The analyses presented here consider equatorial nighttime disturbance conditions as indicated by spread-F traces on ionograms, which record the presence of medium-scale structures (MSSs) created by the passage of atmospheric gravity waves. Relationships between these MSSs and geomagnetic activity (GA) have been examined statistically, particularly considering the ionospheric height (h′F) changes (increases and decreases) associated with the presence or absence of these MSSs. It is shown that for the postsunset height increases with spread-F present, the GA a few hours earlier has well-defined low levels. In contrast, if these height increases are not accompanied by spread-F, the GA a few hours earlier is high. When there are no height rises, similar inverse associations with GA are recorded for both (1) spread-F present and (2) spread-F absent. For the postsunset height decreases, when no spread-F occurs, high GA levels are well defined at times about 7 h earlier. Thus all these results indicate an inverse relationship between the MSSs and GA. The experimental evidence suggests that the presunrise height rises are produced by large-scale traveling ionospheric disturbances generated in auroral-zone regions by substorm onsets.
Fejer et al.  have reviewed equatorial VHF radar recordings at Jicamarca, made over two decades, as they apply to vertical ion drift velocities and ionospheric height changes, and associated equatorial spread-F (ESF). Bertoni et al.  have compared observations of vertical drift velocities using a Digisonde and the Incoherent Scatter Radar at Jicamarca. They found good agreement between these two recording systems. This investigation involves an examination of the height changes and related spread-F, as measured by ionograms at the equatorial station, Huancayo, for a sunspot-maximum (Rzmax) period. The 4 years used were 1979–1982, when the average yearly sunspot number was 141. The data used were obtained from the hourly tabulations of h′F and foF2 as recorded by ionograms. The foF2 data contains a descriptive letter F which indicates spread-F occurrence (SFO). One of the aims of this investigation has been to examine, to what extent, these ionogram-recorded data relate to geomagnetic activity (GA) a few hours earlier, and as identified by the KP indices. Some of the current literature involves examinations, to be discussed later, of relationships between ESF produced by electric fields and geomagnetic storms. In contrast, the analyses here will examine ESF without considering electric fields and also they will examine associations the ESF might have with GA at any level (e.g., GA high or low). The data were suitable for the statistical analyses used which involved the superposed-epoch (SE) method. For the figures to be presented, the SE method produces results for a center day (CD) and 10 d either side, with the standard error of the mean shown for some days. The final values obtained by this method are averages of all the entries which refer to events on particular days. One advantage of the SE analyses is the extent of the data available, as except for a few outages, it is documented at hourly intervals for every day for the 4 years used.
 In equatorial regions, small-scale irregularities (SSIs), recorded after postsunset height rises, are known as equatorial spread-F (ESF). Associated with these height rises are larger structures which are called plasma bubbles. Here these plasma bubbles will be referred to as medium-scale structures, for which the term ESF will also be used. These are found to also be present in the absence of height rises. The inclusion of plasma bubbles as part of the ESF phenomenon allows comparisons to be made with midlatitude disturbance conditions. Structures similar to these plasma bubbles exist in midlatitudes, and also are responsible for the phenomenon of spread-F. The characteristics of these midlatitude events are reviewed by Bowman . For equatorial latitudes, Lin et al.  have found a close relationship between atmospheric gravity waves (AGWs) and plasma bubbles. The central feature of this investigation has been an examination of two interesting aspects of ESF reported by Fejer et al. . One aspect relates to the fact that for some postsunset height rises no spread-F occurs. The other concerns occasions when there are postsunset height decreases. For the postsunset interval, height changes for Rz max will be compared with those for Rz min, and also for the presunrise interval height changes and associated SFO will be examined. Important to this investigation is the literature for equatorial and midlatitude locations which indicates that tilted isoionic surfaces resulting from the passage of AGWs are primarily responsible for the spread-F traces on ionograms. For earlier midlatitude results the term medium-scale traveling ionospheric disturbance (MS-TID) was used. This relates to the ionospheric disturbance conditions created by the AGWs that travel in the neutral atmosphere (see relevant references given by Bowman ). The use of either term is probably appropriate. Since equatorial papers generally use the term AGW [e.g., Argo and Kelley, 1986; Lin et al., 2005], on most occasions it will be used here.
Flaherty et al.  have shown that the SFO recorded by HF-radar measurements in equatorial regions results from strong radio wave signals. They found that the signal strengths of echoes from oblique paths are similar to those for vertical incidence. Flaherty et al.  also found off-vertical radio wave angles as large as 20°, and a range of off-vertical angles were also reported by Reinisch et al. . Sales et al.  tracked a disturbance out to 300 km from the zenith, where the off-vertical angle was 50°. With the aid of four ionosondes at an equatorial location, Whalen  used oblique range spread-F echoes on ionograms to track a single isolated bubble over a distance of over 1000 km. The bubble traveled eastward with a speed comparable with those of AGWs.
Bowman , in a review for the midlatitude results, indicates associations between the recording of spread-F traces on ionograms and the passage at the same time of AGWs [see also McNicol et al., 1956]. Bowman  using mainly spread-F traces on ionograms tracked, at 5-min intervals, three isolated tilted isoinic structures out to horizontal displacements from the zenith by as much as 300 km. Other electronic equipment recorded, for each event, four wave train cycles with periodicities of 12 min and average AGW speeds of 60 ms−1. For midlatitudes it is known from fixed-frequency transmissions that the off-vertical radio wave signals have the same strengths as the vertical-incident signals [McNicol et al., 1956; Bowman et al., 1986; Bowman, 1960, 1988, 1996a]. This is the case even when the ionograms appear diffusely [Bowman, 1960; Bowman et al., 1986; Bowman and Hajkowicz, 1991]. For these larger structures, possibly associated with radio waves of high signal strengths, the analyses here will consider the term medium-scale structures (MSSs), which was used by Argo and Kelley . The acronym ESFm will be used for these MSSs as recorded by spread-F traces on ionograms to distinguish them from the SSIs, (small-scale irregularities), as measured by VHF radars, for which the acronym ESFs is proposed. The term “macroscale” has been used [Bowman, 1990] to refer to midlatitude ionospheric structures, identified by electron density depletions and moderate height rises. They occur during the passage of AGWs that produce spread-F traces on ionograms and last for the duration of the passage, which can extend to 1 h. It seems possible that some of the total electron content depletions recorded during total electron content (TEC) measurements of equatorial plasma bubbles [e.g., Huang, 1990] can also be described as macroscale structures. The annual variation of SFO for midlatitudes [Bowman, 1992a] and also for the postmidnight hours of equatorial latitudes [Bowman, 2001] is different over the year. Thus it has been found convenient to divide the year into three periods. Period J involves months around the June solstice (May, June, July, and August), period D for the December solstice months (November, December, January, and February), and period E for the equinoctial months (March, April, September, and October). Also, to achieve a better understanding of disturbance differences between postsunset and presunrise events, the night has been divided into three 3-h intervals, interval P (INTP), interval Q (INTQ) and interval R (INTR). From the tabulations, for any hour the unit 1 has been used to indicate an ionogram spread-F condition, irrespective of the disturbance level which might be indicated by the extent of the range spreading. For each 3-h interval SFO will be recorded at different levels for which the term 3h-SFO will be used. The 3h-SFO, which can be 0, 1, 2, or 3, has been determined for every interval for every day of the 4-year period. For Huancayo local time (LT) is 5 h earlier than universal time (UT) (i.e., LT = UT − 5 h). Therefore using local times, INTP involves 2000, 2100 and 2200, INTQ 2300, 0000 and 0100, and INTR 0300, 0400 and 0500.
 Using the h′F tabulations, significant increased or decreased levels of h′F have been defined relative to median values for each hour. An event for the interval is recorded if, for any hour of the 3-h interval, the height was ≥40 km above the median value. The term Hh′F is used. For decreases, Lh′F relates to those occasions, when for at least 2 h of the 3-h interval, h′F values occur ≤40 km below the median values. Earlier, for an equatorial region, Bowman  considered an inverse relationship between GA and abrupt changes to no spread-F for intervals when spread-F occurs almost every night. The term “dropout” was used. Here, more generally, for any zero SFO the term Z-SFO will be used. As will be explained later, the analyses need values for GA in the presunset hours. For this, the sum of two adjacent KP intervals 8 and 1 (Kp ∑ 81) has been used covering, in local times, 1600–2200, or universal times 2100–2400 and 0000–0300. Also, for the premidnight period, Kp ∑ 12 is used involving, in local times, 1900–0100, or universal times 0000–0600.
 It has been shown that ionogram-recorded SFO varies significantly over the year for both midlatitudes [Bowman, 1992a] and the AM period of equatorial latitudes [Bowman, 2001]. This variation involves maxima for the June and December solstices and minima for the equinoxes. Bowman [1995a] indicates that for equatorial regions the postsunset SFO is suppressed for GA at 1100 or 1800 (local time). The postmidnight equatorial spread-F is increased for GA around 2200 (local time).
 Earlier investigations have shown the importance of AGW wave amplitudes in the recording of MSSs by ionograms. Experimental evidence exists which shows, for both equatorial latitudes and midlatitudes, that most of the ionogram spread-F traces result from specular reflections from tilted isoionic surfaces because of the large off-vertical angles involved in the radio wave reflections [Bowman, 1990]. Off-vertical angles as large as 40° are not unusual on occasions when the range spread approaches or exceeds 100 km [Bowman, 1981]. However, the ionograms can only detect disturbances of magnitude above a certain level. If the available off-vertical angles are less than about 15° the spread range is so small that the spread cannot be identified [Bowman et al., 1987]. In midlatitudes, Clarke [1971, 1972], using a directional ionosonde, measured the off-vertical angles used to create spread-F traces. Sixty of these angles ranged from 20° to 50°, twenty of them in the range 40° to 50°. Thus the AGW wave amplitudes responsible for the tilted surfaces seem important in assessing the magnitude of any disturbance. Shallow-wave amplitudes, which do not lead to spread-F traces on ionograms, have been shown to occur regularly in equatorial latitudes. Sastri  for the postsunset period and in the absence of spread-F has recorded AGWs with periodicities in the range 5–33 min. The wave amplitudes of these AGWs increased with height [see also Subbarao and Krishna Murthy, 1994]. Similar observations were made in midlatitudes by Bowman [1995b], who detected AGWs with shallow-wave amplitudes, which were often present when spread-F was not recorded on ionograms. Whether or not spread-F is recorded seems to depend on wave amplitudes. The spatial scale of ESFm will depend on medium-scale structures with periodicities in the range 5–33 min [Sastri, 1995] and AGWs with speeds averaging 97 m s−1 [Huang, 1990].
 Earlier results reported by Bowman  have been used in this present paper. The references mentioned therein indicate evidence for ionospheric disturbances which travel toward the equator following GA. The term traveling ionospheric disturbance (TID) can be used for these events. Figures 2 and 3 of Bowman  have been reproduced here as Figures 1 and 2. The SE analyses for Figures 1 and 2 involve the detection of height changes (relative to monthly median values) at equatorial locations following substorm onsets. The substorms were identified as negative bays in the H component of the Earth's magnetic field. The onsets were well defined, and each substorm needed to have some degree of isolation from other substorm activity. An accuracy of 15 min was achieved for both height changes and substorm onsets. Three delays were detected, two for height increases and one for a height decrease. The latter is of direct interest here for a further examination of the height decrease reported by Fejer et al. . As many as 584 substorm events have been used for Figure 1, and results of high statistical significance are shown. The standard deviation displacements range from 6.3 σ to 8.2 σ. A further examination of Figures 1 and 2 will show later that TIDs seem likely to be associated with the delays recorded. In equatorial regions, electric fields are acknowledged as being responsible for postsunset height rises. At other times of the night, electric fields are also thought to produce height rises [Fejer et al., 1999]. However, height rises may also be caused by TIDs.
Table 1 lists the subdivisions used here, Table 2 lists the meanings of the acronyms, and Table 3 lists the geomagnetic coordinates of the stations used in the analyses. The additional acronyms of H3h′F, H0h′F, and L0h′F will be explained in sections 2.3 and 2.4.
INTP, interval P; INTQ, interval Q; INTR, interval R.
June solstice months
December solstice months
Table 2. Abbreviations
SFO level for 3 h
Kp ∑ XX
sum of two adjacent Kp indices
high h′F values
Hh′F for Z-SFO
Hh′F for 3h–SFO = 3
low h′F values
Lh′F for Z-SFO
equatorial spread-F related to MSSs
equatorial spread-F related to SSIs
Table 3. Geomagnetic Coordinates
 The analyses will include a comparison of height increases for this Rzmax period with those of an Rzmin period. The Rzmax height increases or decreases, along with any associated SFO, have been analyzed for possible associations with GA a few hours earlier. The spread-F events (MSSs), when no height increases occur, have also been analyzed similarly. The review by Fejer et al.  for the ESFs is extensive and involves certain features associated with vertical drifts, height changes and recordings of scattering from 3-m irregularities (SSIs). Here, to the extent that ionograms can detect disturbance conditions, data have been examined to ascertain if some of these characteristics, recorded by the VHF radar at Jicamarca, also apply for the larger-size MSSs (ESFm). There are also further analyses of an earlier result.
2.1. h′F Variations for Rzmax and Rzmin
 Figures 1c and 1d of Reddy  indicate that the equatorial electrojet has maximum levels in the equinoctial months (period E) of Rzmax years. In view of this the analyses for this and other sections have concentrated mainly on the period E data. For period E the average monthly median values of h′F at 2000 LT are approximately 100 km greater for Rzmax than for Rzmin, these two values being 372 km and 276 km, respectively. The Rz max data were obtained from the tabulations mentioned in the Introduction. The Rz min data came from similar tabulations for the years 1974–1977 when the average yearly sunspot number was 22. The number of occurrences of the Rzmax Hh′F height values has been compared with that for Rzmin for INTP, INTQ and INTR, with the results shown by Table 4. The percentage occurrence for each Hh′F value is given in brackets. When compared, occurrence levels are quite different for INTP, but essentially the same for INTQ and INTR. Table 5 shows the distribution of event numbers over different height ranges. For Rzmax heights as large as 600 km are recorded, whereas for Rzmin this value is 400 km. The Rzmax events are concentrated in the range ≥400 < 500 km, the range for Rzmin being only ≥300 < 350 km. Thus for the Rzmax years, events occur much more frequently and involve much greater heights.
Table 4. Hh′F Events for Equinoctial Months
Table 5. Ranges for Postsunset Hh′F Events for Equinoctial Months
 In section 1 some detail is given concerning a paper by Bowman , which is relevant to this investigation, particularly for the events analyzed related to height decreases (Lh′F) at the equator. The diagrams illustrated by Figures 1 and 2 show the results of SE analyses involving Δh′F changes relative to the substorm onset times determined at the auroral-zone station, College, for an Rzmax period. Figure 1 shows Δh′F changes for Okinawa, an anomaly crest station, and Figure 2 shows Δh′F changes for Kodaikanal, a station located on the equator. Each division on the ordinate scales of these diagrams represents 2σ. A general assessment of Figures 1 and 2 indicates that three isolated and statistically significant displacements occur with different delays, to be called event 1, event 2, and event 3. For example, for KP = 3, 4, 5 of Figure 1, the delays are 3.25, 5.0, and 8.5 h. One feature is that the delays are shorter for higher levels of GA. Another feature of interest is that event 2 involves a negative displacement.
 Using the rapid onsets of auroral absorption during substorm activity, Hajkowicz  detected possible sources of large-scale traveling ionospheric disturbances (LS-TIDs) at extended regions of longitude. That is, line sources were proposed, rather than point sources. Assuming line sources for the events shown in Figure 1, it is possible to calculate reasonable estimates of the TID propagation speeds for these events taking into account the north/south distance to the auroral-zone station of Tixie Bay which locates approximately at the same longitude as Okinawa. The delay, as mentioned, was measured with an accuracy of 15 min. Table 6 lists the speeds for events 1, 2, and 3. For event 1 the speeds vary from 770 m s−1 to 470 m s−1 depending on the level of GA, and for event 3 the speeds are around 200 m s−1. For event 2, the negative displacement, the speed is 310 m s−1 for KP = 3, 4, 5 for a delay of 5.0 h. Six LS-TIDs, originating at substorm onsets, were tracked by Bowman [1992b] for thousands of kilometers across either Japan or Australia. The average speed determined for these six events was 450 m s−1 when the KP index was 5. When compared with 470 m s−1 for event 1, this seems to be evidence that supports the view that event 1 involves LS-TIDs, which occur regularly at equatorial regions. For LS-TID speeds, see also Tsugawa et al. . An individual LS-TID has been tracked from a significant substorm onset all the way to the equator. Three hours after the substorm onset this disturbance was recorded at Huancayo where spread-F was recorded on ionograms as it was at some of the stations at other latitudes [Bowman, 1978]. The speed was 740 m s−1, and the KP index was 7, which is comparable with 770 m s−1 for event 1 for KP = 7, 8, 9. Event 3 is probably another LS-TID. Table 6 shows that if event 2 is an LS-TID, the speed is 310 m s−1. Figure 2 involves College substorms and displacements (events 1, 2, and 3) with various delays for Kodaikanal, a station close to the equator. It is not reasonable to calculate speeds because of the large longitude separation. However, because for Kp = 5, 6, 7, the delay is 7.0 h for event 2, whereas for Kp = 3, 4, 5 the delay is 8.5 h, it suggests that an LS-TID may be involved. A similar analysis [Bowman, 1978] for the low-latitude station, Concepcion, used substorm onsets for Halley Bay. The LT for Concepcion is 5 h before UT. This result for Concepcion is shown by Figure 3 of Bowman , where significant displacements are recorded for events 1 and 3 (as defined) but there is no displacement for event 2. For these earlier analyses [Bowman, 1977, 1978] some Kp intervals were included in two interval sets (e.g., Kp intervals 3, 4, 5 and Kp intervals 5, 6, 7). Because of the larger number of events used it seems likely that the higher levels of statistical significance for the lower GA levels can be explained. The very disturbed conditions when GA is high (Kp = 7, 8, 9) may produce additional LS-TIDs giving positive displacements (see Figure 1a), which might counteract any possible negative displacement similar to the one recorded on Figure 1c. Bowman [1992b] has detected an LS-TID in each hemisphere originating from the same substorm onset. However, as their speeds are different they may not arrive at the equator at the same time.
2.3. Presence or Absence of Spread-F With or Without Height Increases
 The term H3h′F is used when spread-F is recorded during all 3 h of an interval (i.e., 3h−SFO = 3). If no spread-F is recorded (Z-SFO), H0h′F is used. Figure 3 shows the results of SE analyses that consider the GA (Kp ∑ 81) relative to postsunset height rises. For H0h′F, Figure 3a shows results involving 47 events for INTP of periods D and E, while for Figure 3b, which used H3h′F there were 139 events for INTP of periods D and E. The significant center day (CD) displacements for these two figures oppose each other, Figure 3a being positive and Figure 3b being negative. The CD displacements give results for the day of an event. The Figure 3b displacement is more prominent. The ionogram data show that although Hh′F events are often associated with 3h-SFO, the spread-F can also occur at other times. For example, during INTP of period D there were 48 occasions of spread-F when height increases occurred, whereas the spread-F was also present on 45 occasions when there were no height increases. The state of disturbance conditions for no Hh′F events (whether spread-F is present or absent at these times) is examined, and the results obtained by SE analyses are presented by Figure 4. Here the GA (Kp∑81) is examined relative to 91 occasions of Z-SFO for INTP of period E, and the results are shown by Figure 4a. The Figure 4b results give the average GA levels relative to 84 occasions of 3h-SFO = 3 for INTP and INTQ, period E. Again, the significant displacements on the center days are opposite, Figure 4a being positive and Figure 4b being negative. This result is similar to that for Figure 3 for the absence or presence of 3h-SFO when height increases occur. Fejer et al.  reported that at Jicamarca, as their Figure 3 showed for period E and Rzmax, increases in heights and positive vertical drift velocities were not always associated with the recording of SSIs (ESFs), and the KP index is greater for no SSIs. The results presented in this section for ESFm show that it can also be absent for Hh′F (H0h′F). Relative to average levels of Kp ∑ 81, the GA was enhanced for H0h′F (6.4 for Figure 3a) and suppressed for H3h′F (3.3 for Figure 3b).
2.4. Absence of Spread-F and Height Decreases
 At Jicamarca, Fejer et al.  reported that on occasions the postsunset height was depressed, as their Figures 4 and 10 showed. These height decreases also involved negative vertical drift velocities. They proposed that the F2layer may experience large downward vertical drift velocities due to westward disturbance electric fields. Also, Figure 4 of Fejer  shows that at these times increased GA is involved. It seems likely that the height decreases reported here (Lh′F) from HF radar recordings (ionograms) also relate to this phenomenon. The Lh′F events for INTP, period D numbered 79. For 49 of these no spread-F was present (Z-SFO). The term L0h′F is used for these 49 events, which will be shown to relate significantly to GA some hours earlier. For the remaining 30 events an SE analysis has revealed (not shown) that they have no association with GA.
 In section 2.2 statistical analyses were discussed concerning negative height displacements of some significance which were related to substorm onsets at College. For Kp = 3, 4, 5 these displacements were delayed 5.0 h at the anomaly crest station Okinawa, and 8.5 h at a station on the equator, Kodaikanal. This station is displaced somewhat in longitude from College. The evidence suggests that a traveling disturbance (speed of about 300 m s−1) may be involved. Delays after GA for the L0h′F events have been examined statistically. The KP index was investigated relative to these L0h′F events by five different analyses (KP for intervals 5, 6, 7, 8, and 1, or in UT covering a period from 1200 to 0300 the next day). For all analyses the center day (CD) indicates a significant positive KP level. A measure of this significance is the percentage change of these CD displacements relative to the average of the other 20 d of the final distribution (CD ± 10 d). This percentage change is of the average KP levels as determined by the SE analyses, which involved 49 events for period D and INTP.
Table 7 lists these percentage changes for the five KP index analyses. The result for KP7 is the more important and represents a delay of around 7 h. Figure 5a gives the result of an SE analysis which compares the KP7 index with 83 L0h′F events from both periods D and E, and INTP. The error bars indicate the significance of this result.
Table 7. Positive Percentage Changes for Center Days
2.5. Presunrise Height Increases
 Earlier it was shown that at the equator, GA in the premidnight hours is related to after-midnight height increases, and partly because of an appropriate delay time, it was suggested that LS-TIDs were responsible [Bowman, 1978, 1995a; Bowman and Mortimer, 2000]. Figures 1 and 2 of Bowman  indicated that during Rzmax for 3.5 years, 400 substorm onsets were associated with height rises produced by LS-TIDs. These were tracked at several stations all the way to the equator. Thus, on average, every year more than 100 substorm onsets have an influence at the equator, suggesting that LS-TIDs may occur there regularly. Further, section 2.2 discussed evidence that LS-TIDs are recorded frequently at the equator. Spread-F on ionograms is usually present during the passage of LS-TIDs, but this is not always the case, as was mentioned by Bowman . Fejer et al.  reviewed the presunrise occurrence of ESFs at Jicamarca and listed references on presunrise events. Electric fields were proposed to explain the height increases at these times. For the ESFm considered here it is suggested that LS-TIDs, which travel from auroral-zone regions with delays after GA, are responsible for at least some of the height rises, and any MSSs which occur at the same time.
Figure 5b shows that the 78 presunrise Hh′F events for period E, INTR, are related to the premidnight GA (Kp ∑ 12), and delays ranging from 3–9 h are involved because of the 6 h of KP index used. A CD displacement was achieved by, in the analysis, using the same date (presunrise) for both the parameters. Another analysis produced Figure 6a, which is a plot showing 3h-SFO relative to the 78 Rzmax presunrise height rises. No association is evident. By coincidence, for period E, INTR of 4 years of Rzmin, the Hh′F events also number 78. Figure 6b results from an analysis similar to that for Figure 6a, and shows that for Rzmin 3h-SFO = 3 is associated with the Hh′F events. The results by Fejer et al.  showed that recording of ESFs in the presunrise period also falls from some level of occurrence in Rzmin years to near zero occurrence in Rzmax years.
 Ionogram parameters have been used to investigate the nighttime equatorial disturbance conditions at Huancayo particularly related to height changes as well as the occurrence of MSSs created by the passage of AGWs. The large height increases at these times should favor the breaking of AGWs [Kshevetskii and Gavrilov, 2005], which are known to be present at these times [Rottger, 1973]. In midlatitudes, Kvavadze et al.  reported, at times of SFO, SSIs traveling with AGWs responsible for the SFO, and suggested that these SSIs may result from the breaking of AGWs.
Figure 3 indicates associations between the presence or absence of spread-F and GA in the postsunset period when height rises occurred. For no spread-F the related GA is positive (Figure 3a), whereas for SFO a negative displacement of GA is recorded and is well defined (Figure 3b). Figure 4 shows essentially the same results except that those occasions are considered when there are no height rises. The GA is high for no spread-F (Figure 4a) and low for SFO (Figure 4b). Figure 4 results can be compared with Figure 7 of Bowman [1995a], where it is shown for Huancayo for the period before midnight, that for low SFO GA is high and for high SFO the GA is low. Figure 5a is concerned with the GA related to those occasions when height decreases occur and there is no SFO. The GA is related positively, and again the result is well defined. An examination of some earlier results (Figures 1 and 2) indicated the existence of three LS-TID modes with different speeds, which originated at times of auroral-zone substorm onsets. The second mode (event 2 of Figure 1) has a speed of about 300 m s−1 and the Δh′F displacements are negative for the equatorial station. Figure 5a, which involves height decreases, would seem to relate to the same phenomenon. The negative displacements in Figures 1 and 2 might be recognised as indicating a new phenomenon for which further investigations are probably needed. Figure 5b shows that the presunrise height rises are related to increased GA, and it is proposed that LS-TIDs are responsible. However, other investigators [Blanc and Richmond, 1980; Scherliess and Fejer, 1997; Fejer et al., 1999; Richmond et al., 2003] have given evidence to suggest that electric fields may cause these height rises. As Figure 6 shows, ESFm is associated with these presunrise height rises for Rzmin years but not for the Rzmax years analyzed here. This SFO may be explained by considering the coexistence of an LS-TID related to a fast AGW and a disturbance associated with a slower AGW. Isoionic tilts produced by LS-TIDs are only a few degrees because of the fast speeds of the LS-TIDs [Clarke, 1972; Bowman, 1996a], and therefore their structures are not suitable for the recording of spread-F traces on ionograms. Using an ionosonde and other HF radar equipment, Bowman [1996a] has shown that an LS-TID traveling with a speed of 507 m s−1 (as recorded by Bowman [1992b]), coexists with a slower AGW that is responsible for spread-F traces. These spread-F traces were recorded at the crest, and only at the crest of a related height rise, and the AGW had a horizontal wavelength of 26 km and a speed of 100 m s−1. It seems likely, following a suggestion by Hines , that an AGW with shallow wave amplitudes giving no spread-F traces has its wave amplitudes increased owing to the height rise. It is known that LS-TIDs do not always produce spread-F traces on ionograms [Bowman, 1978], so that the limited occurrence of spread-F during Rzmax years suggests that at these times even AGWs with shallow-wave amplitudes may often be absent.
 The common feature in the five distributions that relate ESFm to GA (Figures 3, 4, and 5a) is the inverse association. The presence of MSSs is related to low GA a few hours earlier, and the GA is high when no MSSs are recorded. As Figures 3 and 4 show, this association is independent of whether or not height rises occur. The early analyses by Lyon et al.  involved an extensive study of the ESF. They detect a strong inverse relationship between SFO and GA for equatorial regions. Even radio-star recordings, which relate directly to ESF, show this inverse relationship [see Lyon et al., 1960, and references given therein]. At an equatorial station, Rao and Rao  reported a similar inverse relationship. Chandra and Rastogi  confirmed that in general spread-F is suppressed by increased geomagnetic activity. More recently, a series of analyses have detected this inverse association. Subdivisions of data were made for this series, which involved PM and AM intervals, Rz max and Rz min periods, equatorial stations across the world and stations at other latitudes. The results for all these analyses were obtained at high levels of statistical significance [Bowman, 1987, 1995a, 1998; Bowman and Mortimer, 2000]. Using airglow recordings for two decades at an equatorial station, near an anomaly crest location, Sobral et al.  have reported decreased SFO for increased GA involving Kp index levels associated with delays >4 h before sunset. Recently, Su et al.  have examined statistically, using satellites, density irregularities associated with ESF for the topside ionosphere at a height of 600 km and found that the density irregularities are more likely during periods of low magnetic activity. Also, using satellite transmissions, at an anomaly crest station in India, and during equinoctial months, when scintillations occur regularly, Ray and DasGupta  found that scintillations are recorded more frequently for low GA levels (Kp = 0−3) than for higher GA levels (Kp = 3 ± 9). The analyses here have shown for Huancayo a well-defined inverse association between ESF and GA, and a review of the literature on the subject has been made.
 The inverse relationship examined here and elsewhere concerns mainly MSSs (ESFm) and GA, and is relevant for any level of GA, that is, for example, low levels of GA lead to the probable occurrence of MSSs (Figures 3b and 4b). The relationship does not depend on equatorial electric fields or the vertical drifts produced by them. Furthermore, this inverse effect has been reported for low latitudes [see Bowman, 1998, Figure 1] and for midlatitudes [see Bowman and Mortimer, 2000, Figure 5]. In contrast, the literature contains many papers that examine, at specific times, relationships between active GA, particularly, geomagnetic storms, and SSIs (ESFs). The analyses concentrate mainly on the variability of the electric fields which are responsible for the SSIs. Factors which might have an influence on this relationship include magnetospheric effects, station longitudes and seasonal and sunspot-cycle variations. These papers have been reviewed by Fejer  and Martinis et al. . In summarizing their review, Martinis et al.  indicate that following geomagnetic storms the occurrence of postsunset ESF can be either inhibited or enhanced and that the postmidnight ESF will be enhanced. Martinis et al.  reported similar results from their own analyses. Figure 8 of Bowman [1995a] shows that, for the AM period of Rzmin years at Huancayo, there is a well-defined direct association between the occurrence of MSSs and GA, and it is suggested that LS-TIDs, associated with increased GA, may be responsible.
 Fluctuations of the upper atmosphere neutral particle density (UA-NPD) have been detected by satellites for the diurnal, annual and sunspot-cycle variations for these geophysical or solar parameters. Other related features are known as (1) the GA effect, (2) the 27-d effect, and (3) the ionospheric height profile effect [Priester et al., 1967; Barlier et al., 1978]. The UA-NPD for the first three parameters has, for a number of investigations, been found to vary inversely with SFO, not only for low latitudes and midlatitudes but also for ESFm after midnight. For example, maxima for the spread-F distributions coincide with minima for the UA-NPD distributions [Bowman, 1964, 1992a, 1993, 1995a]. The 27-d effect involves increased UA-NPD levels after the passage of active solar regions across the face of the sun. Analyses have found for a range of latitudes that SFO levels are reduced considerably as is shown by the high levels of statistical significance [see, e.g., Bowman, 1996b, Figure 13]. The direct association between the UA-NPD and GA [Priester et al., 1967; Prolss et al., 1988; Prolss and Ocko, 2000] is of particular interest here as analyses have shown for an equatorial location an inverse relationship between ESFm and GA. This, in turn, indicates an inverse relationship between the UA-NPD and GA. It might be expected that the UA-NPD changes with height [Priester et al., 1967] might also influence the occurrence of spread-F, as is suggested by Figure 6b for presunrise spread-F in Rz min years. Because of the importance of the magnitudes of AGW wave amplitudes for the recording of MSSs by ionograms, a hypothesis has been proposed suggesting [see, e.g., Bowman, 1992a] that wave amplitudes may be influenced by UA-NPD levels. For example, low levels of the UA-NPD may increase the probability of recording MSSs, and vice versa. Because height increases involve lower UA-NPD levels and referring to AGWs, Hines  suggests that for height increases AGW wave amplitudes will be larger giving an increased probability of spread-F being recorded by ionograms. This observation by Hines  is consistent with the hypothesis that has been proposed.
 Figure 4 of Fejer  indicates that for 20 years of recording at Jicamarca, significant changes to the vertical drift patterns occur relative to high levels of GA (AE index of 400 nT) for (1) positive drift events around 0330 LT and (2) negative drift events around 1930 LT. For height changes detected by ionograms, similar associations have been reported here as shown by (1) Figure 5b for the presunrise height increases and by (2) Figure 5a for the postsunset height decreases. These results also involve delays of several hours.
 The principal result of the present analyses has been the identification of, with or without height increases, an inverse association of the occurrence of ESFm with GA a few hours earlier than the time of SFO. This means, as explained, an inverse association with the UA-NPD. Somewhat similar results for ESF and GA have been reported by other investigators [see, e.g., Lyon et al., 1960]. Thus the absence of spread-F for some height rises [Fejer et al., 1999] can be explained by the occurrence of increased GA (Figure 3a). Also, height decreases are found associated with the absence of ESFm and increased GA about 7 h earlier than the decreases. The absence of spread-F on these occasions can be explained by higher UA-NPD levels expected with the height decreases, as well as higher UA-NPD levels associated with increased GA (Figure 5a). If, as proposed, the magnitude of AGW wave amplitudes is reduced because of the increased UA-NPD levels, it is less likely that ionograms would record spread-F. These height decreases may possibly be associated with the arrival of LS-TIDs. Although usually ESFs is examined relative to high GA (e.g. geomagnetic storms), it has been shown here for ESFm that low levels of GA are also important (Figures 3b and 4b).
 Amitava Bhattacharjee thanks the reviewers for their assistance in evaluating this paper.