2 Observations representing CHVCs over the entire sky

  
2.1 Identification criteria

A full description of the cloud extraction algorithm is given in Paper I; the most salient aspects of the algorithm are the following:

• All pixels in the HIPASS and LDS material above the $1.5 \sigma$level (as appropriate to the particular survey) were assigned to a local intensity maximum; each pixel was assigned to the same maximum as its brightest neighboring pixel. If the local maximum was brighter than 3 $\sigma$, then the local maximum and the pixels which were assigned to it were considered to constitute a cloudlet;
• Adjacent cloudlets were merged into clouds if the brightest enclosing contour for the two cloudlets either exceeds 40% of the brighter peak or if the brightness exceeds 10$\sigma$;
• Those merged cloudlets for which the peak temperature exceeds 5 $\sigma$ were deemed clouds, and were entered in the catalog pending further consideration of their deviation velocity, as described below. The value of the noise that was used for this selection was determined locally, whereas for the other steps a preset noise value was used.

To minimize the influence of noise on the peak detection, the data were first smoothed along both the spatial and spectral axes. The rms fluctation level noted above refers to that within the smoothed data. Once the relevant pixels were assigned to a cloud, unsmoothed data were used to determine the cloud properties.

The Paper I search algorithm led to the identification of sub-structure within extended anomalous-velocity cloud complexes as well as to the identification of sharply bounded, isolated sources. Because of the importance to the present analysis of selecting only isolated objects, we comment on the determination of the degree of isolation. In order to determine the degree of isolation of the clouds that were found by the algorithm, velocity-integrated images were constructed of $10^\circ$ by $10^\circ$ fields centered on each general catalog entry. The range of integration in these moment maps extended over the velocities of all of the pixels that were assigned to a particular cloud. The CHVC classification then depended on the column density distribution at the lowest significant contour level (about 3$\sigma$) of $1.5\times10^{18}$ cm^-2. We demanded that this contour satisfy the following criteria: (1) that it be closed, with its greatest radial extent less than the $10\hbox{$^\circ$ }$ by $10\hbox{$^\circ$ }$ image size; and (2) that it not be elongated in the direction of any nearby extended emission. Since some subjectivity was involved in this assessment, two of the authors (VdH and RB) each independently carried out a complete classification of all sources in the HIPASS and LDS catalogs. Identical classification was given to about 95% of the sample, and consensus was reached on the remaining 5% after re-examination.

A slightly different criterion for isolation was employed by Putman et al. (2002) in their analysis of the HIPASS sample of HVCs. Rather than employing the column density contour at a fixed minimum value to make this assessment, they employed the contour at 25% of the peak $N_{\rm HI}$ for each object. Since the majority of detected objects are relatively faint, with a peak column density near 12$\sigma$, the two criteria are nearly identical for most objects. Only for the brightest $\sim$10% of sources might the resulting classifications differ. We have reclassified the entire HIPASS HVC catalog with the absolute $N_{\rm HI}$ criteria above, and have determined identical classifications for 1800 of the 1997 objects listed. Given that the agreement in classification is better than 90%, we have chosen to simply employ the Putman et al. classifications in the current study. In this way the analysis presented here can be reproduced from these published sources.

High-velocity clouds are recognized as such by virtue of their anomalous velocities. Although essentially any physical model of these objects would predict that they also occur at the modest velocities characteristic of the conventional gaseous disk of the Milky Way, at such velocities the objects would not satisfy our criterion for isolation. In our analysis, only anomalous-velocity objects with a deviation velocity greater than  $70\rm\;km\;s^{-1}$ were considered. As defined by Wakker (1990), the deviation velocity is the smallest difference between the velocity of the cloud and any Galactic velocity, measured in the Local Standard of Rest reference frame, allowed by a conventional kinematic model in the same direction. The kinematic and spatial properties of the conventional Galactic HI were described by a thin gaseous disk whose properties of volume density, vertical scale-height, kinetic temperature, and velocity dispersion, remain constant within the solar radius; at larger galactocentric distances the gaseous disk flares and warps, as described in Paper I, following Voskes & Burton (1999). The gas exhibits circular rotation with a flat rotation curve constant at $220\rm\;km\;s^{-1}$. Synthetic H  I spectra were calculated for this model Galaxy, and then deviation velocities were measured from the extreme-velocity pixels in these spectra for which the intensity exceeded $0.5\rm\;K$. Selection against objects at $V_{\rm dev}< 70$ km s^-1 in the LSR frame introduces systematic effects, as discussed in the following subsection.

Although compactness was not explicitly demanded of these isolated objects, the 67 CHVCs found in the LDS survey and the 179 CHVCs found in the HIPASS data have a small median angular size, amounting to less than $1^\circ$ FWHM.

  
2.2 Selection effects and completeness

The CHVC samples used in this analysis were not extracted from a single, homogeneous set of data, nor are they free of selection effects, nor are they complete. We discuss below how we attempt to recognize and, insofar as possible, to account for some inevitable limitations.

  
2.2.1 Systematic consequences of obscuration by H  I in the Milky Way

The inevitable obscuration that follows from our perspective immersed in the gaseous disk of the Milky Way, and which motivates the use of the deviation velocity, will discriminate against some CHVC detections. We may extend an analogy of the optical Zone of Avoidance to the 21-cm regime. The optical Zone of Avoidance refers to extinction of light by dust in the Milky Way, and thus traces a band with irregular borders but roughly defined by $\vert b\vert<5^\circ$; kinematics are irrelevant in the optical case, since the absorption is broad-band. The HI searches for galaxies in the optical Zone of Avoidance carried out by, among others, Henning et al. (1998) in the north, and by Henning et al. (2000) and Juraszek (2000) in the south, were confined to $\vert b\vert<5\hbox{$^\circ$ }$.

Figure 4: Indication of the degree of completeness of the CHVC catalog extracted from the LDS by de Heij et al. (2002). The solid curve shows the fraction of external galaxies with the indicated peak H  I brightness temperatures that were shown by Hartmann & Burton (1997) to have been detected in the LDS. Dashed lines show the expected completeness for sensitivities that are 25% better or worse, respectively, than that of the LDS. The histogram indicates the fraction of HIPASS sources from the catalog of Putman et al. (2002) within each temperature range that are also found in the LDS, in the declination zone $-30\hbox {$^\circ $ }<\delta <0\hbox {$^\circ $ }$ where the two surveys overlap.

In the 21-cm regime, extinction due to high-optical-depth foreground H  I is largely negligible, but confusion due to line blending occurs at all latitudes. The analogous H  I zone refers to a certain range in velocity, of varying width depending on l and on b, but present to some extent everywhere: near zero LSR velocity, the "H  I Zone of Avoidance" covers the entire sky. The nearby LSB galaxy Cep I was discovered during CHVC work (Burton et al. 1999); although it is at a relatively substantial latitude, $b=8\hbox{$.\!\!^\circ$ }0$, its velocity of $V_{\rm hel}= 58$ km s^-1 locates it within the H  I  obscuration zone. Because of the strong dependence on velocities measured with respect to the Local Standard of Rest, the zone of obscuration is distorted upon transformation to a different kinematic reference frame. (We note that the Magellanic Stream and the HVC complexes plausibly also discriminate against CHVC detections, but because these extended features are smaller in scale and more confined in velocity than the Galaxy, we do not consider them further here.)

The relationship between the different velocity reference systems used to characterize the CHVC kinematics is given by the equations below.

$$v_{\rm LSR}=v_{\rm HEL}+9\cos(l)\cos(b)+12\sin(l)\cos(b)+7\sin(b)$$ (1)

$$v_{\rm GSR}=v_{\rm LSR}+0\cos(l)\cos(b)+220\sin(l)\cos(b)+0\sin(b)$$ (2)

$$v_{\rm LGSR}=v_{\rm GSR}-62\cos(l)\cos(b)+40\sin(l)\cos(b)-35\sin(b).$$ (3)

Note that a typographical error is present (the sign of the coefficient of $\sin(b)$) in the version of Eq. (1) that is published in Braun & Burton (1999).

The influence of the obscuration by the modeled Galaxy is illustrated by Fig. 1, where the integral $\int\exp(-(V - \mu)^2 / 2\sigma^2)$dV is plotted. The range of integration extends over all velocities which deviate more than  $70\rm\;km\;s^{-1}$ from any Local Standard of Rest (LSR) velocity allowed by the Galactic model described above; $\mu$ is the average velocity and $\sigma$ is the standard deviation of the test clouds. The panels in Fig. 1 show the fraction of a population of clouds, homogeneously distributed on the sky and with a Gaussian velocity distribution relative to a particular reference frame, that are not obscured by virtue of being coincident with H  I emission from the Milky Way. The upper panel of the figure represents a model in which the Gaussian velocity distribution is with respect to the Local Standard of Rest frame, with an average velocity of $-50\rm\;km\;s^{-1}$ and dispersion of $240\rm\;km\;s^{-1}$, in rough agreement with the measured CHVC values in this frame. In this case the obscuration is simply proportional to the velocity width of the obscuring emission. The obscuration at high latitudes is quite uniform since the infalling population is always displaced from $V_{\rm LSR}$ = 0 km s^-1, where the obscuring gas resides. The middle panel presents a model wherein the Gaussian velocity distribution is with respect to the Galactic Standard of Rest (GSR) frame with an average velocity of $-50\rm\;km\;s^{-1}$ and dispersion of $110\rm\;km\;s^{-1}$, in rough agreement with the CHVC values in this frame. The low-latitude obscuration is similar to that in the LSR model, although more strongly modulated since the velocity dispersion is smaller. The high-latitude obscuration is quite strongly modulated since the infall velocity in the GSR frame overlaps with $V_{\rm LSR}$ = 0 km s^-1 in the plane approximately perpendicular to the direction of rotation, $(l,b)=(90^\circ,0^\circ)$. Broad apparent maxima in unobscured object density are centered near $l=90\hbox{$^\circ$ }$ and $l=270\hbox{$^\circ$ }$. The lower panel presents a model in which the Gaussian velocity distribution is with respect to the Local Group Standard of Rest (LGSR), again using an average velocity of $-50\rm\;km\;s^{-1}$ and dispersion of $110\rm\;km\;s^{-1}$. The pattern of obscuration is very similar to that of the GSR case, although the maxima in unobscured object density are slightly shifted with respect to $b=0\hbox{$^\circ$ }$. These results indicate that caution must be exercised in interpreting apparent spatial concentrations of detected objects without properly accounting for the distortions introduced by the H  I Zone of Avoidance.

We have also considered how the measured statistics of a distribution, namely the mean velocity and dispersion, are influenced by the non-completeness caused by obscuration. Figure 2 shows the distribution of the errors in the average velocity and dispersions for 1000 simulations, each involving 200 test clouds; one set of simulations was run with the GSR as the natural reference frame, and a second set was run with the LGSR as the natural frame. After removing the test clouds that have an LSR deviation velocity less than  $70\rm\;km\;s^{-1}$, the velocity dispersion of the simulated ensemble was measured for both the GSR and the LGSR velocity systems, and compared with what would have been determined if there had been no obscuration by the Galaxy.

The upper left-hand panel in Fig. 2 refers to test objects with a Gaussian distribution in $V_{\rm GSR}$ with a dispersion of  $115\rm\;km\;s^{-1}$ and average of  $-50\rm\;km\;s^{-1}$. The measured dispersion exceeds the true one by  $9\rm\;km\;s^{-1}$, whereas a more negative average velocity is inferred by  $12\rm\;km\;s^{-1}$. The lower left-hand panel is based upon test samples with a Gaussian distribution in $V_{\rm LGSR}$ with a dispersion of  $105\rm\;km\;s^{-1}$ and an average of  $-55\rm\;km\;s^{-1}$. The differences between the measured and true dispersion and average velocity, of $6\rm\;km\;s^{-1}$ and $-5\rm\;km\;s^{-1}$, respectively, are smaller than for the GSR system. From the 200 clouds which were in the input ensemble, an average of 80 were removed because of obscuration in the GSR model and only 60 in the LGSR model, indicating that the statistical properties of the LGSR model are somewhat better preserved in this case. The particular population attributes chosen above for the GSR and LGSR systems were chosen to match the observed parameters in these systems, as shown below in Sect. 3.

Figure 5:  Spatial deployment of CHVCs over the sky.   Upper panel:  distribution of the cataloged CHVCs, with triangles representing the LDS sample of de Heij et al. (2002) at $\delta >0\hbox {$^\circ $ }$ and diamonds representing the HIPASS sample of Putman et al. (2002) at southern declinations. Filled circles correspond to the Local Group galaxies listed by Mateo (1998). Red symbols indicate positive LSR velocities and black symbols negative velocities. The background grey-scale shows H  I column depths from an integration of observed temperatures over velocities ranging from $V_{\rm LSR} = -450\rm\;km\;s^{-1}$ to $+400\rm\;km\;s^{-1}$, but excluding all gas with $V_{\rm DEV} < 70\rm\;km\;s^{-1}$.   Lower panel:  smoothed relative density field of the CHVCs, accounting for the different observational parameters of the LDS and HIPASS catalogs. The cataloged CHVCs are each represented by a Gaussian with a true-angle dispersion of $20^\circ$; the total flux of the Gaussian is set to unity for the LDS objects and to the likelihood of observing such an object in an LDS-like survey for the HIPASS sources. The grey-scale is calibrated in object number per steradian. Contours are drawn at relative densities of -60%, -30%, 0% (in white) and 30%, 60%, 90% (in black). A significant over-density of CHVCs in the southern hemisphere remains after accounting for the different observational parameters. (This figure is available in color in electronic form.)

Another question that can be addressed with these simulations is whether it might be possible to distinguish between a GSR and an LGSR CHVC population based on a significant difference in the statistical properties. We assessed this by taking 500 populations of 200 objects in both the GSR and LGSR frames, each with a dispersion of 110 km s^-1 and an average velocity of -50 km s^-1. Each of these 1000 populations was analyzed in both the GSR and LGSR frames, both before and after decimation by obscuration. The results are shown in the right-hand panels of Fig. 2 for differences in velocity dispersion (relative to the GSR versus LGSR frames) and mean velocity, respectively. The measured differences in velocity dispersion and mean velocity of our CHVC sample (from Sect. 3) are plotted in these panels as dashed lines. The model results for mean velocity differences form a continuous cloud, for which it is impossible to distinguish between the actual reference frame of the model population. The model results for velocity dispersion differences, on the other hand, are separated into two distinct clouds. The velocity dispersion of each model population is minimized in its own reference frame with a variance of only a few km s^-1, while the dispersions within the GSR and LGSR frames are separated by about 20 km s^-1, both before and after obscuration. The measured difference in velocity dispersion of the CHVC sample relative to the GSR and LGSR frames, of 16 km s^-1, is more consistent with an LGSR reference frame.

Figure 6: Kinematic deployment of CHVCs identified in the LDS (triangles) and in the HIPASS (diamonds) data, plotted against Galactic longitude for three different kinematic reference frames, namely the LSR (upper), the GSR (middle), and the LGSR (lower panel). The filled circles show the kinematic deployment with longitude of the Local Group galaxies listed by Mateo (1998). The mean velocities and the dispersions in velocity of the CHVCs and Local Group galaxies are listed in Table 2 for the three reference frames.

Figure 7: Kinematic deployment of CHVCs identified in the LDS (triangles) and in the HIPASS (diamonds) data, plotted against Galactic latitude in the three different kinematic reference frames, as in Fig. 6. The filled circles show the kinematic deployment with latitude of the Local Group galaxies listed by Mateo (1998). The mean velocities and the dispersions in velocity of the CHVCs and Local Group galaxies are listed in Table 2 for the three reference frames.

Figure 8: Smoothed distributions of velocity and velocity dispersion of the CHVC ensemble. The panels on the left show the average velocity in the LSR (upper), GSR (middle), and LGSR (lower) reference frames, respectively. The panels on the right show the velocity dispersions, similarly arranged. Individual CHVCs in the ensemble were convolved with a Gaussian of true-angle dispersion of $20^\circ$. White contours for the velocity and dispersion fields are at drawn at values of $0,\;50,\;\ldots\ \rm km\;s^{-1}$; black ones are drawn at $-50,\;-100,\;\ldots\ \rm km\;s^{-1}$. These smoothed representations of the observed situation can be compared with similarly sampled and smoothed representations of simulations, as described in the text.

  
2.2.2 Consequences of the differing observational parameters of the LDS and HIPASS

Because the LDS and the HIPASS data do not measure the sky with the same limiting sensitivities, angular resolutions, velocity resolutions, or velocity coverages, the northern population of CHVCs will be differently sampled than the southern one. In particular, the maximum depth of the two samples will be different since the surveys have different limiting fluxes. We describe below how we identify, and compensate for, the differing properties of the two catalogs; we also describe how we sample the simulations using the selection criteria corresponding to the observations. A detailed comparison of objects detected in the two surveys is made in Paper I.

Figure 9: Summary of the observed spatial, kinematic, angular size, and flux properties of the CHVC ensemble. The three panels arranged across the top of the figure show sky projections, as follows: left: smoothed density field of the CHVC population. A Gaussian with a dispersion of $20^\circ$ (true angle) was drawn at the location of each CHVC; the volume of the Gaussian is unity for both LDS and HIPASS sources - thus in this case the observations are shown directly, i.e. HIPASS sources are not weighted by the likelihood with which they would be observed in a LDS-like survey. Middle: smoothed velocity field of the population in the Galactic Standard of Rest frame. Right: smoothed velocity dispersion field. The grey-scale bar for the left-hand panel is labeled in units of CHVC per steradian; the other two bars are labeled in units of km s-1. Contours are drawn at relative densities of -60%, -30%, 0% (in white) and 30%, 60%, 90% (in black). White contours for the velocity and dispersion fields are at drawn at values of $0,\;50,\;\ldots\ \rm km\;s^{-1}$; black ones are drawn at $-50,\;-100,\;\ldots\ \rm km\;s^{-1}$. The two panels in the middle row of the figure show the kinematic distribution of the observed CHVC ensemble, representing $V_{\rm GSR}$ plotted against land b, as indicated. Delta functions at the observed coordinates were convolved with a Gaussian with an angular dispersion of $20^\circ$ and velocity dispersion of $20\rm\;km\;s^{-1}$. The two lower panels show, respectively, the observed peak H  I column density distribution of the CHVC population and the observed angular size distribution.

The LDS covered the sky north of declination $-30^\circ$ (the actual declination cut-off varied between -32 and $-28^\circ$); the angular resolution of the 25-m Dwingeloo telescope was $36^\prime$. The effective velocity coverage of the LDS extends over LSR velocities from $-450\rm\;km\;s^{-1}$ to  $+400\rm\;km\;s^{-1}$, resolved into channels $1.03\rm\;km\;s^{-1}$ wide. The formal rms sensitivity is 0.07 K per 1.0 km s^-1 channel. Stray radiation has been removed as described by Hartmann et al. (1996). Due to the presence of radio frequency interference, it was important that the reality of all CHVC candidates that were identified in the LDS be independently confirmed. Although interference in the LDS often had the shape of extremely narrow-band signals that are easily recognized as artificial, some types of interference were indistinguishable from naturally occurring features. The reality of the CHVC candidates was either confirmed by the identification of the candidates with objects in independent published material, or by new observations made with the Westerbork Synthesis Radio Telescope, operating as a collection of 14 single dishes.

Figure 10:  Variation of heliocentric velocity versus the cosine of the angular distance between the solar apex and the direction of the object; CHVCs from the de Heij et al. (2002) LDS compilation at $\delta >0\hbox {$^\circ $ }$ are plotted as triangles; those from the Putman et al. (2002) HIPASS compilation, as diamonds. The CHVCs with $b~<~-65^\circ$ are plotted in red. Local Group galaxies, from the review of Mateo (1998), are indicated by filled circles. The solid line represents the solar motion of $V_\odot = 316\rm\;km\;s^{-1}$ towards $l=93^\circ,\;b=-4^\circ$ as determined by Karachentsev & Makarov (1996). Dashed lines give the $1\sigma (V)$ envelope ( $\pm60\rm\;km\;s^{-1}$, following Sandage 1986) encompassing most galaxies firmly established as members of the Local Group. (This figure is available in color in electronic form.)

The HIPASS program covered the sky south of declination $+2^\circ$. The survey has been reduced in such a way that emission which extends over more than $2^\circ$ was filtered out. To recover a larger fraction of the extended emission, the part of the survey which covers LSR velocities ranging from  $-700\rm\;km\;s^{-1}$to  $+500\rm\;km\;s^{-1}$ was re-reduced using the MINMED5 method described by Putman (2000), before production of the Putman et al. (2002) catalog. The HIPASS data were gridded with lattice points separated by $4^\prime$ with an angular resolution of $15\hbox{$.\mkern-4mu^\prime$ }5$. The HIPASS velocity resolution after Hanning smoothing is  $26.4\rm\;km\;s^{-1}$, thus substantially coarser than the 1.03 km s^-1 of the LDS. The HIPASS sensitivity for such a velocity resolution is 10 mK for unresolved sources. Because the observing procedure involved measuring each line of sight five times in order to reach the full sensitivity, all HIPASS sources have effectively been confirmed after median gridding.

Figure 3 shows that the LDS and the HIPASS reflect differing measures of the CHVC properties, because of their differing observational properties. The panel in the upper left of this figure contrasts the observed velocity widths of the LDS and HIPASS samples. The velocity FWHM measured in the LDS ranges from about 20 km s^-1 to some 40 km s^-1, with a median of about 25 km s^-1. Only for a few sources were values as low as 5 km s^-1 measured. The relatively high median FWHM likely indicates that most of the observed H  I in the CHVCs is in the form of warm neutral medium. High-resolution observations of a sample of ten CHVCs made with the $3\hbox{$.\mkern-4mu^\prime$ }5$resolution afforded by the Arecibo telescope (Burton et al. 2001) showed warm halos to be a common property of these objects. On the other hand, the median HIPASS velocity width is about 35 km s^-1 FWHM. We can demonstrate that the two observed FWHM distributions are consistent with the same object population by convolving the LDS distribution with the HIPASS velocity resolution. The resulting distribution agrees well with that measured in the HIPASS.

The panel in the upper right of Fig. 3 shows histograms of the angular sizes of the cataloged CHVCs, determined from velocity integrated images of each cloud. A contour was drawn at the intensity of half the peak column density of the cloud. After fitting an ellipse to this contour, the size of the cloud was measured as the average of the minor and major axes. It is clear from these distributions that many of the CHVCs are resolved by HIPASS, but that this is rarely the case for the LDS. Some CHVCs in the LDS catalog were only detected in a single spectrum - giving the peak in the histogram at  $0\hbox{$.\!\!^\circ$ }4$, which is an artifact of the sub-Nyquist LDS sky sampling. After convolving the HIPASS distribution with the LDS beam a more similar distribution of sizes is found, although there remains a small excess of relatively large objects in the north.

The panel on the lower left of Fig. 3 shows the flux distribution for the CHVCs detected by HIPASS and the LDS, respectively. An excess of faint sources is present in the HIPASS sample, even after compensation for the lower LDS sensitivity (as outlined below). Conversely, the LDS may have a small excess of bright objects. If semi-isolated objects are considered (i.e. the :HVC and ?HVC categories discussed by Putman et al. 2002 and de Heij et al. 2002) as in the panel on the lower right of Fig. 3, these differences remain, with the adjusted HIPASS sample showing an excess of faint sources in the south and the LDS sample showing a small excess of brighter sources in the north.

  
2.2.3 Completeness and uniformity of the CHVC samples

The finite sensitivity of the LDS and HIPASS observations results in sample incompleteness at low flux levels in both surveys. The different sensitivities of the two surveys will bias the derived sky-distribution, average velocity, and velocity dispersion towards the more sensitively observed hemisphere, namely the southern one. To compensate for this bias, the objects found in the southern hemisphere were weighted with the likelihood that they would be detected by a survey with the LDS properties. For this likelihood we use the relation plotted in Fig. 4, following de Heij et al. (2002), who assess the degree of completeness of the LDS catalog as a function of limiting peak brightness from a comparison of the detection rates of cataloged external galaxies over the range $-30^\circ<\delta<90^\circ$, and from a comparison with the HIPASS catalog of Putman et al. (2002) for the range $-30^\circ<\delta<0^\circ$. To incorporate plausible uncertainties in this relation, the calculations have also been done for a fictional survey 25% more sensitive, and for one 25% less sensitive, than the LDS, as indicated by the dashed lines in the figure.

Table 1 lists the number of sources with a minimum peak brightness temperature for the northern hemisphere, as observed by the LDS, and for the southern one, as observed by HIPASS. Due to the differences in spectral and spatial resolution, the LDS and HIPASS measure different peak temperatures for the same cloud. For all clouds that are observed in both surveys, the median of the temperature ratio as measured in HIPASS and LDS is 1.5 (de Heij et al. 2002). Applying this temperature scaling to the HIPASS data provides very good agreement with the external galaxy completeness curve of Fig. 4 for declinations $-30\hbox{$^\circ$ }$ to $0\hbox{$^\circ$ }$. However, over the entire HIPASS declination range, the compensated HIPASS data show a strong excess in the source detection rate for sources with an LDS peak temperature in the range 0.2 to 0.4 K. According to Fig. 4, the LDS completeness for these sources should exceed 80%. Therefore the difference in the numbers of relatively faint CHVCs detected by HIPASS and LDS indicates an asymmetry in the distribution upon the sky, with about a factor of two more occurring in the southern hemisphere than in the north. Reducing the sensitivity of the LDS survey by 25% does not change this conclusion.

The CHVC tabulation is probably not incomplete as a consequence of the velocity-range limits of the observational material. Although the part of the LDS that was searched only extended over the range $-450 < V_{\rm LSR} < +350$ km s^-1, de Heij et al. (2002) plausibly did not miss many (if any) clouds because of this limited interval. The high-velocity feature with the most extreme negative velocity yet found is that discovered by Hulsbosch (1978) at $V_{\rm LSR}$ = -466 km s^-1. (This object is listed in Paper I as ?HVC  110.6-07.0-466: being incompletely sampled in velocity, it does not meet the stringent isolation criteria for the CHVC category, and so does not enter this analysis further.) The Wakker & van Woerden (1991) tabulation, which relied on survey data covering the range $-900 < V_{\rm LSR} < +750$ km s^-1, found no high-velocity cloud at a more negative velocity. The HIPASS search by Putman et al. (2002) sought anomalous-velocity emission over the range $-700 < V_{\rm LSR} < +1000$ km s^-1. Of the HIPASS CHVCs cataloged by Putman et al., only 10 have $V_{\rm LSR}<-300$, but the most extreme negative velocity is -353 km s^-1, for CHVC  125.1-66.4-353. Regarding the positive-velocity extent of the ensemble, we note that only 7 objects in the HIPASS catalog have $V_{\rm LSR}$ greater than +300 km s^-1, and only one has a velocity greater than 350 km s^-1, namely CHVC  258.2-23.9+359. All of the 7 CHVCs with substantial positive velocities are near $(l,b)~=~(270^\circ,0^\circ)$, where Galactic rotation contributes to a high positive LSR velocity. Since this extended region has a negative declination, it is sampled with the wider velocity coverage of HIPASS, rather than that of the LDS. In view of these detection statistics, we consider it unlikely that the velocity-range limits of either the LDS or of the HIPASS have caused a significant number of CHVCs to be missed. In other words, the true velocity extent, as well as the non-zero mean in the LSR frame, of the anomalous-velocity ensemble are well represented by the observed extrema of -466 km s^-1 and +359 km s^-1.

The strong concentration of faint CHVCs with an extreme variation in their radial velocity in the direction of the south Galactic pole was already noted by Putman et al. (2002). A complete model for the all-sky distribution of objects will need to reproduce the enhancement in numbers as well as local velocity dispersion in this direction. Much of the north-south detection asymmetry for faint CHVCs remains even after excluding all objects with a Galactic latitude less than $-65^\circ$, as we discuss in detail below.

 

 

Table 1: Number of sources with a minimum peak temperature detected in the northern hemisphere and listed in the LDS catalog of de Heij et al. (2002), and in the southern hemisphere and listed in the HIPASS catalog of Putman et al. (2002). Because of the differing angular and velocity resolutions, the two surveys measure different peak temperatures for the same source. The median ratio of HIPASS to LDS peak temperature of 1.5 determined for the sources in common has been used to resample the HIPASS data in the last column.

minimum	$N_ {\rm LDS}$	$N_{\rm HIPASS}$	$N_{\rm HIPASS}$
$T_{\rm peak}\rm\;[K]$	> $T_{\rm peak}$	> $T_{\rm peak}$	> $1.5~T_{\rm peak}$
1.0	3	5	3
0.5	9	24	9
0.4	12	37	16
0.3	20	56	29
0.2	30	85	56
0.1	38	160	115