Ionospheric imaging with GPS provides a near-global view of the three-dimensional time-evolving ionosphere. This is of particular interest during storms. The focus of this paper is on the height redistribution of the plasma and in particular the longitudinal and latitudinal variations in the time of plasma uplifts. Three storms, 15 July 2000, 30 October 2003 and 20 November 2003, are studied here. Dramatic elevation of the F layer by more than one hundred kilometers was seen in the images during daytime over Europe and the USA for all three storms. All three showed an east-west time delay of around one hour in the peak-height elevation over some 85° longitude. The 20 November 2003 storm also showed a north-south time delay in the change in the F-region height with the uplift seen first at high latitudes and then low latitudes. Independent evidence from other instruments and techniques are provided as supporting evidence that the peak-height uplifts occurred. Candidate mechanisms of the peak height changes are electric fields and neutral winds and the roles of these drivers will be investigated in future modelling studies.
 During geomagnetic storms the ionosphere may be greatly changed in terms of the Total Electron Content (TEC), electron densities, the peak density and peak height of the F region. Traditionally, the response of the ionosphere to storms in terms of the peak density (NmF2) has been studied using ionosonde data [for example, see Fuller-Rowell et al., 1996; Szuszczewicz et al., 1998]. The variations of the F-layer height at storm time can be investigated with ionosondes and Incoherent Scatter Radar (ISR) data. However, it is known that ionospheric storms have a global impact on the ionisation and under very disturbed conditions the ionospheric response to severe storms often presents significant changes in the distribution of ionisation with latitude and altitudes. Unfortunately, ionosondes cannot measure the topside ionosphere and sometimes suffer from absorption during storms, while ISRs have geographical limitations. Here, we monitor the ionospheric height variations over large areas (Europe and North America) using Global Positioning System (GPS) data with a 4-D tomographic algorithm – Multi-instrument Data Analysis System (MIDAS) [Mitchell and Spencer, 2003]. This has the advantage of wide area, continuous ionospheric monitoring throughout entire storm events and provides a complementary technique to other localised ionospheric instrumentation.
 Recent work [Yin et al., 2004] has demonstrated that GPS imaging with tomography can be used to study storm-time disturbances in the ionosphere. In particular, changes in the F-layer height in the GPS imaging were verified in a co-located region with the Millstone Hill, Massachusetts, ISR for the 15 July 2000 storm. In this paper, ground-based GPS dual-frequency observations are used to produce electron density distributions with that same inversion technique–MIDAS. GPS data have been processed and reconstructed to image the storm-time ionosphere over the USA and Europe for three storm events–the 30 October, 20 November 2003 and 15 July 2000 storms. To understand effects of underlying assumptions in the inversion algorithm on results, electron density images were compared to those using another inversion method–Ionospheric Data Assimilation Three Dimensional (IDA3D) [Bust et al., 2004]. This was run with the ASPEN Thermosphere Ionosphere Mesosphere Electrodynamics General Circulation Model (TIMEGCM) [Crowley et al., 1999] as the background and is used here for an example comparison. This paper is concerned with changes in the layer heights and peak densities over the USA and Europe. This study will focus on the period between noon and dusk.
 Ground-based GPS data from a network of nearly 200 GPS receivers (Scripps Orbit and Permanent Array Center, SOPAC) across northern America and Europe were used in the MIDAS reconstruction algorithm to produce images of the electron density. MIDAS was developed to image the ionosphere in a 3-D spatial geometry (latitude/longitude/altitude) with time evolution. The GPS dual-frequency phase observations were first calibrated to derive slant TECs as inputs to the algorithm. The data are used in windows of typically one hour. The ionosphere in a given region is represented as a three-dimensional array of voxels. In this study the voxels are defined at an altitude range from 80 km up to 1080 km, with a size of 1° in latitude, 4° in longitude and 50 km in altitude. Mitchell and Spencer  have given a detailed description about the use of MIDAS for imaging the ionosphere. In this case study, some specific (storm-time) a priori information of the background ionosphere is involved in the MIDAS algorithm when the disturbed ionosphere is investigated. In particular a greater flexibility in vertical profiles of electron density is allowed than would normally be the case for this imaging technique.
 To date, the MIDAS algorithm has been successfully applied to investigate the spatial extent and temporal evolution of electron density altitude/latitude structure and TEC longitude/latitude structure at middle and subauroral latitudes as a function of time and level of geomagnetic activity. Yin et al.  have applied this technique to show variations of the electron concentration in the F2 layer of the ionosphere for the July 2000 major storm. It is interesting to note that vertical changes in the F2 layer, such as the elevation of hmF2, can be reconstructed very well only using ground-based GPS data when compared with the Millstone Hill ISR data. Moreover, the uplifts of the F2 layer are also observed in several major storms, such as the October and November 2003 storms in this study, using the same technique. The IDA3D algorithm is used here to provide a comparison to MIDAS results. This is to help evaluate the effect of the algorithms and their inbuilt assumptions on the images. IDA3D is an ionospheric objective analysis algorithm using a three-dimensional variational data assimilation technique (3DVAR). In this study, IDA3D used the ASPEN TIMEGCM [Crowley et al., 1999] as the background to produce electron-density distributions. GPS data as well as data from other sensors are used in the IDA3D. A detailed description of IDA3D is given by Bust et al. .
3. Results and Discussions
 For the November 2003 storm, images from MIDAS showing electron-density changes with height and geographic latitude at 15°E, 70°W and 120°W longitudes are presented in Figures 1– 3. Figure 4 was produced using the IDA3D algorithms, which ingest more ionospheric data than MIDAS and use a background ionosphere that will have an influence on the vertical distribution of electron density (in this case the ASPEN model was used). Comparing to Figure 2 this confirms the uplift in peak height.
Figure 1 shows MIDAS images of electron density at a longitude over Europe of 15°E for each hour between 15 and 17 UT on 20 November 2003. A depleted region around 55°N at 15 UT is clearly seen in the 15 UT image with a peak height at about 500 km. An hour later (16 UT), this region with a peak height above 500 km extended southward all the way from 60°N to 40°N. Then around 17 UT the trough moves towards lower latitudes and the F-layer height decreased to 350 km except for the narrow trough region still staying at ∼500 km. The elevation of the F-layer up to 500 km between 17–18 UT was also observed by the ionosonde at Rome, Italy, as shown in Figure 5, but unfortunately after 18 UT, ionosonde hmF2 determination failed.
Figure 2 illustrates images over the USA at a longitude of 70°W between 16–18 UT by MIDAS and Figure 4 presents the image at 17 UT by IDA3D. The later time is shown because the elevation in the layer height occurred later over the USA. At 16 UT the peak height is elevated poleward of the trough centered at ∼45°N. However, from 17 UT the ionosphere from 38°N to 63°N was elevated to 550 km, which is also found in the corresponding IDA3D image in Figure 4. Then an hour later (18 UT), the entire layer lifted to 560 km. It should be noted here that similar rises in the peak height were observed with the Millstone ISR at the same time [Foster et al., 2005]. This provides further supporting evidence that the large uplifts seen in the images here are authentic. Figure 3 presents electron density images at 120°W by MIDAS at 16–18 UT. It can be seen that the F region first lifted to ∼600 km above latitudes of 60°N. Then the elevation propagated towards lower latitudes to 35°N around 17 UT. After that, the whole F layer stayed at around 600 km height.
 This summarises the observations of the 20 November 2003 storm. The next stage is to inter-compare the events of three storms: 15 July 2000, 30 October and 20 November 2003, and the focus is put on the first two events. Comparisons of TEC and hmF2, derived from inverted results by MIDAS, over both Europe and the USA between 1500–2330 UT on three storm days are shown in Figures 6 and 7. Figure 6 illustrates peak-height variations with UT at three different locations at storm time. Figure 7 presents TEC changes with UT at these three locations on three days. The location in Europe is at 15°E and 43°N; the location on the east coast of the USA is located at 70°W and 43°N; and the location on the west coast of the USA is at 120°W and 43°N. The locations are chosen to be well within regions of good data coverage for the imaging. The following discussion will concentrate the July 2000 storm and the October 2003 storm since the detailed description for the November 2003 storm has been presented above.
Figures 6 (left) and 7 (left) illustrate hmF2 and TEC variations at three locations for the July 2000 storm. The ionosphere over Europe at 15°E reached an F-layer peak height of around 480 km at 19 UT on 15 July 2000, while the TEC was decreasing. About an hour later at the east coast of the USA, the hmF2 was lifted almost to 460 km at low/mid latitudes around 20 UT (15 LT) in a TEC slightly enhanced region. Meanwhile, the ionosphere was elevated to ∼700 km along the 120°W longitude meridian with increase in TEC. For the 30th October 2003 storm as shown in Figures 6 (middle) and 7 (middle), the uplifts in the European sector began from 19 UT and reached up to 700 km around 20 UT with decrease in TEC (Figure 7, middle). Along the east coast of the USA from 20 to 21 UT (15 to 16 LT) on 30th, the F2 layer peak height lifted to around 730 km and TEC did not exhibit large variations. In particular, TEC increased a lot on the west coast of the USA as the disturbed ionosphere was lifted to 700 km during the period of 20–21 UT.
 The opposite response of TEC to the F-region uplifts over Europe and the USA may be partly explained by the photo-ionization effects. Taking the 20 November 2003 storm as an example, the hmF2 rose to its peak at ∼16 UT that is the dayside of the USA (10 LT) when the photo-ionization is strong, while 16 UT (17 LT) is on the dusk side in the European sector. Hence, the peak density enhanced in the USA but decreased in Europe around 15 UT. For the 30 October 2003 and 15 July 2000 storms, nearly all of the elevated peak heights took place around late afternoon until dusk, when the solar radiation becomes weaker. Therefore for those cases, there should be other mechanisms to explain TEC enhancements. For example, electric fields could play a role in transporting plasma from lower latitudes. Table 1 gives details of the times of the peak height uplifts in terms of the maximum height attained. It can be seen that all three storms exhibit a longitudinal variation in the uplift, starting in Europe and moving west.
Table 1. Summary of the F Layer Peak Heights and Occurring Times at Three Locations for Three Major Storms
15 July 2000
480 km/1900 UT
450 km/2000 UT
680 km/2000 UT
30 October 2003
690 km/1930 UT
650 km/2030 UT
690 km/2100 UT
20 November 2003
540 km/1600 UT
550 km/1630 UT
580 km/1700 UT
 Images of the storm-time ionosphere have been used to study the altitude redistribution of the electron density over Europe and the USA. Three intense ionospheric storm events have been analyzed here: the October and November 2003 storms and July 2000 storm. By comparison with another different imaging algorithm, with ionosonde data and with reference to ISR observations, the elevation of the F layer in the very disturbed ionosphere is confirmed.
 Although individual storms have their own characteristics, some common features such as the F region height changes in the ionosphere, do exist. In summary, the F region height changes here show the following characteristics:
 1. Dramatic elevation of the F layer of the ionosphere over Europe and the USA.
 2. The F2 layer elevation propagates from high latitudes to lower latitudes for the November 2003 storm but for other two storms the elevation is simultaneous across all latitudes.
 3. All three storms show an east-west time delay in the peak height elevation, first, in the European sector, then one hour later occurring over the USA. For the July 2000 storm, the uplifts seem to have simultaneously occurred over the east coast and the west coast of the USA, but for the other two storms, it is apparent that the uplifts first took place over the east coast of the USA, and then less than half an hour later occurring over the west coast of the USA.
 4. The uplifts in the USA sector are always accompanied by TEC/electron density enhancements, but those in the European sector are accompanied by decreasing electron densities/TEC.
 In this paper, our preliminary results on the vertical movements of the F region in the disturbed ionosphere were presented. The results may indicate that different mechanisms may be responsible for the November storm since it shows a latitudinal time dependent uplift. It is also interesting to note the publication by Mannucci et al. , who found unusual large daytime ionospheric uplift with electron content increases in the low-latitude USA sector for the 29–31 October 2003 storm events and it may be that these enhancements are connected with low-latitude events. So far modeling studies have demonstrated similar uplifts with both electric fields and neutral winds [Crowley et al., 2005; Fuller-Rowell et al., 2005; Swisdak et al., 2005]. The new result presented here is the time-dependence of the uplifts. Future studies with physical models will allow greater insight into the mechanisms by attempting to replicate the large-scale observations seen in these images.
 The authors are grateful to the IGS for the provision of GPS data. Support is also acknowledged from the UK Particle Physics and Astronomy Research Council and Engineering and Physical Sciences Research Council. G. Bust's material is based upon work supported by the U.S. National Science Foundation under grant ATM-0228467-1.