The effects of the pre-reversal ExB drift, the EIA asymmetry, and magnetic activity on the equatorial spread F during solar maximum

We use a digisonde at Jicamarca and a chain of GPS receivers on the west side of South America to investigate the effects of the pre-reversal enhancement (PRE) in ExB drift, the asymmetry (Ia) of equatorial ionization anomaly (EIA), and the magnetic activity (Kp) on the generation of equatorial spread F (ESF). Results show that the ESF appears frequently in summer (November, December, January, and February) and equinoctial (March, April, September, and October) months, but rarely in winter (May, June, July, and August) months. The seasonal variation in the ESF is associated with those in the PRE ExB drift and Ia. The larger ExB drift (>20m/s) and smaller |Ia| (<0.3) in summer and equinoctial months provide a preferable condition to development the ESF. Conversely, the smaller ExB drift and larger |Ia| are responsible for the lower ESF occurrence in winter months. Regarding the effects of magnetic activity, the ESF occurrence decreases with increasing Kp in the equinoctial and winter months, but not in the summer months. Furthermore, the larger and smaller ExB drifts are presented under the quiet (Kp<3) and disturbed (Kp?3) conditions, respectively. These results indicate that the suppression in ESF and the decrease in ExB drifts are mainly caused by the decrease in the eastward electric field.


Introduction
Equatorial spread F (ESF) takes its name from the rangedispersed ionogram at stations near the dip equator (Booker and Well, 1938).The ESF usually occurs just after local sunset and the scale sizes of irregularities range from a few centimeters to a few hundred kilometers.Based on the previous investigations, the basic mechanism of ESF onset is the gravitational Rayleigh-Taylor (GRT) instability in conjunc-Correspondence to: C.-C. Lee (cclee@cyu.edu.tw)tion with the E×B drift instability (e.g.Kelley, 1989;Abdu, 2001).The growth rate of instabilities crucially depends on the density gradient that becomes steeper in the nighttime equatorial F-region.Moreover, due to the nonlinear evolution, these instabilities would fully develop into vertically elongated wedges of plasma depletion from bottomside to topside of the F-layer (e.g.Tsunoda et al., 1982;Zalesack et al., 1982;Sultan, 1996).
The scope of this paper is to look for the effects of the first and second mechanisms listed above on the ESF onset.For two decades, the earlier works have investigated the variability of E×B drift on various timescales from day-to-day to season, on longitude, and on solar and geomagnetic activities.Fejer et al. (1979) examined the vertical drift measurements of Jicamarca (12 • S, 76.9 • W) incoherent scatter radar (ISR) during 1968-1976 to find the associated variations in season and solar cycle.Abdu et al. (1983) pointed out that a necessary condition for the ESF occurrence is the E×B drift from analysis of Fortaleza (4 • S, 38 • W) ionosonde data during January-April and September-December in 1978.Recently, Fejer et al. (1999) analyzed the data of the Jicamarca ISR during 1968-1992, and found that the E×B drift and ESF occurrence depend on solar flux, season, and magnetic activity.Whalen (2002) also proposed that the ESF occurrence and E×B drift are a function of magnetic activity and season, using an array of ionospheric sounders in the western American sector during 1958.
Among the studies made on the effects of the second mechanism, most studies employed the short-duration data to explore the day-to-day variability of the meridional wind and associated EIA asymmetry.Maruyama and Matuura (1984) proposed that the asymmetry in the electron density distribution, which is caused by the meridional winds, could suppress the ESF activity.However, they did not show the seasonal variation in the asymmetry in the density distribution.Furthermore, Mendillo et al. (1992) have noted the possible contributions of transequatorial wind in limited samples of ESF day-to-day variability, using an all-sky imaging system at Kwajalen Atoll (9.4 • N, 167.5 • E).Sastri et al. (1997) applied two ionosondes in Brazil to examine the correlation between the ESF occurrences and the associated meridional winds in June of 1978-1981. Mendillo et al. (2000, 2001) utilized the GPS data near the equinox to calculate the asymmetry of EIA and to test the day-to-day variability of ESF.
In the present paper, we first attempt to investigate the seasonal effect of the EIA asymmetry on the ESF development, besides the seasonal effect of the PRE E×B drift velocity and the magnetic activity.We conducted a one-year observation on the equatorial ionosphere during April 1999-March 2000.A digisonde and a chain of GPS receivers on the west side of South America have been applied to this work.The minimum virtual height of the F-layer (h F) and the latitudinal distribution of the vertical total electron content (VTEC) are used to derive the PRE E×B drift and the asymmetry of EIA, respectively.Further, we utilize these data to study the dependence of ESF on the PRE E×B drift velocity, the asymmetry of EIA, and the magnetic activity (K p ).

Experiment setup
Figure 1 shows the locations of a digisonde and all nine GPS receivers.The Jicamarca digisonde (12 • S, 76.9 • W, geomagnetic latitude: 1.28 • N), near the geomagnetic equator, recorded ionograms from April 1999 to March 2000.The presence/absence of ESF and the ionospheric parameters (e.g.h F, foF2 -maximum frequency of F-layer) are determined by both the ARTIST program (Reinisch and Haung, 1983;Reinisch, 1996) and manual work.Notably, we focus on the ESF appearing during 18:00-24:00 LT (LT=UT-5 h).The upward PRE velocity is derived by the temporal rate of of h F, dh F/dt, which is described as the E×B drift velocity (Bittencourt and Abdu, 1981).Since the E×B velocity generally reaches its maximum value before the onset of ESF (e.g.Fejer et al., 1999;Whalen, 2002), we apply the maximum value of dh F/dt between 18:00 LT and the time of ESF onset to the statistical analyses.For example (Fig. 2), as the ESF appears from 19:15 LT on 15 April 1999, the maximum value of dh F/dt, ∼32 m/s, is adopted in the following studies.
The data of the nine GPS sites (Fig. 1 and Table 1) are retrieved from the International GPS Service (IGS) for Geodynamics.Two of the GPS sites (BOGT and RIOP) are situated north of the geomagnetic equator, and the BOGT site is located near the northern crest of EIA.The other seven GPS receivers (AREQ, UNSA, TUCU, CFAG, CORD, VALP, and SANT) are placed south of the equator.The AREQ is close to the equator, while the TUCU is located near the southern crest.These nine GPS sites offer the equivalent VTEC along the common meridian of ∼75 • W (see Mendillo et al., 2000 for reference).Moreover, an asymmetry index of EIA (I a ), which is related to the transequatorial meridional winds, can be derived from the north versus south crest differences.The I a is represented as (N-S)/((N+S)/2) (shown in Fig. 3), where N and S are the VTEC values of the north and south crests, respectively (Mendillo et al., 2000).Based on Abdu (2001), the ESF development would depend upon the symmetry/asymmetry of EIA during the previous 0.5-to 1-h period.Therefore, we examine the I a between 18:00 LT and the time of ESF onset.While I a is positive/negative during this period, the maximum/minimum value of I a is used for the statistical analyses.Notably, I a is not estimated during December 1999-March 2000, because the BOGT data are not available in these months.

Seasonal variations in occurrence probability of ESF
Figure 4 displays the seasonal variation in the occurrence probability of ESF.This seasonal variation in the occurrence probability is the percentage of days on which at least one pre-midnight ESF event is observed daily.Clearly, the highest occurrence probabilities (100%) of ESF is in January, near the summer solstice; while the lowest values (∼32%) is in June, near the winter solstice.During the equinoctial months (March, April, September, and October), the occurrence probabilities are more than 75%.Generally, the occurrence probability is higher in summer (November, December, January, and February)/equinoctial months than in winter months (May, June, July, and August).In this investigation, the ESF is not categorized into three levels proposed by Whalen (2002), because only one ionosonde is used.Thus, the seasonal variation in ESF (Fig. 4) should be compared with that in the total bottomside spread F (BSSF) of Whalen (2002) (his Fig. 4), who analyzed the ionograms at Huancayo (12 • S, 75.3 • W) during 1958.Notice that the total BSSF consists of strong and weak BSSF and bubbles (Whalen, 2002), and Huancayo is located at 160 km east of Jicamarca.The comparison result shows that these two seasonal variations are similar: the highest and lowest occurrence probabilities are in summer and winter, respectively.

Seasonal variations in the PRE E×B drift
We examine the seasonal variation in the PRE E×B drift by deriving the monthly mean value of dh F/dt. Figure 5 shows the seasonal variation in the monthly mean value of dh F/dt (solid line) and the associated standard deviation (error bar), as well as the occurrence probability of ESF (gray histogram in background).It is apparent that two peak values, ∼28 and ∼32 m/s, appear in October and March, respectively; while the smallest dh F/dt, ∼14 m/s, is in June.The seasonal variation in dh F/dt is close to Fejer et al. (1999), who propounded that the largest and smallest vertical drift velocities are at the equinox and the June solstices, respectively.Consequently, the mean dh F/dt are ∼14-18 and ∼20-27 m/s during May-August and November-February, respectively.These mean values are also close to Fejer et al. (1999), in which the maximum values are ∼18 and ∼28-34 m/s, respectively, in May-August and November-February, during a high solar activity period.However, in March, April, September, and October, the values of ∼25-32 m/s are slightly smaller than the maximum value (∼33-48 m/s) of Fejer et al. (1999) during a high solar activity period.
In the equinoctial months, the larger dh F/dt is associated with the higher ESF occurrence.This demonstrates that the larger upward drift lifts the F-layer to higher altitudes, which not only results in favorable conditions for the GRT instability but the larger eastward electric field itself causes an E×B drift instability (Maruyama, 1988;Kelley, 1989).In contrast, the small dh F/dt in the winter months would not raise the F-layer to altitudes where it is high enough to generate irregularities.Thus, the ESF occurrences are lower in the winter months.These results reveal that the PRE E×B drift plays an important role as the seeding mechanism of ESF development (Sultan, 1996;Fejer et al., 1999;Kudeki and Bhattacharyya, 1999;Whalen, 2002).Additionally, there are also other seeding mechanisms such as gravity waves (Sultan, 1996; Kudeki and Bhattacharyya, 1999).
During the summer months, the mean values of dh F/dt, ∼20-27 m/s, are smaller than that at the equinox, but higher than that in the winter.Notice that the occurrence probability of ESF in January is the highest value (100%) between April 1999 and March 2000.Such kind of correlation between the dh F/dt and the ESF occurrence indicates that there would be another mechanism to help the ESF development, besides the larger upward velocity.According to Maruyama (1988) and Fejer et al. (1999), the late reversal time of upward drift should be a considerable reason for the highest occurrence probability at December solstice.

Seasonal variations in the asymmetry of EIA
Figure 6 shows the seasonal variation in the monthly mean value of I a (solid line), the associated standard deviation (error bar), and the occurrence probability of ESF (gray histogram in background).It is found that I a is negative and |I a | is greater than 0.3 during May-August (winter months).
In April, September, and October (equinoctial months), I a is also negative, but |I a | is smaller than 0.3.In contrast, I a is positive and |I a | is smaller than 0.3 in November (summer months).According to Mendillo et al. (2000Mendillo et al. ( , 2001)), the I a index is associated with the transequatorial meridional winds that would distort a presumed symmetrical EIA.Thus, the meridional winds are generally southward and northward in April-October and November, respectively.Moreover, the greater values (>0.3) of |I a | during May-August indicate that the associated meridional winds in the winter months are stronger than that in the equinoctial (April, September, and October) and summer (November) months.These are similar to the meridional winds used by Maruyama (1988), in which the wind is northward and stronger near the winter solstice (in his Fig. 3).
As shown in Fig. 6, the greater |I a | and lower ESF occurrence concurrently appear during May-August.Based on Maruyama and Matuura (1984), Maruyama (1988), and Mendillo et al. (1992), this negative correlation between these two parameters indicates that the asymmetric EIA is related to the inhibition of ESF development.This is because the asymmetry of EIA would form an asymmetry in distribution of conductivity and recombination rates along the field line.Then, the asymmetry in conductivity and recombination rates further suppresses the GRT growth rate (Maruyama and Matuura, 1984;Maruyama, 1988;Mendillo et al., 1992).Besides the suppressive effect of I a , the smaller dh F/dt in the winter months (Fig. 5) is not helpful in developing the ESF either.Therefore, the lower ESF occurrence is caused by the greater |I a | and the smaller dh F/dt.
On the other hand, the smaller |I a | and higher ESF occurrence concurrently appear in April, September, and October.This result shows that the |I a | is negatively correlated with the ESF occurrence, and demonstrates that the asymmetry in conductivity and recombination rates would be not strong enough to suppress the GRT growth rate.Recall that, in the equinoctial months (Fig. 5), the values of dh F/dt are larger.Thus, the higher ESF occurrence is due to not only the larger dh F/dt but also the smaller |I a | during the equinoctial months.Additionally, the smaller |I a | and higher ESF occurrence are also seen simultaneously in November.Similar to the equinoctial months, the higher ESF occurrence is caused by both the smaller |I a | and larger dh F/dt.

Dependence on the magnetic activity (K p )
The ESF is examined on each day in relation to the magnetic activity on that day, which is taken as the average value of K p recorded during the 6 h prior to the onset of ESF (e.g.Fejer et al., 1999;Whalen, 2002).Notice that the average K p is generally derived from the interval of 18:00-24:00 UT (13:00-19:00 LT).
The number of days on ESF are organized by season and plotted versus the average K p in Fig. 7.For the equinoctial months (Fig. 7a), the ESF occurs frequently and rarely when K p is 1 and 5, respectively.Except K p =2, the number of days tends to decrease with increasing K p in the equinoctial months.Moreover, a descending trend from K p =1 to 5 is obvious in the winter months (Fig. 7c), but not in the summer months (Fig. 7b).The decreasing trends in the equinoctial and winter months indicate that the increasing magnetic activity progressively suppresses the ESF generation in these two seasons (Fejer et al., 1999;Whalen, 2002).However, these distributions are not consistent with that of total BSSF in Whalen (2002) (his Fig. 5), which reported that the decreasing trends existed in the equinoctial and summer months, not in the winter months.This might be because the distribution of K p days is different for each seasons between 1958 and 1999-2000.
To further reveal the geomagnetic effect on dh F/dt and I a , we examine the monthly mean values of dh F/dt (Fig. 8a) and I a (Fig. 8b) under quiet (K p <3) and disturbed (K p ≥3) conditions during April 1999-March 2000.In the equinoctial months, the mean dh F/dt as K p ≥3 is less than that as K p <3, and the differences are ∼3-8 m/s.For the summer months, dh F/dt is larger under the quiet condition and the differences are ∼2-5 m/s, except February.Furthermore, for the winter months, dh F/dt as K p <3 is also larger in May and August, and the differences are ∼8 and ∼3 m/s, respectively.Notice that no dh F/dt as K p ≥3 is shown in June, because all K p are smaller than 3. Furthermore, only one dh F/dt as K p ≥3 is in July, and this large upward velocity is related to the prompt penetration effect (Fejer and Scherliess, 1997;Fejer et al., 1999).Generally, the dh F/dt is smaller under the disturbed condition, and these results are similar to Fejer et al. (1999), who proposed that the maximum vertical drift is, respectively, larger and smaller under the quiet and disturbed conditions during a high solar activity period (their Fig. 7).The decrease in dh F/dt indicates that the maximum eastward electric field of PRE is reduced by the disturbed magnetic activity (Scherliess and Fejer, 1997).Moreover, this decrease is due to the disturbance dynamo electric fields which decrease the upward drift velocity near sunset (Fejer et al., 1999).
Regarding the I a (Fig. 8b), it is found that the mean |I a | as K p <3 is larger than that as K p ≥3 in August and September.In contrast, the mean |I a | as K p <3 is smaller than that as K p ≥3 in October and November.Moreover, the mean |I a | is nearly the same between the quiet and disturbed conditions in April, May, and July.These results suggest that the effect of the magnetic activity on I a seems not to be evident.Since the I a is related tothe meridional winds, the disturbed magnetic activity does not obviously affect the strength of the winds.

Summary and conclusion
We have analyzed the ionogram and TEC data to investigate the long-duration effects of the PRE E×B drift, the EIA asymmetry, and the magnetic activity on the ESF generation.The ionogram and TEC data during April 1999-March 2000 are obtained by a digisonde and a chain of GPS receivers, respectively.It is remarkable that this work is a first attempt to examine the seasonal variation in the EIA asymmetry.The results point out that the dependence of ESF on season and magnetic activity can be explained by the corresponding effects on the E×B drift and EIA asymmetry.
For the seasonal variation, the ESF appears frequently in the summer and equinoctial months, but rarely in the winter months.Such kind of occurrence variation is evidently correlated to the magnitude of E×B drift.The correlations in summer and equinoctial months demonstrate that the ionosphere is lifted to higher altitudes where the gravitational drift term is dominant in the GRT growth rate, leading to the development of the ESF, when the upward drift velocities are larger than 20 m/s.Moreover, the influence of the EIA asymmetry (I a ) is also significant regarding the ESF occurrence.The value of |I a | is generally anti-correlated to the ESF occurrence.This anti-correlation indicates that the EIA asymmetry (|I a |>0.3) could suppress the GRT growth rate, because of the asymmetry in distribution of conductivity and recombination rates.
For the magnetic activity, we have shown that the number of ESF days decreases with increasing K p in the equinoctial and winter months, but not in the summer months.This descending trend suggests that the magnetic activity progressively suppresses the ESF generation in those two months.Additionally, our results shows that the E×B velocity is larger and smaller under the quiet and disturbed conditions, respectively.This indicates that this ESF suppression would be due to the effect of the magnetic activity on the E×B velocity.Contrarily, it is found that I a is not obviously related to the magnetic activity.

Fig. 1 .
Fig. 1.Map of a digisonde and a chain of nine GPS receivers.The solid dot represents the Jicamarca digiosonde, while the triangles symbolize the nine GPS stations on the west side of South America.The circles at each GPS site give the field-of-view for a 400-km intersection height at an elevation angle at 30 • .

Fig. 2 .
Fig. 2. Sample for h F (bold line) and dh F/dt (thin line) on 15 April 1999.The maximum value of dh F/dt is ∼32 m/s.

Fig. 3 .
Fig. 3. Schematic illustrating a typical latitudinal distribution of VTEC values.The equation of an asymmetry index I a of the EIA is represented in the upper-right corner.

Fig. 4 .
Fig. 4. Seasonal variation in the occurrence probability of ESF during April 1999-March 2000.The highest value (100%) of the occurrence probability appears in January, while the lowest value (30%) is at June.

Fig. 5 .
Fig. 5. Seasonal variation in the monthly mean value of the dh F/dt (soild line) and the associated standard deviation (error bar) during April 1999-March 2000.The gray histogram in the background represents the occurrence probability of ESF.Two peak values of dh F/dt appear at October and March, while the smallest value is at June.

Fig. 6 .
Fig. 6.Seasonal variation in the monthly mean value of I a (soild line) and the associated standard deviation (error bar) during April 1999-March 2000.The gray histogram in the background represents the occurrence probability of ESF.Notice that I a is not estimated during December 1999-March 2000, because the GPS data of BOGT are not available in these months.

Fig. 7 .
Fig. 7. Occurrence of ESF on each day for each of the equinoctial (a), summer (b), and winter (c) months plotted as a function of K p determined as the average during the 6 h prior to measurement.

Fig. 8 .
Fig. 8. Seasonal variations of the monthly mean values of dh F/dt (a) and I a (b) between April 1999 and March 2000.The black thin and gray bold lines represent the mean values under the disturbed (K p ≥3) and quiet (K p <3) conditions, respectively.The error bars are the associated standard deviation.

Table 1 .
The locations of the GPS stations on the west side of South America.