This paper presents new results about Triton's atmospheric structure f
rom the analysis of all ground-based stellar occultation data recorded
to date, including one single-chord occultation recorded on 1993 July
10 and nine occultation lightcurves from the double-star event on 199
5 August 14. These stellar occultation observations made both in the v
isible and in the infrared have good spatial coverage of Triton, inclu
ding the first Triton central-flash observations, and are the first da
ta to probe the altitude level 20-100 km on Triton. The small-planet l
ightcurve model of J. L. Elliot and L. A, Young (1992, Astron. J. 103,
991-1015) was generalized to include stellar flux refracted by the fa
r limb, and then fitted to the data. Values of the pressure, derived f
rom separate immersion and emersion chords, show no significant trends
with latitude, indicating that Triton's atmosphere is spherically sym
metric at similar to 50-km altitude to within the error of the measure
ments; however, asymmetry observed in the central flash indicates the
atmosphere is not homogeneous at the lowest levels probed (similar to
20-km altitude). From the average of the 1995 occultation data, the eq
uivalent-isothermal temperature of the atmosphere is 47 +/- 1 K and th
e atmospheric pressure at 1400-km radius (similar to 50-km altitude) i
s 1.4 +/- 0.1 mu bar. Both of these are not consistent with a model ba
sed on Voyager UVS and RSS observations in 1989 (D. F. Strobel, X. Zhu
, M. E. Summers, and M. H. Stevens, 1996, Icarus 120, 266-289). The at
mospheric temperature from the occultation is 5 K colder than that pre
dicted by the model and the observed pressure is a factor of 1.8 great
er than the model. In our opinion, the disagreement in temperature and
pressure is probably due to modeling problems at the microbar level,
since measurements at this level have not previously been made. Altern
atively, the difference could be due to seasonal change in Triton's at
mospheric structure. (C) 1997 Academic Press.