4. Discussion

[11]  The calculated thresholds for a given month vary in a wide range so the standard deviations exceed mean values, therefore we were forced to consider 5 (or 10) the smallest values to specify the threshold. However, these (the smallest thresholds) turn out to be relatively low (Table 2), and this looks rather surprising. Some explanations for this effect may be proposed. During our analysis the storms were not distinguished by local time of their onsets. However, the dependence of negative storm onsets on local time is well known. The disturbances most frequently begin in the night-early morning LT sectors, and they are rare during daytime hours [Mednikova, 1957; Prölss and von Zahn, 1978]. This is due to the interaction of background and storm induced thermospheric circulation [e.g., Prölss, 1995, and references therein]. Therefore the ionospheric effect of morning-daytime geomagnetic disturbances may be delayed until nighttime hours when the direction of the background meridional wind changes for the equatorward one, while nighttime geomagnetic disturbances appear in the F2 layer with much shorter time delay. This is one of the reasons for large scatter in time delay between geomagnetic and ionospheric storm onsets (see earlier). Analysis of the smallest thresholds (Table 2) (for instance, 4.70 (Dourbes, March), 4.75 (Tomsk, November), 5.80 (Slough, March), 5.60 (Moscow, February)) has shown that they are due to the following. First, the ionospheric storm may begin with a small (  3 hours, the span for ap index determination) delay with respect to the geomagnetic one, the previous 24-hour period being very quiet (small ap indices). Second, the geomagnetic disturbance may have taken place during the previous day, but because of poleward thermospheric circulation the disturbed neutral composition was restricted to high latitudes [Prölss and von Zahn, 1977] and the ionospheric storm did not begin until the nighttime hours, as mentioned earlier; again the previous 24-hour period was very quiet. In fact, this implies that once the composition perturbation (the disturbance bulge) has been generated, it is pushed around by winds and may move back and forth in latitude [Prölss, 1995]. This effect was confirmed by the storm simulation of Fuller-Rowell et al. [1994] as well as by ESRO 4 data analysis [Skoblin and Forster, 1993]. So, the ionospheric disturbance (of course, with smaller magnitude) may appear at the same location in 24 hours under magnetically quiet conditions. Such a case seems took place at Moscow on 16 February 1963 when the ionospheric disturbance occurred practically under quiet conditions (low threshold of 5.2) but after a preceding prolonged geomagnetic disturbance. The effect of an increase in the interhour correlation coefficients for deviations df_oF2 separated by 24-hour interval was mentioned earlier [Mikhailov, 1990].

[12]  It may seem that small calculated disturbance thresholds (Table 2) present exotic cases of ionospheric storms, therefore the analysis was repeated for strong storms corresponding to daily Ap 30. Seasonal (winter/summer) difference in the thresholds takes place in this case as well. For instance, at Moscow the winter thresholds are 17.9 for December and 14.5 for January, while in summer they are 26.9 for June, 24.4 for July, and 26.5 for August. Similar seasonal difference takes place if the thresholds are calculated over 10 (rather than 5) the smallest values. Therefore seasonal (winter/summer) difference in the disturbance thresholds is a real feature of the F2 -layer negative storms which may be explained by seasonal variations of neutral temperature and thermospheric circulation leading to changes in neutral composition.

[13]  Another interesting result of our analysis is the equinoctial relative minimum or plateau in the threshold annual variations (Figure 1). In fact, one should speak about a plateau if to delete the extreme low points in March which present special storm cases discussed earlier. This equinoctial plateau may be related with the winter/summer transition in the thermosphere manifested by day-to-day changes of the meridional wind at the F2 -region heights [Mikhailov and Schlegel, 2001] as well as with equinoctial transitions observed in the lower thermosphere [Shepherd et al., 1999; Shiokawa and Kiyama, 2000]. Both analyses revealed day-to-day changes in the atomic oxygen abundance during the transition periods, and this may help understand the threshold lowering effect during the equinoxes. General increase of the thermospheric neural temperature from winter to equinox and further to summer provides a steady increase of the threshold as discussed earlier. However, if a geomagnetic storm occurs under summer-type thermospheric circulation accompanied by a decrease in the atomic oxygen [O] abundance, this should decrease the geomagnetic threshold. Indeed, in this case, less O/N₂ decrease is needed to overcome the same N_mF2 disturbance threshold (40% in our case), and this corresponds to lower level of geomagnetic activity. Days with winter-type thermospheric circulation correspond to increased atomic oxygen [O] abundance and positive N_mF2 disturbances [Mikhailov and Schlegel, 2001] are not considered in this paper.