Summary: The original "patches" catalog derived from the Mark IV survey has systematic errors as a function of Declination. I have used the Tycho-2 catalog to remove these errors. Version 2 of the "patches" catalog contains the fixed magnitudes.
In January, 2007, Siegfried Roeser contacted me to describe the work he and a group of collaborators had done. They had compared the TASS Mark IV "patches" catalog to another large catalog, the All-sky Compiled Catalogue of 2.5 million stars. What they found was disturbing: a clear "sawtooth" or "zig-zag" pattern of differences in the V-band magnitudes as a function of Declination. Below is a graph showing the residuals in the range (-6 < Dec < +20).
The period of the sawtooth was about 4 degrees -- corresponding to the spacing of the Mark IV observations. There is apparently a repeated, systematic error in photometry as a function of the position of a star within the large Mark IV field of view. In this Dec range, stars near the North edge of the field are slightly brighter in the "patches" catalog than they should be, and stars near the South edge of the field are slightly fainter than they should be.
This is no great surprise, since we were aware of imperfections in the flatfielding procedures. The paper we published in PASP describing the "patches" catalog mentions their presence; Figure 15 of the paper shows a North-South pattern in residuals for one particular set of observations. I wasn't expecting the errors to be quite so large, or so repeatable.
The most likely cause of the effect is, in my opinion, the following:
The graph above displays the results for regions of the sky near the celestial equator. The camera taking those images pointed South, so the southern edge of each image was close to the horizon and the northern edge of each image was far from the horizon.
We can test this hypothesis; if true, there should be two consequences:
Let's look at the residual errors for other portions of the sky. I'll break the sky up into three regions, based largely on the Mark IV unit doing most of the observing.
Here is the graph of residuals in V-band for region 2,
... and finally region 3:
Yes -- the region overhead has only a weak pattern, and the region to the North DOES show a reversed sawtooth pattern.
So, we know HOW the errors occurred. Can we remove them, or at least reduce their importance?
In order to remove the errors, we need a way to characterize them. Note that the ASCC does not contain I-band magnitudes, nor is it a homogeneous set of measurements. I decided to go back to the original photometric reference for the survey -- the Tycho-2 catalog -- because it provides a uniform, all-sky set of reference stars. I used the relationships described in Table 3 of the survey paper to convert Tycho Bt and Vt magnitudes into a pseudo-I-band, so that I could find systematic errors in the I-band measurements.
I compared the magnitudes of stars in a subset of the Tycho-2 catalog with good photometry to the magnitudes of stars in the original "patches" catalog. I divided the results into bins, of width 0.15 degrees (for Dec < +20) or 0.25 degrees (for Dec > +20). Most bins contained 100-400 stars, though the number fell precipitously at Dec > +80 degrees. I computed the median value within each bin to come up with a set of corrections. Finally, I normalized the corrections (within each of the 3 regions of the sky) so that the mean value would be zero; that way, while the magnitudes of individual stars would be shifted up or down, the overall zeropoint would remain fixed over the entire sky.
Below are graphs showing the corrections in V-band ...
... and in I-band:
You can look at the values in tabular form here:
The corrections should be applied in the sense
corrected V = patches V + median offset
So, how well do this model remove the errors? Here are two ways to evaluate the result. First, we can compare the corrected magnitudes against the ASCC catalog, in V-band only (since ASCC doesn't have I-band values). Below are graphs showing the residuals for the three regions of the sky, (-6 < Dec < +20)
(+20 < Dec < +45)
(+45 < Dec < +90)
Second, we can look at differences between the corrected Mark IV values and the Tycho-2 catalog magnitudes (in both V-band and pseudo-I-band). In each graph, the red points show the V-band residuals, and the black points show the I-band residuals, shifted by -0.5 magnitudes for clarity.
Again, I'll split the sky into three regions, (-6 < Dec < +20)
(+20 < Dec < +45)
(+45 < Dec < +90)
One can still see some sharp features at the edges of each Dec zone in region 1 of the sky in V-band, but the amplitude of the pattern overall is much lower.
I have created a new, version 2 of the Mark IV "patches" catalog. This version is identical to the original in format; the only differences appear in two columns of values.
There are exactly the same number of stars, and all the other columns remain as they were.
You can find a copy of this version 2 at my Mark IV "patches catalog" web site. I will ask the SIMBAD administrators to replace the version hosted on their web site with the corrected version.
Last modified 07/19/2007 by MWR.