The Sun observation by astrolabe has, nowadays, two distinct goals: solar positional astronomy and heliography. In what follows we are concerned only with the determination of corrections to the placement of the reference system (usually the one in which the astrolabe star observations are done) and corrections to the Earth's orbital parameters. All the observations discussed here are placed on the FK5/J2000 reference system.
In a previous paper (Penna et al. 1996 - hereafter Paper I) we presented the results from the analyses of the solar observations with astrolabes, which were obtained at Observatoire de La Côte D'Azur, Observatório Abrahão de Moraes (OAM) and Observatório Nacional (ON). The data set contains 6273 individual observations of transits of both the upper and lower solar limbs, spanning 18 orbital revolutions acquired with nearly uniform distribution, due to the geographical location of the observing sites.
The observed zenith distances range from 
to 
.
The data were divided in 20 subsets, according to the observatory site and
observing zenith distance. Also substitutions of the employed prism or
filter plate, at a given zenith distance and observatory, demanded
the establishment of separate subsets. This last subdivision includes
2 subsets at 
and 
each at OCA, 3 subsets at 
at OAM, and 3 subsets at 
at ON
(Laclare et al. 1981, 1982, 1983; Leister 1989;
Penna 1982).
The investigation of the individual sub-campaigns at each zenith distance
and at each site has led to devise a scale of weights
depending uniquely on the observational zenith distance.
To the subsets that include the smaller zenith distances, 
,
unit weight were assigned. This scheme holds
even for the Southern hemisphere results, where the use of transmission
prisms in the earlier observations is compensated for
by their complementary role in the solution. An intermediate
group, with z ranging from 
to 
, was given half
weight and the 
to 
subsets received one fourth weight.
Table 1: Standard weighted global solution
The results of the Standard Weighted Global Solution presented a 
standard deviation for an observation of unit weight. There were 26
unknowns to be solved for. The typical formal error for them is 
and the pairwise correlations among them are inferior to 0.65 in absolute
value, except for the pair mean longitude and equinox corrections, where it
attains 0.96. These enabled us to assert the significance of the solution.
However, the large corrections found to the obliquity of the ecliptic
(
) and to the eccentricity (
), in comparison to their formal errors, deserve a more detailed
discussion.
In Paper I a global analysis of the observations was presented. Because of the observational conditions, however, the distribution of the observations is not regular, and, moreover, the atmospheric conditions span a large range. In this paper, thus, we discuss the effects of such irregularities. Additional nutation terms are also investigated in order to better analyze the obliquity of ecliptic results.
To further probe on the stability of the results, we analyzed the distribution of the observations and the completeness of the set of unknowns relevant to the problem. Therefore, initially a number of strongly biased sets of data were tried, in order to directly verify the degree of dependence among the results. Then we added up to the complexity of the problem description, incorporating, by steps, terms to account for corrections to 1) the refraction constant; 2) the time variation of the obliquity; 3) azimuthal anomalies of the refraction; 4) a correction of the nutation constant.
This paper brings the results of such investigations, that can be extended and contribute to the effort of combining astrolabe observations in general.