2 Description of the radio surveys

Figure 2: Limiting flux density plotted for all major radio surveys. Lines are of constant spectral indices of -1.3. Note that WENSS, NVSS and FIRST have flux density limits $\sim 100$ times deeper than previous surveys at comparable wavelengths

During the past years, several all-sky radio-surveys have become available (Table 1), which are 1-2 orders of magnitude more sensitive than previous surveys at similar frequencies (Fig. 2). The combination of these new surveys allows us to define for the first time a large sample of USS sources that covers the whole sky in both hemispheres. We list the main survey parameters in Table 1. In this section, we will briefly discuss the usefulness of these new radio surveys for the construction of USS samples.

2.1 WENSS

The Westerbork Northern Sky Survey (WENSS; Rengelink et al. 1997) at 325 MHz is the deepest low-frequency survey with a large sky coverage (3.14 sr). We used the WENSS to define the largest, and most complete USS sample to date, covering the entire sky North of declination $+29\hbox{$^\circ$ }$. We used version 1.0 of the main and polar WENSS catalogs. A small area is covered by both these catalogs; we selected only the sources from the main catalog in this overlapping area.

2.2 Texas

Figure 3: Fraction of sources with "+++'' flag in the Texas catalog (see text) as a function of Texas S365 flux density. Note that the selection of "+++'' sources excludes primarily sources with $S_{365} \lesssim 700$ mJy

The Texas survey, made with the Texas interferometer from 1974 to 1983 (Douglas et al. 1996), covers 9.63 steradians at a frequency of 365 MHz with a limiting flux density about ten times higher than that of the WENSS. The Texas interferometer's 3.5 km maximum baseline provides <1 $^{\prime \prime }$ positional accuracy, but its poor uv-coverage leads to irregular beamshapes and lobe-shifts, hampering accurate modeling of extended sources. A detailed discussion of these complications can be found in Douglas et al. (1996). To minimize these problems, we have selected only the 40.9% sources that are well modeled (listed with a "+++'' flag in the catalog). This selection excludes primarily $S_{365} \lesssim 700$ mJy sources (Fig. 3), but even at $S_{365} \gtrsim 700$ mJy, one out of three sources is excluded by this criterion. Douglas et al. (1996) have calculated the completeness above flux density S of the Texas catalog (defined as the fraction of sources with true flux density greater than S which appear in the catalog) by comparing the Texas with the MRC (Large et al. 1981) and a variety of other low-frequency catalogs. They found that the completeness varies with declination (because the survey was done in declination strips over a large time span), and an expected increase in completeness at higher flux densities. In Table 2, we reproduce their completeness table, extended with the values after the "+++'' selection.

 

 

Table 2: Completeness of the Texas survey

Limiting flux density	all sources	"+++'' sources
250 mJy	0.8	0.2
350 mJy	0.88	0.28
500 mJy	0.92	0.40
750 mJy	0.96	0.51
1 Jy	0.96	0.50

Figure 4: Ratio of integrated Texas over WENSS flux density against integrated WENSS flux density. Only Texas "+++'' sources are plotted. The horizontal line is the expected ratio 0.865 due to the 40 MHz difference in central frequency, and assuming a spectral index of $\alpha _{365}^{1400}=-0.88$. The curved line indicates a 150 mJy flux density limit in the Texas catalog. Note the increasing amount of overestimated Texas flux densities with decreasing flux density $S_{\rm WENSS} < 400$ mJy

To examine the reliability of the listed flux densities, and to check to what extent the "+++'' selection has removed the spurious sources from the catalog, we have correlated the Texas "+++'' sources with WENSS, NVSS and FIRST. In Fig. 4, we compare the Texas flux densities with those of the WENSS. At $S_{325} \gtrsim 500$ mJy, the ratio of the flux densities is closely distributed around 0.9. This ratio is what we expect due to the 40 MHz central frequency difference between the two surveys and assuming a spectral index $\alpha_{365}^{1400} = -0.879$ (the median of the Texas-NVSS spectral indices). At $S_{325} \lesssim 500$ mJy, the number of sources in the Texas catalog which are brighter than in WENSS catalog increases with decreasing flux density. This can be explained by the "up-scattering'' of sources near the flux limit of the Texas catalog (i.e. only sources intrinsically brighter than S₃₆₅ = 150 mJy will be detected, but no S₃₆₅ < 150 mJy sources with a large positive flux density measurement error). The result of this on a USS sample based on the Texas survey and correlated with a higher frequency survey (such as the NVSS), will be that with lower S₃₆₅, we will find more sources whose spectral indices appear steeper than they really are.

We also examined the dependence of the ratio Texas/WENSS flux density on angular size, determined from the FIRST survey (see Sect. 2.1.4). We found no significant residual variation of the flux density ratio at sizes between 5 $^{\prime \prime }$ and 2$^\prime$.

In Fig. 5b, we plot the density of NVSS sources around a Texas source (see also Sect. 2.2.5). The width of the over-density peak ($\sim 10$ $^{\prime \prime }$) is due to the positional inaccuracies in the Texas and NVSS catalogues. However, the very broad tail of sources between 20 $^{\prime \prime }$ and 110 $^{\prime \prime }$ and the secondary peak coinciding with the fringe separation at 73 $^{\prime \prime }$ indicates that the "+++'' selection did not remove all spurious sources from the catalog.

In summary, after the selection of "+++'' sources, the Texas catalog still contains <5% spurious sources (Douglas et al. 1996), probably due to residual lobe-shifted sources. Our comparison of the Texas flux densities with those of the WENSS survey shows that the differences are consistent with the errors quoted in the catalog. The selection of the Texas catalog with only "+++'' sources is thus >95% reliable, but only $\sim$40% complete.

2.3 NVSS

The NRAO VLA Sky Survey (NVSS; Condon et al. 1998) covers the 10.3 steradians north of -40 $\hbox{$^\circ$ }$ at 1.4 GHz, and reaches a 50 times lower limiting flux density than previous large area 1.4 GHz surveys. At the flux density levels we are using ( S₁₄₀₀>10 mJy), the catalog is virtually complete. Because the NVSS resolution is comparable to that of the WENSS and Texas surveys, and its sky coverage is large, we use the NVSS to determine the spectral indices in our USS samples based on the WENSS and Texas surveys. The final NVSS catalog was not yet completed at the time of our USS sample construction. For our final sample, presented in this paper, we use the 1998 January 19 version. This version still lacks data in a small number of regions of the sky (listed on the NVSS homepage). As a result, the sky coverage of the area listed in Table 3 is only 99.77%.

2.4 FIRST

The Faint Images of Radio Sky at Twenty centimeters (FIRST, Becker et al. 1995) survey is currently being made with the VLA in the B-array at 1.4 GHz, and has a limiting flux density three times deeper than the NVSS. We used the 1998 February 4 version of the catalog, covering 1.45 steradians. As noted by Becker et al. (1995), the photometry for extended sources in FIRST might be less reliable than that of the NVSS, due to the $9 \times$ higher resolution, which could underestimate large-scale diffuse radio emission. As the FIRST area is completely covered by the NVSS, we will consistently use NVSS flux densities for our spectral index calculation. The main advantages of FIRST over NVSS for our purposes are the much better positional accuracy (< 0 $.\!\!^{\prime\prime}$5) and the higher (5 $^{\prime \prime }$) resolution. This combination allows the identification of even the very faint (R > 20) optical counterparts of radio sources. Additionally, the fainter detection limit of the FIRST allows an extra check on the flux densities of compact sources.

2.5 MRC

The Molonglo Reference Catalog of radio sources (Large et al. 1981) at 408 MHz is presently the most sensitive low-frequency catalog with reasonable positional accuracy that covers the deep southern hemisphere, $\delta < -35\hbox{$^\circ$ }$. We will use this catalog in combination with the PMN survey (see below) to define the first USS sample at $\delta < -40\hbox{$^\circ$ }$.

2.6 PMN

The Parkes-MIT-NRAO (PMN) survey is a combination of 4 strips observed with the Parkes telescope at 4.85 GHz. The strips cover different parts of the sky, each with a slightly different limiting flux density. The regions are: southern ( $-87\hbox{$.\!\!^\circ$ }5 < \delta < -37\hbox{$^\circ$ }$, Wright et al. 1994), zenith ( $-37\hbox{$^\circ$ }< \delta < -29\hbox{$^\circ$ }$, Wright et al. 1996), tropical ( $-29\hbox{$^\circ$ }< \delta < -9\hbox{$.\!\!^\circ$ }5$, Griffith et al. 1994), and equatorial ( $-9\hbox{$.\!\!^\circ$ }5 < \delta < +10\hbox{$^\circ$ }$, Griffith et al. 1995). For our southern hemisphere sample, we have used the southern and zenith catalogs to find USS sources at $\delta < -30\hbox{$^\circ$ }$.