Sky removal

remsky can not be used to calculate the sky contribution for scan map data because it is no longer possible to select bolometers that are guaranteed to be on sky (since most bolometers will see `source' at some point during the observation).

In order to overcome this problem the source signal must be removed from the data before attempting to calculate the sky. This is achieved with the calcsky task.

For each point in the input datasets calcsky finds the expected flux at that position by comparison with a model of the source and removes that flux from the input data. The source model can be calculated internally by calcsky or an external image can be supplied (usually generated from the same input data using rebin).

Figure: Sky noise calculated by calcsky for a short interval of the M82 data.

The source model is calculated in exactly the same way as for MEDIAN rebinning and despike: the input data are placed in bins related to position on the output image; the median of each bin is then taken to be a good measure of the flux in that region of sky. This approach is an approximation since the bin size (quarter beam width) can accommodate large gradient changes towards point sources but in general these errors are smoothed out by the average taken over the whole array. An alternative approach is to rebin the input data on a fine grid (e.g. 1 arcsec or finer) and use that as the input model (this is especially useful for scan maps since calcsky can add the dual beam response to the data when calculating the model)

Once the source has been removed a sky signal is calculated from the residual signal by finding the average signal across the array for each time. In addition the time series can be smoothed since scan data are sampled much faster (approx. 8 Hz) than the sky emission is expected to vary (a few seconds). These time series are then stored in an extension inside the file (stored in .MORE.REDS.SKY). Once the sky has been calculated it can be removed by using remsky. remsky recognises the presence of a SKY extension and removes this signal from the main data array.

Fig. shows the sky signal for some of the M82 data. In general the sky noise on scan data is below the noise but correlations are visible in the smoothed time series.

This technique is not limited to scan map data. Jiggle maps can benefit from using calcsky in cases where sky bolometers can not be identified or when the sky removal needs to be automated. One caveat is that the quality of the sky removal depends critically on the quality of the sky model. For extreme cases of sky noise in jiggle data where the individual switches are visible as hexagonal patterns across the image, calcsky can not disentangle the source from the hexagonal pattern (no other data are available for that position on the sky) and sky removal will fail.

More information on sky removal for jiggle and scan data can be found in Jenness et al.[13]