GPS-derived TEC measurements (GPSTEC) represent the electron content up to 20,000 km, while a new ionosonde technique estimates the ionospheric electron content (ITEC) from ground ionosondes from the bottom of the ionosphere through the peak up to 1000 km. The two independent techniques were applied to determine ITEC and GPSTEC for Athens (38°N, 23.5°E) using the vertical Digisonde observations at Athens and GPSTEC maps produced by DLR/IKN for Europe using GPS measurements. Comparison of the two quantities over a 12-month time period shows a general agreement at daytime and a systematic deviation of ITEC towards lower values during nighttime. The diurnal variation of their residual values (ΔTEC) exhibits a morning minimum and an evening maximum, which can be explained in terms of ionospheric-plasmaspheric ionization exchange. The prominent evening enhancement, observed in all seasons around 1800–2000 LT, is attributed to the plasmaspheric bulge. The application of the superposed epoch analysis method on the daily ΔTEC values for several geomagnetic storms showed a rapid decrease of ΔTEC just after the storm initiation and a consecutive increase over a time period of 9 days, a behavior that is consistent with the plasmaspheric depletion and successive replenishment following storm activity. The daily variation of the ionospheric slab thickness is compatible with the variation of the thermospheric temperature within a day. Concerning the total slab thickness, this behavior is altered by the nighttime increase, which is most prominent in winter, and it is due to the lowering of the O+/H+ transition height. In summary, this analysis presents additional proof that the residuals ΔTEC = GPSTEC − ITEC provide information about the qualitative characteristics of the plasmaspheric dynamics as deduced from their diurnal behavior and their variation during geomagnetic storms.
 Total Electron Content (TEC) is a key parameter that describes the major impact of the ionized atmosphere on the propagation of radio waves, which is crucial for terrestrial and Earth-space communications including navigation satellite systems such as GPS, GLONASS, and the future GALILEO system. A standard technique to determine TEC is from dual frequency GPS (Global Positioning System) measurements. This technique measures the electron content along a slant signal path, from which a vertical TEC is estimated by simple geometric corrections [see, e.g., Jakowski, 1996]. As the orbit altitude of GPS satellites is ∼20,000 km, GPS-derived TEC corresponds to the total electron content (bottomside and topside ionosphere, and plasmasphere) and it is sensitive to topside ionospheric and plasmaspheric processes. The topside ionosphere is defined as the region of the ionosphere that starts at the height of maximum density of the F2 layer and extends upward with decreasing density to a transition height where O+ ions become less numerous than H+ and He+. The protonosphere or plasmasphere is the region above the O+/H+ transition height. The plasmasphere is normally considered to extend out some 20,000 km to 40,000 km. The shape and density of the plasmasphere is constantly changing, primarily due to changes in the underlying ionosphere (the source and sink of the plasmasphere) and effects of magnetic storms [Webb and Essex, 2002]. During quiet periods the density of the plasmaspheric population drops very quickly for geocentric distances above 4Re [Carpenter, 1970; Huang et al., 2003], resulting in small TEC contributions for heights above 20,000 km (the height of GPS satellites). During periods of high activity, plasmaspheric tails are formed that extend all the way down to the ionosphere affecting the Earth communications [Foerster and Jakowski, 2000].
 Recently, Reinisch and Huang  proposed a new technique to determine the vertical ionospheric electron content from ground-based ionogram measurements. The ionogram provides the information to directly calculate the vertical electron density profile up to the peak of the F2 layer. The profile above the peak is approximated by an α-Chapman function with a constant scale height HT derived from the bottomside profile shape at the F2 peak, according to the equation:
where , NmF2 is the maximum electron density, and hmF2 the height of the F2 layer peak. The ionosonde TEC, or ITEC, is then calculated as an integral from 0 to ∼1000 km over the entire profile.
 The O+/H+ transition height varies but seldom drops below 500 km at night or 800 km in the daytime, although it may lie as high as 1100 km, depending on the geophysical conditions, and particularly on solar activity. Increased solar activity can cause thermal expansion of the atmosphere resulting in an increase in this transition height [Denton et al., 1999]. Above the transition height, the scale height increases rapidly. Our simulations show that such changes affect the total content up to 1000 km by less than 2%. The ITEC parameter could develop into a very useful tool for nowcasting ionospheric space weather for operational applications, as long as some technical problems are worked out first, mainly concerning the performance of the ARTIST autoscaling software [Reinisch and Huang, 1983]; it is therefore important to determine the accuracy of the method. A preliminary ITEC validation was performed by comparing ITEC with TEC values derived from GPS, incoherent scatter radar, and geostationary satellite beacon measurements at middle latitudes and with TOPEX measurement at the equator, showing very good agreement [Reinisch and Huang, 2001; Belehaki and Tsagouri, 2002a]. It was also shown that ITEC is generally within ∼10% of the satellite TEC. As a first approach the difference between GPSTEC and ITEC may be interpreted as the plasmaspheric content, but this interpretation needs to be verified [Reinisch et al., 2001; Belehaki and Jakowski, 2002].
 In this study we are using ITEC values from Digisonde measurements at Athens and GPSTEC data from TEC maps produced by the German Aerospace Center (DLR/IKN) for Europe using GPS measurements, for Athens coordinates. The methodology to produce TEC maps is described in detailed by Jakowski  and Jakowski et al. . After determining the total electron content along a number of ray paths by using a special calibration technique for the ionospheric delay of GPS signals [Sardon et al., 1994], the slant TEC is mapped to the vertical by using a single layer approximation for the ionosphere at 400 km height. Using the GPS ground stations of the European IGS (International GPS Service) network [Zumberge et al., 1994], about 60–100 TEC data points are obtained that cover the area 20°W ≤ λ ≤ 40°E; 32.5°N ≤ ϕ ≤ 70°N. To ensure a high reliability of the TEC maps also in case of only a few measurements or at greater distances from measuring points, the measured data are combined with the empirical TEC model NTCM2 [Jakowski, 1996]. For each grid point value (spacing is 2.5°/5° in latitude/longitude) a weighting process between nearest values and model values is carried out. The data coverage of the European area is illustrated in Figure 1. The circles indicate the half width of the distance-depending weighting function for each data point. Due to the motion of GPS satellites relative to the receiving ground stations the coverage is continuously changing. The TEC estimation is constrained by different factors, such as satellite and receiver instrumental biases, multipath, mapping function, and plasmaspheric contribution errors. The achieved accuracy is of the order of 2 TECU (1 TECU = 1016 electrons m−2 [e.g., Jakowski and Sardon, 1996]).
 The aim of this paper is to determine the cause of the observed difference between ITEC and GPSTEC, performing a statistical comparison between the two independent quantities over a 12-month time period, analyzing the daily behavior of their residuals ΔTEC and studying its variation with season and geomagnetic activity.
2. Difference Between ITEC and GPSTEC
 A comparison between the two independent measurements ITEC and GPSTEC was performed for a 12-month period, October 2000 to September 2001. The ITEC values are 30-min averages calculated from 5-min samples, and the GPSTEC values are estimates for Athens as derived by DLR/IKN, with 30-min resolution. All ionograms from the Athens Digisonde, used to compute the ITEC values for Athens, were manually edited to ensure data integrity and minimize uncertainties especially during periods of high disturbances. According to Reinisch and Huang  the accuracy of ITEC is limited by uncertainties in scaling of the ionograms, nevertheless for manual scaling these errors are generally less than 5%.
 The variation pattern of the two parameters shows general agreement. For illustration we present the time plot of ITEC (dotted), GPSTEC (solid), and Dst-index for March 2001 in Figure 2. During the first 17 days of this month, the geomagnetic activity was at quiet levels, and the ITEC closely follow the GPSTEC especially during daytime. At the second part of the month a series of intense storms occurred and the two TEC estimates are strongly disturbed. However, they follow the same pattern of variation, in general. Figure 3 shows the scatterplot of GPSTEC against the corresponding ITEC for the one-year period, October 2000–September 2001. The scatterplot shows a high degree of correlation between the two independent estimates of TEC. The line drawn corresponds to the best fit line. Examination of the intercept of the best fit line shows that, in average, the GPSTEC values exceed those from ITEC by ∼6.3 TECU at an L shell of 2.9 Re, corresponding to the location of Athens.
 Related comparison studies between GPS and NIMS (Navy Ionospheric Monitoring System constellation in circular orbits at heights of about 1100 km) TEC data, obtained by Lunt et al. [1999a] for ionospheric penetration points in 1° bands centered on four latitudes in the European sector, reported compatible results: for the 50.4°N band the GPSTEC values exceed those from the NIMS observations by some 2 TECU. This value was reduced to 1 TECU at 51.4°N and was close to zero for the 53.4°N band. They found that the decreasing electron content attributable to the plasmaspheric ray paths with increasing latitude agrees with the predictions from the SUPIM model made by Lunt et al. [1999b]. Estimates of the plasmaspheric electron content variation, determined from the difference between GPSTEC and Faraday rotation TEC, were also presented by Breed and Goodwin . The median plasmaspheric TEC over Salisbury at 34.8°S calculated for December 1992 was found equal to 7.2 TECU. This result is in agreement with the general trend reported by Lunt et al. [1999a] of increasing plasmaspheric TEC with decreasing latitude and also agrees with the result from our investigation. Ray paths from Athens intersect lower L-shell flux tubes with higher electron densities than those encountered by ray paths from UK latitudes.
 The diurnal variation of GPSTEC and ITEC parameters for each season is shown in the four panels of Figure 4. It is our hypothesis that the difference between the curves gives the contribution of the plasmasphere. The major difference is observed mainly during nighttime and evening hours and sometimes exceeds 6 TECU showing that the diurnal interchange between the ionosphere and the plasmasphere may be significant particularly for lower L-shell flux tubes at solar maximum in the European sector.
3. Daily Variation of the Electron Content Above 1000 km
 To determine the cause of the difference observed between the two independent TEC estimates, we present in Figure 5 the daily variation of the monthly median values of the residuals ΔTEC for four months (October, January, May and July). These curves correspond to the diurnal variation in electron content between 1,000 km and 20,000 km, which covers the major part of the plasmasphere. The error bars correspond to the statistical standard error of the median monthly values. The daily variation has a minimum between 0800 LT and 1200 LT and a maximum between 1800 LT and 2000 LT, depending on the season. The time of the ΔTEC minimum corresponds to the morning minimum of the plasmaspheric content, following the downward flow from the plasmasphere to the ionosphere at night. The rise of ΔTEC in the afternoon, starting after the time of greatest electron density in the F2 layer, could be due to upward diffusion fluxes from the ionosphere to the plasmasphere. In addition to the statistical standard errors in ΔTEC values (Figure 5), there are also systematic errors caused by the assumptions made in the two techniques. The systematic error for the GPSTEC is 2 TECU, as already reported in the previous section, while for ITEC the systematic error for Athens Digisonde sounding parameters was also estimated equal to 2 TECU. Since during daytime hours the plasmaspheric electron content is almost zero, it is expected that both GPSTEC and ITEC parameters will measure approximately the same electron content that comes from the ionosphere. Nevertheless, the two parameters are using a totally different method to measure the electron content. It is thus expected that the negative peaks appeared during daytime hours in the monthly median ΔTEC time plot, are the result of the relative systematic errors in ΔTEC.
 During the maximum in ΔTEC daily variation, the GPSTEC exceeds the ITEC by 9 to 11 TECU. This value is greater than the systematic errors of ΔTEC, which does not exceed 3 TECU. The time of observation of the maximum refilling is possibly related to the duskside plasmaspheric bulge. This bulge [Carpenter, 1970] results from the opposition at this local time of the corotation flow near the Earth and the sunward convection of the magnetosphere driven by the solar wind.
 To further confirm our findings about the evening maximum in ΔTEC daily variation, we present in Figure 6 the corresponding values calculated for Chilton, UK (51.6°N, 1.3°W, L = 5.0Re) for January and July 2001, together with the corresponding variation estimated for Athens. It should be noticed that GPSTEC for Chilton are TEC values from the GPS receiver operated at the Chilbolton Observatory (51.8°N, 1.3°W), converted from slant to vertical TEC [Ciraolo, 1993; Ciraolo et al., 1994]. The Digisonde ITEC from Chilton is the result of automatic ionogram scaling using the ARTIST program [Reinisch and Huang, 1983], i.e., no manual editing was applied. Despite the different sources of TEC data for Athens and Chilton and the different procedure followed in calculating the ITEC, the evening enhancement in ΔTEC is a persistent feature observed in both sites, at the same local time, around 1700 LT in January and at 2000 LT in July, and has approximately the same magnitude in TECU. The observation of the evening maximum at both sites gives additional evidence that it is attributed to the plasmaspheric evening bulge. The noon minimum at Athens is not observed at Chilton. The different shape of the diurnal curve of ΔTEC for Athens and Chilton might be caused by electrodynamic effects and boundary conditions, as well as diffusion related to the physical processes for the evening bulge. Clearly the variation of the volume of the flux tube with the latitude of its ionospheric footprint greatly influences the behavior of the electron density in both the plasmasphere and the underlying ionosphere; the L-shell of the two sites is very different, L = 2.9 for Athens and L = 5.0 for Chilton. As the latitude of the ionospheric footprint increases, so does the volume of the tube, so that a longer period of dayside filling is required before a tube is in equilibrium with the underlying ionosphere. In the outermost regions of the plasmasphere the volume is so great that the upward flux during daytime cannot produce an equilibrium pressure profile and there is partial depletion at night before filling is resumed the next day [Carpenter and Park, 1973].
 So far we presented evidence that the daily variation of the residuals of GPSTEC from ITEC is compatible with the diurnal variation of the plasmaspheric electron content. To support our findings with additional statistical results, we present in Figure 7 the daily variation of ΔTEC for Athens for the four seasons extracted from the averages of the monthly medians, expressed in percentages. It can be seen that at night the plasmaspheric contribution can exceed the 50% level of the total electron content for the winter months (represented in the plot by the heavy solid line), with smaller contributions (20–30%) in other seasons. The daytime percentages are generally small in winter and fall and can exceed 10% during spring and summer months. This picture is in agreement with the results reported by Breed and Goodwin  and Lunt et al. [1999b]. According to their findings, the plasmasphere may constitute approximately 10% of the total electron content along the GPS ray path in the daytime when the electron density in the ionosphere is high, and 40–50% at nighttime when the ionospheric electron density is low.
4. Relation of Electron Content Above 1000 km With Geomagnetic Activity
 The outflow of plasma from the ionosphere is one of the main factors that determine the size, shape and the dynamics of the plasmasphere, which vary strongly according to the level of magnetospheric activity. During periods of low activity the plasmasphere becomes saturated with upflowing of ionospheric plasma. When the magnetosphere is disturbed by a magnetic storm, enhanced convection erodes the outer plasmasphere, capturing plasma in the afternoon-dusk sector and transporting it outward and sunward toward the magnetopause. Following erosion, which can last hours to tens of hours, plasma flowing upward along magnetic field lines from the conjugate ionosphere begins to refill the depleted plasmasphere. Refilling the plasmasphere typically requires several days. In order for refilling to occur, the counterstreaming plasma must be thermalized and trapped.
 It is considered that evidence in the GPSTEC and ITEC data of depletion and refilling following geomagnetic activity could provide additional and conclusive proof that the residual ΔTEC is a measure of the plasmaspheric contribution. The daily averages of all GPSTEC minus ITEC measurements were subjected to a superposed epoch analysis, based on the start day of 13 storms, determined from the start of decrease in Dst index: October 4, 2000; October 22, 2000; October 28, 2000; November 10, 2000; December 23, 2000; February 13, 2001; March 31, 2001; April 11, 2001; June 2, 2001; June 18, 2001; July 24, 2001; August 17, 2001; September 11, 2001. All selected storms are isolated, in the sense that their initiation occurs during quiet magnetospheric conditions and at least for 10 days after their onset no obvious signature of a new storm was detected by the Dst index. Regarding the mechanism in the solar wind that triggered the storms, they are characterized either as impulse storms, i.e., storms with initial compression phase, or as gradual storms [Belehaki and Tsagouri, 2002b]. In the case of impulse storms, the main phase has a duration of only some hours, whilst during gradual storms the main phase could last for more than three days. The results of the superposed epoch analysis of the daily averages of the Dst index and of the ΔTEC estimates, are presented in the upper and lower panel of Figure 8, with the day of the storm initiation designated as day 0. The error bars correspond to the statistical standard error of the superposed daily averages of ΔTEC. Since the occurrence of the selected storms covers all seasons, the different effects of winter and summer storms in the plasmasphere are expected to be averaged.
 The day of storm initiation is characterized by an increase in ΔTEC with the value rising to over 6.49 TECU. This rise is considered to be statistically significant, since the standard error at this point is ±0.18 TECU and the average ΔTEC is 6.28 TECU, as extracted from the results of the correlation between ITEC and GPSTEC presented in Figure 3. This increase in ΔTEC coincides with an increase in Dst-index caused by storm sudden commencements, which are impulse like disturbances of the magnetic north component at middle to high latitudes caused by the first encounter of an interplanetary shock wave with the magnetosphere. Therefore it is possible that the produced electromagnetic drift will move plasma to the plasmasphere. On the following 3 days after the start of the storm the magnitude of the plasmasphere content drops to 4.5 TECU, although the main decrease occurred the first two days after the start of the storm, giving clear evidence of the depletion of the plasmasphere. According to the picture of the plasmaspheric response to geomagnetic activity determined from measurements of plasmaspheric electron content from the ATS6 satellite [Kersley and Klobuchar, 1980], the plasmaspheric content drops suddenly after the onset of the storm main phase and the depletion phase has a duration of one day. The longer duration of the depletion phase determined in this paper is caused, according to our understanding, by the inhomogeneous type of storms selected for this analysis. Some of them are gradual storms with multiple main phase reactivation events, which in some cases last for three days and consequently alter the typical picture of plasmaspheric depletion. The long duration of the main phase of the storms can also be inferred in average by the Dst index that reached its minimum 2 days after the onset. The gradual refilling of the plasmaspheric flux tubes can be inferred from the rising trend in the content apparent on subsequent days. The filling time of the plasmasphere is generally greater than the recurrence time between storms (approximately 15 days) so that the plasmasphere can be thought of as generally in a state of partial replenishment.
5. Daily Variation Pattern of Ionospheric and Total Slab Thickness
 The slab thickness may be regarded as the depth of an imaginary ionosphere, which has the measured TEC and uniform electron density equal to the maximum electron density of the actual ionosphere [Breed and Goodwin, 1997]. For a given foF2, the slab thickness depends on the electron density versus height profile, for instance, the sharper the peak electron density the smaller is the slab thickness. The slab thickness is calculated using both GPSTEC and ITEC parameters, according to the following formulas: , and , where α is a coefficient equal to 806.45 determined according to the relation between the electron density NmF2 and the critical frequency foF2: NmF2/1 m−3 = , and GPSTEC and ITEC given in TEC units, 1016m−2, foF2 in MHz, and τ in km. According to the definition of Breed and Goodwin , τt includes the electron contents of both the ionosphere and the plasmasphere up to 20,000 km. On the other hand, τi provides information on the bottomside and topside ionosphere only. Figure 9 shows the diurnal plots of slab thickness parameters τt (dotted-black) and τi (solid black), for the four seasons, extracted from the average of the monthly median values.
 We observe a large difference during nighttime hours, which is most prominent in winter. The winter nighttime enhancement in τt is due to a lowering of the O+/H+ transition height [Titheridge, 1973]. The scale height is greater above the O+/H+ transition height. It follows then that the lower the O+/H+ transition height, the greater the average scale height, and hence the greater the slab thickness - which is directly proportional to the scale height [Breed and Goodwin, 1997]. Between ∼0800 LT to 1400 LT the two parameters τt and τi, are almost equal, indicating a state close to diffusion equilibrium between the ionosphere and the plasmasphere. The nighttime difference between the two parameters τt and τi, is getting smaller in fall and spring months, while in summer months, when the solar influence is maximum, the nighttime difference minimizes. The consistent difference between the ionospheric and total slab thickness observed in summer months during night and day hours also can be explained in terms of continuous upward diffusion of ionization during the day. Another effect that might cause the observed behavior in summer is that the replenishment times of the plasmasphere after a geomagnetic storm are significantly longer in summer than in winter [Kersley and Klobuchar, 1980].
 The τi parameter exhibits a diurnal variation indicative of the effect of the diurnal variation of the thermospheric temperature. The variation of the τt parameter can be regarded as a superposition of a diurnal curve with maximum at noon that corresponds to the thermospheric temperature variation and of a second curve which is the result of the diurnal variation between downward (at night) and upward (during the day) diffusion exchange of ionization between the ionosphere and the plasmasphere, resulting in the enhancement of the plasmaspheric content during the day and depletion of plasmaspheric ionization at night.
6. Summary and Conclusions
 The aim of this contribution was to determine the source that causes the observed difference between the Athens ITEC values calculated with the Reinisch and Huang  method and the GPSTEC extracted from the TEC maps produced by DLR/IKN over Europe for Athens. We have shown that the residuals of the two independent quantities, ΔTEC, provide qualitative information about the plasmaspheric dynamics as deduced from their diurnal and seasonal behavior and their variation during geomagnetic storms. In general, the daily variations of electron content above 1,000 km for Athens have the same qualitative characteristics with the plasmaspheric content daily variation. There is some diurnal interchange between the ionosphere and plasmasphere with downward diffusion from the latter helping maintain the nighttime F2 layer [Lunt et al., 1999b; Belehaki and Tsagouri, 2002b] and daytime refilling. The evening peak in ΔTEC residuals, which is the striking feature in ΔTEC variation over Athens and is also observed over Chilton, is attributed to the evening plasmaspheric bulge. The effect of geomagnetic activity in ΔTEC residuals is qualitatively the same as the effect on the plasmasphere, showing a drastic depletion immediately after the start of the storm and a consequent replenishment that lasts for 9 days. The daily variation of the ionospheric slab thickness is compatible with the variation of the daytime thermospheric temperature. There is, however, a significant nighttime increase of the total slab thickness, which is most prominent in winter and is due to the lowering of the O+/H+ transition height.
 The study demonstrates that it is possible to estimate experimentally the effect of the plasmasphere on GPS measurements comparing ionosonde derived TEC (ITEC) with GPS- derived TEC (GPSTEC) at a given location. The magnitude of the plasmaspheric contribution is significant at low L-shells at solar maximum contributing more than 50% of the total at several epochs; therefore it should be taken into account especially for operational applications.
 A. B. wishes to thank Prof. Len Kersley and Dr. Ioanna Tsagouri for useful discussions relating to this study. Thanks are also due to Dr. Ljiljana Cander for providing the TEC estimates from the Chilton and Chilbolton observatories. We would like also to acknowledge the contribution of the WDC for Geomagnetism, Kyoto Dst index service, Dst stations and of the persons who derived the final Dst index used for this work. B.W.R. was in part supported by the contract AF F19628-02-C-0092. This work is part of the European COST271 Action.