Annales Geophysicae (2001) 19: 733–748 c ○ European Geophysical Society 2001

Earlier revealed morphological features of the foF2 and hmF2 long-term trends are interpreted in the scope of the geomagnetic control concept based on the contemporary F2-layer storm mechanisms. The F2-layer parameter trends strongly depend on the long-term varying geomagnetic activity whose effects cannot be removed from the trends using conventional indices of geomagnetic activity. Therefore, any interpretation of the foF2 and hmF2 trends should consider the geomagnetic effects as an inalienable part of the trend analysis. Periods with negative and positive foF2 and hmF2 trends correspond to the periods of increasing or decreasing geomagnetic activity with the turning points around 1955, and the end of 1960s and 1980s, where foF2 and hmF2 trends change their signs. Such variations can be explained by neutral composition, as well as temperature and thermospheric wind changes related to geomagnetic activity variations. In particular, for the period of increasing geomagnetic activity (1965–1991) positive at lower latitudes, but negative at middle and high latitudes, foF2 trends may be explained by neutral composition and temperature changes, while soft electron precipitation determines nighttime trends at sub-auroral and auroral latitudes. A pronounced dependence of the foF2 trends on geomagnetic (invariant) latitude and the absence of any latitudinal dependence for the hmF2 trends are due to different dependencies of NmF2 and hmF2 on main aeronomic parameters. All of the revealed latitudinal and diurnal foF2 and hmF2 trend variations may be explained in the framework of contemporary F2-region storm mechanisms. The newly proposed geomagnetic storm concept used to explain F2-layer parameter long-term trends proceeds from a natural origin of the trends rather than an artificial one, related to the thermosphere cooling due to the greenhouse effect. Within this concept, instead of cooling, one should expect the ther-mosphere heating for the period of increasing geomagnetic activity (1965–1991).


Introduction
The problem of long-term variations (trends) in the ionospheric parameters is widely discussed in recent publications (see reviews by Danilov, 1997Danilov, , 1998;;Givishvili andLeshchenko, 1994, 1995;Givishvili et al., 1995;Ulich and Turunen, 1997;Rishbeth, 1997;Danilov andMikhailov, 1998, 1999;Bremer, 1992Bremer, , 1998;;Upadhyay and Mahajan, 1998;Sharma et al., 1999;Foppiano et al., 1999;Mikhailov and Marin, 2000;Deminov et al., 2000;Marin et al., 2001).On the one hand, the world-wide network of ground-based ionosonde observations provides excellent experimental material for such an analysis, since many of the ionosondes have been operating for 3-5 solar cycles using one method of ionospheric sounding to obtain the set of main ionospheric characteristics.Most of these observations are collected and available from WDC-C at RAL (Chilton, UK) and from NGDC, Boulder, USA.On the other hand, after the model calculations of Rishbeth (1990) and Rishbeth and Roble (1992) who predicted the ionospheric effects of the atmospheric greenhouse gas concentration increase, researchers have been trying to relate the observed long-term trends in the ionospheric parameters to this greenhouse effect (Bremer, 1992;Givishvili and Leshchenko, 1994;Ulich andTurunen, 1997, Jarvis et al., 1998;Upadhyay and Mahajan, 1998).However, an analysis of many European ionosonde stations by Bremer (1998) and an analysis of a global set of ionosonde stations by Upadhyay and Mahajan (1998) has shown that the world-wide pattern of the F2-layer parameter long-term trends is very complicated and can hardly be reconciled with the greenhouse hypothesis.It should be stressed that different authors use different approaches to ex-tract long-term trends from the ionospheric observations and the success of analysis depends, to a great extent, on the method employed.The useful "signal" is very small and the "background" is very noisy, so special methods are required to reveal a significant trend in the observed foF2 and hmF2 variations.An approach being developed by Danilov andMikhailov (1998, 1999), Mikhailov and Marin (2000) and Marin et al. (2001) has allowed us to find systematic variations in foF2 and hmF2 trends unlike the other approaches (e.g.Bremer, 1998;Upadhyay and Mahajan, 1998), which result in a chaos of various signs and magnitudes of the trends at various stations.An application of this approach in foF2 trend analysis resulted in a geomagnetic control concept (Mikhailov and Marin, 2000) used to explain the revealed latitudinal and diurnal variations of the foF2 trends.The efficiency of this approach was also demonstrated by Marin et al. (2001) in the hmF2 trend analysis for many ionosonde stations in the Eurasian longitudinal sector.Briefly, the main results of the analysis by Mikhailov and Marin (2000) and Marin et al. (2001) are the following: 1.The foF2 trends demonstrate a pronounced dependence on geomagnetic (invariant) latitude with strong negative trends at high latitudes and small negative or positive trends at lower latitudes for the period of 1965-1991.Contrary to this, the hmF2 trends show no latitudinal dependence being positive at the majority of the stations analyzed.The foF2 and hmF2 trends are shown to be significant for most of the stations considered.
2. There are well pronounced (especially for foF2) diurnal variations of the trend magnitude, while seasonal variations are rather small and may be ignored compared to diurnal ones.
3. The foF2 trend analysis has shown that there exists periods with negative and positive foF2 trends, which correspond to the periods of long-term increasing/decreasing geomagnetic activity.In particular, the period of 1965-1991 corresponds to the increasing geomagnetic activity, while the geomagnetic activity was decreasing during the 1955-1965 period.
4. The geomagnetic control concept has been proposed to explain main morphological features of the foF2 and hmF2 trends revealed.This newly proposed geomagnetic hypothesis proceeds from a natural origin of the trends rather than an artificial one, related to the thermosphere cooling due to the greenhouse effect.
The aim of the paper is to provide further analysis and physical interpretation of the foF2 and hmF2 trends within the proposed geomagnetic control hypothesis.

Diurnal variations at different latitudes
The final version of the method used for the F2-layer parameter trends analysis is given by Mikhailov and Marin (2000), therefore, only a fragmentary description is presented here.
All available observations at about 30 European, North American and Asian ground-based ionosondes are used in the analysis by Mikhailov and Marin (2000), and Marin et al. (2001) to reveal foF2 and hmF2 trends.The stations are located between 38 Both models were used by Mikhailov and Marin (2000), and Marin et al. (2001) to find the slope K (in 10 −4 per year) of linear trends for each station, for 12 months, and 24 LT moments.Although we are aware of the seasonal variations in trends (Danilov and Mikhailov, 1999), the later analysis has shown that diurnal variations may be much stronger than seasonal ones.Therefore, we analyze annual mean trends for selected LT hours.Averaged over 12 months the δpF2 value is found and this value is considered to be the annual mean value used in the trend analysis.The use of Model 2 was an attempt to exclude the effect of geomagnetic activity after Bremer (1998) and Jarvis et al. (1998).But our analysis (Mikhailov and Marin, 2000;Marin et al., 2001) has shown that such an inclusion of Ap indices to the regression, in fact, does not remove the geomagnetic effect, but only contaminates the analyzed material.Therefore, Model 1 (regression with R 12 ) is used in further analysis.
The magnitude of revealed foF2 tends demonstrates strong diurnal variations depending on geomagnetic (invariant) latitude (Mikhailov and Marin, 2000).No systematic latitudinal variations were found for the hmF2 trends (Marin et al., 2001).Some examples of the foF2 and hmF2 trend diurnal variations are given in Fig. 1 for auroral station Sodankyla ( inv =63.59 • ), sub-auroral station Lycksele ( inv =61.46 • ), mid-latitude station Ekaterinburg ( inv = 51.45• ), and lower latitude station Alma-Ata ( inv = 35.74• ).These stations are in the list analyzed by Mikhailov and Marin (2000), and Marin et al. (2001).Observations for the 1965-1991 period were used in further analysis.As in Mikhailov and Marin (2000) the (m+M) year selection was used for the foF2 trend analysis, where (m) represents the years around solar minima and (M) represents the years around solar maxima; all years were used to analyze hmF2 trends (Marin et al., 2001).
The selected stations demonstrate different diurnal variation patterns of the foF2 trends (Fig. 1, left panel) which are believed (see later) to reflect different physical processes responsible for the F2-layer formation at different invariant latitudes.In general, as it was pointed by Danilov and Mikhailov (1999) and Mikhailov and Marin (2000), the foF2 trends are negative at high and middle latitudes with a tendency to be small or positive at lower latitudes (e.g.Alma-Ata, Fig. 1, bot- tom).Positive significant hmF2 trends for all LT are revealed at most of the stations considered (Marin et al., 2001), but at some stations, negative significant trends take place; therefore, an additional analysis is required to find out the reason.The Shimazaki (1955) formula which converts M(3000)F2 to hmF2, was used in our routine analysis.On the one hand, it was shown by Bremer (1992) and later confirmed by Marin et al. (2001) that hmF2 trends are not sensitive to the formula choice.On the other hand, an insertion of correction terms (depending on the foF2/foE ratio) to the Shimazaki (1955) formula may not be useful for the hmF2 trend analysis, as this ratio itself demonstrates longterm variations.The other problem with using foE is in the absence of observations on many stations as well as during nighttime hours.
The revealed foF2 and hmF2 trends may be explained in the framework of contemporary F2-layer storm mechanisms related to the thermosphere global circulation, neutral composition and temperature perturbations during disturbed periods.The relationship of the F2-layer trends with the longterm changes in the geomagnetic activity is clearly seen from diurnal variation of the correlation coefficients r(δfoF2, Ap 12 ) and r(δhmF2, Ap 12 ) shown in Fig. 2, which are used in further discussion.These variations are seen to repeat the corre-sponding diurnal variation of the foF2 and hmF2 trend magnitudes (Fig. 1), although the correlation coefficients (Fig. 2) are small and insignificant (open squares) at the chosen 95% confidence level for some periods of the day.Usually, as Fig. 2 shows, large correlation coefficients r(δfoF2, Ap 12 ) are significant at the 95% confidence level and the correlation may be of both signs depending on the latitude of the station considered.
Let us consider the obtained latitudinal and diurnal variations of the foF2 and hmF2 trends (Figs. 1, 2) in the framework of the geomagnetic control concept.

Lower latitudes
Positive foF2 and hmF2 trends are revealed both for day-and nighttime hours at the lower latitude station, Alma-Ata.An analysis of the F2-layer storm mechanisms for the lower latitude station Havana, with the same inv = 35 • (L = 1.5) as Alma-Ata, was made by Mikhailov et al. (1995).According to AE-C and ESRO-4 satellite observations, geomagnetic disturbances result in an increase in the atomic oxygen absolute concentration, presumably due to the disturbed thermospheric circulation and downwelling at low latitudes, while the R = (O/N 2 ) storm /(O/N 2 ) quiet ratio remains practically unchanged at the heights of the F2-region (Prölss and von Zahn, 1977;Skoblin and Mikhailov, 1996;Mikhailov et al., 1997).Using the well-known expression by Rishbeth and Barron (1960) NmF2 where ion production rate q m and linear loss coefficient β m are given at the F2-layer maximum, it was shown by Mikhailov et al. (1995) that where all concentrations are given now at a fixed height h 1 .This expression shows that NmF2 will increase provided that the absolute atomic oxygen concentration [O] increases, while [O]/[N 2 ] ratio may remain unchanged at any fixed level (the situation we have according to satellite observations at lower latitudes).Such [O]/[N 2 ] height variations are also confirmed by model calculations (Förster et al., 1999;Rishbeth and Müller-Wodarg, 1999).Thus, an [O] increase due to downwelling motion related to global storm circulation resulting from storm-induced equatorward thermospheric wind can really contribute to the positive NmF2 storm effect, while R(O/N 2 ) ratio remains unchanged.This [O] increase provides a background NmF2 growth (see also Rishbeth, 1991;Field et al., 1998).Additional NmF2 increase is due to enhanced equatortward thermospheric wind (upward plasma drift), resulting from the auroral heating.An increase in neutral temperature and concentrations, as well as in vertical plasma drift (due to the enhanced equatorward wind), usually taking place during disturbed periods, leads to the hmF2 increase.This may be seen from an approximate expression for hmF2 (Ivanov-Kholodny and Mikhailov, 1986) where H = kT n /mg is the scale height and [O] is the concentration of atomic oxygen, β is the linear loss coefficient at a fixed height h 1 , W (in m/s) is the vertical plasma drift, c is a coefficient close to unity, d = 1.38 • 10 19 • (T n /1000) 0.5 is a coefficient in the expression for the ambipolar diffusion coefficient The above scenario takes place in the 'nighttime' (relative to storm onset) longitudinal sector.In the 'daytime' sector, F2-layer positive storm effects with the NmF2 and hmF2 increase are primarily the result from the vertical plasma drift increase without changes in neutral composition and temperature (Prölls, 1995;Mikhailov et al., 1995).The main mechanism of such W increase is the background (poleward during daytime) and the storm-induced (equatorward) wind interaction.Depending on the storm intensity, this interaction may result either in a decrease of the background meridional thermospheric wind or in its reversal.In both cases, we obtain an increase in NmF2 and hmF2.Therefore, one should expect positive NmF2 and hmF2 trends for the 1965-1991 period of increasing geomagnetic activity.Our previous analysis confirms the existence of NmF2 and hmF2 positive trends for the majority of the day (Fig. 1, bottom).
Negative F2-layer storm effects are known to be strongest in the early morning LT sector (Wrenn et al., 1987;Prölls, 1991Prölls, ,1993 and references therein) due to the perturbed neutral composition with the decreased O/N 2 ratio advected towards middle and lower latitudes by the thermospheric circulation.This effect is especially pronounced at middle latitudes (see later), but takes place with strongly decreased magnitude at lower latitudes as well (see Fig. 1, around 07 LT).The area with increased [O] shifts further equatorward in this case.
Interesting results demonstrate the correlation coefficients diurnal variations (Fig. 2, bottom) which support the above discussed scenario.Large and significant coefficients r(δfoF2, Ap 12 ) are found for afternoon and evening hours when foF2 trends are large at Alma-Ata (Fig. 1, bottom).This tells us that the revealed positive foF2 trends are related to geomagnetic activity by the physical mechanism being discussed.Large and significant correlation coefficients r(δhmF2, Ap 12 ) are obtained for all LT moments (Fig. 2, right hand, bottom).This is due to both processes ([O] or/and W increase related to the increased geomagnetic activity) which contribute to the hmF2 increase, as it follows from Eq. (3).
The daytime sunlit F2-region is sensitive to the increase in [O] and W, resulting in the NmF2 increase.Therefore, the correlation coefficients r(δfoF2, Ap 12 ) are largest and significant during daytime hours (Fig. 2, left hand, bottom).The nighttime F2-region formation mechanism is different and NmF2 is less sensitive to the [O] and W variations.This results in small and insignificant (at the 95% confidence level) nighttime correlation coefficients r(δfoF2, Ap 12 ).In contrast, the dependence of hmF2 on [O] and W is practically the same during both daytime and nighttime hours.This gives large and significant r(δhmF2, Ap 12 ) coefficients for the whole day (Fig. 2, right hand, bottom).

Middle latitudes
Typical mid-latitude foF2 and hmF2 trend diurnal variations are presented by the results at Ekaterinburg (Figs. 1,2).Negative (especially at night) foF2 and positive hmF2 (all day long) trends are obtained for most of the mid-latitude stations considered (Mikhailov and Marin, 2000;Marin et al., 2001).For better illustration of the physical mechanisms involved, let us consider Millstone Hill incoherent scatter observations for quiet 17 March 1990 (Ap = 3) and disturbed 21 March 1990 (Ap = 76) days.Millstone Hill and Ekaterinburg have close geomagnetic latitudes; therefore, such a comparison of the two stations is justified.Observed diurnal variations of NmF2, hmF2 and T n at 300 km are shown for the two days in Fig. 3.The observations illustrate well-known and typical negative storm behavior for the mid-latitude F2-layer.When we pass from quiet to disturbed conditions, NmF2 decreases and hmF2 increases during both daytime and nighttime hours.The NmF2 decrease is more pronounced in the nighttime and early morning LT sector.The same diurnal variation is seen in the foF2 trends (Fig. 1, left panel), with the correlation coefficients r(δfoF2, Ap 12 ) being the largest for the same hours (Fig. 2, left panel).Mid-latitude negative F2-layer storm effects are known to be the strongest in the post-midnight-early-morning LT sector and they are much weaker in the afternoon (Wrenn et al., 1987;Prölss, 1991Prölss, ,1993)).As it was pointed out earlier, this is due to the disturbed neutral composition with a decreased O/N 2 ratio, which is advected towards middle latitudes during the night, rotates into the day sector being shifted back to higher latitudes by diurnal varying thermospheric circulation (Skoblin and Förster, 1993;Fuller-Rowell et al., 1994;Prölss, 1995).This effect is clearly seen in the afternoon with a tendency for the foF2 trends to be even positive around 15 LT.
Contrary to the NmF2 behavior, hmF2 is larger for disturbed conditions.This is due to three reasons (see Eq. 3): 1) neutral temperature T n is higher in the perturbed thermosphere.Millstone Hill T n estimations are shown in Fig. 3 (bottom); 2) linear loss coefficient ] is higher for disturbed conditions due to higher molecular concentrations and reaction rate coefficients depending on temperature; 3) vertical plasma drift W is more positive due to an enhanced equatorward thermospheric wind in the nighttime sector, or to a decreased or even a reversal of the solar driven northward wind in the daytime LT sector (Prölss, 1993;Wickwar, 1989).
Let us consider these changes in the thermospheric parameters using Millstone Hill observations for 17 March and 21 March 1990.A self-consistent approach to the ionospheric F2-layer modelling proposed by Mikhailov and Schlegel(1997) with later modifications by Mikhailov and Förster (1999) and Mikhailov and Schlegel (2000) may be applied for daytime Millstone Hill observations to extract the set of main aeronomic parameters for the two days in question.The method uses measured N e (h), T e (h), T i (h), and V z (h) profiles to find the set of main aeronomic parameters responsible for the observed N e (h) distribution in the daytime F2-region.The calculated parameters are given in Table 1.
The results of the calculations are in agreement with the contemporary understanding of the F2-layer storm mechanisms (e.g.Rishbeth, 1991;Prölss, 1995;Field et al., 1998).The calculations show an increase in exospheric temperature T ex (compare to Millstone Hill estimates at 300 km in Fig. 3), a strong enrichment of the thermosphere with heavy molecular species O 2 and N 2 , and an increase in W. The latter results from some damping of the normal solar driven northward thermospheric circulation by the disturbed (southward) one.3).Therefore, the analyzed period of 1965-1991 of increasing geomagnetic activity should result in negative NmF2 and positive hmF2 trends, as our previous analysis has shown (Mikhailov and Marin, 2000;Marin et al., 2001).Unlike the case with lower latitudes (Alma-Ata station) where changes in vertical plasma drift and atomic oxygen concentration are responsible for the positive F2-layer storm effects, neutral composition (O/N 2 ratio) and temperature changes are supposed to be the main physical reason for the F2-layer negative storm effects at mid-latitudes (Prölss, 1995;Field et al., 1998), although the role of vibrationally excited N # 2 is considered in some publications as well (e.g.Pavlov, 1994;Pavlov et al., 1999).The largest neutral composition (O/N 2 ratio) perturbations take place in the post-midnight-early-morning LT sector (Prölss, 1980(Prölss, , 1993) ) and the calculated correlation coefficients r(δfoF2, Ap 12 ) are the largest for this part of the day (Fig. 2, left panel).Similar to the lower latitude case positive hmF2 trends and large correlation coefficients r(δhmF2, Ap 12 ) take place practically all day long.As mentioned above, this is mainly due to the increase in β, T n and W .
Therefore, the revealed mid-latitude foF2 and hmF2 trends may be considered as the manifestation of the storm induced neutral composition, and temperature and meridional wind changes, which should take place for the period of increasing geomagnetic activity 1965-1991.

High latitudes
The situation is more complicated with the high-latitude F2layer where close stations may demonstrate different diurnal variations of the trend magnitude.As an example, Sodankyla (67.40 N; 26.60 E; inv = 63.59N) and Lycksele (64.70 N; 18.80 E; inv = 61.46N) stations are shown in Fig. 1.Very strong negative foF2 trends take place during daytime with the minimum shifted to the morning hours at Sodankyla.The observed foF2 trends are small with a pretty large scatter during morning and evening hours.The correla-tion coefficients r(δfoF2, Ap 12 ) are large and significant during daytime hours (Fig. 2, left-hand, top).On the contrary, at Lycksele, the largest negative foF2 trends are observed in the evening LT sector, while the trends are small during the first part of the day.Corresponding diurnal variation is seen for r(δfoF2, Ap 12 ) in Fig, 2 where large and significant correlation coefficients are found for the second part of the day.
The Sodankyla station also shows an interesting and unusual hmF2 trend diurnal variations (Fig. 1, right-hand, top), when compared to other stations.The trends are positive although small during daytime hours, but they are negative at nighttime.The daytime correlation coefficient r(δhmF2, Ap 12 ) are large and significant, while the nighttime values are small and insignificant at the 95% confidence level (Fig. 2, right-hand, top).Such unusual diurnal variations of the hmF2 trend magnitude are discussed later using EISCAT observations.Large positive hmF2 trends along with large and significant correlation coefficients r(δhmF2, Ap 12 ) take place during the whole day at Lycksele (Figs. 1, 2; right-hand panels).
Let us start with the foF2 trends (Fig. 1, left-hand, top).The Sodankyla station ( inv = 63.59 • ) is located in the plasma ring or FLIZ zone (Thomas and Andrews, 1969;Pike, 1971) where an intensive F2-region ionization is produced by soft electron precipitation (Morse et al., 1971).The equatorial boundary of this zone is located at inv = 61 − 62 • at 00-06 MLT and at inv = 63 • at 18-21 MLT while during daytime, it shifts northward at inv = 70 − 72 • (Sagalin and Smiddy, 1974).This excursion of the precipitation zone explains the appearance of two peaks in the foF2 trend diurnal variation (Fig. 1, left-hand, top).As the intensity of electron precipitation is highly variable in the FLIZ zone, the scatter of the trends obtained is fairly large and the correlation coefficients r(δfoF2, Ap 12 ) are small and insignificant (Fig. 2, left-hand, top) for these two periods of the day.The corpuscular ionization should be strong enough during these periods to compensate large negative effect in NmF2 due to large changes in neutral composition and temperature expected in the perturbed auroral thermosphere.The latter is seen for daytime hours when, despite direct solar photoionization, very strong negative foF2 trends are observed (Fig.

left-hand, top).
The sub-auroral station Lycksele ( inv = 61.46• ) turns out to be in the FLIZ zone in the morning but not in the evening and unlike Sodankyla, there is only one (morning) peak where foF2 trends are small and correlation coefficients r(δfoF2, Ap 12 ) are small and insignificant (Fig. 2, left second panel).The daytime equatorial boundary of the FLIZ zone  intensive electron precipitation took place on 19 November, while both characteristics were small on 17 November (some splashes of electric field took place only after 19 UT).Observed NmF2 are higher and hmF2 are lower on 19 November for the period of 16-22 UT, when an intensive electron precipitation is expected (Fig. 1, left-hand, top).Large scatter in the observed hmF2 is seen on 19 November and is obviously due to a varying precipitation intensity.Median N e (h) profiles found over the 16-18 UT period are given in Fig. 5 for the two days in question.Strong precipitation results in an enhanced electron concentration (especially in the lower F-region) as well as in a decrease in hmF2.Namely, this effect of the electron precipitation is the most important for our analysis.Strong plasma production at lower altitudes shifts normal hmF2 to lower heights (e.g.Torr and Torr, 1969).A similar situation exists for a normal mid-latitude F2-layer when daytime hmF2 is lower than nighttime hmF2dian over one hour for one and the same input parameters.This is due to strong solar photoionization at low F-region heights.As the precipitation intensity increases with geomagnetic activity (Sato and Colin, 1969;Marubashi, 1970), nighttime hmF2 trends are negative at Sodankyla (Fig. 1, right-hand, top) for the period of increasing geomagnetic activity of 1965-1991.Therefore, the revealed features of the NmF2 and hmF2 nighttime trends may be attributed to the electron precipitation effects.
Besides particle precipitation strong electric fields are an inalienable feature of the disturbed auroral F2-region.The observed increase in geomagnetic activity for the analyzed period of 1965-1991 is the manifestation of intensified electric fields in the auroral zone.Joule heating related to the electric fields results in strong perturbations of neutral composition (O/N 2 , O/O 2 decrease) and neutral temperature increase (e.g.Prölss, 1980;Rishbeth and Müller-Wodarg,-1999). Therefore, by analogy with the mid-latitude case, one should expect strong negative foF2 and positive hmF2 trends for the period in question.An additional effect working in the same direction is due to the dependence of the O + +N 2 reaction rate constant (via the effective temperature) on electric field E (Schunk et al, 1975).Strong negative foF2 trends (Fig. 1, left-hand, top) do take place at Sodankyla during daytime hours, the nighttime case was discussed above.But relatively small (although positive) daytime hmF2 trends (Fig. 1, right-hand, top) look rather strange.At least three reasons may be considered: 1) the accuracy of initial experimental M(3000)F2 values, and the M(3000)F2 to hmF2 conversion procedure used; 2) the effect of strong electric fields on the N e (h) height profile; 3) the effect of the auroral thermosphere depletion (due to upwelling) with atomic oxygen (Prölss, 1980).Let us consider EISCAT observations for quiet 04 August 1992 (Ap = 2) and disturbed 05 August 1992 (Ap = 35) days which may help us analyze the problem with the hmF2 daytime trends.The daily mean Ap index was 15 on 04 August due to a disturbance which started late in the afternoon.Yet the first half of the day considered here was very quiet and we accepted Ap = 2 for our model calculations.Observed NmF2, hmF2, as well as electric field diurnal variations, are shown for the two days in Fig. 6.
The selected couple of dates demonstrates the effect of the general NmF2 decrease for the disturbed day which corresponds to the foF2 negative trend (Fig. 1, left-hand, top) for the considered period of increasing geomagnetic activity of 1965-1991.The effect of the electric field switching on and off is also seen in Fig. 6.The median N e (h) profiles taken over a set of one hour observations are shown in Fig. 7 for some UT periods.Two quiet time (04 August 1992) N e (h) profiles are shown for a comparison as well.An abrupt decrease of the layer height down to 200 km (F1-layer) takes place during the morning hours on 05 August (Fig. 6, middle panel) as a reaction to the enhanced electric field (Fig. 6, bottom).Later in the morning, when E decreases and the photoionization rate increases, hm restores back to normal hmF2 values around 350 km.In the afternoon, a moderate E increase again results in the hmF2 decrease, when hmF2 turns out to be close to the 04 August values.
All observed disturbed N e (h) profiles (Fig. 7) show a strongly reduced NmF2 up to a complete disappearance of the F2-layer (0300-0800 UT period), while a pronounced F1-layer appears around 200 km height.It is obvious that the M(3000)F2 parameter, determined from routine groundbased ionosonde observations, is not reliable for such profiles.On the other hand, special care is required when taking into account the effect of the underlying ionization in the empirical formulas relating M(3000)F2 to hmF2.Therefore, hmF2 derived from M(3000)F2 may not be very reliable for such N e (h) profiles, but one may hope that this does not strongly affect M(3000)F2 monthly median values.To illustrate the changes in the thermospheric parameters responsible for the observed NmF2 and hmF2 variations under disturbed conditions, let us consider two sunlit periods around 0630 and 1130 UT for 04 and 05 August.The above mentioned method by Mikhailov and Schlegel (1997) with later modifications is used for this analysis.The two chosen UT periods correspond to the cases of a pronounced F1layer appearance (0600-0700 UT) and to a moderately disturbed F2-layer with a pronounced hmF2 (1100-1200 UT); the corresponding N e (h) profiles are shown in Fig. 7. Calculated thermospheric parameters for the quiet and disturbed days are given in Table 2.The calculated T ex is higher on 05 August (the disturbed day), especially for the morning period when strong electric fields were observed.The enhanced electric field produces an intensive Joule heating and an upwelling in the thermosphere.The latter is seen in the calculated vertical plasma drift W (Table 2).The upwelling motion results in a [O] decrease and a [O 2 ], [N 2 ] increase, which is also seen for the disturbed day with respect to the quiet one.Relatively small [O] decrease at 300 km (58% in the morning and 28% at around noon), in fact, corresponds to a strong decrease in the atomic oxygen abundance, as T ex (and corresponding neutral scale height) is higher on 05 August (Table 2).
The thermosphere heating and upwelling results in the strong increase in [N 2 ] (by a factor of 2.25 in the morning and by 1.71 times around noon), and in the [O 2 ] increase by a factor of 4.12 and 2.77, respectively.This [N 2 ] and [O 2 ] increase, along with the increase in the O + +N 2 rate constant depending on T n , T i and E, results in a very strong β increase by a factor of 8.5 in the morning case, and by a factor of 4.5 around noon.Similar to the mid-latitude case, this increase in the linear loss coefficient β is the main reason for the NmF2 decrease on the disturbed day; the additional negative effect in NmF2 is related to the [O] decrease.This analysis based on EISCAT observations illustrates the physical mechanism of the strong foF2 negative trend obtained for daytime hours at Sodankyla (Fig. 1, left top panel).
Electric fields via the chain of the processes mentioned above strongly affect the N e (h) height distribution and hmF2, accordingly.During nighttime, when direct solar photoionization is absent, or in the morning, when it is not strong enough, the loss coefficient β increase may result in a complete disappearance of the normal F2-layer and formation of the N e (h) profile with maximum around 200 km (Fig. 5, 03-08 UT period).Such a layer is composed of heavy molecular ions, NO + and O + 2 , as model calculations show.Therefore, electric fields along with the earlier discussed electron precipitation effect may really contribute to the negative nighttime hmF2 trends at Sodankyla (Fig. 1, top right panel).During daytime hours, solar EUV ionization becomes strong enough and the F2-layer maximum is formed at usual heights, but a well-developed F1 layer still exists (Fig. 7), with the NmF2 and NmF1 values being close around 08 UT (Figs. 6,7).
Both satellite observations (Prölss, 1980) and model calculations (Table 2) show a decrease in the atomic oxygen concentration for disturbed conditions.According to Eq. ( 3), a decrease in [O] should compensate to some extent for the hmF2 growth, primarily resulting from the β, W and T n increase on the disturbed day.This effect is not strong for the 04 and 05 August case ( lg[O]=−0.108and lgβ=0.65)and disturbed daytime hmF2 values are larger than the quiet time ones (Fig. 6, middle panel).But, depending on the perturbation intensity, the effect may be larger.For instance, an analysis of EISCAT observations for the period of geomagnetic storm on 10 April 1990 (Mikhailov and Schlegel, 1998) has revealed an [O] decrease by a factor of 4.3 at 300 km, with respect to the previous day.In that case, the daytime layer maximum was formed around 200 km height.
Therefore, relatively small (when compared to mid-latitude stations) daytime hmF2 trends at Sodankyla (Fig. 1, right top panel) may be due to a strong decrease in the atomic oxygen abundance in the perturbed auroral thermosphere.

Discussion
Investigation of the ionospheric trends was greatly stimulated by the model calculations of Rishbeth (1990), and Rishbeth and Roble (1992), which predicted the ionospheric effects of the atmosphere greenhouse gas concentration increase.Since then, researchers have been trying to reveal the predicted thermosphere cooling analyzing ionospheric trends (Bremer, 1992;Givishvili and Leshchenko, 1994;Ulich andTurunen, 1997, Jarvis et al., 1998;Upadhyay and Mahajan, 1998).
But the world-wide pattern of the F2-layer parameter trends turned out to be very complicated and cannot be reconciled with the greenhouse hypothesis.On the contrary, the geomagnetic control concept by Mikhailov and Marin (2000), based on the contemporary understanding of the F2-layer storm mechanisms, allows us to explain the revealed morphological features of the F2-layer trends.According to this concept, there are periods of negative and positive F2-layer parameter trends corresponding to the long-term changes in geomagnetic activity shown in Fig. 8. Annual mean Ap indices prior 1932 were reconstructed from aa indices available from 1868.Years of solar cycle minima (m) and maxima (M) are marked in Fig. 8 as well, to show that such long-term variations in geomagnetic activity (presented by 11-year Ap index) are not related to solar cycle variations.
A steady increase in geomagnetic activity took place for the period from 1900 to middle of 1950s followed by a decrease towards middle of 1960s and again an increase towards the end of 1980s.A tendency for a decrease in geomagnetic activity after 1990 is clearly seen in annual mean Ap values.Similar variations of geomagnetic activity can be found in Clilverd et al., (1998, their Fig. 6).Namely, these long-term variations in geomagnetic activity result in the ionospheric F2-layer long-term trends.
An example of such long-term variations is given in Fig. 9 for a mid-latitude station Slough, where ionospheric observations are available from the early 1930s.Variations of the 11-year Ap index are repeated in Fig. 9 (top) for further discussion.The δfoF2 variations are considered for (M + m) and (m) year selections (Danilov and Mikhailov, 1999;Mikhailov and Marin, 2000).Solid and dashed lines are the least squares fitting by the 4th degree polynomial (a higher degree gives practically the same results).Everywhere error bars present the standard deviation over 12 monthly values.An anti-phase type of δfoF2 and Ap long-term variations is seen for the period in question.The periods of increasing geomagnetic activity (before 1955 and after the end of the 1960s) are seen to correspond to negative foF2 trends, while during the decreasing phase of geomagnetic activity (1955 to the end of the 1960s), a small positive foF2 trend takes place.There is also a tendency for the foF2 trend to switch from negative to positive after 1990, in accordance with the change in geomagnetic activity (see Fig. 8).Different signs of the foF2 trends for the periods before and after 1965 were demonstrated earlier by Mikhailov and Marin (2000) for some stations with long observational periods.
On the contrary, δhF2 demonstrates a syn-phase with Ap type of variations (Fig. 9, bottom).The periods with negative hmF2 trends before 1970 and after 1990 correspond to the periods with decreasing geomagnetic activity while a positive hmF2 trend takes place for the period of 1970-1990.In accordance with the analysis by Marin et al. (2001), unlike the foF2 case, all years with M(3000)F2 observations may be used for the hmF2 trend derivation.The period between 1961-1965 is absent in Fig. 9 (bottom), as observed M(3000)F2 values for this period give unreal δhmF2 variations which have not been confirmed by observations at other stations.Physical mechanisms relating foF2 and hmF2 trends Fig. 10.Annual mean Ap 12 and δfoF2 variations at Moscow, 00 and 12 LT.Dashed line is an attempt to remove the dependence on geomagnetic activity using δfoF2 regression with Ap 12 .Note that observed δfoF2 variations are much stronger than model ones especially for daytime.
with geomagnetic activity variations are discussed earlier in the paper.
In the framework of the proposed geomagnetic hypothesis one should expect thermosphere heating rather than the cooling that the researchers are seaking when considering the 1970-1990 period.Indeed, from the Ap variations (Fig. 8), one may accept the Ap index increase from 12 to 16 for the period in question; such an increase in the Ap index, according to the thermospheric MSIS-83 (Hedin, 1983) model, results in the annual mean T ex increase by about 10 K for mid-latitudes (F 10.7 = 140 was used in calculations).But this heating will be followed by the thermosphere cooling, in accordance with the long-term changes in the geomagnetic activity (Fig. 8).
Although there is an obvious relationship between the F2layer parameter trends and the geomagnetic activity, it is im-possible to remove this geomagnetic effect from the trends revealed, using any conventional index (e.g.monthly or annual mean Ap) of geomagnetic activity, and to check if there is any residual trend (of a greenhouse origin, for instance).If it could be accomplished by using the conventional indices, the problem of the F2-layer storm description and prediction would have been solved long ago, but this has not been the case until now.This is not surprising as any global geomagnetic activity index cannot, in principle, take into account the whole complexity of F2-layer storm effects with positive and negative phases depending on the season, longitude, UT and LT of storm onset, storm magnitude, etc.Therefore, an inclusion of the Ap index to the regression, in fact, does not remove the dependence on geomagnetic activity, as supposed by Bremer(1998) and Jarvis et al. (1998), but only contaminates the analyzed data (Mikhailov and Marin, 2000).Indeed, according to Mikhailov and Marin (2000), and Marin et al. (2001), such an inclusion of the Ap index has some effect on the trend magnitude, but without changing, in principle, the main morphological features of the foF2 and hmF2 trends.A similar result was obtained by Ulich and Turunen (1997) who did not include the Ap index in their study for this reason.
Figure 10 illustrates an attempt to remove the geomagnetic effect by the inclusion of the annual mean Ap 12 to the foF2 trend analysis for Moscow, 00 and 12 LT.A two-step procedure was applied.At first, δfoF2 = (foF2 obs −foF2 mod )/foF2 mod values were found and called 'observed' (Fig. 10, solid line).Then a regression (2nd degree polynomial) of these δfoF2obs with annual mean Ap 12 was calculated and called 'model' in Fig. 10 (dashed line).The model curve is seen to follow, qualitatively, the observed δfoF2 variation for 00 LT with a correlation coefficient r = 0.538, which is significant at the 99% confidence level.But the observed δfoF2 variations are much larger and not reproduced completely by the model.The situation is even worse for daytime (12 LT) conditions.In this case, there is not even a qualitative agreement between the two curves (r = 0.227, insignificant).The same result was obtained for Ekaterinburg (Fig. 2, left panel) where significant correlation coefficients were found only for nighttime hours.
The obtained result tells us that, in fact, the geomagnetic effect is much stronger (at least during nighttime) than can be described using the Ap 12 index.Poor δfoF2 correlation with Ap 12 during the daytime confirms the complexity of the F2-layer storm mechanisms as mentioned earlier.For instance, mid-latitude daytime F2-layer storm effects may be due to thermospheric perturbations formed in the nighttime longitudinal sector during the preceding geomagnetic storm (Skoblin and Förster, 1993;Fuller-Rowell et al., 1994;Prölss, 1995).
On the other hand, one should keep in mind that the sunspot number R 12 , usually used in empirical ionospheric models, is far from the best choice (Mikhailov and Mikhailov, 1999) and, in fact, it does not allow us to completely remove the dependence on solar activity being used in foF2 and hmF2 versus R 12 regressions.Despite special methods applied to ex- tract F2-layer parameter trends the imperfection of Ap 12 and R 12 indices results in a fairly large δfoF2 and δhmF2 scatter (e.g.Fig. 10).In practice, it was recommended (Danilov and Mikhailov, 1999;Mikhailov and Marin, 2000) to use only the years around solar minima (m) or the years around solar maximum and minimum (M + m) for the foF2 trend analysis.Such a selection of years allows us to avoid the hysteresis effect which takes place at the falling and rising phases of the solar cycles, and when the foF2 versus R 12 correlation is the worst.With this approach, it was possible to obtain the most consistent pattern of the foF2 trends over all the stations considered.On the contrary, the same approach turned out to be inefficient when applied to the hmF2 trend analysis and all available years were used in the study by Marin et al. (2001).This is rather strange as the hysteresis effect takes place for M(3000)F2 solar cycle variations as well (Rao and Rao, 1969).There is currently no explanation for this effect.A well-pronounced dependence of the foF2 trend magnitude on the geomagnetic (invariant) latitude (Danilov and Mikhailov, 1999;Mikhailov and Marin, 2000) is explained by the perturbed neutral composition and temperature latitudinal dependence, as discussed earlier in the paper.Contrary to this, no systematic latitudinal dependence was revealed for the hmF2 trends (Marin et al., 2001).This result may be explained by the hmF2 dependence on the main aeronomic parameters (Eq.3).Normally, neutral concentrations (O, O 2 , N 2 at a fixed level), temperature T n as well as vertical plasma drift W increase during disturbed periods.According to Eq.
(3), this should result in an hmF2 increase, as all terms in Eq. (3) work in one direction.Therefore, we have positive hmF2 trends at middle and lower latitudes; the high-latitude case was discussed earlier.The meridional wind V nx effect (via W = V nx sin I cos I cos D) becomes efficient at lower latitudes as magnetic inclination I approaches 45 • .As the perturbation in β and T n decreases and the [O] (see above) and W contributions increase towards lower latitudes, no pronounced latitudinal dependence for the hmF2 trend magnitude should take place, in accordance with results of our study.
In principle, some longitudinal effect in the foF2 and hmF2 trends may be expected in the scope of the proposed geomagnetic control hypothesis.A statistical analysis by Ha-jkowicz (1998) of AE-index variations over two solar cycles (1957-1968 and 1978-1986) has shown that the maximum in auroral activity is largely confined to 09-18 UT, with a distinct minimum at 03-06 UT.This means that Eastern Siberia and Japan are primarily at night during the period of maximum auroral activity, whereas Europe and Eastern America are primarily at daytime.This effect, overlapping with the background solar driven thermospheric circulation (equatorward at night and poleward during daytime), may give some longitudinal effects in the F2-layer parameter trends.Another source of longitudinal variation is related to the zonal winds and longitudinally dependent magnetic declination D via the wind term in the vertical plasma drift W . Primarily negative foF2 and hmF2 trends at longitude west of 30 • E, yet positive trends east of 30 • E, were revealed by Bremer (1998).A tendency for similar hmF2 trend separation was reported by Marin et al. (2001).Indeed, the D = 0 line crosses Europe along the longitude λ ≈ 20 • E and the zonal wind effect cannot be excluded.But a preliminary analysis has shown that the situation is not that straightforward.Three pairs of stations with close inv , but different magnetic declination D, are compared in Table 3, where foF2 and hmF2 trends from Mikhailov and Marin (2000) and Marin et al. (2001) are given for 12 and 06 LT.The results in Table 3 show that regardless of different signs of magnetic declination D, foF2 trends are negative and hmF2 trends are positive at all stations considered (the daytime hmF2 trend at Novokazalinsk is insignificant), in accordance with the earlier obtained conclusions (Danilov and Mikhailov, 1999;Mikhailov and Marin, 2000;Marin et al., 2001).The expected longitudinal effect may be due to vertical plasma drift W = (V nx cos D − V ny sin I ) sin I cos I variations, where V nx and V ny are meridional and zonal components of the thermospheric wind.Westward (V ny < 0) zonal wind is strong around 06 LT, but small around 12 LT (Hedin et al., 1991).On the contrary, V nx is small around 06 LT, therefore the V effect may be expected around 06 LT.Table 3 shows that stations with negative D seem to have more positive hmF2 trends at 06 LT, but this is not the case for 12 LT.No definite conclusion can be drawn with respect to the foF2 trends in the results given in Table 3.Therefore, if any dynamical effect due to zonal thermospheric wind exists, it is small when compared to the contribution of other aeronomic parameters.
Some European stations were shown to demonstrate negative hmF2 trends (Marin et al., 2001) and this was the reason to mention a longitudinal effect in the hmF2 trends.Their Table 2 (Model 1) shows that significant negative trends were revealed for some LT moments at Bekescsaba (46.7 N, 21.2 E), Poitiers (46.6 N, 0.3 E), Dourbes (50.1 N, 4.6 E) and Juliusruh (54.6 N, 13.4 E); the other stations may be considered as sub-auroral and auroral ones, with specific mechanisms of the F2-layer formation discussed earlier in this paper.Therefore, an additional analysis is needed for these mid-latitude stations to find out the reason for such hmF2 behavior.Negative hmF2 trends were reported for Southern hemisphere stations in the Argentine Islands and Port Stanley by Jarvis et al. (1998), and for Sodankyla by Ulich and Turunen (1997).The latter result should be discussed as it contradicts our conclusions obtained for the Sodankyla station.
It was stressed by Danilov and Mikhailov(1999), and Mikhailov and Marin (2000) that F2-layer trend results are strongly dependent on the method used to extract the trends from the ionosonde observations.Ulich and Turunen (1997) obtained a negative hmF2 trend, −0.39 km/year for daytime hours over the period of 1958-1994.Unlike our approach, they worked with non-smoothed absolute deviations hmF2 from a model (linear regression hmF2 with monthly averaged F 10.7 ), although they applied to hmF2 a running mean filter with a width of 11 years in order to suppress solar activity effects.We have used a similar approach and did obtain negative daytime hmF2 trends over the period in question.Regarding this,the following should be mentioned: 1) non-smoothed hmF2 (or foF2) values show a very large scatter where a "useful signal" may just be lost.Therefore, smoothing of the initial data and working with relative (not absolute) deviations from a model was recommended for the trend analysis (Danilov and Mikhailov, 1999;Mikhailov and Marin, 2000); 2) it is known that monthly median foF2 and M(3000)F2 parameters correlate better with smoothed (not monthly averaged) indices of solar activity (e.g.Mikhailov and Mikhailov, 1999 and references therein).That is why only 12-month running mean sunspot numbers R or F 10.7 are used in empirical F2-layer parameter modelling.Moreover, a non-linear dependence of F2-layer parameters on solar activity level provides better regression accuracy (e.g.Kouris et al., 1997) than the linear one used by Ulich and Turunen (1997); 3) as the hmF2 trend follows the geomagnetic activity, a separate analysis is required for different periods in the geomagnetic activity's long-term variations; the end of the 1960s and the beginning of the 1990s are the turning points in these variations.Therefore, a trend derived over the whole 1958-1994 period does not correctly present the real hmF2 longterm variations.
Due to these differences in approaches, the daytime hmF2 trends at Sodankyla obtained by Ulich and Turunen (1997) and Marin et al. (2001) have a different sign for the period of 1965-1991.

Conclusions
The foF2 and hmF2 trend morphology earlier revealed by Danilov and Mikhailov (1999), Mikhailov and Marin (2000), and Marin et al. (2001), was interpreted in the framework of the geomagnetic control concept proposed by Mikhailov and Marin (2000).Latitudinal and diurnal variations of the annual mean foF2 and hmF2 trends are the most pronounced features and their analysis was the major concern of the paper.The main results may be listed as follows: 1.The effect of long-term varying geomagnetic activity is very strong in the foF2 and hmF2 trends.But it is impossible to remove this geomagnetic effect from the F2-layer parameter trends using conventional (monthly or annual mean Ap, for instance) indices of geomagnetic activity.An inclusion of Ap 12 to the regression removes only partly the geomagnetic effect, but contaminates the analyzed material, in principle, without changing the obtained result.Therefore, any interpretation of the foF2 and hmF2 trends should consider the geomagnetic effect as an inalienable part of the trends revealed, and this can be done based on the contemporary understanding of the F2-layer storm mechanisms.3.There are periods with negative and positive foF2, and hmF2 trends which correspond to the periods of increasing or decreasing geomagnetic activity.An 11-year Ap index can be used as an indicator of such long-term variations in geomagnetic activity.The turning points are: around 1955, the end of the 1960s and the 1980s, where foF2 and hmF2 trends change their signs.An anti-phase for δfoF2 and syn-phase for δhmF2 type of long-term variations with Ap may be followed for Slough, where ionospheric observations are available from the early 1930s.Such a type of mid-latitude F2-layer parameter variations is due to neutral composition, temperature and thermospheric winds changes related to geomagnetic activity variations.
4. An existence of a pronounced dependence of the foF2 trends on geomagnetic (invariant) latitude and an absence of any latitudinal dependence for the hmF2 trends are due to different dependencies of NmF2 and hmF2 on main aeronomic parameters, the latter being latitudinal dependent during disturbed periods.In particular, for the period of increasing geomagnetic activity of 1965-1991, it may be concluded: (a) at lower latitudes, positive (or small negative) foF2 trends and positive hmF2 trends are primarily due to an increase in the equatorward thermospheric wind and in atomic oxygen concentration; (b) at middle latitudes, the negative foF2 trend is due to neutral composition (O/N 2 ratio decrease) and temperature increase, resulting in the linear loss co- increase.The latter, along with the enhanced T n and equatorward thermospheric wind, determine the positive hmF2 trend; (c) at sub-auroral and auroral latitudes, foF2 and hmF2 trends are determined by strong neutral composition and temperature changes during daytime hours, while at nighttime, soft electron precipitation provides strong contribution.In the auroral zone, electric fields in addition to perturbing neutral composition and temperature via Joule heating, can strongly affect the linear loss coefficient via the γ 1 dependence on E. This results in very strong, negative foF2 and relatively small, positive hmF2 daytime trends.
5. All the revealed morphological features of the foF2 and hmF2 trends may be explained in the framework of contemporary F2-region storm mechanisms.This newly proposed geomagnetic storm concept used to explain the F2-layer parameter long-term trends proceeds from a natural origin of the trends rather than an artificial one related to the thermosphere cooling due to the greenhouse effect.Within this concept, instead of the thermosphere cooling that the researchers are seeking, one should expect the thermosphere heating for the period of increasing geomagnetic activity of 1965-1991.This period will be followed by the thermosphere cooling, in accordance with the long-term changes in geomagnetic activity.

Fig. 1 .
Fig. 1.Diurnal variation of annual mean slope K for foF2 (left panel) and hmF2 (right panel) trends at auroral, sub-auroral, mid-latitude and lower latitude stations for the 1965-1991 period, invariant latitudes are given in brackets.Error bars present the standard deviation of seasonal (over 12 months) scatter in the slope K.

Fig. 2 .
Fig. 2. Same as Fig. 1, but for correlation coefficients r(δfoF2, Ap 12 ) and r(δhmF2, Ap 12 ).Solid squares are correlation coefficients significant at the 95% confidence level, open squares -the coefficients which are insignificant at this level.

Fig. 3 .
Fig. 3. Observed diurnal variations of NmF2, hmF2 and neutral temperature T n estimated at Millstone Hill at 300 km for quiet 17 March 1990 and disturbed 21 March 1990 days.

Fig. 5 .
Fig. 5. Observed with EISCAT median N e (h) profiles calculated over two hours.Note the effect of strong particle precipitation in the N e (h) height distribution on 19 November.

Fig. 6 .
Fig.6.Observed with EISCAT diurnal variations of NmF2, hmF2 and electric fields for quiet 04 August 1992 and disturbed 05 August 1992 days.Note the decrease in the height of the layer maximum after a strong electric field is switched on.

Fig. 7 .
Fig. 7. Observed with EISCAT N e (h) profiles for different UT moments of the disturbed day 05 August 1992.Note the modification of normal F2-layer and formation of the layer maximum around 200 km as a reaction to the strong increase in the linear loss coefficient β.Quiet time N e (h) profiles for 04 August 1992 are shown for a comparison.

Fig. 9 .
Fig. 9. 11-year running mean Ap index along with δfoF2 and δhmF2 long-term variations.Two year selections (M + m) and (m) (see text) are used for the foF2 and (all years) for the hmF2 trend analysis.Least squares fitting curves are a 4th degree polynomials.Error bars present the standard deviation of seasonal (over 12 months) scatter.

2.
Large and significant correlation coefficients r(δfoF2, Ap 12 ) and r(δhmF2, Ap 12 ), as well as similarity in trends and correlation coefficients diurnal variations (Figs. 1, 2) reveals the close relationship of the F2-layer parameter trends with geomagnetic activity.Both diurnal variation patterns (Figs.1, 2) clearly indicate physical processes which are usually used to explain latitudinal and diurnal F2-layer parameter storm variations.This F2layer storm mechanism is based on the background solar driven and disturbed thermosphere circulation interaction, resulting in neutral composition and temperature perturbations.
• N and 81 • N geographic latitude (30 • N and 71 • N geomagnetic latitude) and cover a broad longitudinal range, which provides the possibility to study spatial variations of the trend magnitude.Trends are analyzed for relative deviations of the observed foF2 or hmF2 values from some model δp = (p obs − p mod )/p mod where p is the 12-month running mean of the monthly median foF2 or hmF2.A regression (third-degree polynomial) of p with the sunspot number R 12 is used as a model (Model 1).A regression of p versus R 12 and annual mean Ap 12 index is refered to as Model 2.

Table 1 .
Calculated thermospheric parameters for quiet 17 March 1990 and disturbed 21 March 1990 days at 300 km and 13.5 LT ] abundance in the thermosphere as T ex (and corresponding neutral scale height) is higher on 21 March (Table1).A strong increase in [N 2 ], [O 2 ] as well as in the temperature results in a β increase by more than a factor of 3 and this is the main reason for the NmF2 decrease on the disturbed day (see Eqs. 1,2).The growth of β, W , and [O] on the disturbed day results in higher observed hmF2 (Fig.3, and Eq.

Table 2 .
Calculated at 300 km thermospheric parameters for quiet 04 August 1992 and disturbed 05 August 1992 days and two periods of the day Fig. 8. Annual mean and 11-year running mean Ap index variations.Symbols (m) and (M) refer to years of solar cycle minimum and maximum.

Table 3 .
Annual mean slope K (in 10 −4 per year) for the period after 1965 for the stations with close inv , but different D. Years of solar minimum (m) for foF2 and all years for hmF2 trend analysis are used.