A&A 438, 377-389 (2005)
DOI: 10.1051/0004-6361:20042573
P. Lazorenko 1 - Yu. Babenko 2 - V. Karbovsky 1 - M. Buromsky 2 - O. Denisjuk 1 - S. Kasjan 2
1 - Main Astronomical Observatory,
National Academy of Sciences of Ukraine,
Zabolotnogo 27, 03680 Kyiv-127, Ukraine
2 -
Astronomical Observatory of the Kyiv
National University, Observatornaya 3,
04053 Kyiv-53, Ukraine
Received 20 December 2004 / Accepted 7 March 2005
Abstract
A catalogue of astrometric (positions, proper motions) and photometric
(B, V, R, r', J) data
of stars in fields with ICRF objects has been compiled at the Observatory
of the
National Academy
of Sciences of Ukraine and the Kyiv University Observatory. All fields are
located in the declination
zone from 0
to +30
;
the nominal field size is 46'(right ascension
(declination). The
observational basis of this work is 1100 CCD scans down to V=17 mag which were
obtained with the Kyiv meridian axial circle in 2001-2003.
The catalogue is presented in two versions. The
version KMAC1-T contains 159 fields (104 796 stars) and was obtained with
reduction to the
Tycho2 catalogue. For another 33 fields, due to a low sky density of
Tycho2 stars, the reduction
was found to be unreliable. Transformation to the ICRF
system in the second version of the catalogue (KMAC1-CU) was
performed
using the UCAC2 and CMC13 catalogues as a reference; it contains 115 032 stars
in 192 fields and is of slightly better accuracy. The
external accuracy of one catalogue position is about 50-90 mas
for V<15 mag stars. The average error of photometry is better than 0.1 mag for stars down to 16 mag.
Key words: astrometry - reference systems - catalogs
The instrument was used in two observational projects.
The first long-term project was the astrometric survey of the sky in
the equatorial
zone to extend the Hipparcos-Tycho
reference frame to fainter magnitudes. This programme is still in progress.
The second project, now completed, concerns observations of star fields
in the direction of
192 extragalactic ISRF objects, a list of which, for the declination zone
from 0
to +30
,
was taken from Molotaj (2000).
This declination range was chosen to reduce CCD distortion effects
(Vertypolokh et al. 2001; Vertypolokh et al. 2003).
The project was carried out in the framework of scientific problems: maintenance of the Hipparcos frame of reference and the linking of optical frames to the ICRF. This report describes the data reduction and compilation of the Kyiv meridian axial circle catalogue (KMAC1) of stars in fields of extragalactic radio reference frame sources.
The most important data sources used for the compilation of the catalogue include the major catalogues Tycho2 (Hog et al. 2000); CMC13 (Evans et al. 2003); UCAC2 (Zacharias et al. 2004); 2MASS (Cutri et al. 2003); USNO-A2.0 (Monet et al. 1998) and USNO-B1.0 (Monet et al. 2003). Also, for calibration of the instrumental magnitude scale we used several photometric catalogues of NGC 2264 stars.
The astrometric reduction and source catalogues used for compilation of the KMAC1 are shown in Fig. 1. Compilation of the catalogue followed the following steps of data reduction: image processing (Sect. 2), calibration for instrumental and magnitude-dependent errors (Sect. 3) and correction of the magnitude scale (Sect. 4). Conversion to the ICRF was carried out with the two alternative types of referencing, using the space-based catalogue Tycho2 and the modern ground-based catalogues CMC13 and UCAC2. This resulted in the compilation of two catalogue versions: KMAC1-T and KMAC1-CU. The details of referencing to the ICRF system are discussed in Sect. 5 and the computation of proper motions in Sect. 6. The catalogue description, its properties and external verification are described in Sect. 7.
![]() |
Figure 1: Compilation of the KMAC1: main steps of reduction and source catalogues. |
Open with DEXTER |
The first stage of data reduction began with a search and extraction of data files from the database archive. CCD images of stellar fields were then filtered of various instrumental and noise features that introduce an inhomogeneity in the sky level. The inhomogeneity pattern inherent to a scan mode is dominated by a 1D strip-like structure that changes only along the declination (Dec) direction (the x-axis in the CCD), with a possible weak trend over right ascension (RA), the y-axis of the CCD. The striped structures in the images are formed by increased noise from a few dozen bad bright pixels, which produce vertical pixel-width noisy strips. Images are also contaminated by a number of flares and tracks of radioactive particles of cosmic origin and from Chernobyl and which have coma or star-like shapes. Also, the sky level measured along the x-axis has a large-scale component which under normal observing conditions does not exceed 5% of the total signal level. Some scans also show vertical variations in the sky level related to clouds or changing sky brightness.
All types of background variation were eliminated with a simple correction model that considered these variations to be caused by additive components. While this interpretation is reasonable for a vertical pixel width structure, large-scale variations along the x-axis can also contain a multiplicative flat field component. To investigate this problem, we carried out a study illuminating the CCD with a light source placed at the telescope objective. Using bias information read from the outer calibration regions of the CCD, the flatfield pattern was computed and compared to the systematic trends in the preliminary differences of instrumental v magnitudes and r' CMC13 photometry. Only a partial correlation was found, indicating a possible variation of the bias along the x-axis (a similar conclusion was reached by Evans et al. 2002, for observations at the Carlsberg meridian circle). Considering the small amplitude of variations, they were treated as additive components. However, possible inaccuracies due to omission of multiplicative components in the image analysis is compensated for by the method of calibration for errors dependent on instrumental parameters, in which any residual systematic trend along the x-axis is eliminated using information from an external catalogue (Sect. 3.1).
Thus, scans were filtered by, first, subtracting the local sky large-scale changes in the two directions, and then subtracting a running average taken along each column of 1 pixel width.
Detection of objects in the noisy field was carried out
by application of a smoothing filter whose shape approximately
corresponded to the Point Spread Function,
and by the elimination of bright pixel flares. Detection consisted of a
comparison of the pixel flux with a threshold
defined as
,
where
is the local sky noise. The second term
in this expression ensures approximately constant, independent of
and the sky star density, the number of false
detections
(from 300 to 500 per frame). For faint images, it was required that
an object should fill at least two adjacent pixels.
For bright images a special filtration was applied to avoid
false multiple image detections.
Determination of the x, y positions and fluxes v for each object was performed with the various approaches available for processing of CCD images. These are: 1) the modified Center of Gravity (CoG) method (Irwin 1985) and 2) a group of the full profile fitting methods based on the Gaussian linearized least squares method (e.g. Condon 1997; Viateau 1999).
The modified CoG method used at the CMT (Evans et al. 2002) is based on theoretical considerations by Irwin (1985) who demonstrated that its accuracy is almost equal to that obtained with a full profile fitting. The method, based on profile fitting, provids both for circular and elliptical Gaussians; in the second case, horizontal orientation of semi-axes was considered as adequate. The original non-smoothed scans were used for the image processing. Numerical procedures corrected for the undersampling effect that occurs when the pixel size is large and comparable to the FWHM (Viateau 1999). In bright images, saturated pixels were not used for the fitting.
Centroiding was performed, trying the CoG method and the Gaussian circular and elliptic models in turn. When a solution was not achieved at any step of the computation, the image quality index was flagged as non-standard centroiding. This occured also when a final solution, with reference to the initial approximate position (found from the first CoG iteration) was shifted by more than 1.5 pixels. The image quality index thus marks images that are possibly multiple or of non-standard shape.
Computations made by different methods produced very similar results,
which supports the conclusions of Irwin (1985). Thus, the rms difference of coordinates computed by the CoG and Gaussian methods is
about
0.05-0.06'' for V=15-16 mag stars and is negligibly small
in comparison to the internal random error
0.2-0.3''
of one observation.
The most important feature of the profile fitting methods is
the possibility to change their performance so as to minimize the influence
of systematic errors typical of the CCD used at the MAC and which seriously
degrade the accuracy of the Dec measurements (see discussion in the
next section).
Preliminary processing
showed that these errors appear as a systematic trend
in declination with magnitude, which does not depend on whether computations
are made by the CoG or profile centroiding methods. The largest effect
occurs for bright magnitudes; thus for V=10 mag
stars the systematic effect, measured with reference to V=14 mag stars,
is 0.45''. To reduce this effect, each pixel and the related
equation of the linearized system of equations was weighted by a
factor
where I is the flux received
by the pixel from the star. This modification of the least
squares procedure decreased the amplitude of the error to 0.15''.
![]() |
Figure 2:
Systematic dependence of the image size parameter ![]() |
Open with DEXTER |
![]() |
Figure 3:
Preliminary differences
KMAC1-CMC13 in Dec plotted versus ![]() ![]() |
Open with DEXTER |
Image distortions, by affecting the x positions of stars,
cause a systematic bias in the Dec.
Analysis of the KMAC1 positions, obtained with preliminary
data reduction with reference to 12-14 mag CMC13 stars,
revealed a correlation between
differences
KMAC1-CMC13 and
.
Figure 3 shows the typical systematic trend in
,
which is different for different magnitudes and
normally does not exceed
0.1-0.2''.
The trend is quasi-linear with a slope that depends on vbut that cannot be approximated easily since a more complex
cross-relation between
,
,
x and v occures.
In particular, stars imaged in the 50 px edge area most distant from the
reading register escape this dependency.
To remove the dependence of
on x and
,
we considered a
number of models and found that the best correction is to introduce
directly to the measured x values the factor:
The coefficients Av and
were found
based on a criterion of best convergence of star declinations
computed for the nights when they were observed; the
reduction procedure used the CMC13 catalogue
as a reference. The numerical estimate of 1.11 px obtained for
corresponds to the
value
typical for well-exposed images of v=13.1 mag stars
measured near the CCD reading register (x=0).
A similar calibration procedure, based on a formal reduction to the
CMC13 r' photometry, was applied to instrumental magnitudes v. The
difference of photometric bands is of minor importance here
since color residuals v-r' are not correlated with the image
parameters measured at the MAC. The calibration has a form similar to (1):
Another systematic effect was found considering
preliminary differences KMAC1-CMC13 of positions and photometric
values (formal in the last case)
computed with calibrations (1) and (2).
The differences were found to contain a small
fluctuating component along the x-axis, normally
within 0.04'' in position and
0.03 mag in photometry.
However in the x >1050 px area, the trend in Dec
increased to 0.2''. The origin of this trend is unclear and possibly
can be due to the imperfect pixel geometry of the CCD.
Using KMAC1-CMC13 differences,
the trend was removed (only the variable part, so as not to incorporate
possible systematic errors of the CMC13) and the reduction
procedure, including determination of Av, A'v and
values, was iteratively repeated. In Fig. 4, the
KMAC1-CMC13 residuals in Dec
before and after calibration are shown. Along with a complete remove of
the correlation, the random scatter of
differences
has been noticeably reduced.
![]() |
Figure 4:
KMAC1-CMC13 differences
![]() ![]() |
Open with DEXTER |
The successful refinement of the measured data suffering from various instrumental errors was based on extensive use of the CMC13 as a tool for error calibration. Under the reasonable assumption of no correlation between instrumental errors of the two telescopes, the procedure is correct, and the accuracy of calibration depends on the accuracy of the data in the source catalogue.
In drift scan mode, formation of star images is non-synchronous since the median moment of the image exposure depends on RA. For that reason, the atmospheric turbulent conditions under which the images are formed vary as a function of RA. The measured x, y, v data are therefore affected by a time-dependent component of atmospheric refraction, causing an effect of image motion much larger than is inherent to the astrographic mode of observations. The induced temporal signal is difficult to trace and makes referencing of the observed data to the celestial system more difficult. A number of methods have been proposed to calibrate this effect, see e.g. Evans et al. (2002); Viateau et al. (1999).
In the case of short scans obtained at the MAC, direct calibration of
atmospheric fluctuations with use of the Tycho2 catalogue was found to give
unreliable results since scans often contained few reference stars.
We used a method which consists of substitution of all
individual overlapping (normally to 2') scans available
for the particular ICRF field by a single specially-formed "equivalent'' scan.
For this, each scan of the ICRF field was preprocessed
with the
Tycho2 catalogue so as to determine a zero point of CCD positions and
magnitudes, and to approximately (to
1'') reduce
the relative displacement of individual night scans.
After cross-identification, a compiled list of field objects was formed with x, y, v data averaged.
This procedure is similar to the formation of subcatalogues adopted
at the Valinhos meridian circle (Viateau et al. 1999) but with
no conversion to equatorial coordinates.
The validity of this substitution is based on the linearity of the averaging
operation. Thus, the averaging can be performed either
prior to conversion
to the celestial coordinates (that is, over the CCD measured data),
or after this reduction (over equatorial positions), with equivalent results.
A stringent linearity of the averaging is only achieved, however, when
the star content of individual scans is identical.
This is the case for the bright stars
that are normally detected and measured in each nightly scan.
In the case of omitted (usually faint) object images, corrections allowing for the compensation for the scan system of the "omitted'' star observation should be applied to that object's x, y, v data in the equivalent scan. The information necessary to make this correction is found by obtaining the differences between each nightly scan and the correspondent equivalent scan. Since we consider the measured data, not transformed to celestial coordinates and magnitudes, the differences usually show systematic trends in both x and y directions and due to possible varying magnitude errors in v. The systematic component of these differences was approximated with cubic spline functions.
Calibrations for omitted images started from consideration of major systematic trends along the temporal y-axis. After corrections in the equivalent scan data (x, y, v), these trends were removed from the nightly scans. As a result, the differences between each nightly and "equivalent'' scan became noise-like in shape. Next, similar steps of calibration were applied to the differences of "nightly scan'' - "equivalent scan'' registered along the x and v data axes. To obtain convergence, the whole procedure was reprocessed twice, the outliers removed and all computations repeated again.
The resulting equivalent scans used for transformation to the ICRF are less subject to atmospheric differential image motion due to averaging over a subset of individual scans included in the output. The averaging effect is inversely proportional to the square root of the number of frames, which is 6 on average. The output nightly scans were of less importance since they were tightly reduced to the system of the corresponding equivalent scan by filtering out any systematic differences. The differences between these scans contained only a random noise component which provided valuable information on internal catalogue errors (Sect. 7.1).
An important restriction to the method discussed is
that correct tracing of scan system changes is achieved only with
completely overlapped and co-centered nightly scans. Displacement
of individual scans by 10% of a scan length results
in incorrect extrapolation of the offset scan system in edge areas and
causes the problems discussed in Sect. 5.
It was considered that initial estimates of KMAC1-Tycho2 residuals are somewhat biased due to redistribution of magnitude-dependent errors between reference stars in the field. Therefore, to extract better estimates of the magnitude-related errors from the KMAC1-Tycho2 residuals, calculations were refined in an iterative manner.
After calibration, the residual trend in corrected positions,
estimated using the CMC13, does not exceed 0.04'' for the entire
magnitude range.
![]() |
Figure 5: Systematic differences KMAC1-CMC13 in Dec as a function of magnitude for 10 groups of star fields ordered by RA; before ( upper panel) and after ( bottom) calibration (3). |
Open with DEXTER |
Considering that the picture shown in Fig. 5 may originate from errors
in the CMC13, we used this information
indirectly, to assume a possibility of a specific error
in the MAC declinations (or in x values), and to define a function
The effect of calibrations based on the Tycho2 catalogue is
seen in Fig. 5
where the KMAC1-CMC13 differences after correction
are shown in the bottom panel; the residual
variations of the magnitude-related errors in Dec
are shown to be reduced to 0.03'' or less
at
mag.
Star magnitudes V of the KMAC1 have been computed using measured v values corrected for instrumental and magnitude-dependent errors as described above. The zero point of the V magnitude scale was determined using the Tycho2 photometry of bright V<13 mag stars. The problem consisted of verification of the magnitude scale linearity at its faint end, which cannot be directly controlled due to the absence of faint all-sky standards in the V band. The study and the following calibration used indirect methods relied upon red r'and infrared J data taken from the CMC13 and 2MASS global catalogues respectively. The UCAC2 catalogue, as an alternative r'-like data source, was not utilized since its magnitudes are only approximate and not calibrated. Our attempt to take advantage of this catalogue photometry resulted in a similar but slightly less accurate calibration compared to that provided using the CMC13.
The development of the calibration model and its velidation was based on a photometric study of the open cluster NGC 2264 scans obtained at the MAC, specially for this purpose.
![]() |
Figure 6: Systematic effect in measured V values: a) found from a comparison to photometric data by Sung et al. (2003) (crosses), Kuznetsov et al. (2003) (circles), Internet WEBDA database (inclined crosses) and Tycho2 (squares); b) the bias simulated with the model (5). |
Open with DEXTER |
Examination of residuals, however, revealed a large 0.6 mag
systematic bias of measured data for faint V>12 mag stars,
shown in Fig. 6a. Interpretation of this plot should take
into account that
images of stars brighter than 12 mag are oversaturated
and their fluxes determined by the centroiding
procedure can be systematically biased.
The resulting effect in magnitude is however opposite
because the zero point of the V scale is referred to bright Tycho2 stars.
Another important feature of
the plot is the linearity of magnitude scales in
either bright V<10.5 mag and faint V>12 mag segments of the V-axis,
however, with different zero points.
We tried to simulate this systematic effect using two-color
diagrams that were built for NGC 2264 (Fig. 7a) and for
a complete list of the KMAC1 stars
(Fig. 7b). Star distributions in both plots are clearly
separated depending on V, bright stars being shifted systematically
upward relative to faint stars. The shift
does not depend on V -J color, so we refer it entirely to
magnitude-dependent errors in MAC photometry.
A number of other two-color
diagrams were also tested, including those that incorporate H and K
infrared data
from the 2MASS catalogue; it is the
diagram that ensures the best separation of stars in V.
![]() |
Figure 7:
Two-color diagrams
![]() |
Open with DEXTER |
The differences V-r' for bright KMAC1 stars are approximated
by a function (solid line in Fig. 7b)
The calibration of faint star photometry to the instrumental
magnitude system defined by bright stars is based on the use of the
fitting curve (4) as reference.
Consider the two-color diagram where the
unbiased star location A is a point
,
in the
fitting curve (Fig. 7a).
The star's measured
position, B, is shifted by
in both directions, to
,
.
This geometry allows us to
express the point B distance to the fitting curve (4) in two ways, as
and as
where
f'V-J is the derivative. Hence we derive an estimate
Using Eq. (5), we computed errors
for
each NGC 2264 star with r' and J data available. The results
shown in Fig. 6b match well
the systematic trend found directly on photometric standards
(Fig. 6a), except for a
small systematic discrepancy of about 0.1 mag at the faint V end.
![]() |
Figure 8:
Individual estimates (5) of ![]() |
Open with DEXTER |
![]() |
Figure 9:
![]() |
Open with DEXTER |
Further analysis has shown that the value is to be determined for
each star field individually, by fitting estimates (5) with
the model (6). The calibration of magnitudes therefore was
performed with individual
values that varied from -0.63 to -0.14 (-0.58 for NGC 2264). The scatter of
values,
in particular, is the cause of the large
0.157 mag dispersion of points
in Fig. 8.
The efficiency of corrections is seen from
the two-color
diagram built
with final calibrated V values (Fig. 9). The relative shift
of bright to faint stars like that shown in Fig. 7b
was eliminated; also, the standard deviation of points from the fitting
curve
Figure 10b shows a comparison of the KMAC1 and Valinhos meridian
circle photometry (Camargo et al. 2001) for 1190 stars in 13 stellar fields.
The residuals contain no large systematic trend; the
standard deviation of data points is 0.13 mag.
Consideration of local fields, however, indicates
local systematic discrepancies in magnitudes
sometimes reaching
0.10 mag, with a random scatter of residuals
of about
0.10 mag.
Considering the accuracy of the Valinhos photometry, which is
about 0.10 mag (Viateau et al. 1999),
we estimate that the KMAC1 data
is of the same or better accuracy.
In conclusion, we give a relation between V and the CMC13 r' values
using r'-J colors:
![]() |
Figure 10: Magnitude residuals as a function of V: a) V-r' differences measured with reference to the fitting curve shown in Fig. 9; b) residuals between the KMAC1 and the Valinhos catalogue V values. |
Open with DEXTER |
Compilation of KMAC1 started from astrometric calibration of the measured data as described in the previous sections. Conversion of corrected CCD x, y positions to equatorial coordinates originally was intended to be performed using the Tycho2 catalogue which is the best optical representation of the ICRF. Prior to this conversion we performed a tentative study to determine how well the positions of the Tycho2 catalogue match the modern catalogues CMC13, UCAC2 and our observations. For this purpose we selected Tycho2 stars that:
The above analysis establishes the degradation of the Tycho2 data at the epoch of KMAC1 observations, probably due to uncertanties in proper motions. This problem is often allowed for in different ways when referring CCD observations to equatorial coordinates. Thus, at the Flagstaff Astrometric Scanning Telescope reductions are made by applying weights to the Tycho2 stars depending on their brightness (Stone et al. 2003).
![]() |
Figure 11: Correlation between Tycho2-CMC13 and Tycho2-KMAC1 preliminary differences: a) - in RA; b) - in Dec; correlation between Tycho2-UCAC2 and Tycho2-KMAC1 preliminary differences: c) - in RA; d) - in Dec; only differences exceeding 0.15'' are shown. |
Open with DEXTER |
This discussion suggests
that a reliable reduction to the ICRF using the Tycho2 catalogue requires
the use of a sufficiently large number of reference stars which are to be
first filtered or weighted to eliminate problem stars.
This is especially important in our case since, due to the rather short
scan length,
some fields in sky areas with a low star density are represented only
by 6-8 Tycho2 stars.
Also, the accuracy of the reduction is affected by inhomogeneity in
the sky distribution of reference stars whose images,
in addition, are oversaturated and poorly measured.
At the first stage of referencing we therefore
detected and removed all
Tycho2 problem stars whose positions deviated from
the CMC13, UCAC2 and preliminary KMAC1 data by more than 0.2''.
This greatly improved the reliability of the conversion to equatorial
coordinates and was found to be more efficient than the usual
search for outliers based on an iterative approach to the least-squares
solution. With a truncation limit of
0.2'', reliable results
were obtained, however, for 106 fields only.
For another 53 fields,
a good transformation to the ICRF required a further rejection of
reference stars with Tycho2-CMC13 and Tycho2-UCAC2 differences in
the range from
0.2'' to
0.15''.
For the 33 remaining fields with a low reference star density, the
reduction was found to give
quite unstable and ambiguous solutions highly sensitive to
any changes in the reference star set.
A further comparison with the CMC13 and UCAC2 positions has shown that large systematic deviations are present at the edges of some fields. This concerns those fields that at some nights were observed with incorrect telescope pointings (made by hand since the MAC is not automatic); in a few cases the relative displacement of sky strips exceeds 10' in RA. Individual scans thus were not exactly overlapped as was assumed at the phase of equivalent scan formation. To eliminate this fault, the offset regions were truncated.
A rigorous conversion to the ICRF using the Tycho2 catalogue was achieved for 159 sky fields, most of which had a high star density. Conversion for the complete data array (192 sky fields) required the use of the CMC13 and UCAC2 catalogues which are known to be in the ICRF system. The reduction was performed with well-measured stars not fainter than 14.5 mag and by limiting their number to 170. Reference catalogues were used in a combined form, with equal weights. No truncation of offset scan edges was applied since the large number of reference stars ensured very tight referencing using spline fitting.
Thus, the catalogue KMAC1 exists in the two versions: with reduction to the Tycho2 (KMAC1-T) and to the CMC13 and UCAC2 catalogues (KMAC1-CU). No rejection of stars with large deviations from comparison catalogues was applied. V magnitudes given in both catalogues are identical and based on the Tycho2 photometry with the corrections described in Sect. 4.
The percentage of KMAC1 stars supplemented with proper motions varies from 53 to 97, and on average is 90%. The highest ratio of 93% is obtained in the magnitude range from 13 to 17 mag, dropping to 47% for V<12 mag stars and to 74% for V>17 mag.
Considering the approximate precision of 250 mas,
the mean epoch 1954 of the USNO-A2.0 and the internal positional
precision of the KMAC1 (Table 1), we find a formal estimate
of proper motion errors of 5-6 mas/year.
Table 1: Main characteristics of catalogues KMAC1-T and KMAC1-CU.
The catalogue KMAC1, as explained above, was released in the two versions
and can be obtained in electronic form from the CDS or via
anonymous ftp://ftp.mao.kiev.ua/pub/users/astro/kmac1. The KMAC1-CU catalogue contains 115 032 stars in 192 sky fields and is
referred to the CMC13 and UCAC2; the KMAC1-T contains 104 796 stars (91% of the total star number)
in 159 fields and is referred to the Tycho2 catalogue.
The location of KMAC-CU fields in the
sky is shown in Fig. 12. All fields
are located in a declination zone from 0
to 30
;
the mean epoch of observations
is 2002.33. The main
characteristics of the catalogue are given in Table 1.
![]() |
Figure 12: Distribution of KMAC1-CU fields across the sky. |
Open with DEXTER |
Besides the original positions given at the epoch of observations, proper motions and original V values, the KMAC1 catalogue contains B, R values from the USNO-B1.0 for 83% of the stars identified; r' values are taken from the CMC13 for 67% of stars and J values from the 2MASS catalogue available for 94% of stars. Usual supplementary information, including internal error estimates, the number and epoch of observations, the image quality index, image size for extended objects and cross-identification to the USNO-B1.0 is also given.
Note that the flagged image quality index (see Sect. 2) may indicate centroiding problems of various origins (e.g. binary or unresolved stars); the catalogue positions therefore are probably biased. The unequal number of observations for RA and Dec data means that a rejection of bad measurements was applied which also may indicate certain problems with image quality not marked in the image quality index. These stars should not be used when very high accuracy of positions is required.
The list of sky strips, the IAU designations of the central ICRF object
and star numbers
and
containing, respectively, in the KMAC1-T and KMAC1-CU for each strip,
are given in Table 2.
Note that
is often lower than
due to a truncation of sky strip
edges applied to some fields (Sect. 5).
The star number distribution over
fields is highly inhomogeneous and depends on the Galactic latitude.
This distribution as a function of RA is shown in
Fig. 13.
Table 2: Fields with ICRF objects and star numbers contained in the KMAC1-CU and KMAC1-T catalogues.
The distribution of stars by magnitude (Fig. 14) shows that the catalogue limiting magnitude is near V=17.0 mag. Note that beyond this limit, some of the faint V>17 mag objects in the catalogue may be artifacts appearing due to the low detection threshold. A substantial ratio of very faint stars were not identified with any of the 2MASS, USNO-A2.0 or UNSO-B1.0 objects, which could be related to either variability of stars or false detection. Thus, while 99.3% of stars to 16 mag were found in one of the major catalogues, this ratio drops to 77% for V>17 mag stars, and for yet fainter V>17.5 mag stars the ratio decreases to 56%. Nevertheless, faint stars were not excluded from the catalogue because of the very low probability of false detections since each star was observed in at least two CCD scans. A very powerful indicator of the detection feasibility is the number of times the star was observed. Thus, among V>17.5 mag stars observed at least 3 times, the ratio of identifications with external catalogues is 98.3%.
The internal precision of the catalogue was estimated in a somewhat unconventional way, by comparing CCD positions and instrumental magnitudes of stars in those nights when they were observed. The comparison was made with nightly scans transformed to the equivalent scan system (Sect. 3.2). Note that an important feature of this transformation procedure is a scan to scan fitting that works as a filter which completely removes any systematic differences between scans, leaving only random components. This is the reason why the internal precision can be estimated in the way discussed with no use of equatorial positions and V magnitudes computed for each night (this data was never computed). Results presented in Fig. 15 show the equal accuracy of RA and Dec positions.
External verification of the KMAC1 positional accuracy was performed using the CMC13 and UCAC2 which are the only all-sky sources of present epoch positions available for faint stars. We computed the rms differences of KMAC1 positions with the CMC13 and UCAC2 (Fig. 16). The plots that refer to catalogue versions "T'' and "CU'' are very similar. The increase of errors at the bright V<12 mag end is caused by oversaturation of images and affects primarily declinations. For fainter magnitudes, the precision of RA and Dec is the same, which is evidence of the good efficiency of various calibrations applied to improve declination data.
![]() |
Figure 13: Distribution of KMAC1 star number per field on rightascensions. |
Open with DEXTER |
![]() |
Figure 14: Distribution of the KMAC1-T (solid) and KMAC1-CU (dashed) star magnitudes. |
Open with DEXTER |
![]() |
Figure 15: Internal mean accuracy of one catalogue entry as a function of magnitude. |
Open with DEXTER |
![]() |
Figure 16: Rms residuals of the KMAC1 positions with the CMC13 and UCAC2; upper panel - for the version "T''; bottom - for the version "CU''. |
Open with DEXTER |
![]() |
Figure 17: External errors of the KMAC1 ("T'' and "CU'' versions) positions (curves) and photometry (open circles). |
Open with DEXTER |
We were not able to perform a direct external verification of the KMAC1 magnitudes because of the lack of all-sky referencing available for faint stars in the V band. Comparison to the Valinhos photometry (Camargo et al. 2001) was performed for a 1% subset of catalogue stars and yielded about a 0.1 mag error estimate (see Sect. 4).
External photometric errors (Fig. 17) were found considering the dispersion of points in Fig. 10, which are the deviations of V-r' residuals from the color calibration curve (7). No subtraction of CMC13 photometric errors was applied since for magnitudes fainter than V=16 these errors were found to exceed the rms KMAC1-CMC13 differences. Thus, at V=16.4 mag (16.0 mag in the r' band) the rms KMAC1-CMC13 differences are equal to 0.12 mag while the CMC13 external error is 0.17 mag. Probably, the quality of the CMC13 photometry is better than cited.
Magnitude-dependent systematic errors of KMAC1-CU and KMAC1-T
do not exceed 20 mas and
40 mas respectively
(Fig. 18.). No clear dependency on magnitude is seen
except for a positive hump of plots at
mag, and
a negative downtrend for the version KMAC1-T in Dec at the
bright V<11 end.
![]() |
Figure 18:
Systematic differences
![]() ![]() |
Open with DEXTER |
Individual differences between the KMAC1-T star positions and positions in the comparison catalogues CMC13 and UCAC2 are shown in Fig. 19. We present the worst comparison in Dec and the "T'' catalogue version; slightly better plots can be obtained for RA and the KMAC1-CU catalogue version. No systematic trend of individual differences in magnitude is observed.
![]() |
Figure 19: Individual differences of declinations in the KMAC1-T and in CMC13 and UCAC2. |
Open with DEXTER |
The aim of this work was to obtain a catalogue of faint stars in sky areas with ICRF objects whose declinations are optimal for observations with the MAC. The catalogue contains positions of faint V<17 mag objects referred to the optical Hipparcos-Tycho reference frame and thus presents an extension of the ICRF to the optical domain.
The catalogue described in this Paper is the first catalogue obtained with the Kyiv meridian axial circle after it was refurbished with a CCD camera. Realization of this project involved development of special software for image processing, astrometric calibration for instrumental errors etc. A quite unexpected finding was that the measured data (especially the Dec component) is strongly affected by systematic errors even when star images have a relatively good shape. A solution to this problem was found in extensive use of external astrometric catalogues for calibrations.
Another difficulty arose from underestimation of the Tycho2
errors at the present epoch and from the inhomogeneous sky
distribution of the catalogue stars.
As a result, scan lengths appeared to be too short
to allow rigorous reduction to the ICRF
and forced us to use other catalogues (CMC13 and UCAC2) for referencing.
The use of the Tycho2
catalogue for astrometric work in small fields of about
or less is thus problematic, and feasible only in some sky areas.
Acknowledgements
We acknowledge Dr. D. W. Evans important observation concerning photometric calibrations, V. Andruk for his suggestions on the CCD raw data filtering and image processing, and Dr. A. Yatsenko for valuable remarks about proper motion determination. Technical development of the CCD micrometer was carried out by O. Kovalchuk (Nikolaev Astronomical Observatory, Ukraine). This publication makes use of data products from the Two Micron All Sky Survey, which is a joint project of the University of Massachusetts and the Infrared Processing and Analysis Center/California Institute of Technology, funded by the National Aeronautics and Space Administration and the National Science Foundation; the VIZIER database, operated at CDS, Strasbourg, France; the WEBDA database for open cluster; and NASA's Astrophysical Data System Abstract Service.