For a study of the equatorial ionosphere, ionospheric scintillation data at VHF and L-band frequencies have been routinely collected by ground-based receivers at Ancon, Peru, Antofagasta, Chile, and Ascension Island, UK, since May 1994. The receivers routinely monitor VHF transmissions from two geosynchronous satellites located at 100°W longitude and 23°W longitude, and L-band signals from satellites located at 75°W longitude and 15°W longitude. This combination provides a network of seven usable, reasonably separated links for monitoring ionospheric equatorial bubble activity in the South American longitude sector. A data set of seven years covering the period from 1995 to 2001 was studied to determine the temporal, diurnal, and seasonal behavior of equatorial bubbles. The results of our statistical study are presented here. In general the equatorial ionospheric bubble activity shows a strong systematic and primary dependence in temporal, diurnal, and seasonal variation, and a secondary weak dependence on geomagnetic and solar flux activity. At present, the dependence on solar and magnetic activity is not usable for near-time and short-term prediction of the equatorial bubble activity. Equatorial bubbles usually start 1 hour after sunset, the activity peaks before local midnight, and vanishes by early morning. The activity peaks in the months of November and January–February and is practically absent (weak) from May to August. On a daily basis on the average one sees 1 to 3 bubbles. The duration of bubbles is about 70 min, and the time spacing between the bubbles is 1 to 2 hours. The bubble activity in general follows the phase of solar cycle activity. The observed systematic behavior of the equatorial bubbles allows for a now cast and short-term forecast of the bubble activity in the South American sector.
 Ionospheric bubbles originate in the equatorial region, extend away from the (geomagnetic) equator and are seen up to ±15° latitudes from the geomagnetic equator. The bubble activity most often starts about an hour after sunset, peaks before local midnight, and sometimes continues through the early morning hours before sunrise. The steep walls of equatorial ionospheric bubbles very often severely degrade VHF communications in the equatorial region. The depleted density of the ionosphere inside the volume of the bubble very often produces rapid changes both in amplitude and phase of the VHF signal passing through the bubble. This phenomenon is called ionospheric scintillation. Direct observational evidence connecting the occurrence of a bubble and its effect (scintillation) on the VHF signal was shown by Basu et al. , and by other scientists. Thus ground observations of scintillation suffered by VHF signals from transmitters aboard satellites can be used to study the morphology of the equatorial ionospheric bubbles. It is essential to understand the morphology and long-term behavior of the equatorial ionospheric bubbles as these phenomena affect the space weather [Behnke et al., 1995]. For now casting scintillation activity (and thus the ionospheric bubble activity) in the South American sector, Groves et al.  proposed the Scintillation Network Decision Aid (SCINDA) and deployed a chain of stations for ground-based measurement of ionospheric scintillation. Here we present the morphology of scintillation activity observed in the South American sector of the equatorial region and discuss the reality of now casting the occurrence of the equatorial ionospheric bubbles from the scintillation data routinely recorded at several ground stations. The main emphasis of this study is to apply the observed morphological behavior to improve now casting the occurrence of the equatorial ionospheric bubbles from the ground-based scintillation observations taken from the deployed network.
 The following section describes the scheme for the collection of the scintillation data. The method of analysis and the results are presented in the analysis and interpretation section. That section discusses the morphology of equatorial bubbles and their dependence on geomagnetic activity and on the 10.7 cm solar flux. The comparison of equatorial ionospheric bubbles at pairs of ground stations is discussed. The final section presents the summary of the study.
 Because the ionospheric bubbles are known to originate close to the geomagnetic equator and expand polewards, typically reaching magnetic latitudes of ±15°, Groves et al.  deployed a chain of ground-based stations in the South American sector covering this region for observing the scintillation activity. The VHF (250 MHz) and L band (1.6 GHz) signals are broadcast by transmitters aboard two satellites in geosynchronous orbits. The VHF band transmitters are on satellites at 100°W and 23°W longitudes, and L-band transmitters are on satellites at 75°W and 15°W longitudes. These satellites remain reasonably fixed with respect to the ground stations. The ionospheric scintillation measurements are along the slant path from the satellite to the ground station. The stations shown in Figure 1 are in geographic coordinates, and are listed in Table 1. The figure shows the location of the geomagnetic equator, and of geomagnetic latitudes of ±15°. The locations of ground stations, Ancon, Peru, Antofagasta, Chile, and Ascension Island, UK, along with the ionospheric penetration points (IPP) for the satellites are shown in the figure. In Table 1 the ground stations are arranged by frequency (primary: VHF and L-band) and by decreasing corrected geomagnetic latitudes of the respective links. As the equatorial bubble is a nighttime phenomenon, the UT corresponding to each local midnight is listed in the table. As the ionospheric density peaks at the nominal altitude of 300 km, the geographic and corrected geomagnetic coordinates listed in the table are the coordinates of the Ionospheric Penetration Points (IPP) rather than the coordinates of the ground stations. There are five VHF and two L-band links covering the range from 2° to −17° in corrected geomagnetic latitude, with ground stations located at Ancon (ANC), Peru, Antofagasta (ANT), Chile, and Ascension Island (ASI), UK. Geographically, all the links lie in the southern hemisphere. The month when the given link became operational is listed in the table. Data availability in terms of total months (with at least 8 days of data for a month) and total number of days for each link are also included in the table. For this study scintillation data are available up to the end of September 2001. The periods with no data are: January 1996 for ANT-E, January to August 1996 for ANC-E, February 1996 for ANT-E, ANT-W, and ANT-L, January 1998 for ASI-U and ASI-L, February to August 1998 for ANC-E, December 1999 for ASI-U and ASI-L, May 2000 for ANC-E and ANC-W, and January 2001 for ANC-E, respectively. (The end letters E, W stand for East and West and L and U (really should be V for VHF; did not change the old notation U) stand for the radio bands, respectively. The data refer to VHF band if the letter L is not used.) The shortest distance between links is 524 km, between ANT-W and ANT-L. On the mainland, the longest distance is 1903 km, for the West Coast pair ANC-W and ANT-E. Between ANT-W and ASI-L the distance is 6831 km (longest distance for the deployed network).
Table 1. Ionospheric Penetration Points (IPP) at Nominal Altitude of 300 km for Various Links From Transmitters Aboard Satellites to Ground-Based Receiver Stations
Coordinates of IPP
UT at 0000 LT
 The auxiliary data such as Kp and Dst indices and solar flux at 10.7 cm were obtained from the National Geophysical Data Center (NGDC). The solar cycle had its minimum in October 1996 (monthly averaged flux at 10.7 cm was 69 ± 1 units (10−22 w m−2 Hz) and its maximum in July 2000 (monthly averaged flux at 10.7 cm was 210 ± 44 units). These data are used for determining the dependence of scintillation (therefore of equatorial ionospheric bubble) activity on the geomagnetic and solar activity.
 At each ground station the scintillation data are collected over an interval of 82 s for computing the scintillation index S4. The scintillation index S4 is given by the equation
where I is the instantaneous amplitude of the scintillation signal at each interval of the sampled time width and IAV is the averaged amplitude of the scintillation signal during the sampled time width [Briggs and Parkin, 1963].
 Alternately the scintillation strength is measured in dB and is given by
where Pmax and Pmin refer to the higher and lower amplitude (squared) of the scintillation signal [Whitney et al., 1969]. The scintillation levels measured at 3, 6, and 10 dB correspond to S4 levels of 0.17, 0.3, and 0.45, respectively.
 These data are averaged over a period of 5 min and are available at a remote site/central station at Air Force Research Laboratory (AFRL), Hanscom AFB, Massachusetts, United States. Typically the central station retrieves the data at 15-min intervals. As the equatorial spread F (ESF) bubble is a nighttime phenomenon, at each station the data are collected from local sunset (through night) to local sunrise, as these are very likely the periods for scintillation from the bubble activity in the equatorial region.
3. Analysis and Interpretation
 Here our main interest is to determine the phenomenology that is useful in improving the capability of predicting the occurrence of the equatorial bubbles in the South American sector. Our preliminary study showed that strong scintillations (S4 > 0.3 affecting the operating systems) are rare in the months May to August. Therefore, in general, data for these months are not used in the following study.
3.1. Temporal and Spatial Dependence of Equatorial Ionospheric Bubbles
 The scattering of a HF signal follows Rayleigh's scattering law, that the scattering is proportional to the fourth power of the wavelength. Thus the scale sizes of ionospheric irregularities monitored at 250 MHz and at 1.6 GHz are of the order of 1 km and 300 m, respectively. Therefore one would expect different levels and variations of scintillation for a given station at these two frequencies. Experience has shown that for the VHF band, at a scintillation level of S4 ≥ 0.3, the bubble activity is moderate and at the level of S4 ≥ 0.6, the bubble activity is strong. Therefore, for each link and for each year, the number of days with scintillation exceeding these two levels was counted, along with the total days of data. The results are shown in Figure 2 as percent occurrence. The links are arranged in the same sequence as that listed in Table 1. The numbers of days for which the scintillation data are available for each year are printed above the horizontal axis in each panel. Taking into account the gaps in the data (summarized in the previous section), the figure shows that VHF links (top 5 panels) see at least moderate activity (S4 ≥ 0.3) from 50% to as high as 80% of the days, and stronger activity (S4 ≥ 0.6) for 40% to 70% of the days. The level of occurrence reasonably follows the trend of solar cycle activity, which was at its minimum in October 1996 and maximum in July 2000. Thus equatorial bubble activity that degrades the HF communications is a semi-routine phenomenon. The effect on the L-band (two bottom panels) is relatively weaker, and shows solar cycle dependence.
Figure 3 shows the scintillation dependence on a time interval basis. The 5-min interval nighttime data are used for this determination. In the figure the period of occurrence is expressed in percent (with respect to the observing period of night hours as explained in the data section). On an annual basis, occurrence ranges from 10% to 40% for the VHF-band and about 5% for the L-band for moderate activity (S4 ≥ 0.3), and still shows solar cycle dependence. The comparison shows that for the VHF band the duration of strong events is about half that for moderate events. As shown later, on a seasonal basis these results have to be multiplied by a factor of 2 to 4.
 The scintillation events representing the ionospheric bubble activity were counted by selecting a minimum level of scintillation activity. For the given event the time interval is measured for the period during which scintillation activity exceeds the selected level. Also, the peak value of the scintillation event is determined. If the time of the event is less than 10 min (at least two consecutive intervals) the event is treated as transient or jitter in the data and dropped from the count. For each year the events are divided into four ranges of S4: 0.13–0.3, 0.3–0.45, 0.45–0.6 and >0.6 (corresponding to 3–6, 6–10, 10–15, >15 dB, respectively) at peak activity. These results are presented in seven panels of Figure 4. The year and the number of events are listed in each panel and the percent occurrence over the four intervals is shown in histograms. In Figure 4, the top four panels, with links closer to the magnetic equator, show weak activity (S4 < 0.3) for 10% to 30% of the events; thus at least 70% of the events are of moderate or strong activity. The number of strong events (S4 ≥ 0.6), shows a definite solar cycle dependence. At Ascension (panel 5), away from the magnetic equator and under the anomaly region, the weaker events range from 5% to 40%, and the strong events range from 20% to 80% depending on the level of solar cycle activity. For ANT-L (panel 6), closer to the magnetic equator, most of the events fall in the weak category. At Ascension the L band activity in panel 7 is about half of that for the VHF (panel 5). Thus the VHF band with its larger scale size of irregularity responds more strongly to bubble activity than the L-band.
 For each month the scintillation events are counted. Here, data for all the months are used. Figure 5 presents the seasonal dependence for events exceeding the moderate level of S4 ≥ 0.3. At all the links the activity is minimum in the months of May to August, and peaks in the months February–March and October–November. During the peak activity, VHF links on the average observe 2 to 3 events per day. During the same period the L band sees 1 to 2 events per day.
 While counting the peaks of scintillation activity, the widths (in time interval), and spacing (in time interval) between consecutive peaks were determined. For the VHF band at moderate activity (S4 ≥ 0.3), the average duration of events is about 80 min and the average interval between consecutive events is 160 min. As the activity on L band is weak, we do not feel confident for estimating these durations.
 For each month the number of 5-min intervals with S4 exceeding a given level is counted for determining the time period of the scintillation activity, along with the number of intervals providing scintillation data. Here also data for all the months are used. This seasonal variation of the occurrence of activity during the month is presented in Figure 6 for moderate scintillation activity; S4 ≥ 0.3. The top four panels show that the activity is weak from May to August and is strongest, as high as 40%, from November to February, and shows solar cycle dependence. The fifth panel, for Ascension Island, shows similar activity at a lower level indicating that the activity spreads to at least ±17° corrected geomagnetic latitudes from the equator. The two bottom panels show that the L band sees the same seasonal variation for a reduced time period. Both Figures 5 and 6 show that scintillation activity is negligible in the months May to August. As our main goal is to try to improve prediction capability of equatorial bubbles affecting the operational systems, not using data of these months in this study is justified.
 For each year the diurnal dependence of scintillation activity exceeding a given level of S4 was determined by counting the observations at each 5-min interval of a day (night). Figure 7 shows the diurnal dependence of S4 ≥ 0.3. In each panel a dotted vertical line shows the time of local midnight. All the links show systematic diurnal dependence. The activity starts 1 hour after local sunset, reaches its peak about 2 hours before local midnight and diminishes steadily towards sunrise. The peak scintillation activity and its time period at half the peak width show solar cycle dependence. In general the width is about 5 hours for the VHF-band scintillation and 2 hours for L band scintillation. A check for S4 ≥ 0.6 activity shows that the time width reduces by an hour and the peak activity is at 80% of that at S4 ≥ 0.3. Also for S4 ≥ 0.6 activity the time of peak activity is half an hour earlier than that S4 ≥ 0.3. This very systematic behavior seen at these links is useful for real time prediction discussed later.
3.2. Solar Cycle Dependence of Equatorial Bubbles
 The number of days for the year (from Figure 2), the duration (from Figure 3), number of events (from Figure 5), and time that the diurnal occurrence peaks (from Figure 7), are compared for strong (S4 ≥ 0.6) to moderate activity (S4 ≥ 0.3) for determining the effect of solar cycle dependence on the level of scintillation activity. The results for the respective four items are presented as percent ratios (strong ÷ moderate) in Figure 8. In each panel the averaged annual solar flux (multiply y-scale by 2 for flux) at 10.7 cm is shown by the asterisks. In each panel the notations for the respective links are shown. In each category most of the links show about a factor of 2 increase in scintillation activity (from moderate to strong). Antofagasta experiences stronger activity than Ancon. Mostly the strongest activity is seen at Ascension Island, which is the farthest link from the equator, in the deployed network. In each category the response of the ANT-L link is low (long dashes), though it does show an increase in scintillation activity with increasing solar activity. Thus, in all aspects of occurrence, duration, and number events the scintillation activity exhibits a systematic increase from moderate to strong with increasing solar activity.
 For each year the distribution of the highest level of scintillation from individual events was checked. This distribution did not show any systematic trend. A summation of the scintillation index ΣS4 over the time window of a given scintillation event produced by the equatorial bubble activity is computed. The distribution of ΣS4 is shown in Figure 9. In each panel, the number of events is shown along the y-axis and the ΣS4 intervals are along the x-axis. Each link shows a systematic behavior with a peak and a gradual fall towards higher ΣS4. Also each link exhibits the solar cycle dependence for ΣS4. The ΣS4 would be a measure of the energy expended in the generation of an equatorial bubble. But this relation, though significant, is not useful for improving a prediction capability of the occurrence and the strength of equatorial bubble.
 Although the scintillation activity shows general dependence on solar activity, no significant correlation (useful for prediction of occurrence of equatorial bubble) is observed between daily solar flux (averaged for prior 30 days, because daily solar flux follows solar rotation period of 27 days) at 10.7 cm and scintillation activity observed on the respective day.
3.3. Geomagnetic Dependence of Equatorial Ionospheric Bubbles
 Here we will look at the dependence of bubble activity on geomagnetic indices Dst and Kp for improving prediction capability. For the dependence of equatorial plasma bubbles (EPB), Huang et al.  observed that (1) EPB occurred regularly during geomagnetic storms during the initial and main phase, (2) EPB are usually suppressed during the recovery phase of the storm in the evening sector and increased activity in the dawn sector, and (3) a large fraction of EPB occur during intervals of rapidly changing Dst, independent of whether a large magnetic storm developed.
 In all we have 26 storms (Dst < −130), out of which 16 are sudden storm commencements (SSC). We have 5 VHF and 2 L-Band links. For these links the availability of data varies from 10 (ANC-W) to 21 (ASI) cases, providing a total of 87 (VHF) + 41 (L-Band) = 128 of storm-station cases. Out of these 48 + 6 = 54 cases (54/128 = 42%) showed the presence of S4 > 0.3 scintillation at the main phase of the storm. Considering all the hours (without premidnight or postmidnight condition), 38 + 3 = 41 cases (41/128 = 32%) showed the presence of S4 > 0.3 scintillation at the recovery phase of the storm. Out of these 7 + 2 = 9 cases (22%, no breakdown with respect to links) are premidnight and 29 + 3 = 32 cases (78%, no breakdown with respect to links) are postmidnight.
 Let us concentrate on the SSC storms (16) only: For these links the availability of data varies from 6 (ANC-W) to 13 (ASI). With a total of 54 (VHF) + 26 (L-Band) = 80 SSC-station cases; out of these, 28 + 4 = 32 cases (32/80 = 40%) showed the presence of S4 > 0.3 scintillation at the main phase of the storm. Considering all the hours (without premidnight or postmidnight condition) 27 + 3 = 30 cases (30/80 = 38%) showed the presence of S4 > 0.3 scintillation at the recovery phase of the storm. Out of these, 5 + 2 = 7 cases (23%, no breakdown with respect to links) are premidnight and 17 + 6 = 23 cases (77%, no breakdown with respect to links) are postmidnight. Thus these results show no distinction between SSCs and other Dst storms, and no significant difference in scintillation activity with respect to the phase (initial, main, and recovery) of the storm.
Huang et al.  suggest that rapid changes in Dst are correlated with plasma bubble activity. To check this effect, Dst data for our observation period of 1994–2001 were studied. For a ±2σ level (88%), Dst data fall in the range of −50 to +30 nT. To determine the change, Δ Dst was computed from the hourly observations of Dst. For a ±2σ level (88%), the hourly difference of Dst data fall in the range of −4 to +4 nT range. About 5% of Dst data show an hourly change ≥10 nT and 1% data show an hourly change ≥20 nT. The rapidity of the change was measured as 2(Dsti − Dsti+1)/(Dsti + Dsti+1) from the hourly observations. For a ±2σ level (88%), the rapidity of the Dst change falls in the range of −80% to +80%. Note that a use of this expression for rapidity results in larger changes when Dst values are closer to 0 nT. Using these statistics as a guideline, hourly changes in Dst greater than 20 nT are compared with S4 (>0.3) scintillation data available for the present study. No systematic relationship was observed between these data sets, indicating very little dependence of S4 activity on Dst variation.
Aarons et al.  observed that in the months of May, June and July, the scintillation activity at any hour increases with increasing magnetic activity. For months from August through April the dependence of scintillation activity is more variable and is a function of the local time. They also note that magnetic activity favors an increase in scintillation activity in the postmidnight hours but does not favor it in the premidnight hours. Valladares et al.  note that scintillation data at Ancon show the magnetic control during the June Solstice months and also during the postmidnight hours for the rest of the year. In addition they sometime saw increase in scintillation activity in premidnight hours for the months August through April when magnetic conditions were active. Groves et al.  note that even though scintillations occur during magnetically active periods, major scintillation activity is associated with magnetically quiet conditions. From satellite-based observations of equatorial ionospheric bubbles, Huang et al.  report bimodal dependence on magnetic activity, thus supporting the results of Groves et al. .
 The present large data set was divided in four groups: (1) premidnight and (2) postmidnight hours for months September to April, and (3) premidnight and (4) postmidnight hours for the months May to August. In each group, scintillation activity S4 ≥ 0.3 was compared with Kp at unit intervals from 0 to 9. In general the present results support the above mentioned conclusions. However, no mathematical relation between scintillation activity and Kp index could be derived for improving the prediction capability of the equatorial ionospheric bubbles.
3.4. Real Time Spatial Prediction of Equatorial Bubbles
 An occurrence common (on a daily basis) for a given pair of links was computed in the following way. For each station and for each year, the number of days with observations and days with a presence of scintillation activity are counted. This information is used to compute the days common in both these categories for each pair of links. The ratio of days common with presence of scintillation activity to days common for the pair of links provides the scintillation occurrence common for a given pair of links. This result of scintillation occurrence common for the respective pairs of stations is presented in Table 2. As no common occurrence in seen for 4 months, from May to August, these data are excluded from the computations. In the table in each block, the first number lists the lowest and is followed by the highest magnitude of common occurrence. The lower value observed is around the minimum phase and the highest value is around the highest phase of the solar cycle. On a monthly basis the scintillation activity peaks in the months of November and January–February. This systematic behavior of scintillation activity forms the basis of the now cast and short-term forecast scheme used for the data-driven model in the SCINDA operation.
Table 2. Scintillation Occurrence Common for a Given Pair of Links
Common Occurrence Between Pair of Links, %
 The percent occurrence of scintillation as a function of time delay for various pairs of links is computed in the following way. For a given pair, for each month an occurrence of scintillation at both stations is checked over each half hour interval by using a time delay in steps of half an hour, up to a longest delay of 6 hours. The probabilities are computed on a monthly basis. The results for 3 pairs of links are presented in Figure 10. The pairs of links are along the column and two years, 1997 and 2000, for the comparison are along the row. The curve identification for the month is shown only in the middle left-hand panel. The time delay is along the horizontal axis and the percent occurrence of scintillation is along the vertical axis. The distance separating respective links is shown in the left-hand bottom corner of each panel. Only the periods with scintillation occurrence ≥30% are shown in the figure. Note that in each panel the probability of occurrence of scintillation on the eastern link, after the scintillation has occurred at the western link, is higher for shorter time delay, and it reduces with increasing time delay. The probability is highest in the month of November (dashed line). Also note that the probability increases with increasing solar activity (left-hand panel versus right-hand panel), and that, the period (number of months) of occurrence common to both stations also increases with increasing solar activity. This systematic behavior at pairs of links forms the basis and is the reason of success of now cast and short-term prediction of scintillation in the South American sector by the SCINDA project. Observations at Ascension Island show the longitudinal extent of scintillation and therefore of the equatorial ionospheric bubble activity.
 Based on the scintillation observations over a small network of three stations and two frequencies (one VHF and one L-band), the behavior of equatorial ionospheric bubble activity in the South American sector can be summarized in the following manner. Bubble activity is a nighttime phenomenon, it starts 1 hour after sunset, reaches a peak about 2 hours before local midnight and continuously diminishes till early hours before sunrise. On a daily basis the activity is as high as 80%. On a time interval basis it is as high as 40%. The activity peaks in the months of November and January–February and is practically absent from May to August. For about 70% of the occurrences the activity is moderate-strong. On the average one sees 2–3 bubbles each day. On the average, the duration of a bubble is 80 min, and the time gap is about 160 min. The occurrence of bubbles shows strong solar cycle dependence. The bubbles do not show any statistically significant (for derivation of a mathematical relation) dependence on solar flux, or on magnetic activity indices Kp and Dst. The quantity called scintillation flux, which would be some kind of measure of the energy of the flux tube responsible for the generation of the equatorial bubble, is better organized than the scintillation peak of the bubble event. A very systematic behavior of the bubbles allows a now cast and short-term forecast of the region over which the HF communication is very likely be degraded by the bubble activity.
 The authors thank E. MacKenzie (Institute for Scientific Research, Boston College, Newton, Massachusetts), and W. J. McNeil and R. Caton (Radex Incorporated) for providing the scintillation data.