The first global simulation study and comparison to data of the ionospheric effects associated with the enhanced EUV irradiance of the Bastille Day flare are presented. This is done by incorporating a time-dependent EUV spectrum, based on data and hydrodynamic modeling, into the NRL ionosphere model SAMI3. The simulation results indicate that the total electron content (TEC) increases to over 7 TEC units in the daytime, low-latitude ionosphere. In addition, it is predicted that the maximum density in the F-layer (NmF2) increases by ≲20% and that the height of the maximum electron density (HmF2) decreases by ≲20%. These results are explained by the increased ionization at altitudes <400 km which increases TEC and NmF2 while decreasing HmF2. The results are in reasonably good agreement with data obtained from GPS satellites and the TOPEX satellite.
 The impact of solar geoeffective events on the earth's ionosphere is an important research topic with regard to space weather. There are two main aspects to the effect of solar variability on the ionosphere: enhanced radiation associated with solar flares and storm-time effects (e.g., penetration electric fields and wind-driven effects) as described by Basu et al. . In this Letter we address the former issue for the Bastille Day flare (July 14, 2000). Enhanced X-ray and extreme ultraviolet (EUV) irradiance associated with solar flares directly impacts the earth's ionosphere through increased dayglow emissions and ionization [Mendillo et al., 1974; Meier et al., 2002; Dymond et al., 2004; Liu et al., 2004; Tsurutani et al., 2005] For example, Dymond et al.  reported an increase of 40% in the 911-Å emission during the Bastille Day flare. These observations were made with the Low-Resolution Airglow and Aurora Spectrograph (LORASS) aboard the Advanced Research and Global Observation Satellite (ARGOS). Very recently, Tsurutani et al.  compared the ionospheric response of several flares using Global Positioning Satellite (GPS) data. For the Bastille Day flare, they report an increase in the total electron content (TEC) at the subsolar point of 5–7 TECU (1 TECU = 1016 m−2).
Meier et al.  developed a model of the EUV spectral irradiance for the Bastille Day flare and used SAMI2 to study ionospheric effects at two times: pre-flare (UT 1000) and peak flare (UT 1036). However, these modeling results were limited to a single magnetic longitude plane, did not report changes in TEC, used a simple linear time-dependent EUV spectrum, and did not compare results to data. In this Letter, we substantially improve upon the modeling results presented by Meier et al. . Specifically, we present the first global simulation study and comparison to data of the ionospheric effects associated with the Bastille Day flare. This is accomplished by incorporating a more realistic time-dependent EUV spectrum, based on data and hydrodynamic modeling, into the NRL three dimensional ionosphere model SAMI3. It is found that the TEC increases by ∼7 TEC units in the daytime, low-latitude ionosphere which is consistent with differential TEC measurements obtained from GPS satellites and the TOPEX satellite.
2. EUV Spectrum
 High cadence, spectrally resolved observations of the soft X-ray and EUV irradiance changes during a flare do not exist for most events. To address this limitation we are developing methods for performing numerical simulations of solar flares using a time-dependent hydrodynamic code [e.g., Mariska et al., 1989]. In our simulations we model the flare as a succession of independently heated filaments and derive the energy deposited into each filament from the observed GOES soft X-ray fluxes of the event [Warren and Antiochos, 2004]. Currently we can accurately reproduce the evolution of the high temperature (T ∼ 2 × 106 K) emission that dominates the emission at the shortest wavelengths [Warren and Doschek, 2005]. At transition region temperatures (T ≲ 7 × 105 K), however, the simulations suggest enhancements in the solar irradiance that are much too large. This difficulty is associated with the sharp gradients in temperature and density in the transition region and the complexity of the interface with the chromosphere. Therefore, to develop a meaningful time-dependent EUV irradiance spectrum for the Bastille Day flare we constrain the simulation results at transition region temperatures to agree with the UV and EUV variability observed with the Transition Region and Coronal Explorer (TRACE). However, additional work remains to be done before we can accurately reproduce solar irradiance variations from physical models.
 The modeled spectra are binned in 37 segments, i.e., similar to the binning used in EUVAC [Richards et al., 1994]. The time cadence for the spectra is 5 minutes between UT 1000 and UT 1200. The EUV spectrum is shown in the top panel of Figure 1 at times UT 1000 (black) and UT 1036 (red). The flare began around UT 1000 and reached its peak intensity around UT 1036. In the bottom panel of Figure 1 we show the ratio of the flare irradiance at UT 1036 to the pre-flare irradiance at UT 1000. The largest relative increase, greater than a factor of 9, occurs at ∼150 Å; there is an increase of 1.2–2.0 for the wavelengths between 200 and 1000 Å. The flare irradiance is larger than the pre-flare irradiance between UT 1012 and UT 1200.
3. Simulation Model
 The Naval Research Laboratory has developed a three dimensional low-latitude to midlatitude ionosphere model: SAMI3. This model is based on the two dimensional model SAMI2 [Huba et al., 2000, 2002, 2003] SAMI2 describes ionospheric dynamics in a magnetic plane (i.e., latitude and altitude); SAMI3 contains the same fundamental physical processes as SAMI2 but extends the grid in longitude and includes zonal drifts. Additionally, SAMI3 uses an IGRF-like magnetic field model where the IGRF field is approximated as a dipole field line at each longitude in the simulation grid.
 The simulation results presented in this paper use a grid (nz, nf, nl) = (121, 150, 75) where nz is the number of points along a dipole field line, nf is the number of points in altitude along the magnetic apex (i.e., number of magnetic field lines), and nl is the number of points in longitude. The grid is nonuniform in a magnetic plane (nz, nf) and uniform in longitude (nl). We only consider the four majority ion species: H+, O+, NO+, and O2+. The geophysical parameters used are the following: day 196, year 2000, F10.7 = 231.9, F10.7A = 186.3, and Ap = 51. The simulations were started at UT 0000 and run until UT 1300. Lastly, we note that the results presented do not adjust either the EUV spectrum or SAMI3 parameters to obtain agreement with the data.
 We performed two simulations: one simulation used the pre-flare spectrum at UT 1000 throughout the run, the other simulation used the time-dependent flare spectra previously described. To quantify the impact of the enhanced flare irradiance on the ionosphere, we have calculated the difference in TEC, NmF2, and HmF2 between the two simulations. We show only the TEC results here. In Figure 2 we show the difference in TEC at time UT 1100. The TEC increases by ∼7 TEC units over a wide area in the equatorial region, and by ∼4 TEC units at higher latitudes towards sunrise and sunset. For example, at UT 1100 the TEC is increased by 4 TECU over the latitude range −40° < θg < 30° and the longitude range −50° < ϕg < 100°. These values are consistent with the TEC increases reported by Tsurutani et al. . In addition, we find that the maximum value of NmF2 is ∼2.5 × 106 cm−3; the maximum increase in NmF2 is ∼3 × 105 cm−3. Thus, the maximum increase in NmF2 is roughly 10% of the pre-flare value. However, throughout most of the sunlit ionosphere, the value of NmF2 is increased by ∼5%. The HmF2 decreases by almost 100 km at ∼10° latitude in the longitude range ±5°.
 In Figure 3 we plot the time dependence of the EUV irradiance at 150 Å (solid red) and 400 Å (dashed red). We also plot the difference in electron density between the flare and no-flare cases (Δn = nflare − nno flare) at altitudes of 172 km (solid blue), 222 km (dashed blue), and 382 km (blue dashed-dot-dot). The geographic latitude is θg = 0° and the geographic longitude is ϕg = 0°. The values of the plotted quantities are normalized to unity at their maximum values so that the temporal relationships are elucidated. The temporal relationships between the 150 Å irradiance and the Δn at z = 172 km, and the 400 Å irradiance and the Δn at z = 222 km agree extremely well until shortly after the EUV peak values. Thus, the increase of the electron density is directly proportional to the enhanced solar irradiation during the onset of the flare for altitudes <230 km. Following the peak intensity of the flare, the electron density decays primarily because of chemical processes and recombination; however, there is still additional photoionization relative to the no-flare case because of the flare enhanced irradiation. In contrast to these results, the increase in the electron density at higher altitudes (e.g., z = 382 km) is primarily due to the enhanced photoionization flux but occurs at a slower rate and is more long-lived. The plasma at higher altitudes persists for longer times because the lower neutral atmospheric densities slow chemical/recombination processes. The long-lived plasma at high altitudes (≳350 km) leads to the enhancement of TEC even after the flare has subsided.
5. Data Comparison
 We first compare our model results to GPS data. GPS receivers provide dual-frequency phase and range measurements between the observing station and the GPS satellite. Line-of-sight TEC data are derived from the phase and range measurements, with transmitter and receiver instrumental biases removed. Such line-of-sight TEC is then converted to a vertical TEC measurement at the assumed sub-ionospheric point (450 km) along the radio link. In Figure 4 we compare our results to data obtained from the GPS station at Malindi, Kenya (MALI) for three satellites (GPS25, GPS31, and GPS33). The MALI site is located at −3° latitude and 40.2° longitude. The sub-ionospheric latitude of the satellite tracks for the MALI station during the interval of interest is from ∼−2° to ∼−10°, and in longitude from ∼39° to ∼42°. The GPS data are denoted by dashed lines and represent the difference in vertical TEC between July 13, 2000 (no-flare day) and July 14, 2000 (flare day) as a function of time. The solid lines are the difference in TEC between the flare and no-flare simulations versus time. The differential TEC calculated using SAMI3 does not vary very much over the range of latitude and longitude associated with the GPS data so we use a single position to calculate TEC: −4° latitude and 40° longitude. The simulation results have been offset so that at flare onset (UT 1000) the simulation results have roughly the same value as the GPS data to facilitate the comparison. The increase in TEC because of the enhanced flare radiation is roughly 7 TECU for both the data and the simulation results. However, the data shows a somewhat sharper increase in TEC at flare onset and a more abrupt decrease in TEC after ∼UT 1030 than the simulation results. The persistence of the larger values of TEC in the simulation results is attributed to the increased electron density being lifted to higher altitudes (i.e., >300 km) and being long lived.
 In Figure 5 we compare TOPEX TEC data (red lines) [Coker et al., 2001] with SAMI3 model results (blue lines). The solid curves are for July 14, 2000 at time 1112 UT and longitude 81E; the dashed curves are for July 13, 2000 at time 1050 UT and longitude 90E. The TOPEX data is also a function of longitude while the SAMI3 results are for constant longitude. The positions are coincident at the geographic equator (0° latitude). The salient points are that both TOPEX and SAMI3 observe an increase in TEC from July 13 to July 14 due to the solar flare and that the model results are in reasonably good agreement with the data. The largest discrepancy occurs for latitudes <0°; one explanation for the lack of agreement between data and model results is that the E × B drift assumed in the model was not accurate. Simulation model results in the low-latitude ionosphere are very sensitive to the E × B drift and neutral wind models.
 We have presented new results of the effect of the Bastille Day flare on the low-latitude to midlatitude ionosphere using the NRL ionosphere model SAMI3 in conjunction with a time-dependent EUV spectrum based on hydrodynamic modeling and data. We find that the total electron content increases by ∼7 TECU. In addition, it is found that NmF2 increases by ≲20% and that HmF2 decreases by ≲20%. These results are explained by the increased ionization in the altitude range ≲400 km which increases TEC and NmF2 while decreasing HmF2. Overall these results are consistent previous observational measurements [Tsurutani et al., 2005; Dymond et al., 2004].
 We have compared the model results with GPS satellite data and TOPEX data. An underlying assumption in the comparisons is that the ionosphere is similar, except for the flare, on the pre-flare day and flare day. The purpose of the comparison is not to obtain precise agreement between the model results and data but to demonstrate that the flare-enhanced ionization predicted by the model is consistent with observations. We obtain good agreement with respect to the magnitude of the TEC increase in the GPS data; the agreement with the TOPEX data is also reasonably good with respect to the absolute value of the TEC and differential value of the TEC. However, one significant difference between the GPS data and simulation results is the more rapid increase in TEC at flare onset. We note that the current SAMI3 model does not account for ionization processes for spectral lines below 50 Å. For instance, ionization produced by photoelectrons produced by soft X-rays in the altitude range 100–200 km. This additional ionization mechanism may explain this discrepancy. One future improvement of the model will be to extend the irradiance spectrum down to ∼5Å and to include the relevant physics. Also, the temporal cadence of the modeled flare data is 5 min; a faster cadence rate may lead to a faster increase in TEC. A second difference is the persistence of enhanced TEC in the model as compared to the GPS data. The TEC decrease or variations seen in the temporal sampling is likely due to the ionospheric spatial variations in the structured equatorial anomaly region, which often shows day-to-day weather variability. This variation is not included in the model results at a fixed location as plotted in Figure 4. We note that Tsurutani et al.  reported a persistence in the TEC for several hours following several solar flares consistent with the model prediction. Finally, we intend to model the ionospheric response to more recent flares for which we can obtain spectral data from the TIMED SEE instrument [Woods et al., 2005] for a more accurate representation of the irradiance spectrum. We intend to address these issues and elaborate on the details in a future publication.
 This research has been supported by ONR and NASA.