A 1420 MHz Catalog of Compact Sources in the Northern Galactic Plane

The CGPS DRAO image products are 512 × 512 images mosaicked from overlapping Synthesis Telescope fields. The images are available through the Canadian Astronomy Data Centre.⁶ The complete survey is comprised of 85 such mosaics. We constructed the radio source catalog from these images using an automated source finding algorithm, FINDSRC, developed as part of the DRAO export software package (Higgs et al. 1997). FINDSRC uses a wavelet filter convolution to identify compact source candidates in the image. This convolution effectively filters out emission structures larger than a few beam sizes. A second DRAO software program, FLUXFIT, reads in the candidates identified in the filtered image by FINDSRC and fits the signals in the original, unfiltered image with a two-dimensional Gaussian function plus a "twisted" plane fit to the local background level. To avoid the potential problem of nearby sources within the fitting region, if a nearby sources is detected, FINDSRC expands the area of the region to include that source and carries out a simultaneous multiple Gaussian fit to all sources in the region. The output of FINDSRC are fit parameters and errors for the peak and integrated flux density of the source above the local background, as well as the major and minor axes and the major-axis position angle of the best-fit Gaussian approximation to the source response function. For each source an RMS background level was calculated from the variance in the image around the twisted plane background fit. We retained a source in the catalog if the peak amplitude of the best-fit Gaussian is greater than five times the background RMS.

Following the automated source extraction, an annotation file marking the position of each detected source was overlaid on the mosaic images and each image was examined visually to identify spurious detections arising, for example, from knots on filaments of highly structured diffuse emission in the Galactic Plane. On a few occasions a source was missed by the automated detection algorithm. In this case Gaussian fits were run "manually." The final source list contains 72,758 sources down to a minimum Stokes I peak intensity of ∼1 mJy beam⁻¹.

Because the DRAO ST is an east–west aperture synthesis array, the major axis of the synthesized beam has position angle of zero (oriented north–south) and its dimension is a function of decl. The angular resolution of the ST observations is given by

$$\begin{eqnarray}&&{b}_{\mathrm{major}}\times {b}_{\mathrm{minor}}=49^{\prime\prime} \mathrm{cosec}\delta \times 49^{\prime\prime} ,\end{eqnarray} \tag{ 1 }$$

where b_major and b_minor are respectively in the decl. and R.A. directions. A Gaussian synthesized beam is deconvolved from the fitted two-dimensional Gaussian to estimate the observed angular size of each source. A source is considered unresolved in either the major or minor axis dimension if the deconvolved angular size is less than three times the 1σ error on the solid angle. Unresolved dimensions are listed in the catalog with zero dimension.

A measure of the polarized intensity for each source was obtained from intensity of the Stokes Q and U images at the Stokes I position of each source after removing a mean background level determined in a 25 × 25 pixel region around each source. A provisional polarization detection was flagged if either the Q or U signal was greater than four times the local RMS, σ_Q and σ_U, in the region surrounding the source. For provisional detections an estimate, ${\hat{p}}_{o}$, of the intrinsic polarized intensity, p_o, was calculated as

$$\begin{eqnarray}&&{\hat{p}}_{o}=\sqrt{{p}^{2}-{\sigma }_{{QU}}^{2}},\end{eqnarray} \tag{ 2 }$$

which is a good estimator of the polarized intensity when p/σ > 4 (Simmons & Stewart 1985). A source was retained as a polarized detection if in addition ${\hat{p}}_{o}\gt {\epsilon }_{p}{S}_{p}$, where _p is the instrumental polarization error, taken to be 0.3% (Taylor et al. 2003), and S_p is the peak Stokes I intensity. The error on the polarized intensity is given by

$$\begin{eqnarray}&&{\sigma }_{{\hat{p}}_{o}}=\sqrt{2{\sigma }_{{QU}}^{2}+{({\epsilon }_{p}{S}_{p})}^{2}}.\end{eqnarray} \tag{ 3 }$$

We detect polarized signals from 12,368 sources. The polarized intensity should be treated with caution for resolved sources. The polarized intensity structure for extended sources is not necessarily the same as that of the total intensity. In such cases, the polarized intensity at the position of the total intensity peak may not represent the true peak of Q and U. Those interested in the polarized emission from extended sources are advised to examine the CGPS images directly.

Table 2 lists the first 30 entries of the CGPS 1420 MHz catalog. The columns contain: (1) the source name CGPS JHHMMSS+DDMMSS, where HHMMSS are the hours and minutes and seconds of time for R.A. and DDMMSS are the degrees, minutes, and seconds of arc of decl. (both in J2000); (2) the J2000 R.A. (degrees) and error (seconds); (3) the J2000 decl. (degrees) and error (arcseconds); (4) the integrated Stokes I flux density and error (mJy); (5) the peak Stokes I intensity and error (mJy/beam); (6) the average rms around the position of the source in the Q and U images (mJy); (7) the bias corrected polarized intensity in the case of detected polarized emission (mJy/beam); (8) the polarization position angle (degrees); (9 and 10) the deconvolved dimensions of the major and minor axes of the Stokes I emission with errors (arcseconds); and (11) the position angle of the source major axis (degrees) with respect to north (J2000). The deconvolved source dimensions are listed as zero if the source is classed as unresolved. The beam dimensions at the location of the source can be calculated using Equation (1). All errors are 1σ.

The complete version of Table 2 is available in machine-readable format through the CGPS data repository at the Canadian Astronomy Data Centre (DOI 10.11570/17.0003).

Our survey is comparable in resolution and sensitivity to the NRAO VLA Sky Survey (Condon et al. 1998); the flux density scale and source positions of the CGPS at 1420 MHz were tied to the NVSS. Figure 1 shows the distribution of source flux densities for the CGPS and for the NVSS over the region of the CGPS survey. These curves give an indication of the relative depths of the catalogs. The NVSS flux distribution peaks at 3 mJy and drops to zero by 2 mJy. The CGPS source numbers continue to increase to 2 mJy, with 20% of the sources having flux density below 2 mJy. This is consistent with the relative noise levels of ∼0.45 mJy for the NVSS and ∼0.24 mJy for the CGPS. Moreover, the CGPS, with an almost filled u − v plane coverage, has a much more uniform image background noise distribution, with very little contamination from sidelobes of the synthesized beam. Given the strong and highly structured diffuse emission that is ubiquitous in the plane of the Galaxy, the depth of the survey will vary significantly with location in the plane, but should be complete to about 2 mJy in quiet regions.

Zoom In Zoom Out Reset image size

From the 72,758 sources in the CGPS catalog we identified 52,876 that have counterparts in the NVSS within 60'' of the CGPS position. Figure 2 shows a plot of the position differences between the CGPS and NVSS catalog sources for 34,449 sources with CGPS flux density above 5 mJy. The individual CGPS images are registered in position and flux density against the NVSS images using a small number of strong sources within each field (Taylor et al. 2003). Using the very large number of sources in the catalog comparison, we measure an inverse signal-to-noise weighted mean difference between the two catalogs of Δα = 0.011 ± 0003 and Δδ = −0.024 ± 0005. The difference is small, but significant relative to the error. To make the CGPS catalog positions formally consistent on average with the NVSS to within 0005, we have corrected for this small difference for the source coordinates listed in Table 2.

Zoom In Zoom Out Reset image size

To compare the CGPS and NVSS flux densities we calculate for all sources detected in both surveys the fractional difference in flux density, F, defined as

$$\begin{eqnarray}&&F=2\left(\displaystyle \frac{{S}_{\mathrm{CGPS}}-{S}_{\mathrm{NVSS}}}{{S}_{\mathrm{CGPS}}+{S}_{\mathrm{NVSS}}}\right).\end{eqnarray} \tag{ 4 }$$

The result is plotted in Figure 3 for sources detected with a position difference between the two surveys of less than 10'' and CGPS major-axis diameter less than 20''. The solid line in the figure, which delimits the upper bound of the data points, shows the selection effect of a 2 mJy detection threshold in the NVSS, which is approximately the NVSS minimum detectable flux density. The absence of points above this line is due to noise variations in the NVSS data reducing the NVSS signals of CGPS sources below the NVSS detection limit. We used 370 unresolved sources with S_CGPS > 40 mJy and position differences less than 10'' to derive a median fractional difference of $\langle F\rangle$ = −0.0289 ± 0.005. The raw CGPS flux densities are in the median 2.9% lower than the NVSS catalog flux densities. We have corrected for this in the CGPS catalog. The CGPS and NVSS flux density scales are thus aligned with an error of 0.5%.

Zoom In Zoom Out Reset image size

Figure 4 shows the fractional difference in peak polarized flux density, F_p, between the CGPS and NVSS for polarized sources in the CGPS. The data show that for the CGPS, polarized intensities are on average substantially larger than the NVSS. This arises from the effect on the NVSS polarized intensities of the high levels of Faraday Rotation from propagation through the magneto-ionic medium of the Galaxy at the low Galactic latitudes of the CGPS.

Zoom In Zoom Out Reset image size

As noted by Condon et al. (1998), the widely separated dual-band structure of the NVSS survey produced significant depolarization in the band-average polarized intensities for Rotation Measure (RM) magnitudes larger than about 100 rad m⁻¹. Over the area of the CGPS, RM values of this order and larger are common (Brown et al. 2003). The depolarization effect is illustrated by the data points in Figure 5, which show, as a function of source RM, the mean ratio of NVSS to CGPS peak polarized intensity for CGPS polarized sources having ${p}_{\mathrm{CGPS}}\gt 10$ mJy. RMs were derived from the two bands of the NVSS data (Taylor et al. 2009). The ratio recovers the theoretical depolarization curve of the NVSS band structure (shown by the dashed line in the figure). The narrower band CGPS data suffer less than 10% depolarization up to Rotation Measures of ∼1000 rad m⁻². The CGPS polarization data thus provide a low-latitude complement to the NVSS polarization catalog.

Zoom In Zoom Out Reset image size

The CGPS observations were taken over the period 1995–2009, while the NVSS data were taken between 1993 and 1996. The similar resolutions and sensitivities of the two surveys allow for the identification of sources in the plane of the Galaxy that have variable flux density over time intervals of a few to several years to sensitivities of a few mJy by comparison of the catalog data.

Several candidate variable sources are visible in Figure 3 as outliers to the noise-broadened distribution of F. However, some of these are spurious, resulting from differences between the NVSS and CGPS algorithms for identifying and fitting sources as either close double or extended. In some cases, objects fit as single extended sources in the NVSS are fit as multiples in the CGPS, and vice-versa. Extended single component sources that are split into multiple components in the other catalog will have very different flux densities and each component will be offset in position from the single fit. To remove this effect from the variability analysis, we have restricted the search for variability to unresolved CGPS sources that are isolated from their nearest CGPS neighbor by at least 90'', and whose NVSS counterpart also classed as unresolved and agrees in position within 5''. We also restrict the analysis to sources with S_CGPS > 4.0 mJy. There are 10,897 CGPS sources that meet these criteria.

As shown in Equation (4), the quantity F measures the change in flux density between CGPS and NVSS as a fraction of the mean flux density measured between the two surveys. A change in flux density by a factor of two results in F = ±0.67. The significance of the change is quantified by a comparison to the uncertainty on F. We calculate this uncertainty, σ_F, by propagating the errors on the individual flux densities ${S}_{\mathrm{CGPS}}$ and S_NVSS. We then identify as variables sources those for which

$$\begin{eqnarray}&&V=\displaystyle \frac{F-{F}_{\mathrm{MED}}}{{\sigma }_{F}}\gt 4,\end{eqnarray} \tag{ 5 }$$

where F_MED is the median value of F for population. In the absence of variable sources, if σ_F is an accurate measure of the error, the variance in F will be determined by noise alone and the distribution of V over the ensemble of sources will be Gaussian with unit standard deviation. Figure 6 shows the distribution of V for the CGPS variable search sample. The central peak of the distribution is well fit by the Gaussian. Evidence for a population of variable sources appears as excess over the Gaussian distribution in the wings, beginning at about $| V| \sim 2.$ However, we take a conservative approach and class as variable sources those for which $| V| \gt 4$. Of the 10,897 sources examined, 146 (1.3%) satisfy this criterion. For a Gaussian distribution we expect less than 1 object at $| V| \gt 4$ by chance.

Zoom In Zoom Out Reset image size

Figure 6 shows an asymmetry in the tail of the distribution, with more objects with high positive V. There are 111 sources with V > 4, compared to 35 with V < −4. This is a result of the threshold in S_CGPS, which results in a bias against large negative values of F. A source with S_CGPS > 4 mJy and F < −1.0 must have S_NVSS > 12 mJy, which represents a small fraction of the NVSS sources. Conversely, F > +1.0 requires S_NVSS > 1.25 mJy, sampling the entire NVSS catalog.

While not a complete statistical study for sources of variable emission in the Galactic plane at 1420 MHz, this simple comparison does identify a small sample of isolated, unresolved radio sources whose flux density exhibits a large fractional change between the two catalogs. The resulting list of variable source candidates is given in Table 3.

In addition to searching for variable sources that appear in both the CGPS and NVSS catalogs, we also looked for candidate transient objects that appear in the CGPS but have no counterpart in the NVSS catalog. There are 19,911 sources in the CGPS without NVSS counterparts. The vast majority of these are faint objects below ∼3 mJy that are missed by the NVSS due to the high level of NVSS imaging artifacts arising from the lack of low spatial frequency data in the presence of the strong extended emission at low Galactic latitudes. This effect is illustrated in Figure 7, where we compare CGPS and NVSS images for one of the CGPS 512 × 512 mosaics. These images include Cas A and artifacts from this strong source are visible in both. The diffuse structure away from Cas A visible in the CGPS is not well represented in the NVSS image and results in lower level artifacts in the NVSS that extend throughout the image. The CGPS fully samples the structure of the diffuse emission resulting in much greater intensity dynamic range.

Zoom In Zoom Out Reset image size

To avoid missing sources due to the effect described above, we limited the analysis for candidate transients to unresolved sources stronger than 10 mJy in the CGPS. The resulting 71 objects were each examined in the NVSS images. In most cases, the missing detections in the NVSS could be attributed to strong image artifacts obscuring the source. The final list of 13 transient candidates includes sources that should have been visible in the NVSS if present at the CGPS flux levels. These sources are listed in Table 4. Figure 8 shows an example of one transient candidate, CGPS J234153+610329, that is detected at 12.6 mJy in the CGPS but is not visible in the NVSS down to the NVSS image RMS at the source location of 0.58 mJy.

Zoom In Zoom Out Reset image size