Photometry for extended sources is less accurate than for point sources with the same magnitude due to the use of more pixels to obtain the flux measurement and due to the error in determining the Atlas Image background level for those pixels. Most of the time, for most extended sources, the photometric uncertainties have been demonstrated to be close to theoretical expectations calculated solely from the absolute background measured in the frames, the operations used to construct the Atlas Images, and the number of pixels in the aperture used to determine the flux measurement (see Analysis of Photometric Noises for 2MASS Galaxies).
Thus, the flux uncertainties quoted for all sources in the sampler release solely derive from the theoretical flux uncertainties based on those three quantities.
However, there are three known sources of additional error beyond the theoretical expectations for 2MASS extended sources:
i. Error in Determining the Aperture Size
The aperture size must be determined for each source for isophotal magnitudes. As a result, there is an additional flux error of:
df = df/dr * dr
which depends on the variation of the source flux with radius. 2MASS cannot independently determine df/dr, and hence the user must be aware that the quoted flux uncertainties for isophotal apertures are somewhat larger than the quoted errors. There is no additional flux error from this source for fixed-aperture fluxes.
Error Analysis For Circular Isophotal Magnitudes discusses this effect further. Because of the enforced minimum isophotal radius of 7", many of the fainter 2MASS galaxies have in fact no additional error, due to this cause since the radius is fixed at 7" for their "isophotal" magnitudes. However, nearly all 2MASS galaxies brighter than 14.5, 13.8, 13.0 mag at J, H, and Ks have additional error due to this cause, which can contribute an additional error as large as 0.1 mag.
ii. Small-Scale Airglow Variation
Even on perfectly photometric nights, H band frequently exhibits significant small-scale airglow variation that cannot be removed with the 2MASS background-removal algorithm (see Analysis of Noise In The 2MASS Atlas Images). The 2MASS background-removal algorithm works very well most of the time, and removes nearly all background variation with spatial scales larger than 4-5', including nearly all of the natural airglow variation. The spatial scale of 4-5' was chosen as the best compromise between fitting the background without affecting extended sources of size 1-2'. In the dataset we worked with to tune the algorithm, no cases of background variation at shorter spatial scales were seen.
However, the much larger 2MASS data sets revealed that a significant fraction of the data contain a small amount of power in the airglow variation at spatial scales smaller than 4-5', causing significant additional error in the photometry of extended sources. The additional error is statistically correlated with the background-removed noise as measured in each 2MASS Atlas Image, which is quoted for each extended source. H Photometric Error Due To Airglow gives quantitative measures of this effect for a galaxy with H = 13.8 mag. The additional error ranges from a negligible error for a background-removed noise of 1 DN or smaller, but can be as large as 0.15 mag for 10% of 2MASS scans and 0.27 mag for 1% of scans. The additional error of course also depends on the magnitude of the extended source and the aperture used for the flux measurement.
Visual examination of the Atlas Images usually quickly reveals whether airglow is a problem for that Image. If present, it is relatively easy by hand to determine the level of the background noise at the position of an individual source, and correct the reported 2MASS H flux. We will explore automatic methods of removing the residual airglow, but are not optimistic that we can do so without significantly affecting source fluxes in other ways.
iii. Electronic Noise
The 2MASS camera electronics produces additional noise present in the individual camera frames that maps into noise that varies on a spatial scale of less than the 4-5' that cannot be removed by the extended source background-removal algorithm. The noise varies in frequency and amplitude, and hence the resulting noise in the Atlas Images varies dramatically due to phasing between the frequency of the noise and the fixed 1.5-s separation of the six frames that are combined to produce the Images.
We have only recently begun to study the additional noise caused by the remaining electronic noise in the Atlas Images. We have found evidence of J band electronic noise with dominant periods of around 50" spatial scale in the northern data, and 80-100" in the southern data. For scan 116 of 971116n, the Sampler night, electronic noise is found only on the left-hand side of the Atlas Images, with peak amplitudes of 0.16, 0.12 and 0.15 DN at J, H and Ks (see Pickup Noise In 971116n Scan 116. These values are very much larger than the level of 0.05 DN at which errors in the background do not contribute extra error to extended source photometric uncertainties. Typical additional errors are 0.05 - 0.10 mag.
When further analysis is available, we will revise this documentation.