Ionospheric foF2 anomalies during some intense geomagnetic storms

. The global evolutions of f o F2 anomalies were examined for three very intense geomagnetic storms, namely the Halloween events of October–November 2003 (Event X, 29–30 October 2003, D st − 401 nT; Event Y, 20–21 November 2003, D st − 472 nT), and the largest D st storm (Event Z, 13–14 March 1989, D st − 589 nT). For Event X, troughs (negative storms) were clearly seen for high northern and southern latitudes. For northern midlatitudes as well as for low latitudes, there were very strong positive effects on 29 October 2003, followed by negative effects the next day. For Event Y, there were no troughs in NH high latitudes for morning and evening hours but there were troughs for night. For midlatitudes and low latitudes, some longitudes showed strong negative effects in the early morning as expected, but some longitudes showed strong positive effects at noon and in the evening hours. Thus, there were many deviations from the model patterns. The deviations were erratic, indicating considerable local effects superposed on general patterns. A disconcerting feature was the presence of strong positive effects during the 24 h before the storm commencement. Such a feature appears only in the 24 h before the geomagnetic storm commencement but not earlier. If genuine, these could imply a prediction potential with a 24-h antecedence. For Event Z (13–14 March 1989, equinox), all stations (all latitudes and longitudes) showed a very strong “negative storm” in the main phase, and no positive storms anywhere.


Introduction
The ionospheric F2 region has average patterns of daily and seasonal variations.These patterns have considerable dayto-day variations, but spectacular changes occur (positive or negative anomalies) during geomagnetic storms, when there 5. Low latitude and equatorial zone: The E×B drifts are affected by prompt penetration of magnetospheric convection electric fields, as well as by longer-lived dynamo electric fields from the disturbance neutral winds and storm-related changes in ionospheric conductivity (Fejer, 1997).In addition to the drifts caused by electric fields, TADs and also longer duration disturbances in the global thermospheric circulation with resulting neutral composition changes have important effects on the low latitude region during storms.
6.Even under geomagnetically quiet conditions, electron density is extremely variable in the equatorial zone between sunset and midnight due to the presence of irregularities with scale sizes ranging from less than 1 m to greater than 200 km.How geomagnetic storms affect the development of equatorial irregularities depends on longitude but varies considerably from storm-to-storm.
Ionospheric storms associated with geomagnetic storms have been studied copiously in the past, for individual locations, for groups of locations, and on a global basis, for one or many storms (Prölss, 1997, and references therein;Szuszczewicz et al., 1998, and references therein).Since IGY, the largest geomagnetic storm occurred on 13 March 1989 (D st −589 nT).However, two very intense storms occurred recently in quick succession, namely Halloween events of 29-31 October 2003 (D st −401 nT), and 20 November 2003 (D st −472 nT).In the present communication, the morphology of ionospheric f oF2 anomalies is illustrated for these storms.

Data
All data were obtained from the NGDC SPIDR website http://spidr2.ngdc.noaa.gov/spidr/.Data quality and continuity were not always good.The f oF2 values could have errors due the presence of spread F and many other factors.In the SPIDR data, values of f oF2 are given as simple numbers, with no qualified coding.Hence, effects due to spread F, etc., cannot be ruled out and could be important, particularly for low latitudes.However, scrutiny of these would need access to original, detailed data from individual locations, which is a laborious process.These cannot be considered in detail in a general analysis like the present one, and no scrutiny of data of any kind was done.We expect (hopefully) that errors would be minimized in averages.With f oF2, data for hmF2 would be of great importance, but these were mostly meagre or absent and hence, are not considered here.
3 The Halloween events of October-November, 2003 (Events X and Y) Figure 1a shows a plot of hourly D st values during the 27-day interval 28 October-23 November 2003.The first storm (henceforth called Event X) started at ∼06:00 UT on 29 October, reached a maximum depression of −363 nT at 00:00 UT of 30 October (main phase of 18:00 h), recouped but had a second maximum depression of −401 nT at 22:00 UT on 30 October, and then recovered, first rapidly and then slowly.Thus, this was a complex storm.The triangles indicate solar flare occurrences.There were two strong solar flares, one on 28 October and another on 29 October.The second storm (henceforth called Event Y) started at ∼11:00 UT of 20 November, reached a maximum depression of −472 nT at 19:00 UT of 20 November (main phase of 9 h), and then recovered first rapidly, then slowly.There was a strong solar flare on 18 November.In between, there was a small storm on 4 November (D st −89 nT).There was a very, very intense solar flare on 4 November (largest in known history, so far), but it was a limb flare, without emissions (CMEs) directed towards the Earth and no terrestrial disturbances were produced.The mild storm of 4 November was caused by less strong solar flares, which occurred on 2-3 November.
Figure 1b shows a plot of hourly f oF2 (MHz) at the location Juliusruh/Rügen (54.6 • N, 13.4 • E) in European midlatitude, LT about 1 h ahead of UT.There is a substantial daily variation, with a maximum of ∼7-10 MHz at about noon and a minimum of ∼1-2 MHz soon after midnight.The storm effects are superposed on this background daily variation.To isolate the storm effects, the background daily variation needs to be subtracted.In conventional methods, the background is estimated as a monthly mean.However, this may become polluted by storm days.In the present case, the interval 7-16 November was almost geomagnetically quiet (except for a mild, extended storm during 11-15 November).Hence the average daily variation for these 10 (or less, as available) days was considered as a reasonable estimate of the background.(This does not ensure that the pattern would be representative of absolutely quiet conditions, but mild storm effects are not similar on successive days.Hence, averaging over several quiet and even some mildly disturbed days could be considered as a reasonably good background.This point will always remain subjective and debatable, but nothing much better can be done about it.)Then, two methods were employed.In one method, the background was subtracted from the actual hourly values.The deviations f oF2 minus f oF2 (average) were considered as anomalies (in MHz) and will be called henceforth as Anomalies, plotted in Fig. 1c.Positive deviations are painted black and negative deviations are shown as hatched.This location has some anomalies during 7-16 November, but the deviations are small as compared to those of other intervals.The storm effects are mostly positive, during 29-30 October and 20-21 November, but also during 4 November, when there was a small storm (D st −89 nT).In the second method, the ratio of hourly f oF2 to f oF2 (average) was calculated.Henceforth, these will be called Ratios and are shown in Fig. 1d.The fluctuations (anomalies and ratios) in Figs.1c and 1d are very similar, so any one of these can be used for the study.Small differences are mainly at low values of f oF2.Thus, a f oF2 value of, say, 8 MHz increasing to 9 MHz, would imply an anomaly of +1 MHz and a ratio of 1.125 (12.5% increase).However, a f oF2 of, say, 2 MHz increasing to 3 MHz, would also imply an anomaly of +1 MHz, but an enormously large ratio of 1.50 (50% increase).
The above procedure could be adopted only for data of 41 locations (out of 211) in which data were available at the website for Event X and/or Event Y, as listed in Table 1.Figures 2 and 3 show the plots, for Event X in the left half and Event Y in the right half.The top plots are for D st .In Fig. 2, other plots are for f oF2 anomalies at (a) 10 stations (latitudes from Thule in the north to Port Stanley in the south) in longitudes near about −65 • (i.e.65 • W) and (b) another 10 stations (latitudes from Tromsø in the north to Grahamstown in the south) in longitudes near about 12 • (i.e. 12 • E).Similarly, in Fig. 3, plots are for f oF2 anomalies at (a) 18 stations (latitudes from Manzhouli in the north to Christchurch in south) in longitudes near about 135 • (i.e.135 • E) and (b) for another 3 stations (College, King Salmon, Dyess, latitudes north) in longitudes near about 225 • (i.e.135 • W).The following may be noted: Event X (28-31 October 2003): 1.In Fig. 2a, left half, there is considerable latitudinal variation in the patterns, with roughly negative deviations in high latitudes and positive deviations in middle and low latitudes, but there are negative deviations at middle latitudes also.There is no systemetic movement of troughs from higher to lower latitudes as envisaged in the "average" pattern of the various models, indicating that localized electric fields rather than general global fields may be dominating and producing narrow troughs.Also, positive deviations seem to occur interspersed with negative deviations in an irregular way.Thus, ionospheric storm-time anomalies do not seem to have any reliable general pattern in individual storms.Patterns seem to vary largely from storm to storm.As such, predictions based on general patterns could be grossly inadequate and misleading for users like aviation pilots.
2. The most striking feature is the large positive deviations on 28 October, the day before the D st storm commencement.Such pre-storm anomalies were pointed out earlier in Kane (1973 a,b;1975), but do not seem to have received much attention by other workers, except by Danilov andBelik (1991, 1992).(These pre-storm positive anomalies are different from the F2-layer stormlike phenomena during geomagnetically quiet times observed by some Russian scientists in the 1980s.)If true, these could have very important implications, namely, these could be considered as precursors of geomagnetic disturbances.Such pre-storm increases can be seen in some of the plots in Araujo-Pradere and Fuller-Rowell (2002) also but have been ignored by them, and matching is discussed only starting from the main phase onwards.Thule/Qanaq THJ77 77.5 Northern Hemisphere (NH) middle latitudes (30 NH and SH low latitudes (30   missing for many locations, particularly in Australia (which is a pity as their network generally has very good continuous data), and for 30 October onwards, severe negative effects are seen, probably because the end of October is almost summer for these locations.
5. In Fig. 3b, there are mainly positive deviations for all 3 locations in the Northern Hemisphere, before, during and after the main phase.
Event Y (19-22 November 2003): 1.In Fig. 2a,b, right half, the anomalies are mostly positive, though one would have expected strong negative effects at least for Port Stanley (52 • S), where 20 November is almost summer.Instead, Jicamarca (12 • S) in low latitude shows large negative deviations before and during the storm commencement, and positive effects thereafter.The reliability of the reported values is not known and spread F is very frequent at Jicamarca.Also, since 9-16 November was not completely quiet, the use of the average for these days as background might not be fully adequate.However, since the same background is used for subtraction for the X event as well as the Y event, and the X event (Fig. 2a, left half, Jicamarca plot) does not show large negative deviations before the storm, the large negative deviations before and during the storm commencement in the Y event (Fig. 2a, right half, Jicamarca plot) could be genuine.
However, since we have not examined the original data for spread effects, etc., a doubt will remain about this feature.
2. In Fig. 3a, right half, very strong negative effects are seen during the main phase in the Australian region as expected, but positive effects before the storm are embarrassing.In the north, effects are mostly positive.
3. In Fig. 3b, right half, for College and King Salmon, effects are small, but Dyess (32 • N, 100 • W) shows strong negative effects not expected for a northern midlatitude station in winter.A few hundred kilometers away, Eglin (30 • N, 87 • W) (Fig. 2b, right half) showed no such strong negative effects.
To bring out the latitude and longitude dependence more clearly and with more confidence, data for nearby locations were averaged.The plots are shown in Fig. 4, for Event X 1.In Fig. 4, left half, the top plot is for D st .The next four plots are for average latitudes 65 • N, 37 Event Y: 1.In Fig. 4, right half, the longitude group A has mostly positive effects at all latitudes, though the storm is during the morning hours.A strong negative effect was expected in southern high latitudes, because of local summer.Thus, the behavior is not consistent with any storm model.
2. For the longitude group B, the storm occurred at about noon, and effects were positive to start with (and even before the storm commencement) and negative in the evening and night hours.
3. For the longitude group C, the storm occurred in the evening, and effects were mostly negative.
4. For the longitude group D, the storm occurred at about midnight, and effects were small at high northern latitudes and negative for a northern midlatitude.
Thus, whereas some effects are as per model prediction, considerable disagreements or distortions (deviations not conforming with models) occurred in many instances.However, these distortions did not have any systematic dependence on latitudes or local times.On the whole, it looks like local electric field perturbations and composition changes are more dominant than the general patterns envisaged in models.In particular, positive deviations seem to occur more frequently than expected, particularly in the 24-hour pre-storm interval.A question that may arise is.How magnetically quiet is the pre-storm period?We have used here the D st index and it shows sharp changes (depressions exceeding 350 nT) only on 29 October and 20 November.D st values are available hourly and we consider these better than K p values available every 3 h.On 28 October, the D st depressions were less than 50 nT.However, some workers use K p .In the present case, the K p values were about 9 (highest possible value) on 29-30 October and 20 November, while values a few days earlier were 5 or less, considered only as weak or moderate.On 28 October, the 3-hourly K p values were, 3, 5−, 4−, 5−, 3−, 4, 3+, 4. In principle, one can argue that the positive ionospheric anomalies on 28 October could be associated with the moderate geomagnetic disturbance of some K p values of 5−, present even on 28 October, but we feel that just the two stray low 5− values of K p could not have produced so strong ionospheric positive anomalies.This is, however, a subjective judgement, and in geophysics, strange things can and do occur, so all possibilities need to be considered.All that we can say is, the association of the strong positive ionospheric anomalies of 28 October with moderate K p is possible but not probable.
Figure 5 shows two examples from the plots in Araujo-Pradere and Fuller-Rowell (2002), where their (empirical) model estimates were grossly different from the observations, both qualitatively and quantitatively.The interval shown is 5-9 April 2000 and the storm started at about 18:00 UT on 6 April (marked by vertical line).The expected (empirical) STORM model values (thick lines) show a negative storm, starting at the geomagnetic main phase and lasting for almost 48 h, with a minimum ratio of 0.6 (40% decrease) for Boulder (northern midlatitude) and 0.8 (20% decrease) for Port Stanley (southern high latitude).Actually, the observed values for Boulder showed a decrease (marked hatched) of ∼60% (instead of 40%) but for only for the first 12:00 UT hours of 7 April, and large positive effects for the rest of the time, including much before and much after the storm interval.Port Stanley showed large positive effects before and during the storm, and a negative effect (40%) only in the latter half of 7 April and small negative effects thereafter.We do not know for certain how the model prediction information is used by aeroplane pilots, but in the present case (storm of 6 April 2000), the model estimates (thick plot) would have certainly misled considerably the pilots overflying Boulder or Port Stanley.The positive effects before the storm are quite large (30-40%) and the pilots would have been perplexed, as these do not appear, in the predictions.In their paper, Araujo-Pradere and Fuller-Rowell ( 2002) have presented 75 panels (for 15 stations for 5 storms in the year 2000, in their Figs.4a-e, like those shown here in our Fig.5).In all of these, their gray lines represent the outputs of their STORM model and these are mostly depressions (negative storms), starting at the geomagnetic storm commencement and intensifying in the next few tens of hours to as much as −30%.Only 10 (out of 75) show positive storms in the model values of southern midlatitudes, with increases of only about 10%.Thus, a negative storm seems to be a more certain feature, while positive storm effects seem to be small and uncertain.In their 75 panels (15 stations, 5 storms), more than half show substantial observed positive effects (∼20% increases above normal) before the storm commencement, but these have been ignored by those authors.

The giant event of 13 March 1989 (Event Z)
The D st magnitude −589 nT of this event was the largest ever recorded since IGY, when the index D st was formulated.The event had severe effects on the terrestrial environment (Allen et al. 1989).For this event, ionospheric effects have been reported in many publications (e.g.Batista et al., 1991;Greenspan et al., 1991;Huang and Chang, 1991;Lakshmi et al., 1991;Morton et al., 1991;Binachi et al., 1992;Rich and Denig, 1992;Yeh et al., 1992;Rasmussen and Greenspan, 1993; and probably many others).Many of these refer to a few stations in the equatorial and low latitudes in the American and Asian sectors and report large decreases or up and down oscillations.However, among these, Yeh et al. (1992) analyzed data from 52 ionosonde stations and 12 total electron content observing stations.Their global data showed a longitudinal dependence of the storm behavior, a worldwide depression of diurnal maximum f oF2 (sometimes accompanied by a large rise in h F2), TIDs, large-scale standing oscillations, hemispheric asymmetry, and suppression of equatorial anomaly.Thus, almost every ionospheric feature showed large deviations from normal ionospheric patterns.In the present communication, a similar analysis is presented, illustrated in a slightly different way, namely, anomalies.For this event, data were available on the website for only 52 locations (out of 211) and only 10 of these were common to those for the Halloween events.The plots for anomalies (MHz) only (not ratios) are shown in Fig. 6, not for individual locations but for avearges for nearby locations (the number of stations used for each plot is mentioned in circles).The whole period 8-17 March 1989 is plotted so that the effects of the minor storm of 8-9 March can be compared with those of the giant event of 13-14 March 1989 and with the quiet period in between.Anomalies (MHz) are plotted separately for the Northern Hemisphere (NH) high (>50 • N) latitudes and middle (30 • N-50 • N) latitudes, NH and SH combined low latitudes (30 • N-30 • S) and for the Southern Hemisphere (SH) high (>50 • S) latitudes and middle (30 • S-50 • S) latitudes.In each, successive plots are for progressive longitudes (A1, A2, B1, B2, C1, C2, D1, D2, each of 45 • range), so that LT effects can be distinguished.The following may be noted: 1.The mild storm of 8-9 March seems to have substantial storm effects (1-4 MHz), mostly negative, with some positive effects interspersed.There is no clear latitude or longitude (LT) dependence.
2. The giant storm of 13-14 March seems to be predominantly a very strong negative storm which would gladden the hearts of the modelers.It started at the geomagnetic main phase (marked by a vertical line), was intense during the next ∼24 h, irrespective of latitude and longitude (LT), recouped to almost the zero level (or even slightly positive at some middle latitudes) and then had a second negative swing lasting for another ∼24 h.Later, some positive effects appeared but on 17 March, some negative effects are seen even though geomagnetic activity was quiet.Thus, storm effects lingered for 2-3 days before disappearing, and did not have any clear relation to LT.
3. Since this was an equinox period, no hemispherical differences were expected and none were observed.The negative storm started at the main phase commencement in both hemispheres, the maximum depressions of f oF2 were also comparable, but the evolution was not similar.Some locations showed two swings, some showed three, and others only one.
4. The positve effects were not seen near the storm commencement at any latitude or longitude and were seen at midlatitudes only after ∼24 h.In the case of this storm, no positive efects were seen at the main phase or before.
These results are roughly similar to those mentioned by Yeh et al. (1992), except for a few slightly different details.Also, altitude effects (hmF2) are not considered here, not because these are not important but because data were not available.This is a lacuna of this analysis.On the whole, this storm mostly conformed to the model expectations.

Conclusions and discussion
The global evolutions of f oF2 anomalies were examined for three very intense geomagnetic storms, namely the Halloween events of October-November 1.For Event X (29-30 October, slight winter in NH and summer in SH), the troughs (negative storms) were clearly seen for ∼65 • N at nighttime, but not at any other LTs.Troughs were strongly seen in high southern latitudes, as if this was a summer storm for SH (see also Pincheira et al., 2002).For northern midlatitudes as well as for low latitudes, there were very strong positive effects on 29 October, followed by negative effects the next day.The results for this storm are uncertain because firstly, it was a mixed, double storm (one on 29 October at 05:00 UT and another 36 h later on 30 October at 17:00 UT) and secondly, data for some locations were missing for 29 October.
2. For Event Y (20-21 November, winter in NH and summer in SH), there were no troughs in NH high latitudes for morning and evening hours but there were troughs for night.For midlatitudes and low latitudes, some longitudes showed strong negative effects in the early morning as expected, but some longitudes showed strong positive effects at noon and in the evening hours.
3. A striking feature was the presence of strong positive effects in the 24 h before the storm commencement, often continuing in the storm interval.This pre-storm feature was pointed out in earlier papers (Kane 1973 a, b;1975) but does not seem to have attracted much attention.It is seen clearly, for example, in the plots of Araujo-Pradere and Fuller-Rowell (2002) (sample shown in our Fig. 5, where the positive effect is seen strongly before the storm and spilling into the storm interval).Such a feature appears only in the pre-storm 24 h but not earlier (for example, the positive deviations were not there on 27 October, but only on 28 October, one day before the storm day, 29 October 2003).If genuine, these would have a very important implication, namely a prediction potential with a 24-h precedence.
4. For Event Z (13-14 March 1989, equinox), all stations (all latitudes and longitudes) showed a very strong "negative storm" (and no positive storm at all) in the main phase.Also, the magnitudes (5-7 MHz) were consistently far greater than those for Event X or Y (hardly 5 MHz).True, the D st for Event Z was large (−589 nT), but the D st for Event Y was also large (−472 nT).Incidentally, the anomalies for a weak storm (8-9 March 1989) were also large (1-5 MHz).Thus, the magnitude of D st does not seem to be exactly proportional to the anomaly magnitudes of f oF2.(This is understandable as D st reflects the low latitude, high altitude currents at several Earth radii, while f oF2 changes are due to auroral high latitude ionospheric phenomena, with expansion towards low latitudes.)While all f oF2 depressions started at the storm commencement of Event Z, the further evolution was different at different longitudes (one swing, two swings, three swings) but not in any systematic way.
On the whole, whereas the March 1989 storm (Z event) conformed to the model expectations, the Halloween events of October-November 2003 showed ionospheric anomalies considerably different from the expected average patterns, and the differences seemed to be erratic, indicating strong local effects.The positive deviations seen before the geomagnetic storm commencement are intriguing.Some explanations can be examined.Ionospheric parameters are known to have a high variability, even in quiet geomagnetic conditions.Forbes et al. (2000) estimated the Nmax variability for annual, semiannual and 11-yr solar cycle variations.Under quiet geomagnetic conditions, the standard deviations of Nmax variability were 25-35% at high frequencies (periods of a few hours to 1-2 days) and 15-20% at low frequencies (periods 2-30 days).This quiet-day ionospheric variability could be considered as random or could be due to "meteorological influences".Ionospheric variability increased with geomagnetic activity, increasing from low to high latitudes.This is the geomagnetic effect.Changes due to variations in solar photon flux are reported to be rather small by these authors.
However, Mendillo et al. (1974) showed for the Total Electron Content (TEC) measured at 20 locations during the great solar flare of 7 August 1972 that there were 15 to 30% TEC increases with a rise time of about 10 min, with larger increases at lower latitudes.This is the solar flare effect.Such effects can last several hours (but less than ∼6) due to the slow recombination rate of the F region plasma.Recently, Tsurutani et al. (2005) examined the global ionospheric effects (TEC enhancements) of the 28 October 2003 solar flare and found 30% increases within a few minutes, lasting for ∼3 h.Thus, some of the pre-storm positive changes (increases in f oF2) in the 24 pre-storm hours could be due to lingering effects of the additional ionization caused by strong solar flare effects.Rishbeth and Mendillo (2001) gave estimates of ionospheric variability as 20% by day and 33% by night.They found that a large part of F2-layer variability was linked to that of geomagnetic activity, and the rest to "meteorological" sources at lower levels of the atmosphere.As such, the positive effects before the geomagnetic storm commencement could be partly of meteorological origin.However, strong positive effects a few tens of hours before the beginning of the geomagnetic disturbances could not all be meteorological effects or natural quiet time day-to-day variability.Recently, Danilov (2001) has discussed this problem of positive phases which are sometimes observed several hours before the beginning of a magnetic disturbance (e.g. during 13-14 September 1973, observed by Danilov andBelik, 1991, 1992).The ionospheric positive storm during magnetic disturbances is attributed to the F2 layer uplifting due to vertical drift, plasma fluxes from the plasmasphere, and downwelling of the gas as a result of the storm-induced thermospheric circulation (Danilov and Belik, 1992;Prölss, 1995).But for positive phases occurring sometimes before the beginning of the magnetic storm, this scheme does not work, as there is still neither the depleted [O]/[N 2 ] nor storm-induced circulation.So, some other channel of penetration of the disturbed solar wind energy to ionospheric heights, other than the usual one which leads to the Joule heating and auroral precipitation, is needed.It could be the effect of soft particle precipitation (emanating from solar flares but reaching the Earth a few hours later) in the region of the dayside cusp, as the cusp is the only formation which starts to react to the coming geomagnetic disturbances before any geomagnetic index does: the cusp begins to move equatorward a few hours before the beginning of the D st depletion (Danilov and Belik, 1992).However, no quantitative evaluation has been done so far and the role of electric fields, particularly its B y component, needs to be considered.Danilov (2001) concludes that for F2 region responses to geomagnetic disturbances, there are still unsolved problems, the most acute ones being: appearance of positive phases before the beginning of the magnetic storms, the occurrence of strong negative phases at the equator, the role of vibrationally excited nitrogen in the forming of the negative phase, and the relation of positive phases to the dayside cusp.Further investigations are needed to resolve these problems.

Fig. 1 .
Fig. 1.Plots for the 27-d interval 28 October-23 November 2003 of the hourly values of (a) geomagnetic D st , (b) ionospheric f oF2 at the midlatitude European location Juliusruh/Rügen (54.6 • N, 13.4 • E), (c) f oF2 anomalies, (d) f oF2 ratios.Positive deviations and ratios above 1.0 are painted black, negative deviations and ratios below 1.0 are shown as hatched.The triangles indicate solar flare occurrences.

Fig. 2 .
Fig. 2. Plots for Event X (28-31 October 2003) in the left half and Event Y (19-22 November 2003) in the right half, for D st (top plots) and the f oF2 anomalies (MHz) for stations in different latitudes (north to south, indicated on the right) and longitude belts, (a) 45 • W-90 • W, (b) 15 • W-30 • E. Vertical lines mark the storm commencements.Positive deviations are painted black, negative deviations are shown as hatched.

Fig. 4 .
Fig. 4. Plots for Event X (28-31 October 2003) in the left half and Event Y (19-22 November 2003) in the right half, for D st (top plots) and the f oF2 anomalies (MHz) and ratios (one below the other) for averages of stations in different average latitudes (north to south, indicated in the middle) and average longitudes (A) −66 • i.e. 66 • W, (B) +12 • i.e. 12 • E, (C) +135 • i.e. 135 • E and (D) −135 • i.e. 135 • W. Vertical lines mark the storm commencements.Positive deviations are painted black, negative deviations are shown as hatched.

Fig. 5 .
Fig. 5. Plots of f oF2 ratios for the interval 5-9 April 2000 (storm occurred during 6-7 April) for Boulder and Port Stanley (read out from Araujo-Pradere and Fuller-Rowell, 2002).The thick line is their (empirical) STORM model prediction and full lines are observed values.Positive deviations are painted black, negative deviations are shown as hatched.
• N, 8 • S and 33 • S, for longitude group B around 12 • E. Here, effects are mostly positive on 29 October but negative on 30-31 October.Thus, a mixed effect is seen, probably because of the two separate storms of 29 October and 30 October, but strong positive effects are seen before the storm.