Peter Yoachim presented on the cloud model, data available so far and how he's been analyzing it, and what we might be able to learn about the cloud model and cloud prediction from that data. We had a lively discussion, including several very useful suggestions as to what to do next with the available data.
Action items:
create cloud mask (note: not a transparency map, just a binary cloud/no cloud mapping of where on the sky clouds exist – we have trouble getting data to do transparency maps), from daytime all-sky camera records. The daytime records are useful in that large clouds with small breaks are easily distinguishable from small clouds in a clear sky (not necessarily possible during nighttime, as clouds can be faint spots or bright spots on the sky, depending on moon illumination).
create maps of stellar density in 'chunks' across the sky, using the nighttime all-sky camera data.
evaluate requirements for future all-sky camera, if cloud transparency maps are required. making cloud transparency maps using the existing all-sky camera is difficult or impossible, as stars are simply not bright enough and not high enough SNR to measure cloud extinction beyond the very faintest level. what sensitivity would be required in a future all-sky camera to enable the creation of all-sky cloud transparency maps (rather than just cloud masks)?
As a follow-up, here's a plot of the number density of stars produced by the all-sky camera photometry pipeline:
There are actually 4 pixels there that have NO STARS at all, and many more where there are 1-5 stars. Since the photometric precision is very limited (typically 0.2 mag precision), and light clouds can make many of the stars undetectable, there's not much that can be done with the photometry.
And just for good measure, here's a pretty movie. Upper left is the median-binned all-sky image, upper right is the image differenced with the previous image, and lower left is the all-sky frame differenced with the dark-time median image. A pretty contrail goes by.
