4 Discusion

In Fig. 4 we plot monthly mean values of the solar radius given by the astrolabes of Rio de Janeiro, Antalya and Santiago. The monthly means are represented by the center of circles with diameters equal to their mean errors. For Rio de Janeiro they were computed from the homogeneous data set discussed above, and for Antalya from the individual daily measurements that are available at the home page of Tubitak Observatory (Golbasi et al. 2001). On average, each monthly mean is based on 273 and 43 CCD observations for Rio de Janeiro and Antalya respectively, and on 38 visual observations for Santiago. These figures and the dispersion of the circles and the magnitude of its diameters in Fig. 4 do not show higher precision of the CCD astrolabes over the visual one. For further discussions about visual and CCD observations of the solar radius see Wittmann (1997), Wittmann & Bianda (2000) and Noël (2001) (see also Laclare et al. 1999).

Considering that the observational period at Antalya is rather short in comparisson with those of the other two astrolabes, we have plotted also the curve of a least squares second order fit applied to the set of individual observations at Antalya. This fit, with a standard deviation of $\pm0\hbox{$.\!\!^{\prime\prime}$ }36$, gives a similar drift of the solar radius as that shown by the astrolabes of Rio de Janeiro and Santiago.

The curve in Fig. 4 that represents the variation of sunspot number is based on the smoothed monthly means of sunspot number available at the home page of SIDC.

There is a clear agreement between the variations in time of the homogeneous solar radius observed at Rio de Janeiro, the solar radius observed at Santiago and Antalya and the sunspot number variation. From our point of view, it is quite difficult to believe that the improvement of the internal consistency of the results of Rio de Janeiro after our procedure of homogenization, and its agreement with the results of Santiago and Antalya, could be due only to coincidence. Rather, we think that the data displayed in Fig. 3B show the real variation in time of the solar radius observed at Rio de Janeiro.

Figure 4 shows also that there are significant systematic differences between the mean values of the solar diameter obtained by the three astrolabes. The mean values for the periods of observations displayed in Fig. 4 are:

Antalya	Rio de Janeiro	Santiago
$958\hbox{$.\!\!^{\prime\prime}$ }57\pm0\hbox{$.\!\!^{\prime\prime}$ }01$	$959\hbox{$.\!\!^{\prime\prime}$ }76\pm0\hbox{$.\!\!^{\prime\prime}$ }01$	$960\hbox{$.\!\!^{\prime\prime}$ }56\pm0\hbox{$.\!\!^{\prime\prime}$ }01$.

Since the astrolabes of Antalya and Rio de Janeiro are equipped with CCD cameras, their results are independent of the observer. Therefore, one should expect a rather low difference in their mean results and not the large difference that they actually give. In our view, this confirms that solar radius measurements with CCD cameras are strongly affected by effects not intrinsic to the Sun (Wittmann 1997, 2000; Noël 2001). The results are very sensitive to instrumental modifications, as shown in Fig. 3A by the large jump in solar radius observed with the CCD astrolabe of Rio de Janeiro after a change in its instrumental system.