All users should download the data release paper. This paper describes the design of the survey, the steps used to validate it, it's completeness and reliability, and a discussion of caveats to the survey.

There is a systematic error in the faint fluxes reported in the SWIRE version 1 release. Although this flux distortion is mentioned in Section 7.2.1 of the data release paper, we feel that it is sufficiently significant as to warrant further discussion here. During validation a point-by-point photometric comparison was made between catalogs extracted from the SWIRE survey data in ELAIS N1 and from a much deeper pointing the same region used to validate the GOODS survey strategy and released as part of the extragalactic First Look Survey. This comparison is shown below for each of the four IRAC channels. Note particularly for the 4.5 micron data the SWIRE fluxes are underestimated by on order of 20 percent relative to the FLS data, and at relatively high flux levels (20 microJy).

It is now understood that this error originated with the way the MOPEX software was used to mosaic the IRAC data. The default single frame outlier rejection was used (in conjunction with MOPEX's other outlier rejection schemes) and this choice was later found to be the source of the problematic fluxes. Data reprocessed without any outlier rejection lacked this problem, and the specific issue was isolated by further investigation and comparison to the FLS pipeline. Since the error occured during coaddition, the error lies in the images themselves, and not in the photometric extraction. Although the FLS data were also coadded using MOPEX, the depth of coverage of the FLS data was noticeably higher, minimizing this error. The Version 2 release of the SWIRE data will have this effect corrected.

For clarity we show here two important ways that this error will manifest itself. The first is a distortion in the number counts at faint flux levels. Due to the flux underestimation, objects are shifted from one bin down into lower flux bins. This results in a notch or shoulder in the number counts.

4.5 micron number counts, showing the distortion in the counts due to objects being shifted to lower flux bins. Shown are the released survey in N1 vs. counts in XMM, which have been processed with the new software.

The second important effect is on the observed colors. The following figure presents color-magnitude diagrams for data processed in the same manner as in the version 1 data release, and using new processing with the processing error fixed.

Distortion in 3.6/4.5 colors processed as in Version 1 (left) and with corrected processing (right).