In this paper, new sophisticated computing formulas for astronomical r
efraction have been developed based on the generator function method o
f atmospheric refractive integrals proposed by Yan & Ping (1995, AJ, 1
10, 934). The accuracies of the expressions have been numerically impr
oved about one order of magnitude in comparison with those commonly us
ed. A significant advantage of using a calibrated continued fraction m
apping function for the astronomical refraction correction is to exten
d the coverage of the formulas to rather lower elevation observations
with high accuracy. The corrections related to the finite distances of
objects have been considered in detail. We have briefly analyzed the
influences of the errors of atmospheric profile and of the deviation o
f atmospheric parameters. In order to match different requirements in
astrometry and geodesy, we have considered the corrections at both rad
io and optical frequencies, respectively. The dispersion of signals at
optical frequencies plays a primary role in astronomical refraction c
omputation. (C) 1996 American Astronomical Society.