6 Conclusions

The Stone ([1996]) code calculating mean refraction has been adjusted to the case of the 10'' LX200 telescope installed at the Lohrmann Observatory in Dresden. We have estimated how accurate the appropriate stellar and atmospheric parameters should be to provide desirable accuracies in calculated refraction. The conclusion has been made that the accuracy of calculated refraction of 0.02 arcsec per one stellar parameter (spectral type, luminosity class, degree of interstellar reddening) should be achievable if the best available spectral classification systems are applied. The special case of computing differential refraction has also been discussed. The usefulness of photometric V-R and V-I indices in estimating mean refraction has been justified.

The dependence of atmospheric refraction on stellar metal abundance (it is the third parameter in spectral classification, side by side with spectral types and luminosity classes) has been estimated. We have found that the effect of metal abundance does not exceed 0.02 - 0.03 arcsec in calculated refraction.

We underline that accurate spectral classification of stars or color photometry should accompany any exact determinations of stellar positions. The use of objective prisms in spectral classification seems to be very efficient. Objective prism camera is the most efficient stellar spectrograph, because it makes use of the full telescope field. No light losses occur at a slit and those in the optic are small in comparison to slit spectrographs. The available classification methods based on photographic objective prism spectra can equally well be applied to CCD objective prism images. Some specific problems will arise because of the use of non-coinciding wavelength ranges but much higher accuracy and quantum efficiency parallel with linear energy response of CCD detectors are large gains.

In the above discussion we have referenced many classification methods. They were developed for different tasks and the used resolution as well as penetrating abilities are very differing. The problem of choosing optimal spectral resolution of objective prism CCD images aimed at a concrete spectral classification task should be set up. Under optimal resolution we mean the resolution which is as low as possible (to reach fainter stars) but which yet allows to perform a spectral classification with the accuracy required for the concrete task. The idea is to take one sample of stars observed with one relatively high spectral resolution, to classify these stars with a chosen classification method and to estimate the respective classification accuracy. Then the spectral resolution should artificially be decreased step by step and at every stage the accuracies of spectral classification should be estimated. We should stop the process at the stage when further deterioration of spectral resolution makes classification accuracy too bad for any reasonable classification task. The spectral resolution at the stage when the classification accuracy is yet good enough for the concrete classification task, should be considered to be optimal for this task. The results of this investigation can be applied to some astrometric, astrophysical and galactic studies.

Acknowledgements

We are very grateful to Ronald C. Stone of the U.S. Naval Observatory for sending us his code developed for computing refraction as well as for encouraging comments what made this investigation possible. We are thankful to a referee for his careful reading of the paper and for helpful suggestions. We are very grateful to Prof. M. Soffel who has suggested the idea of the present investigation and to Dr. K. Kurzynska for comments on this paper.