The Quality Assurance of the Gsii Database
THE QUALITY ASSURANCE OF THE GSCII DATABASE:
FIRST RESULTS OF THE QA_2PLATES PROCEDURE
by
R.Pannunzio, R. Morbidelli, M. G. Lattanzi and A. Spagna
Osservatorio Astronomico di Torino
Internal Report no. 57/01
Pino Torinese, February 2001
ABSTRACT
This Report is the continuation of the study described in the previous OATo Internal Report no. 53/00, in which were showed the procedures used for the evaluation of the reliability of the data produced both by the pipeline and by the matching procedure on the GSCII plates processed at Torino.
In particular, preliminary results, obtained by selected samples of GSCII plate pairs of the CRA list, loaded in our local COMPASS Database and processed with the QA_2PLATES procedure (see Internal Report no. 53/00), are given in this Report.
The first statistical results obtained on the whole sample of data produced by the QA_2PLATES till now seem to show a good agreement with the expected astrometric and photometric accuracy for most pair of plates.
The goodness of these results is a preliminary indication that both pipeline and object matching procedures are reliable even if future astrometric and photometric calibrations could improve the astrometry and photometry in order to give to scientific community the best that can be obtained from the analysis of these GSCII plates.
INTRODUCTION
Some months ago at the Astronomical Observatory of Torino (OATo), an IDL procedure called QA_2PLATES has been realized (see OATo Internal Report no. 53/00).
The purpose of this procedure was the extraction of the parameters of two overlapping plates loaded in our local COMPASS Database, by means of a C++ interface (CIQA), in order to perform a selected series of statistics with the parameters obtained by the previous steps of the GSCII project (pipeline and object matching procedures).
The QA_2PLATES procedure has been refined and checked several times in OATo and now is in exportable version. Recently a copy of this procedure has been delivered and implemented in the COMPASS database of the Space Telescope Science Institute (STScI) of Baltimore (U.S.A.).
PRELIMINARY QA RESULTS ON GSCII PLATES REDUCED AT TORINO
The output of the QA_2PLATES procedure are essentially reports files and statistical graphics on the results of the pipeline and object matching of the plates processed and contained in the local COMPASS Database.
In order to have a global check about the quality of the results furnished by the QA_2PLATES procedure we have realised a mock-up IDL software (GLOBAL_QA) that collects all the information contained in the report files of the previous software. The GLOBAL_QA software works on the basis of the structure of the output produced in batch mode by the QA_2PLATES procedure.
On the other hands, the QA_2PLATES runs on all pair of plates starting from a list of source plates.
The batch procedure finds all the overlap plates of each source plate and for each pair performs all the statistics foreseen by the routines of the QA_2PLATES.
For each examined pair of plates this code creates a directory of first level labelled with the name of the source plate and each ones of these directories gives origin to a series of subdirectories of second level labelled with the names of the overlap plates, where each ones of these contains all the report files and figures produced by the routines of the code.
In this context the QA_2PLATES has been ran in batch mode on a sample of 67 GSCII source plates of the CRA list.
For each considered source plate, several overlapped plates have been found and for all 67 source plates, 840 combination of plate pairs are resulted. In this context we want to remark that in our COMPASS Database, at present, not all the overlap plates are connected with the GSCII project, because at the beginning the Database has been populated with GSCI plates only , while the GSCII plates are loaded sequentially only after the pipeline processing.
The fist result obtained by this software has put in evidence that not all the plate pairs have given reliable results both in pipeline and in the objects matching, for which in some cases the parameters extracted from the Database are resulted completely wrong (null values or inconsistent values). In a such way the plate pairs having reliable results are 708 only.
If we take into account the GSCII pair of plates only , the number furthermore decreases to 335 pairs.
All the above numerical considerations are reported in the following Table 1. In addition the same Table1 also gives the number of pair of plates having same area but different colour (usually plates quasi-completely overlapped) or plates having same colour but different area (usually contiguous plates in the same pass-band with a small percentage of overlap ) .
With the same criteria the combination of pair of plates having both different colour and area or having same colour and area, are taken into account. For each one of these combinations in Table 1 two columns ,are reported .The first column gives the global number of coupling (GSCI and GSCII plates), while in the second column the GSCII coupling only, are shown .
How it is possible to see from Table 1 the last four particular combinations are much reduced in number with respect to the total amount of data but however are sufficient to put in evidence possible systematic and accidental errors in astrometry or photometry depending on the colour or on the percentage of overlap.
Even if the number of reliable plates is only 708 against 840 (see column 1 of Table1 ) the effective number of wrong plates is lower, because usually a wrong plate is combined with several overlapped plates.
The adopted criteria for the selection of reliable plates can be summarized in the following points:
- only pairs of plates having at least 50 matched objects have been selected
- the pairs of plates with () , () and(mag) having null values or values greater than 2 arcsec or 2 mag for both stars-stars and non stars-non stars combination of classifications, have been rejected
We remember that all the astrometric statistics performed by the QA_2PLATES and consequently by the GLOBAL_QA are in standard coordinates. In addition when in the text is indicate , and mag we want to remark that these variables are the average values of all the differences in , and mag between each pair of matched objects of a single pair of plates (intended as source – overlap plate), while the () , () and (mag) are the standard deviations with respect to the average values of , and mag forasingle pair of plates.
On the basis of the constraints above quoted only two GSCII plates are really wrong, in particular the plate A343 (survey : XO323) and the plate G0R9 (survey : S020), while the other wrong plates are combination of GSCI plates (usually with larger errors and few objects) with GSCII plates (more accurate and populated) .
The sample of processed plates
•NUMBER OF SOURCE PLATES 067
•NUMBER OF PLATE PAIRS 840
•PAIRS OF GSC I-II/GSCII PLTS WITH RELIABLE RESULTS708335
•PAIRS OF GSC I-II /GSCII PLTS WITH SAME COL.DIFF.AREA127059
•PAIRS OF GSC I-II /GSCII PLTS WITH SAME AREA DIFF.COL.091064
•PAIRS OF GSC I-II /GSCII PLTS WITH DIFF.COL.- DIFF.AREA479212
•PAIRS OF GSC I-II /GSCII WITH SAME COL. AND SAME AREA011000
The first figure gives the spatial distribution of the source plates with their overlap plates on a planisphere in galactic coordinates. How it is possible to see in Fig. 1 the source plates (+) are preferentially distributed around the galactic plane and to the galactic poles together to the corresponding overlap plates (). In this way we have the possibility to compare the accuracy of the QA results in two peculiar part of the sky.
In the following two histograms of Fig. 2 are presented the distributions of the outliers percentage with respect to the whole matched objects found in each pair of plates both on the whole sample of reliable plates (GSCI+GSCII) and on the GSCII pairs only, respectively.
We have considered as outliers all the pair objects whose , and mag in absolute value are greater than 4 arcsec and 4 mag.
The analysis of the first histogram of Fig. 2 shows that most outliers are confined in less of one percent of the matched objects, while a residual part of plates has a percentage included between 1% and 5%. The situation is decidedly better for the GSCII plates only, because there are very few outliers with a percentage greater than 1% . The explanation of the low outliers percentage in the GSCII pairs with respect to the whole sample of plates is in the different procedure of the astrometric and photometric calibrations adopted for GSCI and GSCII plates.
In order to have reliable statistics from all the parameters collected by the report files of the QA_2PLATES procedure we have decided to take into account only the parameters connected to the stars-stars and non stars-non stars combination of classifications.
The histograms of Fig. 3, show the frequency of the plate pairs both for all the whole sample of plates (GSCI and GSCII plates) and for GSCII plates only as a function of the () found for the stars-stars combinations on each pair of plates .
In both cases the average () is about 0.5 arcsec while their standard deviation() are about 0.2 arcsec even if the global distribution (left histogram) seems slightly better than the GSCII distribution (right histogram).
We want to remark that the average () are approximatelylarger than the average () of a single plate by a factor 2, because the average () is the combination of the average () of the source and overlap plates.
Of course the same conclusions seen above can be extended to the average () and (mag) and to the () of the same parameters too.
A similar distribution is shown in Fig. 4 in which are represented in abscissa the ().
Comparing Fig. 3 with Fig. 4 we can remark that the () are slightly better than the corresponding () both for the left and for the right histogram. The greater accuracy showed in the coordinates with respect to is explained by the fact that in there are probably residual errors in the telescope driving.
If we consider the same distributions of Fig. 3 butreferred to the non stars-non stars pairs we have the situation shown in Fig. 5.
In this case the average () are practically coincident in both histogram but if we compare these values with those of Fig. 3 we note that the average values of the non stars-non stars are about one and half times greater than the corresponding stars-stars pairs.
The reason of this is probably due to the irregular shape of this images that in many cases are not perfectly circular.
are very different from the corresponding values of Fig. 3 (0.7 arcsec against 0.5arcsec
In order to have a direct comparison of the previous figures, in Fig. 6 the distribution of the () for GSCII plates only for stars-stars and non stars-non stars combination, are repeated. In this case it is clear that the average () is betterfor the stars-stars combinationsfor the reasons above seen, but the comparison with the first plots of Fig. 3 and 5, where the whole sample of plates (GSCI + GSCII) has been considered, shows no evident variations .
Fig. 7 shows the distribution of the () and () for GSCII plates only, for stars-stars combination but referred to plates having same colour and different area . The average values of the two histograms shows a better accuracy in the coordinates as already seen in the above statistics. On the other hands when overlapping plates of the same colour are considered, the percentage of overlap between this kind of plates is in general small, as the most plates considered in our statistics. In fact a large number of plates pair has different area and colour, and consequently a small percentage of overlap (see Table 1), for which the statistics of Fig. 7 are very similar to the previous one on the global data.
Fig. 8 represents the distribution of the () for the GSCII plates only having same colour but different area for the stars-stars and non stars-non stars combination.
The results of these statistics are very similar to those obtained with the previous global statistics for the same reasons given for the Fig. 7. In particular is confirmed that the stars-stars combination is slightly better than the non stars-non stars combination.
A very interesting statistics is shown in Fig. 9 in which are compared the distributions of () for the GSCII plate pairs of the stars-stars combinations both for plates having same area but different colour (left histogram) and plates having same color but different area (right histogram)
It is clearly evident that the first histogram of this figure (different colour and same area) has an average () lower than the corresponding average () of the second histogram (same colour and different area). The motivation of this appreciable difference is probably due to the fact that the astrometric accuracy is function of
the position on the plates of the matched objects. In other words, matched objects near to the plate centres show a lower dispersion of the () or () with respect to the same dispersions of the matched objects at the edges of the plates, probably due to residual distortions far away by the plate centres, not completely corrected by the astrometric calibrations.
This situation is really verified when the plates are partially overlapped, namely when the plates have same color but different area.
In Fig. 10 the same distributions of Fig. 9 are shown, but referred to the non stars-non stars classification. The average () in both histograms are very similar among them and consequently seems that the astrometric accuracy is independent by the colour of the plates or by the percentage of overlapped area.
A possible explanation of this should be in the low astrometric accuracy found in the position of the non stellar objects that hides the greatest accuracy usually given by the plates with a total overlap with respect to the other ones with a partial overlap (even if of the same colour).
As far as the standard deviations by the average () or () are concerned, it is possible to see that in all the previous figures the () are all included in the range 0.15 – 0.20 arcsec. This indicate that the astrometric quality of all the considered plates is comparable.
If we take into account the statistics on the average (mag) for the stars-stars case both for all the sample of plates and for the GSCII plates only , we have the distribution represented in the following two histograms of the Fig.11, respectively
The average values of (mag) are practically coincident in both distributions around to about 0.5 mag while the () of about 0.2 mag is slightly better on the whole sample of data.
A little bit different is the distribution of the average (mag) of Fig. 12 . The histograms show the comparison between the whole sample of plates and the GSCII plates only for non stars-non stars pairs.
The average value of (mag) is greater than the values of the corresponding histograms of Fig. 11 .
The motivation of this lower accuracy is still due to the shape of the images that for non stellar objects is much irregular with the consequent imprecision in the magnitude determination.
The first histograms presented in Fig. 13 is referred to statistics conduced on the (mag) of the GSCII plates only for the stars-stars combination of the plates having same colour but different area, while the second histogram gives the same statistics but for plates having same area and different colour.
It is evident that the plates of the same colour show an accuracy on the average (mag) double with respect to the plates having the same area.
The reason of this difference between the two histograms depends on the fact that the accuracy in magnitudes is correlated with the colour of the plate pairs .
On the other hands if the plate pairs have the same colour, the photometry of the objects of the source and the corresponding objects of the overlap plates should be theoretically equal , apart the accidental errors , while for plate pairs having different colour each matched object shows a colour index depending on the intrinsic colour of the objects and on the colour difference between source and overlap plate.
If we take into account that these systematic effects in (mag) are alternatively positive and negative (in the case where the source becomes the overlap plate and the overlap becomes the source) the average value of (mag) will be greater than the corresponding value of the plate pairs with the same colour.
The same statistics of Fig. 13 are conduced in Fig. 14 where the non stars-non stars case is represented.