As part of a scientific investigation into the colours of SDSS BOSS training sets QSOs we have compared the photometry derived from the VDFS list driven photometry tool implemented at the WSA on UKIDSS LAS data with the VDFS catalogue photometry for the same images and find some significant systematic differences. Below we show the results for K band analysis. The systematic offset was on the scale of 0.08 magnitudes as shown below [Figure 1] .
On the assumption that this could be correlated with our own input, one of the photometric correction factors, or pipeline version we did some trend analysis.
First we looked at our input data ra, dec distribution [Figure 2] . This indicated that the problem was restricted to a narrow range of RA, Dec superposed with UKIDSS chip boundaries. This implied that there could be a temporal link with the problem since adjacent fields are often imaged consecutively. In the absence of mjdobs we used the multiframeid as a proxy modulo that this is an ingestion order which might not always coincide with mjdobs order. The look at the whether apercor, extinctionCorr, distortionCorr are illuminationCorr where correlated with the systematic difference.
This analysis showed that the problem was restricted to specific set of multiframes and that the atmospheric extinction correction was the likely guilty parameter. By the power of deduction and persusal of Hodgin et al, 2009 we deduced that the atmospheric extinction correction was possibly been incorrectly applied in the wrong direction.
We therefore corrected the output by subtracting twice the extinction correction which had been provided in the list photometry output. This appears to give results that agree better with the magnitude derived from VDFS CASU catalogue products.
A hisogram of the magnitude differences is shown in [Figure 3]. The difference histograms show the effects of using different input positions; (i) the SDSS positions (ii) UKIDSS LAS bandmerged Source positions; these are based on the bluest detected band (iii) K detection position. Using the K band catalogue position on the same K band image gives the best results.
A normalised error plot is in preparation. We will look at the outliers next. Some initial analysis indicates that the largest outliers are due to edge effects or cases where the sources are blended.
We have not been able to verify that the other corrections such as apercor3, distortionCorr,illuminationCorr have been applied correctly since the raw fluxes and x and y pixel values are not provided.
We request the following features in order of priority:
The first two which relate to the position on the chip are the most urgent since they are needed for data flagging. The information is within the fits file analyzed.
The aperflux and aperfluxerr are needed to derive upper limits. At the moment we can reverse engineer this from the luptitudes but it is quite convoluted since luptitudes do not differentiate between -ve and +ve fluxs.
|
|
| Figure 1 Comparison of K band LAS DR5 LasDetection magnitudes with out from VDFS List driven photometry using positions from LasDetection. [Top of page] |
|
|
| Figure 2 Extinctioncorr corrected magnitude comparison. [Top of page] |
|
|
| Figure 3 Extinctioncorr corrected magnitude comparison. [Top of page] |
