We have seen that all the objects studied in the present paper
did show changes in the shapes
of one or more of their P Cygni profile up to levels of 
over timescales of years. It is remarkable that in two (NGC 40 and BD +30
3639) out of the four stars presently studied, the addition of few spectra
has led to the discovery of changes in the shapes of their P Cygni
profiles, while they did not exhibit changes in the spectra available at
the time of Paper I. Combining the results of Paper I with those
of the present paper, we find that seven out of fourteen observed
central stars did exhibit changes in their P Cygni profiles over
timescales of years.
They are: NGC 40, NGC 1535, NGC 2392, NGC 6543, NGC 6826, IC 4593
and BD + 30 3639. Those where changes have not been revealed
are: NGC 246, NGC 6210, NGC 6572, NGC 7009, IC 418, IC 2149 and
Lo 8. In NGC 1535 and NGC 6572 observations
only in two epochs are available, and this is also true
for NGC 246, considering that two of the three available spectra
are "adjacent''.
We then suspect that, if more data would have been secured, the phenomenon of the P Cygni line variability in central stars within timescale of years might have revealed to be more common than the 50% resulting from the presently available data.
We have made an effort to clarify whether the data indicate any
correlation of the observed variations of the P Cygni profiles
with the fundamental parameters of the stars. We did not see
any indication of this type. Specifically the property
"variations detected vs. variations not detected'' does not
exhibit any tendency to favour higher or lower values of the
following parameters: 
, 
, 
, 
,
of the wind.
Neither the available data suggest a tendency for objects closer
to the Eddigton limit to favour the variability of the P Cygni
profiles, as one might suspect. For instance the two O(H) central
stars of NGC 1535 and NGC 7009 are quite close to each other in the
diagram (McCarthy et al. 1990;
Mendez et al. 1988). Yet we noted variations in the first
star, not in the second one. As with the two WR stars BD +30 3639 and NGC
40, according to the recent detailed study by Hamann and collaborators
(see Hamann 1996), the first star is very close to the
Eddington limit, while NGC 40 is quite far from it. Yet both exhibit
variations in their P Cygni line profiles.
Certainly the amount of data secured, while useful to prove the existence and extent of the phenomenon, is not sufficient to clarify any possible link to the fundamental physical parameters.
We observed variations in the edge velocity of CSPN up to about 200
km/s, i.e. 10% of 
. This occurred in the resonance lines
NV in NGC 6826 (Paper I), CIV in NGC 6543 and NGC 6826,
and in the subordinate line OV in NGC 6826.
In population I OB stars velocity variations have been seen all across the wind profile, i.e. from zero velocity to the edge velocity, with shifts in velocity on a 10% level (16% in the case of HD 203064) (Kaper et al. 1996). All of them have been interpreted as manifestations of the DAC (Discrete Absorption Component) phenomenon. Due to the more pronounced saturation effects in the resonance P Cygni lines of CSPN relative to population I OB stars and to the lower signal-to-noise of CSPN spectra, we were not able to detect the DAC phenomenon in CSPN. We cannot then conclude whether also in the CSPN the edge velocity variations are manifestations of the DAC phenomenon.
As with variations in the intensity level inside the profiles, we observed changes from 10 to 30%. This is comparable with the analogous variations in OB stars (cf. Kaper et al. 1996). We recall however that in the case of population I OB stars, the wind variability concentrates on the absorption part of the P Cygni profiles, whereas the emission peak is constant in time.
The IUE data base does not allow to investigate whether the important smaller timescale variations (DAC) recognised to occur in all the population I OB stars are indeed present in the quite higher gravity CSPN. The new generation STIS instrument to be installed on board of HST will provide enough throughput and spectral resolution to pursue this important issue.
In conclusion we have revealed that significant variations in the shape of the P Cygni profiles over timescales of years are a common phenomenon in CSPN. But to further investigate the real meaning of these variations along the evolution history of a central star, quite more data are needed.
To examine how much the detected variations can affect the mass loss of
the star we have considered the case of NGC 1535. This star has P Cygni
profiles only in NV and OV (Cerruti-Sola & Perinotto
1989). While the profile of OV did not change appreciabily in
the examined spectra (December 1980, March 1981), the NV profile
did vary in its absorption component (see Paper I). Using the
SEI method (Lamers et al. 1987) we have fitted both NV
profiles. From the fitting parameters we have estimated 
,
where 
is the mean ionization fraction across the wind
whose mass loss rate is 
. The above quantity
did increase by a factor 
. This does not necessaryly imply a
corresponding variation in the mass loss rate, also considering the
constancy of the OV profile. In Fig. 10 (click here) the observed
profile of March 1981 is compared with the best fit profile (thick line) and
with a model profile obtained
decreasing only the most important parameter, i.e the total optical depth
in the wind, from 8 to 2 (thin line). The other parameters are:
, 
, 
,
, 
(cf. Lamers et al. 1987).
To understand the meaning of the above
result it is essential to have an accurate knowledge of the ionization
structure, whose determination is quite difficult and out of the
purposes of the present paper.
Figure 10: SEI model fit (thick line) of SWP 13495 spectrum of NGC 1535.
For the meaning of the thin line see text
Acknowledgements
We are grateful to the referee, L. Kaper, for useful criticisms and suggestions also with the analysis procedure.