Time variations of the ionosphere at the northern tropical crest of ionization at Phu Thuy, Vietnam

. This study is the ﬁrst which gives the climatology of the ionosphere at the northern tropical crest of ionization in the Asian sector. We use the data from Phu Thuy station, in Vietnam, through three solar cycles (20, 21 and 22), show-ing the complete morphology of ionosphere parameters by analyzing long term variation, solar cycle variation and geomagnetic activity effects, seasonal evolution and diurnal de-velopment. Ionospheric critical frequencies, fo F2, fo F1 and fo E, evolve according to the 11-year sunspot cycle. Seasonal variations show that fo F2 exhibits a semiannual pattern with maxima at equinox, and winter and equinoctial anomalies depending on the phases of the sunspot solar cycle. 1 fo F2 exhibits a semiannual variation during the minimum phase of the sunspot solar cycle 20 and the increasing and decreasing phases of solar cycle 20, 21 and 22. 1 fo F1 exhibits an annual variation during the maximum phase of solar cycles 20, 21 and 22. 1h ′ F2 shows a regular seasonal variation for the different solar cycles while 1h ′ F1 exhibits a large magni-tude dispersion from one sunspot cycle to another. The long term variations consist in an increase of 1.0 MHz for fo F2 and of 0.36 MHz for fo F1. fo E increases 0.53 MHz from solar cycle 20 to solar cycle 21 and then decreases − 0.23 MHz during the decreasing phase of cycle 21. The diurnal variation of the critical frequency fo F2 shows minima at 05:00 LT and maxima around 14:00 LT. fo F1 and fo E have a maximum around noon. The diurnal variation of h ′ F2 exhibits a maximum around noon. The main features of h ′ F1 are a minimum near noon and the maximum near midnight. Other minima and maxima occur in the morning, at about 04:00 or 05:00 LT and in the afternoon, at about 18:00 or 19:00 LT but they are markedly smaller. Only during the maximum phase of all sunspot solar cycles the maximum near 19:00 LT is more pronounced.


Introduction
Since the first ionosonde sounding in 1925 (Breit and Tuve, 1926), large data bases of ionosonde data covering several sunspot solar cycles were built. It is now possible to analyze several different time variation patterns of the ionosphere: long term, sunspot solar cycles, annual and diurnal. The present work gives a general description of ionospheric layers variations at Phu Thuy, a station located at the northern crest of ionization in the Asian sector.
The different time variations of the ionosphere have been studied by various authors.
-The long term variations observed by ionosonde located in different regions were previously analyzed by Elias and Ortiz de Adler (2006), Bremer (2008) and Ouattara et al. (2009). Three factors were proposed to explain the long term variation (Lastovicka et al., 2006;Elias, 2009): (1) green house gases (Bremer, 2008), (2) changes in the Earth's magnetic field (Elias and Ortiz de Adler, 2006;Cnossen and Richmond, 2008) and (3) geomagnetic activity. Concerning the solar cycle variation, it is well known that ionization of the ionospheric layers is controlled by solar UV and EUV radiations which follow the solar sunspot cycle (Rishbeth and Gariott, 1969).
-The variations of the F1 layer, the transition layer between the E and F2 layers, are explained by the combination of F2 and E layers variations, related to photochemical rates and changes in atmospheric composition or temperature (Rishbeth and Kervin, 1968).
We analyze data recorded since 1962. The paper is organized in several sections. The second section presents the data set and data analysis. Sections 3, 4 and 5 are, respectively, devoted to the long term and solar sunspot cycle variations, yearly and diurnal variations. Then we discuss our results and conclude in Sect. 6.
In this study, we analyze long term variations, solar cycle, seasonal and diurnal variations of the ionospheric parameters for different magnetic activity given by the am index: magnetically quiet days with am < 20 and disturbed days with am ≥ 20.
-Arithmetical mean values obtained from day-time hourly values of parameters are used for studying diurnal variation. The daily mean values have been derived from the available hourly data (for foF2 and h ′ F1 with maximal 24, for the other with maximal 13 values).
-Seasonal variation and yearly variation is obtained by using respectively arithmetical mean values of monthly and yearly values.
-Our seasons are the following: winter (November, December, January, and February), summer (May, June, July, August), autumn (September, October), spring (March, April). Error bars (σ = √ V , V : variance is provide an estimation of the uncertainty in the ionospheric parameters. -The study is performed for the different phases of solar cycles. We distinguish four parts: the minimum phase, years with Rz < 20, the increasing phase, years with 20 ≤ Rz ≤ 100, the maximum phase, years with Rz > 100 and the decreasing phase, years with 100 ≥ Rz ≥ 20. Table 1 gives the years for each sunspot cycle phases. Figure 1 illustrates the solar cycle variation of critical frequencies and virtual heights of ionospheric layers F2, F1, E and Es, during sunspot cycles 20, 21, 22. On the left side are shown the critical frequencies. On the right side are drawn the virtual heights. The panels from the top to bottom correspond respectively to F2, F1, E and Es layers. On each panel is superimposed the yearly mean value of the sunspot number. Figure 1 shows a good correlation between the critical frequency of F2 layer (foF2), F1 layer (foF1) and E layer (foE) and the sunspot cycle. On the contrary, the critical frequency of the Es layer (foEs) and the virtual heights h ′ F2, h ′ F1, h ′ E and h ′ Es are poorly correlated with the sunspot number. From 1980 to 1994, due to technical reasons, the virtual heights h ′ E and h ′ Es are not reliable and are not plotted on Fig. 1. Figure 1 also shows a long term variations of the critical frequencies foF2 and foF1 which are increasing through the 3 solar cycles. The critical frequency foE exhibits another pattern: it increases from 1962 until 1984 and decreases after. These observations will be discussed later in Sect. 6.

Solar cycle and long term variations
The long term variations are computed by using the linear regression technique: where Rz is the sunspot number. Then we estimate where X i correspond to observations of critical frequencies.  199 1967 1974 1981 1988 1995 2002 Year 0 Figure 3 presents, from the top to the bottom, the yearly variation of foF2, h ′ F2, foF1 and h ′ F1 during quiet days with am < 20 (blue curve), during magnetically active days with am ≥ 20 (red curve) and for all the days (violet curve). This figure clearly shows that the geomagnetic effect is very small on the average values. Tables 2 and 3 give the correlation coefficients between the ionospheric parameters and the Rz index for magnetic quiet days with am < 20 (Table 2) and for all the days (Table 3). The correlation coefficients of foF2 during solar cycles 20, 21 and 22 are rather similar between 0.835 and 0.867. There is no difference between the two samples with quiet days and all the days. Concerning the critical frequency foF1 the correlation coefficient decreases from solar cycle 20 to solar cycle 21 and increases for solar cycle 21 to solar cycle 22, both for quiet days (Table 2) and all the days (Table 3).
The correlation coefficient for foE increases with solar cycle from 0.612 (cycle 20) to 0.754 (cycle 22) for the quiet days sample and from 0.611 to 0.739 for all the days.

Seasonal variation
In this section, for each month a mean regression equation (X = aRz + b) has to be derived. Then by use of a constant value of Rz (can be zero or another value as e.g. the mean Rz value during the investigated time interval) for each month the corresponding ionospheric parameter X(Rz) can easily be calculated. In these X(Rz) data the solar influence is more or less eliminated. And then we analyze the seasonal variation foF2, foF1, h ′ F2, h ′ F1. Figure  The second maximum appears generally in October.
We observe that the two equinox maxima are asymmetric: the autumnal maximum is smaller than the spring one during minimum and decreasing phases, whereas the spring maximum is smaller than the autumnal phase during increasing and maximum phase of the sunspot cycle 22. During the decreasing and maximum phases of solar cycles 21 and 22, the critical frequency foF2 in December is greater than in July.    During the increasing phase h ′ F1 is rather the same for solar cycles 20 and 21. Figure 8 shows the mean diurnal variation of foF2 observed during the different phases of the sunspot cycles, the red curve corresponds to cycle 20, the blue one to cycle 21 and the violet one to cycle 22. foF2 follows the same variation: before sunrise, foF2 decreases and is minimum at 05:00 LT. Then it increases to reach a maximum at 14:00 LT and decreases again later. Figure 9 shows the mean diurnal variation of foF1, observed from 06:00 LT to 18.00LT, during all the phases of the sunspot cycles 20, 21, 22. foF1 increases from 06:00 LT to 12:00 LT and then symmetrically decreases. This pattern is observed for all the phases except for decreasing phases, during which the curves are different for different sunspot cycles (bottom right panel). Figure 10 shows the mean diurnal variation of h ′ F2 observed during daytime for the different phases of the sunspot cycles 20, 21 and 22. During all the phases of the sunspot 1964 1968 1972 1976 1980 1984 1988 1992 1996 Year 200 1964 1968 1972 1976 1980 1984 1988 1992 1996 Year 3 4 5 6 7 1964 1968 1972 1976 1980 1984 1988 1992 1996 Year 4 6 8 10 12 14 1964 1968 1972 1976 1980 1984 1988 1992 1996 Year 200   cycles h ′ F2 increases in the morning, is maximum at about 12:00 LT and decreases in the afternoon. The mean diurnal variation of h ′ F2 varies from 260 km to 360 km in the daytime. h ′ F2 is greater during daytime than during night time. This is a particular feature of the F-layer near the tropical crest of ionization. We also notice that the diurnal variation of h ′ F2 is rather the same for all the phases of the sunspot cycle, but from one solar cycle to another the amplitude of the virtual height h ′ F2 changes. Figure 11 shows the mean diurnal variation of the virtual height h ′ F1 for the different phases of sunspot cycles 20, 21 and 22. The main features are the minimum near noon and the maximum near midnight. Other minima and maxima occur in the morning, at about 04:00 or 05:00 LT and in the afternoon, at about 18:00 or 19:00 LT but they are markedly smaller here, except during the sunspot maximum (top right panel) where the afternoon maxima are greater than the night maxima for solar cycle 21 and 22.

Discussion and conclusion
In Table 4 we recapitulate our main results. The columns correspond to the time scale variations of the data (long term, solar, annual and diurnal) and the rows to the ionospheric parameters (critical frequencies and virtual heights). We will analyze in detail the characteristics and properties of the ionospheric parameters for different time scales.

Long term variations
The critical frequencies foF2 and foF1 show a long-term increase of 1.0 MHz and 0.36 MHz, respectively, during the period here analyzed. foE increases by 0.53 MHz from sunspot cycle 20 to 21 followed by a decrease of −0.23 MHz. Bremer (2008), explained the long term changes in the parameters of the E-and F1-regions by an increasing atmospheric greenhouse effect (increase of CO2). Bremer (2008) observed different long term variations from one station to another station. Elias and Ortiz de Adler (2006) explained the long term changes in the amplitude of foF2 observed at Tucuman at the southern crest of the equatorial anomaly by the trend of the magnetic dip angle which increased during the analyzed 30 years interval with a rate of 0.35%/year. Mikhailov andMarin (2000, 2001), Mikhailov and Morena (2003), Mikhailov (2008), Elias and Ortiz de Adler (2006) explained the long term variations observed at mid and high latitudes in the Northern Hemisphere by geomagnetic activity effects. Finally three main factors must be considered: greenhouse gases, geomagnetic activity and Earth's magnetic field (Lastovicka et al., 2006;Elias, 2009). Cnossen and Richmond (2008) modelled the effects of the changes in the Earth's magnetic field from 1957 to 1997 on the ionopsheric hmF2 and foF2 data using model results from the TIEGCM. They can explain a variation of hmF2 of 20 km and of foF2 of 0.5 MHz over the atlantic Ocean and South America. These values are of the same order as the Phu Thuy observations in Asia.
In our case we observe similar long term variations of the F1-and F2-layers: increase of the critical frequencies. Concerning the E-layer the critical frequency increases and decreases. It is necessary to analyse more deeply the Ann. Geophys., 29,[197][198][199][200][201][202][203][204][205][206][207]2011 www.ann-geophys.net/29/197/2011/ geophysical conditions at the northern crest of equatorial anomaly in the Asian sector which are different from the geophysical conditions in the European and American longitude sectors.

Solar cycle variations
In Table 4, column 2, there is given the evolution of the correlation coefficients of ionospheric parameters with sunspot number. The correlation coefficient is rather good for foF2, foF1 and foE, and less so for the foEs and virtual heights. Ta- ble 5 gives the confidence intervals [99% (column 3)] for the correlation between sunspot number Rz and foF2, foF1, foE. The correlation coefficients are statistically significant with 99% confidence level by the Fisher's t-test. The good correlation seen above is explained by the ionization of the Earth's atmosphere. The solar X-ray and extreme ultraviolet (EUV) radiation control the Chapman layers (Rishbeth and Gariott, 1969). The difference in the correlation coefficients from one sunspot cycle to another can be explained by changes in the solar cycle intensity (Balan et al., 1993). Ouattara et al. (2009) found for Ouagadougou, located at the magnetic equator in the Northern Hemisphere, a good correlation between the sunspot number and foF2, foF1, and h ′ F1, but no correlation for foE, h ′ F2 and h ′ E. The absence of correlation between the sunspot number and foE at the equator can be explained by the existence of the equatorial electrojet, which drives instabilities and generates strong plasma waves which strongly affect the electron density (Farley, 2009).

Annual and semiannual variations
The annual and semiannual variations of all layers are given in column 3 of Table 4. foF2 exhibits the well known equinoctial pattern with two maxima at the equinox (Fig. 4). But we observe also an asymmetry between the two maxima which changes with the phase of the sunspot cycle (row 1):    sometimes the autumnal maximum is greater than the spring one (increasing phase of cycle 21 and 22 and maximum of cycle 22, see Fig. 4). But also the inverse behaviour could be found (minimum phase of cycle 20, decreasing phase of cycles 20 and 22, see Fig. 4). The foF2 semi annual variation is controlled by three mechanisms: (1) seasonal change of O/N 2 (Rishbeth and Setty, 1961;Rishbeth and Müller-Wodarg, 1999), (2) changes in the Sun-Earth distance (Yonezawa, 1959;Rishbeth et al., 2000a) and (3) the wave and tide forces transmitted to the thermosphere from the lower atmosphere (Zou et al., 2000). The asymmetry between the two equinoctial peaks results from the asymmetry of thermospheric parameters which influence the foF2 by neutral wind and composition (Balan et al., 1998). We observe in Fig. 4 also some small indications of the winter anomaly of foF2 (smaller foF2 values in summer than in winter) during the maximum phase and the decreasing phase of solar cycles 20 and 22. This is related to winter maximum of O atoms, implying the relative increase of light gases above the winter hemisphere (Scialom, 1974), due to dynamic influences in the thermosphere (Rishbeth and Müller-Wodarg, 2006). Ouattara et al. (2009) propose to explain the absence of the winter anomaly during cycle 20 by the influence of fluctuating wind stream activity. foF1 (Fig. 5) exhibits a semiannual variation during the minimum phase of solar cycle 20 (top left panel) and an annual variation during the maximum phase of solar cycles 20, 21 and 22 with a morning maximum (top right panel). During the increasing and decreasing phases of the sunspot cycle (bottom panels) foF1 exhibits a semiannual pattern, nevertheless the amplitude of the variation is small. This complex structure is not surprising because the F1-layer is the transition from the E-layer to the F2-layer, with a combination of both variations due to photochemical rates, and changes in atmospheric composition or temperature (Rishbeth and Kervin, 1968).
The annual variation of the h ′ F2 (Fig. 6) results from the solar cycle variations of thermosphere winds (Rishbeth et al., 2000b).

Diurnal variation
The diurnal variation of foF2 (Fig. 8), exhibits the same pattern for all the phases of all the solar cycles, i.e. a minimum at 05:00 LT and a maximum at 14:00 LT. This is explained by the photochemical processes and the transport.
-foF1 diurnal variation (Fig. 9) has a maximum at 12:00 LT. This is explained by the control of the solar zenith angle (Rishbeth and Gariott, 1969). We can observe a large dispersion of this parameter during the decreasing phases of the three solar cycles (see bottom right panel of Fig. 9) -h ′ F2 diurnal variation has a maximum at 12:00 LT (Fig. 10). The dispersion is large for all the solar cycle phases, except the decreasing phases.
-h ′ F1 diurnal variation (Fig. 11) exhibits a minimum near noon and the maximum near midnight. Other minima and maxima occur in the morning, at about 04:00 or 05:00 LT and at about 18:00 or 19:00 LT but the afternoon maximum is markedly smaller (except during the maximum phase of solar activity).
The different changes of h ′ F1 and h ′ F2 are attributed to temperature changes (Appleton, 1935;Lawden, 1969;Lejeune, 1972). Concerning the F1-layer it is due to its transition between the E-region where temperature oscillations are mainly semidiurnal, and the F2-region where the diurnal component prevails (Fontanari and Alcayde, 1974). The second peak of h ′ F1 during maximum phase seems to be the well-known post sunset peak usually attributed to the ExB ionospheric electric field pulse (Fejer et al., 1979;Adohi et al., 2008).

General conclusions
The results from our ionosonde series provide the longest database (40 years) yet available on ionospheric layers variations at the northern tropical crest of ionization in the Asian longitude sector. This unique series of results is presented with abundant curves and tables. The main characteristics of the ionospheric parameters observed at Phu Thuy are: -The critical frequencies foF2, foF1 increase with solar cycle.
-The correlation coefficients between foF2, foF1, and foE with solar sunspot number Rz are rather good but markedly smaller for foEs and virtual heights.
-The critical frequency foE increases from solar cycle 20 to solar cycle 21 and then decreases during the decreasing phase of cycler 21. The geomagnetic influence should be mentioned by an own point.
-The geomagnetic influence on mean values of different ionospheric parameters is very small at Phu Thuy as shown in Fig. 3.
-Seasonal variation of foF2 parameter shows semiannual, winter anomaly (the winter anomaly is only a small and not always detected phenomenon at Phu Thuy) and asymmetry between the two equinoctial peaks changing with the solar phases.
-The virtual heights of the F1 and F2 layers are highest in summer months.
-The diurnal variation of foF2 exhibits a minimum at 05:00 LT and a maximum at 14:00 LT. h ′ F2 has a maximum at 12:00 LT.
This work constitutes a starting point for the study of average variations. Further investigations with models are needed to explain all these characteristics, and particularly those related to the long term variations of the layers and the variations related to solar sunspot cycle phases.