MegaPrime/MegaCam Observing Statistics II
Table of contents:
This is a summary of observing statistics for the years 2019-2021. It is similar to
these observing stats, but with a focus on the IQ
and FWHM (Full Width at Half Maximum) of point sources. The previous statistics page covers the first 3 years of MegaCam, from
2003 to 2006. This page covers from 2019 to 2021. There are a couple of differences
to note between the 2 sets of statistics:
- MegaCam started using all 40 chips (instead of 36) in 2015.
- The filters were changed out in 2015 - they are similar but not identical, and
were change out to fully cover the 40 chips of MegaCam.
- There are minor differences in the data selection (cuts on exposure time for 2003-2006, cuts on whether an exposurse is sidereal or not for 2019-2021)
- There are minor differences in what is displayed in the histograms in addition to the differences in data selection - 2003-2006 counts exposures, 2019-2021 counts exposure time.
2. Image quality and seeing
2.1 Selection of data :
In recent years, CFHT has observed a lot of fields using non-sidereal tracking. The IQ for these exposurse will be large, because the stars and other objects on these images will appear as streaks, which Elixir will then try to determine the size. Therefore, we exclude from the data sets any exposures that requested non-sidereal tracking. In the time period between Jan 1, 2019 and Jan 14, 2022, there were 43135 science exposures taken with MegaCam. Of those exposures, 10227 were non-sidereal, 32908 were sidereal. When counting by exposure time, ~204 hours (12% of the time) were non-sidereal, and ~1477 hours (88% of the time) were sidereal. For the histograms and statistics that follow, only science images with sidereal tracking are included.
2.2 Image quality histogram per filter :
2.3 Comparison to 2003-2006 statistics :
For 2019-2021, the mode is taken as the highest histogram bin - the phase is off by half a bin from the 2003-2006 data. However, certain trends can be seen. First, u and r band are the primary filters used in recent years, while it was more evenly spread out in 2003-2006. Also, u and r currently have a slightly better IQ than in 2003-2006. Otherwise, the numbers are generally quite similar.
3. Comparison of Elixir IQ to true FWHM
Elixir's "IQ" metric used for monitoring image quality and seeing on MegaCam exposures is an estimate of the Point Spread Function (PSF) Full Width at Half-Maximum. It is based on SExtractor's
FWHM_IMAGE measurement. Being derived from isophotal areas it is extremely fast to compute and reasonably robust towards source crowding. However it slightly overestimates the true FWHM by about 7%:
The figure above shows Elixir's IQ parameter as a function of the PSF FWHM for 34,079 sidereal-guided MegaCam exposures.
PSF models and FWHM measurements were obtained using the PSFEx software package. PSFEx' PSF models are tabulated at an arbitrary resolution (it is a small image sampled in such a way that it meets Nyquist's criterion). When the PSF FWHM drops below 4 pixels, a custom super-resolution algorithm is applied that recovers aliased signal components. For this comparison we restricted the measurement to MegaCam's (central) CCD#13 to make the processing faster, and we assume that PSF variations within the CCD can be modeled with a first degree polynomial.
PSFEx' FWHMs are computed by fitting a 2-dimensional Moffat function to the PSF model at the center of the image. The Moffat function is convolved with the Intra-Pixel Response Function, which we assume to be a perfect square boxcar with 100% fill-factor. PSFEx' FWHMs are therefore expected to be those of the optical PSF, corrected for the effect of pixelation.
To test whether PSFEx could itself underestimate the FWHMs, we ran through the same measurement pipeline a set of SkyMaker simulations of MegaCam images with seeing FWHMs ranging from 0.4 to 2.3". As one can see below, the recovered FWHMs match the ground truth within ~0.5%. Hence one expects PSFEx' pixel-free FWHM to be a good proxy for the actual seeing FWHM.