Global morphology of night-time NmF2 enhancements

An overall statistical study of night-time enhancements ofNmF2 has been carried out. All available foF2 observations since 1955 at 53 ionosonde stations distributed worldwide in the latitude rangeφgeom = 15 − 60 were used in the analysis. More than 200 000 station-nights of data were analysed. This large data base allowed us to study seasonal, solar cycle and spatial variations of the NmF2 nighttime enhancements. Both pre-midnight and post-midnight NmF2 peaks demonstrate distinct variations with geophysical conditions, indicating different physical mechanisms responsible for their formation.


Introduction
Night-time enhancements of electron concentration are typical phenomena of the F2-layer in middle latitudes that can be observed both in the maximum of electron concentration in the F2-layer (NmF2) and total electron content (TEC) (Arendt and Soicher, 1964;Evans, 1965;Da Rosa and Smith, 1967;Bertin and Papet-Lepine, 1970;Young et al., 1970;Titheridge, 1973;Tyagi, 1974;Davies et al., 1979;Ivanov-Kholodny and Mikhailov, 1986;Balan and Rao, 1987;Joshi and Iyer, 1990;Lois et al., 1990;Jakowski et al., 1991;Jakowski and Förster, 1995;Mikhailov and Förster, 1999;Mikhailov et al., 2000a, b).Although NmF2 and TEC variations usually exhibit similar behaviour in the night-time ionosphere, there are some differences in the occurrence of these variations and in the treatment that has been applied to their analysis (Tyagi, 1974;Lois et al., 1990).Two peaks (preand post-midnight) in the NmF2 and TEC daily variations are considered here, analysing such characteristic parameters as: occurrence probability, time of occurrence and amplitude, to-Correspondence to: M. Herraiz (mherraiz@fis.ucm.es)gether with their geographic, seasonal and solar activity dependences.
Different mechanisms have been proposed to explain the observed variations.However, the results of previous analyses have often been contradictory.While Rao et al. (1982) found that for stations in Asia the amplitudes of both enhancements are higher during solar maximum; according to Titheridge (1973) and Tyagi (1974), the opposite tendency is observed.Jakowski et al. (1991) pointed out a higher occurrence probability of night-time enhancements of electron concentration during winter and solar minimum for observations in Havana, and a reversal of this behaviour during solar maximum with higher probabilities in summer, which agrees with Rao et al. (1982) but contradicts Titheridge (1973) and Tyagi (1974).Different conclusions are also reached when the local time of occurrence and the duration of NmF2 enhancements are studied.
It should be stressed that in previous publications devoted to this problem the authors considered either one or two increases of electron concentration at night but in quite different ways.In some papers no distinction was made between pre-and post-midnight enhancements and only one peak (the one with the higher amplitude) was considered, even though the presence of both was mentioned (Young et al., 1970;Titheridge, 1973;Tyagi, 1974;Balan and Rao, 1987;Joshi and Iyer, 1990).Other researchers treated these increases separately in their statistical studies, but only one was considered for each night (Jakowski et al., 1991;Jakowski and Förster, 1995).In our study both increases are considered separately, following a method of data analysis similar to one used previously by Mikhailov et al. (2000a, b).Comparisons with results of previous studies based on other methods should be made with caution.
Another important point to be considered is that previous analyses are quite limited either geographically or temporally, lacking complete seasonal and solar activity coverage or being restricted to a single station.Only Mikhailov et al. (2000a) made an extensive study over four solar cycles for four stations in the Eurasian region.Consequently, the contradictions between the results of previous analyses may be due to geographical variations in the morphology of enhancements and to the different methods of data treatment used by the different authors.These contradictions point to the necessity for a broader investigation of the phenomenon that considers a wide range of seasonal and solar conditions and mid-to low-latitudes.
Various physical processes have been suggested to explain the formation of night-time electron concentration enhancements.The most important are: (1) plasma fluxes from the plasmasphere (Hanson and Ortenburger, 1961;Evans, 1965Evans, , 1975;;Titheridge, 1968;Jain and Williams, 1984;Förster and Jakowski, 1988;Jakowski et al., 1991;Jakowski and Förster, 1995;Mikhailov and Förster, 1999); (2) raising of the F2layer to higher altitudes (where the recombination rate is smaller) by electric fields and thermospheric winds (Young et al., 1970;Standley and Williams, 1984;Hedin et al., 1991;Titheridge, 1995;Mikhailov et al., 2000a, b).Other less important processes are plasma transfer from conjugate points (Wickwar, 1974;Balan et al., 1994) and night-time ionization at the top of the ionosphere at high latitudes (Titheridge, 1968;Leitinger et al., 1982).All of these physical processes are also among the main mechanisms that generate the night-time F2-layer.Therefore, a proper understanding of the night-time F2 enhancements can make a significant contribution to knowledge of how the night-time F2-layer is formed.
The main goal of this paper can be defined as a morphological study of the night-time NmF2 enhancements using all available worldwide, ground-based ionosonde observations for the last 3-4 solar cycles in both hemispheres.No detailed physical interpretation of the morphological features revealed is attempted here.This will be done elsewhere.

Data analysis
The method used to select and specify each of the two nighttime peaks is similar to the one described by Mikhailov et al. (2000a, b).All hourly values of the F2-layer critical frequency (foF2) available at the World Data Center for STP at Chilton (WDC) from 53 ionosonde stations for the period since 1955 were analysed.This covers four solar cycles at many of the stations used.The locations of the ionosonde stations selected cover the 15 • to 60 • range of geomagnetic latitude in both hemispheres, and all longitudinal sectors.Coordinates of the stations used are listed in Table 1.Night-time observations were selected according to solar zenith angle.Each NmF2 enhancement was identified by the presence of a relative maximum in hourly foF2 values, as shown in Fig. 1.Maximum electron concentration, NmF2, is known to be related to the critical frequency, foF2, by the expression Each NmF2 peak revealed in this way was specified by three parameters: occurrence probability, amplitude (relative to the minimum electron concentration for the particular night) and local time of occurrence.Amplitude of the enhancement was defined as Amplitude where foF2 max1, 2 are the critical frequencies at the two possible peaks: pre-midnight (foF2 max1 ) and post-midnight (foF2 max2 ), and foF2 min is the minimum critical frequency through the time between sunset, defined by solar angle > 95 • , and 04:00 LT.
One-hour gaps in the data were filled in using neighbouring values.Nights with gaps lasting two or more hours were rejected.To avoid the effects of solar illumination during summer nights, only periods when the solar zenith angle was greater than 95 • were considered.Only magnetically quiet days (A p < 12) were analysed, though night-time enhancements have also been observed during disturbed periods (Mikhailov and Förster, 1999).These filtering processes reduced the number of available observations (nights) to those shown in Table 1.

Results
Two different kinds of analysis were carried out for the selected stations.First, seasonal and solar variability were studied in different bands of geomagnetic latitude.Sec-ond, spatial variations of selected features were analysed.It should be stressed that only by using a sufficient amount of data for each of the conditions considered is it possible to obtain reliable results.This was achieved by using 53 stations, most of which cover a wide temporal range.Even after fil-tering for disturbed days and data gaps, the total amount of data is over 200 000 station-nights.This gives an indication of the reliability of the results obtained.The main results are summarized in Tables 2 and 3.

Solar cycle and seasonal variation
To analyse the seasonal and solar cycle variations of the six parameters selected (occurrence probability, relative amplitude and local time of occurrence of each of the two peaks), nights were classified according to their season and solar activity level.Average values of the parameters for each of the gradations considered are shown for the first (pre-midnight) enhancement (Figs.2a-c) and for the second (post-midnight) enhancement (Figs.3a-c) for all 53 stations grouped into six geomagnetic latitude ranges.The corresponding numerical values are listed in Table 2. Data being grouped by latitude allows us to study the latitudinal dependence of seasonal and solar cycle variations.

First peak (pre-midnight)
The occurrence probability of the first peak (Fig. 2a) shows a clear seasonal dependence, tending to be higher in winter than in summer.There is also a clear dependence on solar activity level, with a higher probability during the years around solar minimum.However, in summer the peaks are more frequent during solar maximum.This is related to an upsurge in the occurrence probability associated with the months around the summer solstice (June-July in the Northern Hemisphere, December-January in the Southern).This was noted earlier by Mikhailov et al. (2000a) for the Northern Hemisphere.The effect is especially pronounced at high latitude stations.A latitudinal variation can be seen in Fig. 2a, with maximum occurrence probability at lower geomagnetic latitudes.
The variation of the relative amplitude (N peak /N min ) for the first peak (Fig. 2b) does not show such a clear pattern as does that of the occurrence probability.The graphs show a complicated pattern; the amplitude of winter peaks increases along with solar activity, whereas in summer the behaviour is the reverse.The largest peaks occur at low latitudes, especially in winter and near solar maximum.Large amplitudes are also observed during equinox at lower latitudes.It should be noted that relative amplitudes usually do not exceed 3.5, except for lower geomagnetic latitudes (see later).
The third characteristic parameter studied (Fig. 2c) is the local time of occurrence of the peak.At all latitudes, the time of occurrence tends to be later in summer than in winter.At middle latitudes, the enhancements occur later in the evening as solar activity increases, but no clear seasonal variation can be noted.The opposite behaviour is found at low latitudes, with later enhancements during solar minimum.At high latitudes the variability is more complicated, with different behaviour depending on the season.In winter the enhancements occur later at low solar activity level, while in summer the same happens during solar maximum.Early enhancements are found at high latitudes during equinoxes.
The time of occurrence shows small changes with geomagnetic latitude.Very small variations of this parameter are found for different seasons and solar activity levels, the time being always close to 21:50 LT.As it will be shown later, the occurrence time of both peaks mainly reflects its dependence on longitude.

Second peak (post-midnight)
Night-time enhancements of electron concentration are more common after midnight than before midnight, as is clearly shown in Fig. 3a.Post-midnight enhancements occur on about 80% of all the nights, while the pre-midnight occurrence probability is about 50%.The occurrence probability shows a clear seasonal dependence at all latitudes, with more enhancements in winter than in summer.Variations with solar activity are less clear, though in general a higher occurrence probability is observed during solar minimum.
In general, at middle latitudes the amplitude of the postmidnight peaks (Fig. 3b) is slightly smaller than that of the pre-midnight ones, with values being less than three.At higher and lower latitudes we don't see any extreme behaviour of these amplitudes (see later).
In general, the highest relative amplitudes are found in winter and during solar minimum.A physical mechanism to explain this behaviour was proposed by Mikhailov et al. (2000b).This pronounced seasonal and solar variability occurs at high and middle geomagnetic latitudes only.
Post-midnight enhancements usually occur between 01:50-03:50 LT and are later in winter than in summer.No clear dependence of any parameter on solar activity is apparent, except for a small decay of amplitude during solar maximum, which was explained in detail by Mikhailov et al. (2000b).A small variation of these parameters with geographic location occurs and is described later.

Geographical morphology of night-time NmF2 enhancements
A similar analysis was applied to each of the 53 ionosonde stations (Table 1) to determine spatial variations of the F2layer night-time electron concentration enhancements.The variations were analysed and presented in geomagnetic coordinates.
Figures 4a-c and 5a-c show the spatial variations of each of the previously defined parameters in panels displaying changes with season and solar activity level.Numerical values are listed in Table 2.In general, latitudinal variations are the most important and longitudinal variations are negligible.Only the time of the peak's occurrence shows a small longitudinal dependence displayed in Figs.4c and 5c.No latitudinal variation of time of occurrence is apparent, and there is no significant difference between hemispheres for any of the parameters studied.Only cases with more than 100 valid nights for a station under a particular solar activity condition and season were used for the graphs in order to provide statistical reliability.

First peak (pre-midnight)
Figures 4a-c show the geographic morphology of the first enhancement of electron concentration.The occurrence probability (Fig. 4a) shows a maximum around φ geomag = 35 • and a noticeable latitudinal variation.The lowest occurrence probabilities take place at high latitudes.This behaviour is most pronounced in winter, while in summer, and especially under solar maximum conditions, it distorts at high latitudes due to the influence of the summer solstice feature mentioned earlier.
The amplitude (Fig. 4b) shows quite a different type of morphology, with a relative minimum around 40 • and very high values in lower latitudes.The minimum in latitudinal variation shifts equatorward in summer and poleward in winter.Not much can be said about the latitudinal variation of the local time of the first peak's occurrence.However, though smaller and less clear than in the case of the second peak, an irregular longitudinal variation is apparent (Fig. 4c).

Second peak (post-midnight)
The geographic morphology of the second peak of electron concentration is displayed in Figs.5a-c.The occur-rence probability (Fig. 5a) shows a clear maximum around φ geom = 40 • for all solar activity conditions, with values very close to 100% in winter.For other seasons, occurrence probability is higher around solar minimum.A great reduction in occurrence probability is seen around φ geom = 25 • .As in the case of the first peak, the amplitude of the postmidnight enhancement shows the reverse behaviour, with a minimum in amplitude around φ geom = 40 • and the greatest amplitudes at latitudes < 25 • .This is not surprising, since the occurrence probability depends only on the existence of relative increases in electron concentration, while the amplitude is strongly related to the minimum electron concentration during the particular night.
Finally, the time of occurrence shows both a latitudinal and longitudinal dependence.Latitudinally (see Fig. 3c), the variation of the occurrence time of the second enhancement is almost constant, while longitudinally (Fig. 5c) it shows a sinusoidal type of variation with period 360 • .Stations at low latitudes (φ geom < 25 • ) show quite a different behaviour for both enhancements, indicating a different mechanism of formation at these latitudes.

Discussion
The method of data analysis used here allows one to investigate the NmF2 night-time enhancements morphology, revealing that the two peaks have different behaviours and indicating different formation mechanisms.From this perspective it follows that previous studies analysing only one peak omitted the characteristics of the "weaker" peak, which will be pre-or post-midnight, depending on season, solar activity and geographical location.This is believed to be the main reason for the differences between our results and those of other authors.Balan and Rao (1987) found that for low latitudes and winter solar minimum two peaks are common, while only a post-midnight peak is present at middle and high latitudes.This contradicts our results.However, they only considered one peak each night, missing the presence of pre-midnight peaks at low latitudes, whose amplitude is small under these conditions.The difference in methods of analysis does not allow us to make a comparison with their results for geographical variation.Jakowski et al. (1991) found for observations at Havana a higher occurrence probability in winter during solar minimum and in summer during solar maximum.The joint consideration of the occurrence probabilities of both peaks in our graphs (Figs.2a and 3a) explains this change, which is due to the increased occurrence of the first peak at high solar activity during summer solstice.They also found that the occurrence of two peaks on the same night occurs only during winter and solar maximum.This can be explained by the fact that the amplitude of both peaks is similar under solar maximum conditions for middle latitude stations like Havana, which makes both enhancements "visible" to their method of data analysis, while under other conditions only the (larger) second peak is found.Other authors also miss one of the peaks for similar reasons.
The distinct and systematic behaviour of each of the two enhancements indicates that different physical mechanisms lead to their formation.For the first (pre-midnight) peak several mechanisms appear to act, depending on season and solar activity.At middle latitudes the highest amplitudes are found during summer and low solar activity.This has been attributed to a collapse in the F2-region as electron temperature decreases after sunset, producing large downward ion fluxes.This, along with the contribution of equatorward meridional thermospheric winds uplifting the F2-layer to regions with low recombination rates, causes the NmF2 increase.
During solar maximum, however, the amplitude of the summer peak decreases.The occurrence probability increases during summer solstice months, especially during solar maximum.This effect was attributed by Mikhailov et al. (2000a) to a direct solar photoionization as the night-time ionosphere rises to higher sunlit altitudes.This effect is especially apparent at high latitudes.The time of occurrence of these summer solstice peaks shifts to later hours as solar activity increases.This is consistent with the hypothesis of a photoionization origin.It should be noted that higher amplitudes are often associated with low occurrence proba-bility, while the highest amplitudes are always observed at lower latitudes.This may indicate that the relative amplitude of night-time peaks is controlled mainly by the value of minimum electron concentration used for a scale (N min ), which determines the background level above which the enhancements are counted.This was checked by studying the dependence of amplitude on N min .The importance of this background level specification was pointed out by Joshi and Iyer (1990).It may also be the cause of high amplitudes found at lower latitudes for both peaks.
The formation of pre-midnight peaks in winter is due mainly to a strong equatorward thermospheric wind raising the F2-layer to heights with a lower recombination rate (Young et al., 1970;Standley and Williams, 1984;Mikhailov et al., 2000a, b).
Large amplitudes at high-latitudes during equinox can be associated with the highest efficiency in the interaction between the Earth and the solar wind that occurs during these periods (Hargreaves, 1992).The behaviour found at lower latitudes is opposite to that commonly found at middle latitudes and may explain the contradiction between our results and those by Lois et al. (1990) and Jakowski et al. (1991) at Havana, Cuba.
The season-dependent behaviour of the first peak's amplitude that can be seen clearly in Fig. 2b, has a different and unknown physical mechanism.The second enhancement shows a clear seasonal variation, being much more likely to occur in winter than in summer, both around solar maximum and solar minimum.The highest amplitudes also occur in winter, being higher during solar minimum.As with the first peak, the highest amplitudes occur at lower latitudes, which also demonstrate different behaviour compared with middle and high latitudes.
The physical mechanism proposed to explain the formation of the second enhancement is again electron fluxes from the plasmasphere.However, observed fluxes are smaller than those required to explain the peaks (see Mikhailov et al. (2000b) and references therein).This problem was solved by Mikhailov and Förster (1999) and Mikhailov et al. (2000b), who proposed that the observed night-time NmF2 variations were due to the uplifting of the F2-layer by the equatorward thermospheric winds, along with the observed night-time plasmaspheric fluxes into the F2-region.The increase in the night-time height of the maximum electron concentration in the F2-layer, hmF2, and the corresponding decrease in the recombination rate is then sufficient to explain the night-time NmF2 increases with the observed relatively small plasma fluxes from the protonosphere.
The critical role played by thermospheric winds in the development of night-time peaks of NmF2 is evident from the geographic variation of the parameters studied, which show relative extremes (maximum in occurrence probability, minimum in amplitude) around φ geom ≈ 40 • .Vertical drift of plasma (W ) due to the meridional component (V nx ) of the neutral winds depends on magnetic inclination (I ) as which is a maximum when I = 45 • .Additionally, the flux tube content is proportional to L 4 , which gives a maximum of electron content in the tube at φ geom ≈ 60 • (Carpenter and Park, 1973) due to the partial filling of tubes with L > 3.
Tubes at lower latitudes are filled, but their volume is insufficient to produce the necessary plasma influx to the F2-layer.The action of meridional thermospheric winds at the equator may be another cause for the different behaviours, especially for the high amplitudes observed (Titheridge, 1995).A better knowledge of the global pattern of thermospheric winds is desirable in order to improve the understanding of the mechanism of NmF2 peak formation.

Conclusions
A detailed study of the morphology of NmF2 night-time enhancements was carried out at 53 ionosonde stations worldwide for different seasons and different levels of solar activity.The main results of our analysis are the following: 1.There are two distinct (pre-and post-midnight) NmF2 peaks, which can occur for any season and solar activity level.
2. All the characteristics of the night-time enhancements that were analysed demonstrate a pronounced dependence on geomagnetic latitude, with distinctive behaviour at lower (φ geom < 25 • ) latitudes, indicating different formation mechanisms.
3. In general, the occurrence probability is higher for the second (post-midnight) peak.
4. The occurrence probability of the first peak shows a clear seasonal dependence, with a maximum in winter solar minimum.The greatest occurrence probability is at φ geom ≈ 35 • , and the least is at high latitudes (except for summer and high solar activity, due to the summer solstice occurrence upsurge).
5. The occurrence probability of the second peak shows a similar seasonal pattern, but without a dependence on solar activity.The greatest occurrence probability is observed at φ geom ≈ 40 • , and the least at high latitudes.
6.In general the amplitude N peak /N min of the first peak is higher than for the second peak.The largest enhancements are observed at low latitudes in winter during solar minimum.Appreciable enhancements also take place at high latitudes during equinoxes.At middle latitudes peaks show various patterns depending on the season.
7. The amplitude of the post-midnight peak shows a clear seasonal and solar activity dependence, being larger in winter and during solar minimum.In general, postmidnight enhancements of electron concentration are smaller than the pre-midnight ones.The geographical morphology of the NmF2 enhancements shows a relative minimum around φ geom ≈ 35 • , with a large upsurge at lower latitudes.
8. The time of occurrence of the first peak is between 20:00 and 22:50 LT and shows a clear dependence on solar activity.Enhancements occur later at middle latitudes during solar maximum.The dependence of the time of occurrence on solar activity is different at different geomagnetic latitudes.A small longitudinal effect is also present.9.The time of occurrence of the second peak is between 01:50 and 03:00 LT.Enhancements occur later in winter than in summer.There is a small but distinct longitudinal variation.

Fig. 1 .
Fig. 1.Definition of each of the two peaks and their characteristic parameters from the original foF2 data.Shadowed rectangles show sunlit time.

Fig. 2 .
Fig. 2. Seasonal and solar variations of characteristic parameters of the first (pre-midnight) peak.Solar activity is classified in three levels: high solar activity (High), medium solar activity (Medium) and low solar activity (Low).Seasonal variability is arranged in three categories: winter, summer and equinox.See the text for details.(a) Probability of occurrence, (b) relative amplitude, (c) local time of occurrence; for 6 geomagnetic latitude ranges.Bars represent means.Negative latitudes represent Southern Hemisphere positions.

Fig. 4 .
Fig. 4. Geographic variation with of characteristic parameters for the first peak (a) probability of occurrence, (b) relative amplitude, and (c) local time of occurrence with geomagnetic longitude, under different seasonal and solar conditions.Categories are the same as in Figs. 2 and 3.For example, "Low winter" refers to winter data under low solar activity conditions.Points indicate means.Bars indicate 95% confidence limits.

Table 1 .
List of stations and data samples used in alphabetical order

Table 2 .
Numerical results: mean, standard deviation and number of valid cases for each variable and condition Table2.Numerical results: Mean, standard deviation and number of valid cases for each variable and condition.

Table 3 .
Summary of results