Документ взят из кэша поисковой машины. Адрес оригинального документа : http://www.adass.org/adass/proceedings/adass96/reprints/smirnovo1.pdf Дата изменения: Thu Jan 15 01:03:40 1998 Дата индексирования: Tue Oct 2 15:30:32 2012 Кодировка: Поисковые слова: http astrokuban.info astrokuban

Astronomical Data Analysis Software and Systems VI ASP Conference Series, Vol. 125, 1997 Gareth Hunt and H. E. Payne, eds.

Refining the Guide Star Catalog: Plate Evaluations
Oleg M. Smirnov and Oleg Yu. Malkov Institute of Astronomy, Russian Academy of Sciences, 48 Pyatnitskaya St., Moscow 109017, Russia Abstract. This is a preliminary rep ort on an investigation of the quality and prop erties of plates digitized during creation of the Guide Star Catalog (GSC). The results will b e vital for reclassifying GSC ob jects.

1.

Introduction

The GSC was created at STScI to supp ort Hubble Space Telescop e observations. It contains ab out 20 million ob jects, making it the largest all sky photometry source to date. The GSC was created by digitizing 1593 plates and by including bright stars from the HIPPARCOS INCA database. Many ob jects are measured on more than one plate, and thus have multiple catalog entries. We call such ob jects multiple-entry ob jects, or MEOs. Among the shortcomings of the GSC are: biased magnitudes and incorrect classifications for certain ob jects, presence of artifact ob jects. These shortcomings hinder many interesting applications of the GSC, e.g., those involving stellar counts. The Refined GSC (RGSC) pro ject currently in progress at the Center for Astronomical Data, Institute of Astronomy, Moscow,1 is an attempt to rectify these shortcomings. 2. RGSC: a Refined GSC

The ongoing RGSC pro ject is aimed at creating a new catalog, RGSC, based on the GSC. RGSC will contain all GSC data, plus, for many ob jects, corrected magnitudes and more detailed and (hop efully) correct classifiers, complete with confidence-of-classification ratings. The primary effort involves verification and reclassification of GSC ob jects. The following approaches are used: 1. Cross-identification with other catalogs and databases (Malkov & Smirnov 1997). These results are likely to b e quite valuable on their own. 2. Multiple-plate analysis (MPA). A large numb er of ob jects (MEOs) are registered on more than one plate, and thus have several catalog entries. Each plate has, in effect, its own "opinion" on the nature of a MEO, expressed in that plate's entry for the ob ject. Comparative analysis of multiple entries usually yields insight into the true nature of an ob ject.
1

http://www.inasan.rssi.ru/CAD

426

© Copyright 1997 Astronomical Society of the Pacific. All rights reserved.


Refining the Guide Star Catalog: Plate Evaluations

427

3. Probability maps showing the average density of ob jects of a particular class for given coordinates and magnitude will b e built. 4. An exp ert system will b e develop ed that evaluates results of approaches 1­3 and produces final classifications and confidence estimates. For any MEO, prop erties of the individual plates on which the ob ject is registered have at least as much of an effect on the resulting GSC entries as the nature of the ob ject p er se. Our initial studies show that these prop erties vary a great deal from plate to plate. Therefore, detailed analysis of individual plate characteristics is a prerequisite to carrying out accurate multiple-plate analysis of GSC ob jects. 3. Plate Analysis: Issues and Metho ds

To determine the likelihood of an ob ject of magnitude M app earing on a given plate, we compute the luminosity function for the plate, calculating limiting magnitude, saturation magnitude, and maximal p opulation magnitude. To determine the likelihood of an ob ject b eing misclassified on a given plate, and whether some plates are "sp ecial" in that they tend to bias classifications, we calculate 3 0 and 0 3 tendencies, or P {0|3} and P {3|0}. The former reflects the tendency of a plate to misclassify extended (class 3) ob jects as stellar (classifier 0), and the latter the reverse. The 3 0 tendencies are computed using an iterative process. Initial studies demonstrate that 3 0 tendencies of the ma jority of plates are a very significant factor in GSC classifications. In particular, as seen in Figure 1, the Nnon-stellar to Nstellar ratio dep ends on galactic latitude. The b est fit to the average ratio as a function of galactic latitude is a sum of a Gaussian and a very small linear comp onent: Nns /Ns = 0.372 · e-
0.5(|b|+1.44/15.7)
2

+0.0187 + 2.98 · 10-3 ·|b|

To determine whether the magnitude of ob jects is biased when measured near a plate edge, we calculate an average magnitude and flux as a function of distance from the plate center (separately for ob jects of b oth classes). To determine whether some plates bias magnitudes, for overlapping plates we compute the mean magnitude discrepancy among ob jects that app ear on b oth plates. To determine whether an ob ject's classification is biased if the ob ject is near a plate edge, we calculate the average density of ob jects of b oth classes as a function of distance from the plate center. To determine whether GSC plate quality codes are meaningful, we look for a correlation b etween plate quality codes and the parameters mentioned ab ove. We examined whether an ob ject at given coordinates should b e exp ected to app ear on a given plate (i.e., determining the area of the sky that the plate really covers). Plate centers are listed in the GSC, and plates are supp osed to have a regular (square) shap e of a known size. However, most of them have "dead zones"--irregular sections with no ob jects registered, e.g., clamp marks (evidently the result of scanning technique), broken-off corners, circular or rectangular areas near bright stars and globular clusters (manually removed during GSC production), etc. We produced a "plate atlas" (plots of all the ob jects registered on a plate), and develop ed algorithms to determine the "true" b oundary of a plate using (a) the plate atlas, (b) actual GSC data, and (c) lists of bright stars, clusters, and other sp ecific ob jects.


428

Smirnov and Malkov

Figure 1. GSC plates: classifier ratios. Solid line is smoothed data, dashed line is the b est fit. 4. Plate-related Effects

A numb er of problems can b e traced to the plate-related effects discussed ab ove. A stationary ob ject overlapp ed by a given plate is not measured on the plate: Possible causes: Affects: Hit a "dead area." Brightness is b elow the limiting magnitude. Brightness is b elow the saturation magnitude. Multi-plate analysis.

The magnitude of an ob ject (as measured on a given plate) is biased: Possible causes: Affects: The plate tends to bias magnitudes. The ob ject is near the edge of the plate. Stellar counts.

An ob ject is misclassified (and its magnitude is p ossibly biased): Possible causes: Affects: The plate biases classifications. The ob ject is near the edge of the plate. Multi-plate analysis, stellar counts.

Acknowledgments. This presentation was made p ossible by financial supp ort from the Logovaz Conference Travel Program. OM is grateful to the Russian Foundation for Fundamental Researches for grant No. 16304. References Malkov, O. Yu., & Smirnov, O. M. 1997, this volume, 298