Backgrounds in Herschel maps

Following several questions and comments about backgrounds in Scanamorphos maps, I felt the urge to provide this short explanation, that I hope will be useful:

Before wondering whether the background or sky brightness was incorrectly set in your maps, please remember that the bolometer arrays of Herschel are not absolute photometers, contrary to those of Planck. As a consequence, the map produced by any software is the superposition of an estimate of the true sky emission and an unknown (large) offset. There is absolutely no way of determining this general offset from Herschel observations alone.

In addition to this, the background level that you have to use in photometric measurements will depend on the exact astronomical application that you have in mind. Suppose that you want to measure the flux of an extended feature (a cloud, or a nearby galaxy, for instance); then you will define the background level as the average brightness of pixels located well outside this feature, and subtract it from the map before doing the photometry. Suppose now that you want to measure the flux of a compact source on top of the same extended feature (a protostar, or a star-forming complex, to follow the previous examples). Then, obviously, the definition of the background has to change: the extended feature that was the source of interest in the previous case now becomes the local background, whose brightness will be estimated in a small area around the compact source.

For these two reasons, it is impossible to perform a generic background subtraction that would be meaningful, without having a pre-established knowledge of the nature of the field and sources. Therefore, it is the task of the astronomer to define and measure the background, before subtracting it from the map.

Scanamorphos performs a crude background subtraction, but the offset that is subtracted is not exactly the same in maps processed with different options (for example with or without the deglitching), since in this case the brightness series are not identical. Conclusion: always estimate a background for each map, even if the purpose is only to compare two maps produced by the same software.