This study presents the Global Self-Consistent Model of the Thermosphere, Ionosphere and Protonosphere (GSM TIP) numerical simulations of the 9–14 September 2005 geomagnetic storm effects in the middle- and low-latitude ionosphere. Recent modifications to the GSM TIP model include adding an empirical model of high-energy electron precipitation and introducing a high-resolution (1 min) calculation of region 2 field-aligned currents and a cross-cap potential difference. These modifications resulted in better representation of such effects as penetration of the magnetospheric convection electric field to lower latitudes and the overshielding. The model also includes simulation of solar flare effects. Comparison of model results with observational data at Millstone Hill (42.6°N, 71.5°W, USA), Arecibo (18.3°N, 66.8°W, Puerto Rico), Jicamarca (11.9°S, 76.9°W, Peru), Palmas (10.2°S, 48.2°W, Brazil), and San Jose Campos (23.2°S, 45.9°W, Brazil) shows good agreement of ionospheric disturbances caused by this storm sequence. In this paper we consider in detail the formation mechanism of the additional layers in an equatorial ionosphere during geomagnetic storms. During geomagnetic storms, the nonuniform in height zonal electric field is generated at the geomagnetic equator. This electric field forms the additional layers in the F region of equatorial ionosphere.
 The majority of numerical modeling studies of the ionospheric response to geomagnetic storms takes into account the changes of a cross-polar cap potential difference. Besides this, in model calculations it is necessary to account for the changes of energy and flux energy of high-energy particle precipitation and spatial-temporal variations of the Region 2 field-aligned currents (R2 FAC). The inclusion of such inputs in global numerical models allows more accurate description of the effects of penetration of magnetospheric convection electric field to lower latitudes, the overshielding effects, and the effects of the disturbed dynamo electric field. Most modeling studies of ionospheric response do not consider accumulative effects of all possible drivers. For example, studies of ionospheric response to geomagnetic storms do not include effects of solar flares, though such effects are studied separately [Meier et al., 2002; Huba et al., 2005; Smithtro et al., 2006; Pawlowski and Ridley, 2009].
 The Global Self-Consistent Model of the Thermosphere, Ionosphere, Protonosphere (GSM TIP) [Namgaladze et al., 1988, 1991] allows modeling studies with all the drivers described above. This study presents the middle- and low-latitude ionospheric effects of geomagnetic storms of 9–14 September 2005 simulated by the GSM TIP. The use of numerical modeling for the description of the ionospheric behavior during geomagnetic disturbances allows investigation of the physical mechanisms responsible for the ionospheric disturbances. In this paper, we discuss the role of each possible mechanism of ionospheric disturbance during the sequence of geomagnetic storms. In addition, we use the observational data from ionosondes and Incoherent Scatter Radars (ISRs) for comparison of our modeling results with experimental data. The paper also considers in detail the formation mechanism of the additional layers in the equatorial ionosphere during geomagnetic disturbances.
2. Description of the Modeled Event and Statement of the Problem
 The 9–14 September 2005 period was marked by a series of consecutive geomagnetic storms. On 7 September a huge X17 class solar flare occurred. Although not directed at the Earth, this solar flare caused a weak geomagnetic storm with storm sudden commencement (SSC) at 1401 UT on 9 September. On 9 September at 0243 UT Sunspot 798 hurled a magnitude X17 flare into space. This solar flare was one of the ten most powerful flares ever registered. It was accompanied by a coronal mass ejection, and the resultant shock wave has reached the Earth on 10 September, causing a weak geomagnetic storm with SSC near 0600 UT. This weak storm was followed by a strong magnetic storm on 11 September with SSC at 0114 UT. This storm which lasted until 15 September was caused by the second shock wave from the following flare on the Sun. This storm has resulted in the increase of auroral activity, the radio blackout and the ionospheric storm. Figure 1 describes the behavior of Kp, AE and Dst indices of geomagnetic activity and index of solar activity level, F10.7, for the period of 9–14 September 2005. The event under study occurred at average level of solar activity, with F10.7 index varying from 101 up to 120. It is necessary to note the high flare activity during the considered period, in which there were 5 flares on the Sun (see Table 1).
Table 1. Solar Flares During Geomagnetic Storm Sequence on 9–14 September 2005
 The observational data of the ionospheric effects of these geomagnetic storms at various midlatitude stations have been presented and analyzed by Goncharenko et al.  and Kurkin et al. . Ionospheric effects of these storms have been also analyzed with two global numerical models: TIEGCM (Thermosphere-Ionosphere Electrodynamics General Circulation Model) [Lu et al., 2008] and GSM TIP model [Klimenko et al., 2009, 2010, 2011]. The comparison of model calculated results of different ionospheric parameters with the observational data, presented by Klimenko et al. , has revealed the satisfactory qualitative agreement. Suggested reasons for model/data differences include the coarse temporal resolution of the model input parameters (e.g., 3 h Kp index), the use of the dipole approach of geomagnetic field in the GSM TIP model, and the absence of solar flare effects in the model. The use of the dipole approach in the GSM TIP model does not allow considering the distortions of the Earth's magnetic field during geomagnetic storms: the compression of geomagnetic field lines on the magnetosphere dayside and the expansion on the nightside. The compression (expansion) of geomagnetic field lines decreases (increases) the volume of plasma tubes that should lead to the increase (decrease) in electron density. The use of the real geomagnetic field would allow taking into account these effects during geomagnetic storms.
 At the present stage of GSM TIP model development the use of realistic geomagnetic field represents a difficult problem which requires the development of an absolutely new model. The current study attempts to address two other reasons of data/model differences in the storm time ionospheric effects on 9–14 September 2005. First, instead of dependence of the model input parameters, such as a cross-polar cap potential difference and R2 FAC, from a 3 h Kp index, we used their dependence from AE index with 1 min temporal resolution. Second, we examined the ionospheric effects of the solar flares included in Table 1. During simulation, the parameters of solar flares were considered according to Koren'kov and Namgaladze  and Leonovich and Taschilin . In this case, we took into account the changes in solar radiation in the X-ray and UV parts of the spectrum only.
 We investigate the ionospheric effects of geomagnetic storm sequence using the GSM TIP model [Namgaladze et al., 1988, 1991], developed in WD IZMIRAN (West Department of N.V. Pushkov Institute of terrestrial magnetism, ionosphere and radio wave propagation RAS). This model calculates time-dependent global three-dimensional distributions of temperature, composition (O2, N2, O) and velocity vector of neutral gas, the density, temperature and velocity vectors of atomic (O+, H+) and molecular (N2+, O2+, NO+) ions and electrons and two-dimensional distribution of electric field potential, both of a dynamo and magnetospheric origin. All model equations are solved by the finite difference method. The Earth's magnetic field is approximated by the tilted dipole. Thus discrepancy of geographical and geomagnetic axes is taken into account. The calculation of electric fields of dynamo and magnetospheric origin has been recently modified by Klimenko et al. [2006, 2007]. Thermospheric and ionospheric parameters are calculated in the altitude range from 80 km up to 15 Earth's radii with 1–2 min temporal resolution and spatial resolution 5° in latitude and 15° in longitude.
 The simulation uses as initial condition the values of ionospheric parameters at 0600 UT on 9 September 2005, obtained for quiet geomagnetic conditions (Kp = 0.7, AE = 1). To calculate the quiet time behavior of ionospheric parameters, we varied only the F10.7 index, while the values of Kp and AE index remained fixed at the level of 0.7 and 1, respectively. For quiet conditions, the cross-polar cap potential difference ΔΦ was set at geomagnetic latitudes ±75°, and R2 FAC, j2, at geomagnetic latitudes ±70°. The cross-polar cap potential difference was set according to Feshchenko and Maltsev  (see Table 2). The changes of the polar cap sizes were not taken into account.
Field-aligned currents of the second region (R2FAC).
Time delay of R2 FAC variations with respect to the variations of cross-polar cap potential difference.
Quiet conditions and recovery phase: 0000 UT 9 Sep to 1401 UT 9 Sep; 1800 UT 9 Sep to 0600 UT 10 Sep; 2000 UT 10 Sep to 0114 UT 11 Sep; 1100 UT 11 Sep to 2400 UT 14 Sep
38 + 0.089 × AE
3 × 10−9 + 6 × 10−12 × AE
SSC phase: 1401 UT 9 Sep to 1600 UT 9 Sep; 0600 UT 10 Sep to 1300 UT 10 Sep; 0114 UT 11 Sep to 0500 UT 11 Sep
38 + 0.089 × AE
3 × 10−9 + 1.5 × 10−11 × AE
Main phase: 1600 UT 9 Sep to 1800 UT 9 Sep; 1300 UT 10 Sep to 2000 UT 10 Sep; 0500 UT 11 Sep to 1100 UT 11 Sep
38 + 0.089 × AE
3 × 10−9 + 3.6 × 10−11 × AE
 Studies of the R2 FAC behavior during geomagnetic disturbances have demonstrated that the R2 FAC amplitude varies in different ways depending on the storm phase [Iijima and Potemra, 1976; Snekvik et al., 2007; Cheng et al., 2008; Kikuchi et al., 2008]. We split the considered period into intervals with conditional names (see Table 2). Using experimental results of Snekvik et al.  and Cheng et al. , we have constructed the empirical dependences of R2 FAC amplitudes from AE index at different phases of geomagnetic storm (see Table 2). We also have included the 30 min time delay of R2 FAC variations with respect to the variations of cross-polar cap potential difference during the SSC phase [Kikuchi et al., 2008] (see Table 2). In addition, according to Sojka et al.  we varied the position of R2 FAC maximum depending on changes of a cross-polar cap potential difference such as it is shown in Table 3. Figure 2 shows the resulting cross-polar cap potential difference, and amplitude and latitude of R2 FAC used in the model.
Table 3. Geomagnetic Latitude of R2 FAC
GMLat of R2 FAC
ΔΦ ≤ 40 kV
40 kV < ΔΦ ≤ 50 kV
50 kV < ΔΦ ≤ 88.5 kV
88.5 kV < ΔΦ ≤ 127 kV
127 kV < ΔΦ ≤ 165.4 kV
165.4 kV < ΔΦ ≤ 200 kV
ΔΦ > 200 kV
 Current simulation also uses the empirical model by Zhang and Paxton  for high-energy particle precipitation. In this model, the energy and energy flux of precipitating electrons depend on a 3 h Kp index, instead of functional dependences based on the morphological representations in earlier GSM TIP studies. The values of energy and energy flux, calculated in GSM TIP model according to this empirical model, are shown in Figure 3 for various Kp indices.
3. Calculated Results
 To study the ionospheric effects of geomagnetic storms of 9–14 September 2005 period, we used the modeling results obtained by GSM TIP model, the experimental data of ISRs at Millstone Hill and Arecibo, and ionosondes at Jicamarca, Palmas, and San Jose Campos (see Table 4). In observational data, we selected 8 September 2005 to demonstrate quiet time behavior.
Table 4. Ionospheric Stations and Their Coordinates
Millstone Hill, USA
Arecibo, Puerto Rico
Palmas (PAL), Brazil
San Jose Campos (SJC), Brazil
3.1. The Ionospheric Effects Above Arecibo and Millstone Hill
 To show the improvements in the updated version of the model, we have compared the storm time ionospheric effects above Arecibo on 10 September, obtained in the earlier version of the model [Klimenko et al., 2009, 2011], and in a modified model. Figure 4 shows the calculated results of the NmF2, hmF2 and the meridional component of thermospheric wind velocity at the altitude of 380 km, together with the observational data above Arecibo. It is obvious that the new simulation results are in a much better agreement with observations than the earlier version. We interpret it as a cumulative result of more accurate description of such inputs as cross-polar cap potential difference, R2 FAC, and auroral electron precipitation [Zhang and Paxton, 2008]. They lead to a more realistic calculation of Joule heating in the auroral zone due to the energy dissipation during storm time and, as consequence, to more precise calculation of the disturbances in thermospheric parameters and, in particular, in the meridional component of the thermospheric wind. As was discussed by Lu et al.  and Klimenko et al. , the positive disturbances in NmF2 and hmF2 on 10 September are caused by the meridional component of thermospheric wind velocity, shown in Figure 3. While in the earlier version of the calculation the disturbances of meridional component of thermospheric wind velocity did not exceed 20 m/s, in the new version they reach 150 m/s. The effects in NmF2 and hmF2 also increase in appropriate ways.
 The new calculation of NmF2 indicates the formation of small positive disturbance at 1100 LT, the negative disturbance at 1400 LT, the large positive disturbance with a maximum at 1700 LT, and the negative disturbance at 1900 LT. All these disturbances are in a good agreement with observations and have only minor differences. According to the observational data, at 1100 LT more appreciable positive effect is formed, at 1400 LT the negative disturbance is not formed, and instead of it the reduction of positive disturbance is observed with a minimum at this time. All these ionospheric disturbances are the effects of a geomagnetic storm on 10 September. It is necessary to note that the inclusion of solar flares in the model leads to a greater positive effect in NmF2 in the afternoon, which is consistent with the observational data.
 The positive and negative disturbances in NmF2 manifest in hmF2 1 h earlier. This behavior is related to the formation mechanism of these disturbances. As plasma at heights of F region and higher is magnetized, it can move across geomagnetic field lines only under the action of an electric field, effects of which in this case are insignificant. Because of the inclination of geomagnetic field to the Earth's surface, meridional component of thermospheric wind can move plasma due to ion-neutral collisions, transporting it upward (downward) along geomagnetic field lines into the regions of smaller (larger) chemical loss rates and leading to the growth (decrease) in electron density [Brjunelli and Namgaladze, 1988]. Thus, in the afternoon the additional equatorward wind will lead to the upward plasma transport, i.e., to increase in hmF2 without time delay. The upward plasma transport means the plasma is moving into the region of smaller rates of chemical losses. Adjustment of plasma density to lower rates of chemical losses requires additional time and determines the time delay of effects in NmF2 relative to the effects in hmF2. Exception from this behavior is the positive disturbance in hmF2 during the period 0300–0500 LT, which does not appear in NmF2. It is explained by the competing action of the meridional component of thermosphere wind velocity, directed to the equator, and the eastward component of the electric field, leading to the positive effects in NmF2 as a result of plasma transport to the larger heights, and changes in the neutral atmosphere composition, leading to the negative effects in NmF2 owing to the reduction in the n(O)/n(N2) ratio during geomagnetic disturbances. The neutral atmosphere composition changes affect the behavior of electron density in the ionosphere F region as follows. As the atomic oxygen is the basic source of ionization at heights of the ionospheric F region, and the molecular nitrogen is the basic source of recombination, the growth (reduction) of the ratio n(O)/n(N2) leads to the positive (negative) disturbances in electron density at heights of the ionospheric F region.
Figure 5 presents a comparison of calculated results obtained by the updated version of the model with the observational data of the F2 layer parameters above Millstone Hill. In general, the model well reproduces the ionospheric disturbances in NmF2 observed above Millstone Hill. The exception is noted on 14 September when instead of the observed small positive disturbances, the model predicts a negative disturbance. According to calculations, negative disturbances above Millstone Hill are formed throughout all the considered period in the afternoon and in the evening, and the positive disturbances are formed in the afternoon on 9 and 10 September. All these positive and negative disturbances were observed by the Millstone Hill ISR. The positive disturbance in NmF2 on 10 September, described by Goncharenko et al. , was much stronger than on 9 September.
 As for the behavior of hmF2 above Millstone Hill, the model calculations and the observational data are in a good agreement both in disturbances and in absolute values. Large increases in hmF2 (∼200 km) in presunrise hours from 11 September until 14 September present special interest and will be discussed in a future work. It is necessary to note that the increases in hmF2 in presunrise hours also become apparent in calculated results on 9 and 10 September, but their values are smaller. The increase in height of the F2 layer maximum on 13 September is in a good agreement with observational data. We do not discuss increases in hmF2 on other days due to the absence of observational data.
 In this study we did not consider the formation mechanisms of disturbances in NmF2 and hmF2 above Millstone Hill. As it was noted by Lu et al.  and Klimenko et al. , the positive disturbances in NmF2 and hmF2 in the afternoon on 10 September are caused by the meridional component of thermosphere wind velocity. This conclusion is confirmed by a new simulation. The hmF2 growth in the afternoon on 10 September testifies to it. Thus, the new simulation adequately reproduces the ionospheric effects of the considered geomagnetic storms at middle latitudes, and improves the data/model agreement in comparison with an earlier simulation.
3.2. Effects in Brazil Region
Figure 6 shows the comparison of the F region maximum parameters obtained from the simulation and from observational data above two Brazilian ionospheric stations in Palmas (PAL) and S.J. Campos (SJC). While the PAL station is in close proximity to the geomagnetic equator, the SJC station is in the region of the equatorial anomaly crest on the southern hemisphere. The location of the stations explains many features of their differing ionospheric F region behavior. Although the qualitative agreement between foF2 simulations and observations is satisfactory, the absolute values of these parameters obtained in model calculations are smaller than were observed above the Brazilian stations. The reason for this is the overestimated neutral atmosphere density in the GSM TIP model that leads to the overestimated rates of ionization losses in the ionospheric F region. It in turn leads to the underestimated values of electron density. However, the goal of this paper is the research of the ionospheric disturbances related to the sequence of geomagnetic storms, instead of the absolute values of ionospheric parameters. Once again we shall note that the disturbances in critical frequencies of the F region obtained in calculated results and from observational data are very similar.
 Above stations PAL and SJC the simulation results and the observational data foF and foF2, accordingly, reveal positive disturbances in the afternoon during storms on 10–14 September, and the negative disturbance in postsunset hours on 10 September and at night on 11 September. The model calculated results and observations above station PAL reveal differences during the transition from 11 to 12 September and from 12 to 13 September. If observations show the positive disturbances in foF, the model results do not reveal any effect. In contrast, both simulations results and observations of foF2 above station SJC do not differ during transitions from 11 to 12 September and from 12 to 13 September. Both observations and model calculations show small negative disturbances in foF2 at this time.
 Comparison between the hmF behavior calculated in the model with the observational data above station PAL shows good agreement in disturbances and in the absolute values of hmF and hpF2. All sharp increases of hmF, both in quiet and in the storm time conditions, are coupled, as will be shown later, with the appearance of the F3 layer formed in the vicinity of geomagnetic equator at heights larger than F2 layer maximum. When the critical frequency of the F3 layer, foF3, becomes greater than foF2, a jump in the F layer maximum height occurs. Sharp decreases in hmF are coupled with the transition from the case when foF3 > foF2 to the case when foF3 < foF2. It is important to note the formation of the F3 layer in our calculated results, both in quiet conditions and during storm time. During disturbed time, the time periods when foF3 > foF2 occur more frequently and last longer than during quiet conditions. Experimental data demonstrates sharp increases and decreases of hpF2 approximately at the same time as modeling results.
 The hmF2 behavior above station SJC, calculated in the model, is in a good agreement with the observational data, both in absolute values and in variations during the disturbed times. However, not all sharp increases of hmF2 are related to the formation of the F3 layer, but to the increase of the F2 layer maximum height under the action of meridional component of thermospheric wind and eastward electric field in the equatorial ionization anomaly crest on the southern hemisphere, where the SJC station is located. It is important to note very good agreement of hmF2 calculated results with the observational data of h′F at nighttime on 10 to 11 September and on 12 to 13 September.
 To illustrate these statements, we show in Figure 7 the GSM TIP model vertical profiles of electron density above these stations for selected time periods on 11 September. The F3 layer is formed during this storm time period above the station PAL at height of ∼400 km, with electron density during the period of 1400–1530 UT exceeding NmF2. During the quiet conditions and the same local time, the F3 layer above station PAL is not formed. Balan et al.  noted that the F3 layer is formed above the Brazilian stations in these local hours and at the same heights. During the same period the F3 layer above station SJC is not formed during storm time or in quiet conditions. As discussed by Balan et al. [1997, 1998], the F3 layer is formed only in the vicinity of geomagnetic equator in a band of geomagnetic latitudes from −10° up to 10°. However, Fagundes et al.  observed, and Lin et al. [2009b] predicted the formation of additional layers in the region of equatorial ionization anomaly crests. They connected the formation of such additional layers either with internal gravity waves of tropospheric origin [Fagundes et al., 2007], which were not taken into account in our model, or with thermospheric wind surge during storm time [Lin et al., 2009b], which is reproduced in our model, but does not lead to the formation of additional layers in the region of equatorial anomaly crests in the Brazilian region. In this paper we do not investigate the details of the formation mechanisms of additional layers above stations in the Brazilian region. Such research is planned for the near future. In this paper we have considered the stations in the Brazilian region to demonstrate the good agreement for ionospheric disturbances at the low latitudes between the modeling results and observational data.
3.3. Effects Above Jicamarca
 In this section, we will consider in more detail the ionospheric effects of geomagnetic storms at the magnetic equator using simulations and observations above Jicamarca station. Figure 8 presents the F layer maximum parameters, zonal and meridional components of electric field, meridional component of thermospheric wind velocity and neutral atmosphere composition at height of 350 km above Jicamarca station on 9–14 September. In Figure 8, the calculated results are also compared to the observational data foF and hmF. The absolute foF values obtained in the model considerably differ from the observational data. The reason for this is, as was discussed above, the overestimated neutral atmosphere density in the GSM TIP model. However, the disturbances in foF, caused by the geomagnetic storm sequence, are very similar in calculated results and observational data. We also note the good qualitative and quantitative agreement of absolute values and effects in hmF.
 The electric field of magnetospheric origin and the disturbed dynamo electric field, together with variations of meridional component of thermospheric wind velocity and neutral atmosphere composition during geomagnetic storms are the basic formation mechanisms of the ionospheric disturbances at the lower and equatorial latitudes. Thus, the additional zonal electric field causes the vertical E × B drift of magnetized thermal plasma at heights of ionospheric F region. The plasma is transported to the region of different loss rates in ion-molecular reactions that leads to the disturbances in foF. However, at the geomagnetic equator the role of the diffusion along geomagnetic field lines becomes more important for the vertical distribution of thermal plasma. Therefore, near the geomagnetic equator the diffusion complicates the interpretation of zonal electric field effects. It is important to note that effects of zonal electric field in our calculated results are clearly seen in hmF disturbances. The hmF decreases with the appearance of additional westward electric field, and hmF increases with the appearance of additional eastward electric field. The meridional electric field in the Earth's ionosphere causes E × B drift of magnetized thermal plasma in longitudinal direction. It leads to the transport of features in latitude-altitude plasma distribution from one longitude to another. The additional thermospheric wind along magnetic meridian, as was discussed above, due to ion-neutral friction leads to the upward or downward transport of plasma along geomagnetic field lines at some distance from the geomagnetic equator. At the equator and in its vicinity, the meridional component of thermospheric wind leads to the transequatorial plasma transport from one hemisphere into the other. The storm time variations in neutral atmosphere composition as described above in the section 3.1, affect the behavior of electron density in the ionospheric F region.
 As shown in Figure 8, modeling results for Jicamarca predict an additional thermospheric wind directed either from the northern hemisphere into the southern (mostly in daytime hours), or in the opposite direction (mostly at night). At the beginning of the considered geomagnetic storm sequence the model predicts only the negative disturbances in n(O)/n(N2) ratio. Positive disturbances are starting to appear on 12 September, and on 14 September the negative disturbances disappear in general. The positive disturbances in n(O)/n(N2) ratio are the main reason for positive disturbances in foF, which are formed in the afternoon from 12 September till 14 September.
 The only essential discrepancy between the calculated results and observations above Jicamarca during the geomagnetic storm sequence is the formation of negative effects in foF on 11 September from 0100 till 1100 UT in calculated results, whereas the observational data specify positive disturbances. During this period, spread F was registered by ionosondes (see Figure 9). Spread F is the phenomenon when the trace of reflection in ionograms is degraded. This phenomenon is connected either with scattering of radio waves on small-scale irregularities, or with reflection from mesoscale irregularities [Abdu, 2001]. In both cases spread F either distorts the information about Ne profile, or in general makes the reception of this information impossible. During the same time the calculated results specify the formation of at least two additional layers in the ionosphere above Jicamarca, where the electron density in the maxima of these layers is greater than in underlying layers. The sharp rises in F layer maximum height in calculated results indicate that. The sharp rises of the F layer maximum height are formed during the same time in the observational data too. It is important to note that simultaneous occurrence of spread F in observational data and additional layers in model calculations do not contradict each other, as spread F and the formation of additional layers are related to the action of the same mechanism, E × B drift [Surotkin et al., 1985; Balan et al., 1998; Abdu, 2001].
 We will consider in more detail the effects in the geomagnetic storm on 11 September, when several sharp rises and decreases in hmF are observed in the calculated results and in observational data. To demonstrate the formation of an additional layer in the calculated results, we show the vertical profiles of electron density above Jicamarca at different UT epochs on 11 September (Figure 10). According to the model, during the period from 0000 till 0300 UT, the F3 layer appears and the F2 layer disappears, at that foF3 > foF2 from 0100 till 0300 UT. It also leads to the sharp increase of hmF above Jicamarca at this time. The vertical profiles of electron density at 0200, 0230 and 0300 UT show the E layer, the F1 layer, the disappearing F2 layer as a bend below the formed F3 layer, and the additional layer in an external ionosphere at heights of ∼1700 km, formed by H+ ions. To be consistent with the well-established nomenclature of using capital letters to denote different regions of the ionosphere, e.g., D, E, and F, the additional layer in an external ionosphere has been named the G layer [Klimenko and Klimenko, 2011]. The G layer occurrence in quiet geomagnetic conditions was predicted in model simulations by Klimenko and Klimenko . From the electron density profile at 0630 UT we see that during the geomagnetic storm at this time, the normal F2 layer is absent, but the E layer, F1 layer, and also F3 and F4 layers have formed resulting in the stratification of the equatorial F layer with significant plasma tube depletion. The H+ ion density sharply decreases, and the transition height from ions O+ to ions H+ considerably grows. The jump in hmF up to the height of the F4 layer also occurs at this time, as the electron density in the maximum of the F4 layer is greater than the electron density in the maximum of the F3 layer. Observational data during the period from 0400 UT till 1100 UT show the formation of spread F (Figure 9). At 1506 UT in a vertical profile of electron density, an E layer, F1 layer and F2 layer below the additional F3 layer are visible. The formation of the F3 layer at this time is also observed in experimental data and can be seen in the ionogram (Figure 9) and in vertical profiles of plasma frequencies and electron density above Jicamarca at 1545 UT, shown in Figure 11. Ionogram data have been manually scaled using an interactive ionogram scaling software, SAO Explorer and the Lowell DIDBase [Reinisch et al., 2004; Khmyrov et al., 2008].
 The appearance of the additional layers on 11 September above Jicamarca in modeling results and spread F in the observational data can be explained by the formation of strong electric fields on 10 and 11 September, obtained in model calculations. How do the additional layers form in the calculated results? Figure 12 shows the calculated latitudinal profiles of zonal component of electric field at geomagnetic longitude of Jicamarca at the different UT epochs of the quiet and disturbed conditions. Figure 12 also presents the calculated vertical profiles of zonal component of electric field above Jicamarca, obtained from latitudinal profiles shown above.
 The zonal electric field above Jicamarca is nonuniform in height and at 0000 UT directed westward with a maximum in vertical profile at a height of ∼300 km, that leads to the nonuniform plasma drift downward. As a result, the region of the lowered plasma density is formed at the height of a maximum of the westward electric field, and two maxima are formed in a vertical profile of electron density at the geomagnetic equator: the bottom maximum, the F2 layer, and top maximum, the F3 layer. The resulting irregularity in vertical distribution of electron density is relatively small. However, it becomes stronger due to the prolonged presence of westward electric field with maximum at the height of 175 km, as the F2 layer maximum descends downward faster than the F3 layer maximum. The changes of chemical loss rates with height also become important. These changes affect the F2 layer maximum, which is located at lower altitudes and descends into the region of larger chemical loss rates faster than the F3 layer maximum. As the result of this, the plasma density in the F2 layer maximum decreases much faster and becomes eventually considerably reduced in comparison with the density in the F3 layer maximum. The nonuniform in latitude and altitude zonal electric field turns eastward at 0200 UT, has a minimum at the geomagnetic equator, and grows with height above Jicamarca location. This eastward electric field causes nonuniform in height upward plasma drift, with minimal velocity at the height of 175 km and larger velocity at higher altitudes. As a result of this plasma drift, the F3 layer maximum is uplifted to higher altitudes, where recombination rates are slower. The F2 layer maximum is less affected, as it is located at lower heights where chemical loss rates are larger, and upward drift velocity is smaller. Therefore, the plasma density in the F2 layer maximum continues to decrease. It is also important to note, that at the same time a reduction of the n(O)/n(N2) ratio takes place above Jicamarca (Figure 8), which affects the F2 layer maximum more strongly than the F3 layer maximum. Eventually the F2 layer practically disappears. The processes of disappearance of the F2 layer and the formation of the F3 layer at this time are well seen in Figure 10.
 With the progress of time, the vertical distribution of the zonal electric field at the geomagnetic equator becomes even more nonuniform. At 0600 and 0700 UT, as it shown in Figure 12, the electric field is directed to the West at heights of ∼500–600 km and changes sign at larger and smaller heights, forming a local maximum of eastward electric field at the geomagnetic equator at the height of 175 km. Such an electric field causes the plasma drift to be nonuniform in height at the geomagnetic equator, directed downward at heights of ∼500–600 km, and upward at larger and smaller heights. This drift results in: the increase in plasma density in the F3 layer maximum because of plasma transport into this region from larger and smaller heights; the decrease of electron density at heights of ∼500–600 km because of joint action of upward plasma drift under the action of eastward electric field at higher altitudes and downward plasma drift under the action of westward electric field at these heights. Such nonuniform change of electron density leads to the formation of the F4 layer at the height of ∼700 km, which is seen in Figure 10 at 0630 UT. It is also seen that by this time the F2 layer is absent.
 With the progress of time, the zonal electric field at the geomagnetic equator becomes completely directed westward, and its amplitude grows with height. This field causes nonuniform plasma drift downward, that leads to the plasma density decrease at the geomagnetic equator due to plasma sinking to the region of larger chemical loss rates (0900 UT). It is necessary to remind that at the same time there is a strong decrease in n(O)/n(N2) ratio.
 After the onset of solar ionization at the geomagnetic equator, the growth of electron density in the bottom part of F region exceeds the growth at larger heights. Once again the F2 layer appears, electron density in this layer quickly increases, with time the F3 layer absorbs the F4 layer (at ∼1100 UT), and then F3 layer at ∼1200 UT itself is absorbed by the F2 layer. Thus, there is a disappearance of additional layers in the equatorial ionosphere. However, the strong eastward nonuniform in height electric field appears at 1300 UT at the geomagnetic equator, with a minimum at the height of ∼300 km. It again leads to the occurrence of the F3 layer exceeding the F2 layer maximum above Jicamarca location, as seen in electron density profile at 1506 UT in Figure 10. With decrease of amplitude of eastward electric field as seen from Figure 12 at 1600 UT, the F3 layer disappears.
 Based on the above considerations, we conclude that the basic formation mechanism of additional layers in the F region of the equatorial ionosphere during geomagnetic storms is produced by a nonuniformity in height of the zonal electric field in the vicinity of the geomagnetic equator, which leads to nonuniformity in height of the E × B plasma drift.
Surotkin et al.  and Balan et al.  discussed the formation mechanisms of stratifications (bifurcations) of the equatorial F2 layer and an additional F3 layer during quiet geomagnetic conditions. They demonstrated that, in general, stratifications (bifurcations) of the equatorial F2 layer and an additional F3 layer are formed when zonal component of electric field (e.g., meridional component of E × B plasma drift) is included. Surotkin et al.  noted that the possible factors assisting the occurrence of stratifications in the morning are (1) displacement of maximum of upward E × B drift related to the action of zonal electric field in the morning sector; (2) the presence in the afternoon sector of the meridional electric field directed upward at the equator, causing zonal plasma drift from the dayside to dawnside; and (3) the phase delay of daily variation of zonal electric field at the geomagnetic equator with regard to the variation of electric field at nearby latitudes. The zonal plasma drift was found to be the most essential process in the formation of stratifications in the morning. The mechanism of the F3 layer formation, suggested by Balan et al. , looks as follows. Early in the morning there is an usual F2 layer. With the progress of time it becomes wider due to the effect of photoionization and unique dynamic effects in the equatorial region. Due to the dominance of E × B drift at the geomagnetic equator, the peak of the layer at the equator moves up faster than at other latitudes. While being transported upward, this maximum passes through the region in which chemical and dynamic processes are equally important and gets into the region where dynamic processes dominate. After some time and below this peak, a new maximum is formed in the region of balance of chemical and dynamic processes. The top maximum becomes a maximum of the F3 layer which after some time disappears due to chemical losses and diffusion, which dominate over ionization processes at these heights. The Balan et al.  study concludes that the F3 layer is formed in the morning and daytime sector, when the ionization processes dominate over chemical losses due to upward plasma transport under the action of E × B drift and neutral wind.
 Further studies of formation mechanisms of additional layers in the equatorial ionosphere have been performed during strong geomagnetic disturbances [Balan et al., 2008; Lin et al., 2009a]. These studies indicate that the formation mechanism of an additional layer in the disturbed conditions is the same as in quiet conditions, but this additional layer becomes more prominent and more distinct during geomagnetic disturbances.
 Our research shows that the basic formation mechanism of the F3 layer is nonuniform in height distribution of zonal component of an electric field. It causes the nonuniform vertical plasma drift that leads to plasma convergence at some altitudes and to its rarefaction at others. The formation mechanism of the F4 layer is the same, but this layer is formed at the larger heights. It is important to note that the mechanism of stratification of the equatorial F2 layer and the formations mechanism of the F3 layer suggested by Surotkin et al. , Balan et al. [1998, 2008], and Lin et al. [2009a] also take place.
 At the end of our research, we will consider in more detail the observational data and calculated results on 11 September above Jicamarca. Before proceeding to the description of experimental data, we shall remind the terminology used in the description. The F layer is considered to have single layer if electron density Ne in the ionospheric F region monotonously grows from some value up to the main maximum, and thus its height gradient dNe/dh monotonously falls from some value up to zero at a point of the main maximum. The F layer is considered multilayered if in the ionospheric F region the Ne behavior reveals even one of two properties.
 1. A local maximum is observed in the vertical profile of Ne (dNe/dh = 0). In this case, a typical breakup is registered in the ionogram.
 2. The local maximum in vertical Ne profile is absent, however on an ionogram just needs a short variation in electron density with height to produce an additional ionogram cusp.
Figure 13 presents the behavior of the F layer maximum parameters above Jicamarca during the storm on 11 September 2005 in Universal Time, as obtained from Digisonde data and from model calculation, and our interpretation of observational data after the analysis of model calculated results and observational data. The analysis of Figure 13 is presented in Table 5.
Table 5. The Vertical Structure of F Region Above Jicamarca
Time Interval (UT)
Model Calculated Results
The distinct single-layered pattern: F2 layer
A three-layered pattern: F1, F2, F3 layers; foF1 < 1 MHz; foF3 arises at 0100 UT
F2 or F3 layer
Spread F, but the distinct traces show a single-layered pattern: F2 or F3 layer
The spread F at frequencies above the critical value. It does not distort the Ne(h) profile information. The distinct traces show the one-layered pattern: F2 layer
The two-layered pattern: F1, F2 layers
A distinct two-layered pattern: F1, F2 layers
A three-layered pattern: F1, F2, F3 layers
A three-layered pattern: F1, F2, F3 layers
A distinct two-layered pattern: F1, F2 layers
A two-layered pattern: F2, F3 layers; due to the absorption the F1 layer is not visible
A two-layered pattern: F1, F2 layers
A distinct two-layered pattern: F1, F2 layers
A distinct three-layered pattern: F1, F2, F3 layers
A two-layered pattern: F1, F2 layers. The model does not describe the appearance of F3 layer.
A distinct two-layered pattern: F1, F2 layers
A two-layered pattern: F1, F2 layers
A distinct single-layered pattern: F2 layer; due to the spread F the F1 layer is not visible
 As was shown above in model calculations, the nonuniform in height westward electric field is formed at the geomagnetic equator (0000 UT), with significant magnitude in the bottom part of F region and smaller magnitude at higher altitudes. It causes nonuniform in height downward plasma drift at nighttime. This drift leads to the reduction of electron density in the F2 layer maximum and its eventual full disappearance, and to the formation of a new maximum in the top part of vertical profile (at higher altitudes than maximum existing at this time in quiet conditions). This layer cannot be considered as the F2 layer, as in the morning, when the photo ionization becomes effective, the missed F2 layer appears again. First, it appears as a bend in the bottom part of the profile, with the density in the bend increasing rapidly and much faster than in the maximum above it. After some time, the bend becomes a local maximum, and later it becomes the main maximum as it absorbs the maximum above it. Thus, we make the following conclusion. The presence of a single-layer in the nighttime equatorial ionosphere, at least during geomagnetic disturbances, does not mean the existence of the F2 layer, as at this time the F2 layer can be absent, and instead of it we can observe the additional layer formed in the top part of the F2 region. To identify this layer, one must carefully trace the temporal behavior of the existing layers. This will clearly define whether we are dealing with the F2 layer, or observe an additional layer in the absence of the F2 layer.
 As it is visible from Figure 13 and Table 5 the F1 and F2 layers sometimes were not observed in ionograms (if they really existed) because of low foF1 or/and foF2 values, which are less than the threshold sensitivity of the ionosonde. With regard to the main peak, we cannot identify it as the F2 layer. From the results of our calculations, we suggest that this is sometimes an F3 layer or even an F4 layer. In other cases, the main peak is an F2 layer.
 Multilayer pattern in vertical profile in our calculations and spread F in observations are explained by large disturbances of electric field above Jicamarca station (Figure 8), obtained in our calculations. In our opinion the observation interval of the F3 layer should be increased according to the results of model calculations. However, available observational data does not permit unambiguous determination due to the presence of spread F. The observations indicate the stratification of the equatorial F2 layer during 1415–1545 UT and 1845–1900 UT. During the period 1415–1545 UT, the stratification of the F2 layer also is presented in the calculated results.
 This study demonstrated the following.
 1. The combined use of several updates in the GSM TIP can significantly improve the agreement between the results of calculations of various ionospheric parameters with observational data at the middle and low latitudes. These updates are (1) AE index with 1 min temporal resolution as an independent variable instead of the 3 h Kp index for modeling the time dependence of cross-polar cap potential difference; (2) the new empirical model of high-energy particle precipitation, depending on the Kp index; (3) description of the R2 FAC according to the currently available experimental data and theoretical concepts; and (4) inclusion in the model the effects of solar flares.
 2. Geomagnetic storms affect the formation, existence, lifetime, and the number of additional layers in the equatorial ionosphere.
 3. During geomagnetic storms, a zonal electric field is generated which is nonuniform in height at the geomagnetic equator. This electric field forms the additional peaks in the upper part of the electron density profile (above the maximum that existed at that time in quiet conditions) through the nonuniform in height vertical plasma drift.
 We acknowledge the Arecibo and Millstone Hill ISR teams and Jicamarca, Palmas, San Jose Campos Digisonde teams for processing the data and making the experimental data available. The authors acknowledge the developers of the Lowell DIDBase project for access to the ionosonde database. Authors thank Iurii V. Cherniak of WD IZMIRAN for useful discussions. These investigations were carried out at financial support of Russian Foundation for Basic Research, grant 08-05-00274.