As seen in the previous item, the quantity of data is overabundant, therefore it is possible to examine in detail the standing of the hypotheses implicitly adopted in building the Standard Weighted Global Solution. That is, the residuals from the solution behave as white noise, of null average and uniform variance.

Paper I brings the histograms of the residuals inside each subset. No important anomaly (skewness or curtosis) is apparent on the plots. In general, the central portion of the histograms depicts a normal distribution of zero average and uniform standard deviation.

There is, however, in the most critical cases - OCA at and of zenith distance - a significant tail of negative residuals. The fact corroborates the need to assign smaller weights to these subsets.

A possible time evolution of the residuals was also studied. Figures 1 (click here) to 14 (click here) show the plotting of the residuals against the Julian date of observation, for each subset. Altough some trends might be suggested, they do not repeat for neighbouring zenith distances nor for similar epochs. Their amplitude is allways smaller than the scattering inside a campaign (about 1.5 arcsec), thus, even if they were real their cause could hardly be disclosed and their effect is too small to trouble.

One of the most conspicuous of these trends is shown in Fig. 1 (click here). The downstepping of the average from 1983 onwards is probably associated with the tightening of the campaigns' scatter. The feature is believed to be due to an improvement of the observational routine and apparatus, leading to higher quality of the results. Figures 3 (click here) and 8 (click here) show a smooth downwards trend. Altough barely significant, vis-a-vis each campaign scatter, as discussed above, they might indicate a slow deformation of the corresponding prisms ( and ). A longer time interval would be required to confirm the feature.

**Figure 1:** Residuals Distribution / OCA

**Figure 2:** Residuals Distribution / OCA

**Figure 3:** Residuals Distribution / OCA

**Figure 4:** Residuals Distribution / OCA

**Figure 5:** Residuals Distribution / OCA

**Figure 6:** Residuals Distribution / OCA

**Figure 7:** Residuals Distribution / OCA

**Figure 8:** Residuals Distribution / OCA

**Figure 9:** Residuals Distribution / OCA

**Figure 10:** Residuals Distribution / OCA

**Figure 11:** Residuals Distribution / OCA

**Figure 12:** Residuals Distribution / ON

**Figure 13:** Residuals Distribution / OAM

**Figure 14:** Residuals Distribution / OAM

In some of the plots, for , specially in Figs. 5 (click here) and 7 (click here), it could be distinguished an upside loop for the residuals spanning the years 1985 to 1987 at the OCA. The changing in the method employed to measure the atmospheric parameters used for calculate the refraction could be a cause of such effect.

To check upon this possibility, we prepared a run of the Standard Weighted Global Solution, including as additional unknown an additive constant to the zenith distance. This new unknown applies for all the OCA subsets, between 1985 and 1987, for which . The results are presented in Table 5. There are no important differences on the results compared to the Standard Solution displayed in Table 1. The general standard deviation, however, drops from to , which is marginally significant and seems to confirm the effect. The additional unknown () takes the value of with standard deviation , i.e. it is hightly significant. Even so, as the reference system and Earth's orbit orientation parameters are not changed by the inclusion of the additional unknown and as its definition is empirical, we choose to rather not include it in the Standard Weighted Global Solution.

**Table 6:** The output from the modified models of the Standard Weighted
Global Solution. The additional unknowns (U1, U2, U3, U4, see text) enter in
each solution as follows:
1 - None, standard model.
2 - Unknown (i), constant of refraction for the three sites.
3 - Unknown (i), only for OCA.
4 - Unknown (ii)
5 - Unknown (iii), harmonic of azimuthal frequency 2*Z*.
6 - Unknown (iii), as solution 5, but only for OCA and
z 60.
7 - Unknown (iii), harmonic of azimuthal frequency 3*Z*.
8 - Unknown (iii), harmonic of azimuthal frequencies 2*Z* and 3*Z*.
9 - Unknown (iii), as solution 8, but only for OCA and
.
10 - Unknown (iv)

web@ed-phys.fr