It is my understanding that 2MASS currently uses the observed scatter from the calibration stars during a night as the measure of the uncertainty in the zero-point calibration for that night. Based upon an analysis of the observed zero-point uncertainties for all southern hemisphere K nights, a better model would be to use a constant zero point uncertainty for every night.
Roc Cutri has histogrammed the Nightly Photometric Zero Point Uncertainties for each band and each hemisphere. Here I have analyzed the data for the southern hemisphere at K.
Figure 1 shows the data, as read off Roc's plot, as well as a model which assumes that:
For each of 569 nights (the number in Roc's histogram), I generated six random zeropoints drawn from the population with mean = 0.00 and true sigma = 0.010 mag. I calculated the observed standard deviation from that sample about the observed mean. The histogram of the simulated results matches the observed zero point uncertainties extremely closely.
Thus I conclude that, in this case at least, it is far better to simply quote 0.010 mag as the uncertainty in the zero point correction for a given night, instead of using the observed values (from the simulation) of 0.0023 to 0.0220 mag, which are factors of 4.4 times too low to 2.2 times too high.
The only significant difference between the model and the data is for a single one of the 569 points. The observed histogram has a single point in the 0.030-0.032 bin, which is much larger than the largest point in the simulation, found in the 0.022 - 0.024 bin. I suggest that this night be reexamined for some other problem, such as clouds.
http://spider.ipac.caltech.edu/tchester/2mass/analysis/other/zero_point_uncertainties.html
Comments and feedback: Tom Chester
Last update: 22 January 2001.
