magnitude scale is not defined in the MKO system, so 2MASS colours are used instead. Transforms to the 2MASS system are given by:

JMKO - J2MASS = -0.03 - 0.03 (J - Ks)2MASS
HMKO - H2MASS = -0.01 + 0.05 (H - Ks)2MASS
KMKO - Ks2MASS = -0.01 - 0.01 (J - Ks)2MASS

The difference between IRIS2's Ks magnitudes and 2MASS Ks magnitudes was found to be <0.01 over the full colour range.

Transforms to the LCO system are given by:

JMKO - JLCO = -0.02 - 0.05 (J - K)MKO
HMKO - HLCO = -0.02 + 0.06 (H - K)MKO
KMKO - KLCO = 0.00 - 0.03 (J - K)MKO

Transforms to other systems can be derived with the help of Leggett et al. 2006, MNRAS, 373, 781 (astro-ph/0609461).

When absolute photometric accuracy is paramount, we still recommend observers determine their own extinction corrections each night, and when transformation to other infrared photometric systems is required, that a suitable colour range of standards be observed to calibrate their own data.

Dome flats - J,H,K,Ks

Experience has shown that by far the best flatfields for imaging in both broad- and narrow-band filters are those constructed from median-filtering the dithered on-sky images of your target field (or interleaved sky frames if imaging extended sources). However, where circumstances preclude this, it may still be possible to remove the pixel-to-pixel variations in sensitivity (but not the broad-scale illumination pattern) using dome flats. The best procedure for these is to take 'lamp on'-'lamp off' sequences with the telescope pointed at the white patch on the dome. The dome flat patch on the windscreen can be illuminated using lamps mounted at the Prime Focus access area. There are two lamps, which can be plugged into a socket, the voltage for which can be controlled from the control room. The two lamps are a desk lamp, and a much brighter flood lamp. If the telescope is not positioned for dome flats, ask the night assistant, support astronomer or afternoon technician to put it there (Dome=0o, Windscreen=21o).

Change the lamp at Prime Focus access to the desk lamp. Adjust the dome flat lamp intensity (from the knob in the control room) to get ~9000 ADU with the lamp on at Ks in 1.5s. Then check that with the lamp off, there is a difference in recorded counts from the lamp on of at least 4000 ADU / pixel (this may get hard to achieve in summer, especially for K band flats - in this case you'll just have to take more flats to beat down noise).

Recommended exposure times are Ks=1.5s, H=3s, J=4s, (and do about 30 cycles). Then for each filter repeat with the lamp off.

There is a sequence to do this. Set up the lamp to get the right counts in Ks (above) in 1.5s. Then start tcl_jan03/dome_broad_on.tcl. Then turn off the dome flat lamp, and start tcl_jan03/dome_broad_off.tcl. Each sequence takes about 7 minutes to run.

Dome flats - Narrowband filters

For narrow band filters change the desk lamp at prime focus for the large flood-lamp. Turn the control room dial right down to its minimal setting. You should get ~9500 ADU in a 1.5s exposure with the H2 v=2-1 filter. Then the following table indicates rough count levels for all the filters in 1.5s.
Filter Time Desk Lamp on full
Desk Lamp off
H2v=1-0 1.5 11000
Br gamma 1.5 10000 300
H2 v=2-1 1.5 9500 500
K cont 1.5 10000 1000
CO 2-0 1.5 7000 800
Pa beta 1.5 6000
H cont 1.5 8000
Fe II 1.5 12000

There are sequences to take all these. Set up the main lamp (as dim as it will go) to get the right counts in H2 v=2-1 (as above) in 1.5s. Then start tcl_jan03/dome_narrow_on.tcl. Then turn off the dome flat lamp, and start tcl_jan03/dome_narrow_off.tcl. Each sequence takes about 10 minutes to run.

Return to the main IRIS2 page.

These pages contain information on the functionality of the IRIS2 Infrared Imager and Spectrograph. Pages maintained by Stuart Ryder (sdr -@- aao.gov.au) and Paul Dobbie (pdd -@- aao.gov.au). Page last modified by Stuart Ryder.