Why thin high cloud ruins long exposures

Cirrus you can barely see will halo your bright stars, lift your background and vary between frames. Visual observers shrug it off. Long exposures integrate it, so it accumulates.

The key difference is integration. Your eye works in something like a tenth-of-a-second refresh, so thin cloud drifting across just makes the view flicker slightly. A four-minute sub-exposure adds up every photon that arrived during those four minutes, including all the scattered light from the cirrus.

Three things go wrong. Background level rises, costing contrast. Bright stars grow soft halos that no amount of processing removes cleanly. And because the cloud is moving, the effect varies frame to frame, so your subs no longer calibrate consistently against each other — which is a far worse problem than a uniformly slightly worse night.

This is why the forecast breaks high cloud out as its own track rather than folding it into a total, and why the long-exposure and wide-field modes in the window finder penalise it much harder than the visual modes do. A night that reads "mostly clear" on a single-figure forecast can be unusable for imaging.

If you are stuck with thin cirrus, narrowband is the escape hatch. A filter with a passband in the 3 to 12 nm range transmits the emission line you want and rejects most of the surrounding continuum, including the scattered light the cirrus is adding, and tighter filters at 3 to 5 nm suppress it further still. Shorter subs and more of them also help, by giving you more frames to reject.