Here is a histogram of r and i exposure minor-axis FWHMs for all SV nights through Dec 16, 2012. FWHMs are computed by fitting sech^2 profiles. A few nonsense-values
have not been purged (e.g., there are no actual exposures with FWHM < 0.7 arcsec).
Showdown: DECam v. Kolmogorov¶
Turbulence theory predicts that the seeing FWHM should depend on wavelength as (lambda)^(-0.2). Let's see if that's true.
I queried the firstcut database for all SV images with exposures 3-60 seconds for ugriz and 3-44 seconds for Y, sorted by filter, and computed the median. To accommodate the occassional bad tracking, I use the minor-axis width of the seeing profile as the measure of seeing. These "short" exposures are mainly standard star fields and other test exposures. There is a presumption here that by the ensemble of all nights represents a fair sample of the seeing at any time.
I also queried the database for all SV images with exposures 90 seconds (griz) and 45 seconds (Y), sorted by filter. These are all the main survey field exposures (mini-Survey, SPT West, SPT East, and the cluster fields).
The following figure gives log(median FWHM) v. log(wavelength) for the two datasets. Wavelengths are from David James docdb article on the Asahi filters. The points are measured; the cyan line is the prediction of Kolmogorov. (I have offset the short and long exposure points for clarity). Basically, it all looks pretty good. The median FWHM is about 1.1 arcsec, so it is going to be dominated by the atmosphere and not the camera optics or focus. Kolmogorov is validated.
Kolmogorov also says that seeing varies as (airmass)^0.6. The following figure is now a plot of FWHM v. air mass (except I use a linear scale for FWHM this time). The points are from just standards stars, and all bands are used, with the FWHM being adjusted for wavelength as above.