Ionospheric disturbances during a magnetic storm on 6 November 2001 were analyzed using total electron content (TEC) calculated from measurements made with a dense GPS receiver network, GEONET, which covers the whole of Japan and F-layer peak parameters obtained by a meridional ionosonde chain which consists of ionosondes at Wakkanai (45.39°N), Kokubunji (35.71°N), Yamagawa (31.20°N), and Okinawa/Ogimi (26.68°N). Maps of TEC as a function of latitude and time were compared with NmF2 and hpF2. A weak to moderate ionospheric positive storm in terms of fOF2 was associated with the magnetic storm. On the other hand, TEC was nearly doubled at all latitudes during the daytime. This event was the effect of a prompt penetrating eastward electric field in the presence of an enhanced equatorward neutral circulation, which was set up prior to the electric field penetration and persisted for more than 24 hours. The ionosphere was raised simultaneously by ∼100 km at the four ionosonde stations. The small ion-neutral collision frequency at high altitudes results in loading of plasma into the plasmasphere. The significant difference in storm signatures between NmF2 and TEC was interpreted as increased upward plasma diffusion, which worked as a sink for the plasma at the F layer peak. The increase in TEC in the plasmasphere, however, was the order of 10 TEC units, which is insufficient to cause the large observed TEC enhancement but was responsible for maintaining the nighttime TEC enhancement. On the bottomside the plasma distribution departed significantly from the photochemical equilibrium due to the upwelling, and the photochemical production tended to adjust it, providing the major source of the great increase in TEC. At night, positive storm conditions both for NmF2 and TEC persisted at latitudes higher than 33°N, which was caused by the downward plasma flux and the equatorward neutral winds. At latitudes lower than 33°N, negative perturbation was observed in the evening hours, caused by the suppression of the evening enhancement of the eastward electric field.
 Ionospheric storms are disturbances persisting from several hours to days and are associated with magnetic disturbances [e.g., Prölss, 1993, 1995; Fuller-Rowell et al., 1994, 1996], which are distinguished from the traveling ionospheric disturbances (TIDs) that are ionospheric manifestations of atmospheric gravity waves. Ionospheric storms have been recognized in terms of the ionospheric critical frequency (fOF2); phenomena in which fOF2 is significantly reduced are referred to as negative storms, while those in which fOF2 is increased are referred to as positive storms. The negative storms are ascribed to traveling atmospheric disturbances (TADs) accompanied by atmospheric composition changes which originate at polar latitudes and are due to energy deposition associated with magnetic activity, and which then propagate to lower latitudes [Prölss, 1987]. The composition changes, characterized by enhanced [N2]/[O], increases the loss rate, and hence decreases the ionospheric electron density. At the initial phase of a TAD, an equatorward surge pushes the ionosphere up to high altitudes where the electron dissipation process is ineffective, which results in increases in the ionospheric electron density under sunlit conditions. Thus negative storms are often preceded by positive storms depending on the local time when the equatorward surge passes [Prölss, 1993]. Another agent in positive storms is the upward/poleward E × B drift driven by eastward electric fields, which pushes the ionosphere up to high altitudes and results in increases in the ionospheric electron density. Two different origins of such electric field disturbances have been considered. One is prompt penetration of magnetospheric convection electric fields into the low-latitude ionosphere [Kelley et al., 1979]. The other is disturbance dynamo fields generated by a modified circulation of thermospheric wind fields caused by energy inputs into the high-latitude thermosphere [Blanc and Richmond, 1980; Fejer et al., 2000].
 In actual ionospheric storms, the disturbances are a mixture of several processes and two or more processes may compete with each other [Fuller-Rowell et al., 1994]. Forbes et al.  and Forbes  used data from a latitudinal ionosonde chain to separate these processes. The leading mechanism might vary from hour to hour, and the significance of each process might differ from event to event. Thus it is not always easy to identify the determinant of the disturbance in individual storm phases.
 Although fOF2 is the most important ionospheric parameter characterizing ionospheric storms, changes in fOF2 are not sufficient to understand the whole process. When the F layer is raised by eastward electric fields or equatorward thermospheric winds, fOF2 could be decreased or increased depending on the rate of rise and the local time. As reviewed by Rishbeth , plasma flux into the plasmasphere may deplete NmF2 (fOF2) appreciably, and its effect depends on the loss and transport terms, being strongly modified by drift due to the meridional wind and by drift due to the electric field. The decrease in fOF2 in the uplifted F layer peak is due to the redistribution of plasma or a change in the vertical profile of the electron density. The plasmasphere works as a sink for the plasma at the F layer peak during the daytime, while it contributes to maintaining the F layer at night. The increase in fOF2 in the uplifted F layer peak is due to the reduced recombination loss rate of plasma in the lower part of the ionosphere under sunlit conditions, as has often been cited in explaining positive ionospheric storms [Prölss, 1995]. The net change in NmF2 could be either an increase or a decrease depending on the overcoming effect. Likewise, when the F layer is depressed by westward electric fields or poleward thermospheric winds, fOF2 could be decreased or increased. The increase in fOF2 in this case would be due to the changes in the vertical profile caused by the increased ion-neutral collision frequency. The decrease in fOF2 would be caused by the increased recombination loss rate. The increases in fOF2 caused by a depression of the ionosphere never last long, however, because the chemical recombination effect soon becomes dominant.
 Recently, measurements of the ionospheric total electron content (TEC) have become available by using dual frequency GPS radio signals. The behavior of TEC during ionospheric storms is basically similar to that of fOF2, but it is not necessarily the same. When fOF2 is changed by the redistribution of plasma along the magnetic field line, TEC does not change significantly because the redistribution does not result in either the loss or production of plasma in a magnetic flux tube. When fOF2 is reduced by the chemical loss process, TEC will also decrease. In this case, however, the decrease in TEC depends not only on the reduction in plasma at the F layer peak height but also on the density of the plasmasphere, which does not immediately respond to the increase in the chemical loss rate in the lower part of the ionosphere. Thus combined use of TEC data with fOF2 is of great advantage in interpreting ionospheric storms.
 Another advantage of the TEC data derived from GPS is the denser data points as compared with ionosonde networks. Foster et al.  analyzed data from more than 120 GPS sites in North America and the Caribbean. The resulting TEC maps provided snapshots of TEC perturbations, which revealed the latitude-longitude position of storm enhanced density (SED) and plumes with greatly elevated TEC associated with the erosion of the outer plasmasphere. A similar event was analyzed by Vlasov et al. [2003b]. GEONET is the GPS receiver network covering Japan, and it consists of more than 1000 receivers. The dense data points allow for depicting detailed latitudinal changes in the phase and strength of individual storms, enabling us to identify the causes of the disturbances. TEC measurements include a contribution from the plasmasphere, since the satellites' orbit is at 20,000 km. Thus comparing the TEC with other ionospheric parameters may give us some insight into the plasmasphere-ionosphere coupling.
Section 2 describes the data used for the study. Section 3 presents the outline of the storm. Section 4 discusses the processes of the disturbances in chronological order. Section 5 summarizes the results.
2. Data Used for the Study
 The ionospheric storm on 6 November 2001 was analyzed by using data gathered from the dense GPS receiver network, GEONET, and a meridional ionosonde chain (Figure 1). GPS data from ∼200 of the 1000 GEONET receivers were used, for which the P code was available at two frequencies for evaluating TEC [Ma and Maruyama, 2003]. The four ionosonde stations were Wakkanai (45.39°N, 141.69°E; magnetic dip 59.8°), Kokubunji (35.71°N, 139.49°E; magnetic dip 49.3°), Yamagawa (31.20°N, 130.62°E; magnetic dip 44.5°), and Okinawa/Ogimi (26.68°N, 128.16°E; magnetic dip 37.8°). The Okinawa station was moved northeast by about 50 km in November 2001; the new station is referred to as Okinawa/Ogimi or simply Okinawa.
 To evaluate the absolute values of TEC, the instrumental biases of the individual receivers and satellites had to be determined. We determined those biases and TEC self-consistently by using a data set for 24 hours. For this purpose, the area of the data set was divided into 32 cells, with each corresponding to two degrees of latitude and longitude, as shown by the dashed lines in Figure 1. The vertical TEC giving the propagation delay of the GPS radio waves, whose ray paths intersected at a reference altitude within each cell, was assumed to be constant at a given time. The instrumental biases were assumed to be constant during the 24-hour period. Thus we determined TEC values in 32 cells at every quarter hour. The details of this method and an evaluation of its accuracy are given by Ma and Maruyama . The two-dimensional TEC distributions as a function of latitude and time were generated by applying surface harmonic function fitting. As we were interested in the TEC perturbation caused by a magnetic storm specified by universal time, JST (= UT + 9 hours) was used irrespective of the longitude of cell.
 Ionograms were obtained every quarter hour, and the basic parameters were scaled by eye. The virtual height, h′F, directly scaled from the ionogram was not appropriate as an indicator of the layer height for the analysis through the day because h′F is strongly affected in the daytime by production and loss processes at the bottom of the layer irrespective of the F layer dynamics. Instead, the peak height of the F layer, hpF2, was utilized, and it was evaluated based on an empirical formula. A relationship between hpF2 and the ionospheric transmission factor, M(3000)F2, was first introduced by Shimazaki  for a parabolic distribution of electron density. The relationship was later refined by Bradley and Dudeney , who added a correction for retardation by the underlying layers as
Here, no F1 region parameters explicitly appear, but they are accounted for in the numerical coefficient. Berkey and Stonehocker  showed that the formula is highly accurate by comparing the empirical estimates and the real height analysis. fOE appears in the formula, but its scaling was often affected by the sporadic E and interferences, and empirically determined values of fOE [Muggleton, 1975] were adopted instead of observed values.
3. Perspective of the Disturbances on 6 November 2001
Figure 2 shows the interplanetary magnetic field and geomagnetic conditions during the period from 5 to 8 November 2001. The top panel shows the z-component of the interplanetary magnetic field observed by the ACE satellite at the L1 point. The curve is shifted by 1 hour later to compensate for the propagation time from L1 to the Earth. The second panel shows the asymmetric disturbance index, ASY-H, which is a good indicator of auroral substorm activity [Iyemori and Rao, 1996]. The third panel shows the symmetric disturbance index, SYM-H, which is essentially the same as the Dst index except for the slightly different magnetic stations from which the data are gathered [Iyemori and Rao, 1996]. The bottom panel shows the Kp index. The storm sudden commencement was observed at 0151 UT (1051 JST) on 6 November, and an asymmetric ring current immediately developed. The Kp index reached 9_. Prior to the SC, the southward condition of IMF Bz persisted for ∼7 hours, and the ring current (third panel) gradually developed from 1900 UT on 5 November following irregular disturbances. Corresponding to these disturbances, the Kp index increased from 3_ to 5.
 The variations in the F layer peak electron density, NmF2, at the four stations, Wakkanai (WAK), Kokubunji (KOK), Yamagawa (YAM), and Okinawa (OKI) are shown in Figure 3 (upper panel), along with the Kp indices (lower panel), for the period from 2 to 7 November 2001 (JST). The quarter hourly values of NmF2 are shown by the dots. The mean diurnal variations in NmF2 over the period from 2 to 4 November are taken as a reference representing quiet days (shown by the solid curves). Corresponding to the magnetic disturbance on 6 November, NmF2 fluctuated at all stations. At sunrise it started to increase, but soon after, it decreased. At lower latitudes the negative turn occurred earlier and the amplitude of the perturbation was larger. The NmF2 perturbation turned positive again after 1200 JST at the three lower-latitude stations and after 1330 JST at Wakkanai. At the two higher-latitude stations, Wakkanai and Kokubunji, the positive perturbation persisted throughout the whole night. At Yamagawa and Okinawa, NmF2 perturbation turned negative at around 1500 JST. In the evening hours at Okinawa, NmF2 was deeply depressed and remained so until midnight. The perturbation turned positive again at 2200 JST at Yamagawa and at 0130 JST of the next day at Okinawa. Although the fluctuation in NmF2 was complicated, as described here, its amplitude was not very significant, except for the large depression during the evening hours at Okinawa, as compared with the disturbances often observed during severe magnetic storm periods.
 TEC maps as a function of latitude and time are shown in Figure 4. The upper panel shows the reference for the quiet period from 2 to 4 November 2001 (the same period as for the fOF2 analysis). The numbers on the contours are in TEC units, or 1 × 1016 el/m2. TEC started to increase after sunrise and reached its peak value at noon at 45°N and at 1430 JST at 27°N. At latitudes lower than 30°N, a secondary TEC peak was observed after sunset, which corresponds to the evening enhancement of the upward E × B drift. The second panel shows the TEC on the storm day of 6 November 2001. An extreme increase was observed during the daytime. TEC peaked near 1345 JST at 27°N and 1415 JST at 45°N. To highlight the TEC perturbation on the storm day, the reference was subtracted in the third panel of Figure 4. This panel (ΔTEC) clearly depicts a large enhancement in TEC, which started between 1100 and 1200 JST almost simultaneously at all latitudes. The enhancement was larger at the lower latitudes, and its peak shifted to later times with increasing latitude, as shown by the dotted curve: 1345 JST at 27°N and 1445 JST at 45°N. Prior to the anomalous enhancement, TEC was depressed from 0900 to 1130 JST. This depression also was larger at the lower latitudes. After the enhancement peak, a moderately high ΔTEC continued until midnight at latitudes higher than 33°N. The bottom panel shows the ratio of TEC on the storm day to the reference. Hereafter, we refer to this quantity as the relative perturbation. The relative perturbation is large at nighttime, since the background TEC is small then. Moreover, the time of peak daytime enhancement is shifted later at higher latitudes as compared with the diurnal peak of ΔTEC, which is shown by the dotted curve in the third panel and repeated here. The delay increased with latitude: the relative perturbation peaked near 1600 JST at 45°N.
 In general, the electron density at the height of the F layer peak, NmF2, contributes largely to TEC. Thus both quantities can be expected to behave similarly during ionospheric storms. The upper two panels of Figure 5 compare ΔTEC and ΔNmF2 for the period from 1200 JST on 5 November to 1200 JST on 7 November. At Wakkanai, the variations in the two quantities were fairly dissimilar, contrary to expectations, during the major disturbance period as follows. In terms of ΔTEC, a gradual increase started at 0600 JST on 6 November and the rate of increase abruptly became large at 1100 JST. After reaching the maximum at 1445 JST, ΔTEC monotonically decreased. For ΔNmF2, there were two broad maxima near 1000 and 1530 JST, and between them, ΔNmF2 dropped to a negative value despite the large enhancement in TEC. In contrast, at the lower three stations, from Kokubunji to Okinawa, the two quantities' trends resembled each other from 0900 to 1500 JST, with a negative value centered at 1030 JST and a positive one centered near 1400 JST for both ΔTEC and ΔNmF2. We find quantitative dissimilarities, however, in comparing the amplitudes of the positive and negative perturbations. For ΔTEC the positive perturbation (100 TEC units at the latitude of Okinawa) was quite large compared with the negative perturbation (20 TEC units at the same latitude), while for ΔNmF2, the positive perturbation was comparable to or smaller than the negative perturbation. These dissimilarities were prominent at lower latitudes. Another large quantitative dissimilarity between the behavior of ΔTEC and ΔNmF2 was the negative value centered at 2100 JST at the latitude of Okinawa, which was quite deep for ΔNmF2 but shallow for ΔTEC.
 The ionospheric disturbances on 6 November 2001 were quite complicated, as described in the previous section. Although the significance of ionospheric storms can be quantified in terms of fOF2 and TEC, the variations in ionospheric height are important for interpreting the processes governing storms. This is because these variations are direct manifestations of the underlying dynamics, such as movements of the layer owing to meridional winds and zonal electric fields. Figure 6 shows the diurnal variation in hpF2 for Kokubunji. The thin lines are for 2 to 5 November and the thick line is for 6 November. The day-to-day variation was large during the nighttime but small during the day except for an uplift of ∼100 km after 1100 JST on 6 November. Figure 7 shows individual ionograms at Kokubunji at each hour from 0800 to 1500 JST on 6 November (solid curves), before and after the onset of the uplift, along with ionograms for the same local time on the previous day as a reference (dotted curves). Only the O mode traces of the F layer are shown. For the sake of comparing hpF2 to the TEC and NmF2 perturbations, a quiet condition reference is taken from 2 to 4 November (the same period as for the other parameters) and subtracted from the individual hpF2 values. These results (ΔhpF2) are shown in the bottom panel of Figure 5. As the leading mechanisms of the disturbances are considered to vary with the time and latitude, we have divided the whole term into several periods of disturbances, and we examine them here in chronological order. The times separating each period are not very strict because they may vary with the signature and parameter under discussion.
4.1. Before 0700 JST on 6 November
 On the first day (5 November) and before 0700 JST on the second day (6 November), TEC and NmF2 did not differ significantly from the reference, as seen in the upper two panels of Figure 5. On the other hand, hpF2 varied significantly from 0000 to 0600 JST on 6 November. The changes in the height clearly revealed the nature of TIDs. The phases of positive and negative ΔhpF2 shifted later with decreasing latitudes, as shown by the long dashed lines. The estimated velocity was 740 m/s, which is a typical value for a large-scale TID. Similar TIDs occurred from 1800 JST on 6 November to 0600 JST on 7 November, with velocities between 350 and 700 m/s. Despite the large amplitude of ΔhpF2, the corresponding changes in ΔTEC and ΔNmF2 were not significant. Thus when we connect hpF2 variations to TEC and NmF2 disturbances, the TIDs appear as noise to be distinguished from meridional wind effects associated with thermospheric circulation changes and E × B drifts by zonal electric fields causing hpF2 variations. We do this distinction by analyzing the data from the meridional ionosonde chain, as demonstrated above.
4.2. From 0700 to 1100 JST on 6 November
 The TEC perturbation simultaneously turned negative from 0900 to 1100 JST at latitudes lower than 38°N, as shown in the upper panel of Figure 5. The magnitude of the depression was larger at lower latitudes. Although ΔTEC remained positive at higher latitudes, there still seems to be a component of TEC depression like that at lower latitudes, with another component of a slowly varying increase superposed on the sharp depression. Corresponding to the TEC depression, ΔNmF2 was depressed at Okinawa, Yamagawa, and Kokubunji, but the time of maximum perturbation was delayed by about half an hour from that for ΔTEC. At Wakkanai, ΔNmF2 also increased, similarly to ΔTEC. Although ΔhpF2 during this period (the bottom panel of Figure 5) was not very large, there was a general tendency of a decrease with time at the three higher-latitude stations, as well as a turn from a weakly positive to a weakly negative ΔhpF2 at Okinawa. The geomagnetic conditions during this period were moderately disturbed. The Kp indices for the 1800–2100 UT and 2100–2400 UT periods on 5 November (0300–0600 JST and 0600–0900 JST periods on 6 November) were 5+ and 5, respectively. The Kp index for the 0000–0300 UT period on 6 November (0900–1200 JST period) reached 9_, but SC at 0151 UT (1051 JST) strongly contributed to the high Kp value.
 The individual ionograms for 0800, 0900, and 1000 JST at Kokubunji in Figure 7 indicate that the decrease in fOF2 was associated with the depression of the layer. At 0800 JST the traces for 5 and 6 November were identical, but 1 hour later the lower frequency part of the trace on 6 November was lowered and fOF2 slightly decreased. At 1000 JST, fOF2 was further depressed.
 Putting the observations of TEC, NmF2, and hpF2 together, we can see that there were two competing factors determining the TEC variations. One was a gradual increase in TEC, which started at 0700 JST or earlier and was identifiable at higher latitudes. The other was an abrupt decrease in ΔTEC, which started at 0800 JST (most evident in the bottom panel of Figure 4) and was prominent at lower latitudes. The former can be attributed to the setup of equatorward neutral winds, whose strengths are reduced toward lower latitudes [Fuller-Rowell et al., 1994; Fejer et al., 2000] under moderately disturbed geomagnetic conditions. The latter can be attributed to an increase in the westward electric field. There are two candidates for the source of this electric field. The short-term effect of the ionospheric disturbance dynamo driven by the high-latitude geomagnetic activity during this time period is westward electric field perturbation [Scherliess and Fejer, 1997]. The quicklook AE index (not shown here) distributed by the World Data Center for Geomagnetism, Kyoto, showed around 500 nT after 1900 UT on 5 November (0400 JST on 6 November) and increased to ∼1200 nT near 2240 UT on 5 November (0740 JST on 6 November). Thus the disturbance dynamo is a probable cause of the TEC depression. The latitudinal variation in the perturbation electric field as calculated by Blanc and Richmond , however, is contrary to the observed latitudinal variation in TEC depression, even though the difference in the magnetic dip angle is taken accounted of. This might be explained by the coexisting equatorward neutral wind effect that was weakened at the lower latitudes. The other candidate source of the westward electric field is prompt penetration of magnetospheric electric fields. This may not be a major source, and it will be discussed later, along with the more prominent but eastward prompt penetrating electric field perturbation.
 Another fluctuation source is the TIDs, which made interpretation of the height variation difficult during this period. In the middle panel of Figure 5, the small dips in ΔNmF2 around 0700 JST delayed with decreasing latitude, as shown by the long dashed line, which indicates that the TID activity was continuing beyond the sunrise. If we consider that the negative ΔhpF2 of the TID near 0500 JST was the largest at Okinawa, TID-related fluctuation cannot be ruled out for the negative ΔhpF2 at Okinawa near 0900 JST.
 The onset time of the equatorward thermospheric circulation is not easy to determine precisely because of the preceding wavy fluctuation in hpF2 caused by TIDs and the small ionization rate in the early morning. The relative perturbation in the bottom panel of Figure 4 suggests that it began near 0700 JST on 6 November (2200 UT on 5 November), when a moderate magnetic disturbance persisted for two 3-hour periods of Kp (Kp = 4 and 5+ for 1500–1800 UT and 1800–2100 UT periods, respectively). Theoretical calculations with an idealized stepwise storm input by Fuller-Rowell et al. [1994, 1996] predict equatorward neutral winds. In their calculations an equatorward surge precedes the circulation changes. In our present case, no indication of such equatorward surge was observed and the setup of the equatorward neutral wind appears to be gradual. This might be expected from the gradual increase in the magnetic disturbance prior to the SC, as seen in Figure 2.
4.3. Enhancement of TEC From 1100 to 1600 JST on 6 November
 The TEC enhancement during this period was highly prominent throughout the whole storm event. At the latitude of Okinawa (27°N), TEC rapidly increased by 100 TEC units in 2 hours, and at the latitude of Wakkanai (45°N) it increased by more than 50 TEC units. The start of the enhancement was simultaneous at all latitudes but the time of the peak was delayed with increasing latitude: ΔTEC reached its peak value at 1345 JST at 27°N and at 1445 JST at 45°N. Despite the spectacular increases in TEC, the corresponding ΔNmF2 (the second panel of Figure 5) were not nearly as significant, and the signs were not necessarily the same as ΔTEC. At Wakkanai, during the period from 1000 to 1300 JST, ΔTEC monotonically increased, while ΔNmF2 decreased. The ionospheric height perturbation, ΔhpF2, started to increase just before 1100 JST, which nearly coincided with the start of the increase in ΔTEC, as shown by the short dashed line in Figure 5. It also coincided with the SC within the 15-min time resolution of the ionosonde observations. The rapid increase in ΔhpF2 continued for 2 hours, and the rate of the increase was larger at the lower latitudes. After that, ΔhpF2 remained at a high level for 24 hours, with several TID fluctuations superimposed on it.
4.3.1. Prompt Penetrating Electric Fields
 The simultaneous increases in ΔTEC and ΔhpF2 strongly suggest that the disturbance was caused by prompt penetrating magnetospheric electric fields directed eastward [e.g., Kelley et al., 1979; Fejer et al., 1979; Spiro et al., 1988]. Before the storm onset at 0151 UT on 6 November (the plot of the ACE data is shifted later by 1 hour), a southward Bz condition continued for ∼7 hours, with a short impulsive reversal as depicted in the top panel of Figure 2. Near 0200 UT on 6 November, the southward Bz abruptly increased. This coincided with the SC, an abrupt increase in the asymmetric ring current index, ASY-H, and the layer height increase observed by the ionosondes. Although the detailed process of the penetration of magnetospheric electric fields into the lower latitudes will not be discussed in this paper, there is little doubt that there was a strong eastward prompt penetrating electric field. The sign of the disturbed electric field is consistent with a theoretical calculation by Spiro et al. . Analyses of low-latitude and equatorial magnetograms also support the penetrating electric fields hypothesis (T. Kikuchi, private communication, 2003).
 Just before the storm onset, the southward Bz was temporarily reduced from −12 to −4 nT. The corresponding variations in the ASY-H and SYM-H plots are visible for the period from 0030 to 0150 UT on 6 November. Ionospheric responses corresponding to the anticipated westward electric field perturbation, however, were not observed. As the ΔTEC and ΔNmF2 depression discussed in section 4.2 started earlier (near 2300 UT on 5 November) than the change in the magnetic field condition, prompt penetrating westward electric fields are ruled out for a source of the depression.
 For the major electric field disturbance starting with the storm onset, the strength of the perturbation electric field was estimated from the upward drift velocity of the F layer peak and the magnetic dip angle. At Okinawa the upward velocity inferred from the slope of the curve during the first half an hour in Figure 5 was 28 m/s, which corresponds to the electric field of 1.4 mV/m. This value is somewhat of an underestimate, however, because the apparent vertical drift of the ionospheric peak height in response to the stepwise increase in the eastward electric field slows with the departure from the equilibrium peak height and thus with time [Maruyama, 1990].
4.3.2. Upward Plasma Flux
 The insignificant increases and even the negative perturbation in NmF2 at Wakkanai compared with the large TEC enhancement at the same latitudes can be explained as follows. In the topside ionosphere well above the F2 peak, the general solution of the diffusion equation for the equilibrium state in isothermal atmosphere is approximated as
where N is the electron density at height h and H is the scale height of the neutral atmosphere. The first term on the right-hand side of equation (4) represents the dominant diffusive equilibrium distribution. The second term represents the flux solution [Rishbeth, 1986], which specifies departures from diffusive equilibrium, and its contribution to the vertical profile of the electron density quickly diminishes with increasing height. Lockwood and Titheridge  showed departures of the plasma scale height from that expected for the diffusive equilibrium, using the topside ionograms obtained by the Alouette 1 satellite. The plasma scale height often decreased, i.e., the plasma gradient increased, from the diffusive equilibrium below ∼700 km (e.g., their Figure 2). At great heights (say at an altitude between the F2 peak and the ion-transition height) the field aligned flux is approximately given by
where Dm is the plasma diffusion coefficient at the F2 peak and I is the dip angle of the magnetic field [Rishbeth, 1986]. The plasma diffusion coefficient depends on the ion-neutral collision frequency, which decreases with the scale height, H, and it thus increases exponentially with height. When the topside ionosphere is in a diffusive equilibrium state, N2 is zero. Generally, however, there are departures from the equilibrium condition, and the flux is upward (N2 > 0) by day and downward (N2 < 0) at night, causing the diurnal cycle of plasmasphere replenishment. Rishbeth  estimated N2 to be roughly 30 percent of a typical midlatitude, mid-solar cycle daytime value of NmF2.
 We next consider the ionosphere when it is not in diffusive equilibrium with positive N2. When the layer is abruptly uplifted by 100 km, which is slightly higher than the scale height of the neutral atmosphere (∼80 km), without any change in the shape of the vertical distribution of ions, the upward flux given by equation (5) increases by more than a factor of e. We will estimate the contribution of the upward flux to TEC enhancement in the plasmasphere, using following parameters: NmF2 = 2.3 × 1012 m−3, H = 80 km, N2 = 0.3NmF2, I = 50° (Kokubunji is assumed), and Dm = 1.5 × 106 m2s−1 (at 350 km), 5.4 × 106 m2s−1 (at 450 km). The diffusion constants and scale height were calculated with parameters obtained from the MSIS model [Hedin, 1991] and standard values used commonly [e.g., Bailey and Sellek, 1990]. The resulting upward field aligned flux is Φ = 5 × 1012 m−2s−1(hpF2 = 350 km), and Φ = 1.8 × 1013 m−2s−1(hpF2 = 450 km). The upward flux increased by 1.3 × 1013 m−2s−1 with the uplift of the F layer. The observed TEC increased by 70 TEC units in 3 hours. At a very rough estimate, assuming that the above flux persisted for 3 hours, we obtain 1.4 × 1017 m−2 or 14 TEC units. Because the cross section of the magnetic flux tube increases with height and flux tubes at higher altitudes are connected to the ionosphere at higher latitudes, contribution of the upward flux to the increase in TEC would be smaller than this.
 The upward flux given by equation (5) depends not only on the increase in the peak height but also on the departure from the diffusive equilibrium, or N2/N1 in the initial state. Lockwood and Titheridge  investigated departures from diffusive equilibrium in the topside ionosphere in terms of the ratio of the field-aligned flux of O+ (ϕ) to its limiting flux value (ϕL) due to frictional drag with the ambient neutral atmosphere. Generally, the departures were large at low latitudes (10–30° geomagnetic latitude range) and near 1400 LT during magnetically quiet periods. Additional evidence for the departures from diffusive equilibrium can be found in the slab thickness (τ) of the ionosphere, which is given by the ratio of TEC to NmF2. Titheridge  found that τ is abnormally small compared with that corresponding to an equilibrium profile between sunrise and noon in winter at midlatitudes. The small τ could arise from the slower rate of increase in TEC than that in NmF2 when the new ionization replenishes the ionosphere after sunrise. Thus the topside ionosphere at the same local time as the uplift event but on the reference days was expected to depart significantly from the diffusive equilibrium, and the insignificant increase, or even decrease, in NmF2 was anticipated during the uplift event.
 Another possible mechanism for the upward plasma flux that causes the decrease in NmF2 is a change in the magnetic flux tube. If the volume of the magnetic flux tube significantly increases, upward plasma flux will be induced to fill the thinned flux tube. However, what we are discussing here is at the lower midlatitudes, and the magnetic field configuration does not change very much even during a severe magnetic storm event. The volume of the magnetic flux tube that passes the peak electron density height above Kokubunji increases by 6% when the height increases by 100 km. Thus this mechanism is not the case here.
4.3.3. Increase in TEC
 When the production and loss terms are considered at a fixed height on the bottomside, plasma is first depleted by the uplift of the layer, but soon the depletion is adjusted by the photoionization, resulting the increase in TEC. The timescale of this adjustment is approximately 1/β (β is the recombination rate), which is ∼103 s at h = 250 km and 104 s at h = 350 km. When the layer is uplifted, the reduced recombination rate at the layer peak causes an increase in NmF2, which competes and compensates for the depression of the plasma density due to the increased upward field aligned flux. There is a time lag of 2 ∼ 3 hours for β ∼ 10−4 between the NmF2 reduction and the compensation, and we may observe a transient negative disturbance prior to the positive storm, as seen at Wakkanai. A similar transient negative NmF2 perturbation associated with a height increase prior to a positive storm can be found in published data (Figure 8.3.9 of Prölss  and Figure 3 of Prölss ). Unfortunately, no TEC data were available for this event. The peculiarity of the 6 November event was that the uplift of the layer was quite large during 1100–1300 JST, with the elevated hpF2 remaining high till beyond sunset (Figure 6). As a result, the compensation might have been insufficient, and NmF2 increased insignificantly, in contrast with the widely accepted positive storm scenario.
 The above explanation is consistent with the individual ionograms shown in Figure 7. At 1100 JST the uplift of the F layer started, and the trace shifted slightly upward at all frequencies. One hour later, the trace shifted further upward except for the low-frequency part, where the photochemical adjustment proceeds quickly. At 1300 JST the F peak height was near the maximum, but the increase in fOF2 during this hour was insignificant in spite of the large F layer peak altitude. The kink in the trace near 7 MHz had become prominent by this time. Finally, at 1500 JST, the low-frequency part of the trace was adjusted while maintaining the large F layer peak height (see Figure 6 also) and fOF2. This discussion is based on the virtual heights, but it also holds qualitatively for the real height profile.
4.3.4. Delayed Enhancement in TEC
 The large increase in TEC has been ascribed to the prompt penetrating magnetospheric electric fields. Such electric fields should not differ much at different latitudes. The peak of the TEC enhancement, however, shifted to later times with increasing latitude, as shown in the third and last panels in Figure 4. When the layer is uplifted, the ion composition would depart from chemical equilibrium near the ion-transition height, and O+ ions are converted to H+ by the charge exchange reaction. The H+ flux is limited by Coulomb drag with O+. The limiting flux of H+ through O+ is an order of magnitude smaller than that on O+ flowing through the neutral atmosphere [Lockwood and Titheridge, 1982]. As a result, the ion transition height is raised, and the H+ ion distribution departs from diffusive equilibrium where O+ predominates. The H+ flux will therefore persist for longer times, as compared with the upward O+ flux, before diffusive equilibrium reestablished. The delay in ΔTEC reaching its peak at higher latitudes may imply that a long time was required to fill the plasmaspheric magnetic flux tube having a large volume at higher latitudes.
4.4. Recovery of TEC Enhancement From 1600 JST to Midnight
 The TEC perturbation recovery at latitudes higher than 33°N greatly differed from that at lower latitudes during the period following the large enhancement. At the higher latitudes, ΔTEC, as seen in the upper panel of Figure 5, remained nearly constant from 1700 to 2300 JST after the initial rapid decay from 50 to 30 TEC units. At the lower latitudes it continued to decay until 2100 JST. The decay became significant with decreasing latitude. The variation in NmF2 was qualitatively similar to that of TEC. At Wakkanai and Kokubunji, a weak positive storm condition persisted. At Yamagawa and Okinawa, however, the variations in ΔNmF2 and ΔTEC quantitatively differed from each other. The difference was significant at Okinawa, where the duration of the period of negative ΔNmF2 was much longer than that of ΔTEC, and its maximum depression was substantially larger.
 During this period the ΔhpF2 values were persistently positive at the four stations. Several TID activities overlapped in the plots of ΔhpF2, but the corresponding changes in ΔTEC and ΔNmF2 were not very significant. The general tendency of high layer heights and positive ΔNmF2 and ΔTEC at the latitudes higher than 33°N does not conflict with the widely accepted ionospheric storm mechanism [e.g., Prölss, 1995], though the magnitude of ΔNmF2 was much smaller compared with that of ΔTEC. On the other hand, the high layer heights and negative ΔNmF2 and ΔTEC at the latitudes lower than 33°N after sunset might be incompatible because the chemical recombination process becomes less effective when the layer height is high.
4.4.1. Latitudes Lower Than 33°N
 The uplifted layer height and the large negative ΔNmF2 and ΔTEC at Okinawa centered at 2100 JST do not amount to a contradiction if we consider an equatorial anomaly developed by the evening enhancement of the eastward electric fields. The time of the large negative ΔTEC coincided with the time of the secondary TEC peak in a quiet period attributed to the equatorial anomaly caused by the evening electric field enhancement, as seen in the top panel of Figure 4. Thus the negative ΔNmF2 and ΔTEC could have been caused by a suppression of the evening enhancement of the eastward electric fields.
 Associated with geomagnetic disturbances, middle latitude ionospheric electric fields are often changed by the modified thermospheric circulation, termed a disturbance dynamo [Blanc and Richmond, 1980]. The relationship between the suppression of the evening enhancement and the disturbance dynamo is not very clear because the evening enhancement is a result of F region dynamo action driven by the prevailing eastward neutral wind [e.g., Farley et al., 1986; Eccles, 1998]. Observations, however, show that the equatorial perturbation electric field is directed westward with a large scatter during evening hours [Scherliess and Fejer, 1997]. The suppression or total absence of evening enhancement inferred here is similar to the events previously observed during magnetically disturbed periods at equatorial stations [Abdu et al., 1995, 1997].
 The lack of the fountain effect causes the equatorial anomaly crests to recede back to the equator and results in negative ΔNmF2 at the high-latitude edge of the crests. The amplitude of the evening enhancement or vertical plasma drift velocity can be greatly reduced with altitude and hence with latitude [Murphy and Heelis, 1986; Pingree and Fejer, 1987]. Thus the manifestation of the evening enhancement in the ionosphere is limited to low latitudes. This fact leads to two consequences regarding the observed ΔTEC and ΔNmF2 signatures. One is the latitudinal extent of the negative ΔTEC, which differed between the two events centered at 1030 and 2100 JST. The former, which was attributed to the westward disturbance dynamo electric field, extended beyond 40°N, while the latter, which was caused by the receding of the equatorial anomaly, was confined to latitudes lower than 35°N. Another signature is the difference in significance between the negative ΔTEC and ΔNmF2 at the latitude of Okinawa at 2100 JST. The electron density along the magnetic field line passing through the F layer peak near Okinawa was greatly affected, and NmF2 over Okinawa was also greatly perturbed. In contrast, TEC over Okinawa included a major contribution from the topside ionosphere, for which magnetic field lines pass through the F layer peak at higher latitudes, where the effect of changes in the evening enhancement is negligible.
4.4.2. Latitudes Higher Than 33°N
 The continuing positive ΔhpF2 over the four stations, irrespective of the weakening or reversal of the eastward electric field at lower latitudes, implies that the uplift was caused by the equatorward thermospheric neutral wind. The equatorward circulation started in the morning at 0700 JST, as already discussed. According to an analysis of a large database of wind measurements gathered by the Upper Atmosphere Research Satellite (UARS), the meridional perturbation winds during magnetically disturbed periods are directed equatorward throughout the day, and their amplitudes increase with latitude and decrease with local time [Fejer et al., 2000]. In the present case the equatorward winds seemed to be sustained beyond the evening hours. In the study by Fejer et al., data for the condition of Kp > 5 were excluded. In the present case, Kp reached 9_, and it is plausible that a large persistent equatorward wind perturbation occurred.
 The nighttime positive perturbation of TEC and NmF2 at latitudes higher than 33°N can be explained by the combined effects of the downward plasma flux [e.g., Vlasov et al., 2003a] from the plasmasphere and the sustenance of the ionosphere at a higher level by the equatorward neutral wind. The value of ΔTEC at Kokubunji after sunset was ∼20 TEC units, which is on the order of the TEC enhancement in the plasmasphere as estimated in the previous section. In the bottom panel of Figure 4, the decay rate of the relative perturbation during the period from 1500 to 1700 JST was gradual compared with that of ΔTEC in the third panel of Figure 4, suggesting that the plasmasphere acted as a plasma reservoir after the time of the TEC enhancement peak.
4.5. From 0000 to 1200 JST on 7 November
 During this period the perturbations of TEC and NmF2 were consistent with each other and hpF2 was still at a high level. The general tendency was for ΔTEC to be larger at lower latitudes. However, ΔhpF2 was larger at higher latitudes. The rapid recovery of the negative ΔNmF2 just before midnight and its positive turn at Okinawa suggest the effect of the disturbance dynamo electric field, which is generally eastward and increases with time toward sunrise [Scherliess and Fejer, 1997]. Although it is not very easy to determine the dominant process maintaining the high layer heights, both the equatorward neutral wind and the eastward disturbance dynamo electric field may be involved.
 Studies of ionospheric storms have long been based mostly on fOF2 (NmF2) variations [Forbes et al., 1988; Forbes, 1989; Fuller-Rowell et al., 1994, 1996; Prölss, 1995]. There are, however, limitations to interpreting storm activity from ionosonde observations only. In particular, ionospheric positive storms involving dynamical processes, such as layer height changes caused by thermospheric neutral wind effects and E × B drifts, are complicated because fOF2 may increase or decrease even when the layer height increases. Furthermore, it is often difficult to determine the leading mechanism of ionospheric height change and thus the causes of disturbances of fOF2.
 We have shown that the combined use of a dense GPS receiver network and an ionosonde chain was of great advantage in interpreting an ionospheric disturbance on 6 November 2001, when there was a severe geomagnetic storm. A prominent feature of the disturbance was a great increase in total electron content (TEC) over Japan, when the F layer was uplifted by 100 km. TEC increased by more than 100 TEC units at the latitude of Okinawa (27°N) and by 50 TEC units at the latitude of Wakkanai (45°N), that is, it nearly doubled from the quiet level everywhere. In spite of such a large increase in TEC, NmF2 did not change very much, being slightly greater than the range of day-to-day variability, except during the evening hours at Okinawa.
 The discrepancy between the TEC and NmF2 disturbances was interpreted as increased upward plasma diffusion, when the layer was raised, which worked as a sink for the plasma at the F layer peak. The upward flux loaded plasma into the plasmasphere by the order of 10 TEC units, which is insufficient to cause the large observed TEC enhancement but was responsible for maintaining the nighttime TEC enhancement. On the bottomside, the plasma distribution departed significantly from the photochemical equilibrium due to the upwelling, and the photochemical production tended to adjust it, providing the major source of the great increase in TEC. At the F layer peak, however, it took many hours to compensate for the departure from the photochemical equilibrium, which resulted in the insignificant increase in NmF2. This large disturbance in TEC was due to prompt penetration of magnetospheric electric fields, together with an equatorward thermospheric neutral circulation, which was set up prior to the penetration of the electric fields. If we had had only ionosonde data to analyze, the disturbance would not have been recognized as a severe ionospheric storm in terms of fOF2. The term, TEC storm, may properly describe the disturbance.
 The geomagnetic disturbance indices were provided by the World Data Center for Magnetism, Kyoto University. The authors are grateful to Takashi Kikuchi for his comments on the electric field disturbances. The authors thank the ACE MAG instrument team and the ACE Science Center for providing the ACE data.
 Arthur Richmond thanks Mangalathayil Ali Abdu and Michael N. Vlasov for their assistance in evaluating this paper.