An analysis of thermospheric density response to solar flares during

JOURNAL OF GEOPHYSICAL RESEARCH, VOL. 117,, doi:10.1029/2011ja017214, 2012 An analysis of thermospheric density response to solar flares during 2001 2006 Huijun Le, 1,2 Libo Liu, 1 and Weixing Wan 1 Received 30 September 2011; revised 1 February 2012; accepted 3 February 2012; published 6 March 2012. [1] Previous studies show there are significant thermospheric responses to the two great solar flares on October 28, 2003 (X17.2) and November 4, 2003 (X28). In the present study, we further explored the thermospheric response to all X-class solar flares during 2001 2006. The observed results show that X5 and stronger solar flares can induce an average enhancement of 10 13% in thermospheric density in latitude 50 S-50 N within 4 h after the flare onset. Many important lines and continua in solar EUV region are optically thick, thus EUV enhancements are smaller for flares located near the solar limb due to absorption by the solar atmosphere. Limb flares induce smaller thermospheric responses, due to the limb effect of solar EUV. The thermospheric density enhancement is much more correlated with integrated EUV flux than with peak EUV flux, with a high correlation coefficient of 0.91, which suggests that thermospheric response is strongly dependent on the total integrated energy into the thermosphere. Citation: Le, H., L. Liu, and W. Wan (2012), An analysis of thermospheric density response to solar flares during 2001 2006, J. Geophys. Res., 117,, doi:10.1029/2011ja017214. 1. Introduction [2] Solar extreme ultraviolet (EUV) and X-ray photons are the primary energy source of the ionosphere and thermosphere of the Earth. A solar flare is a sudden eruptive solar phenomenon, associated with significant enhancements in EUV and X-ray radiations, with larger enhancements in X-rays and relatively smaller enhancements in long wavelength EUV. These emissions cause sudden and intense increases in ionization at various levels in the Earth s ionosphere [e.g., Afraimovich, 2000; Leonovich et al., 2002; Liu et al., 2004, 2006; Tsurutani et al., 2005; Wan et al., 2005; Zhang et al., 2002; Zhang and Xiao, 2005; Le et al., 2007]. [3] Most previous studies related to solar flares have so far focused on the ionospheric responses [Afraimovich, 2000; Zhang et al., 2002; Le et al., 2007; Liu et al., 2011]. Only recently have studies been performed to quantify the thermospheric response to flares using observations and modeling [e.g., Sutton et al., 2006; Liu et al., 2007; Pawlowski and Ridley, 2008]. These studies show that in addition to the great disturbances in the ionosphere, solar flares can also induce significant responses in the thermosphere. Sutton et al. [2006] reported the first measurements of the thermosphere density response to two great flare events on 28 October 2003 and 04 November 2003. They found the thermosphere density increases associated with the flares are about 1 Beijing National Observatory of Space Environment, Institute of Geology and Geophysics, Chinese Academy of Sciences, Beijing, China. 2 Also at State Key Laboratory of Space Weather, Center for Space Science and Applied Research, Chinese Academy of Sciences, Beijing, China. Copyright 2012 by the American Geophysical Union. 0148-0227/12/2011JA017214 50 60% and 35 45%, respectively, at low to midlatitudes. Liu et al. [2007] examined the thermospheric and ionospheric responses to the solar flare on 28 October 2003 by using measurements from the CHAMP satellite. They found the contrasted behavior between the thermosphere and ionosphere: the neutral density was enhanced by 20% almost homogeneously at all latitudes below 50 S-50 N; the plasma disturbance exhibited a distinctive latitudinal structure, with largest density enhancements of 68% at the dip equator, moderate increase of 20% at midlatitudes. Pawlowski and Ridley [2008] simulated the thermospheric responses to the solar flare on 28 October 2003 using the Global Ionosphere- Thermosphere Model. The results show that the thermospheric density at 400 km can increase by as much as 14.6% in about 2 h and takes 12 h to return back to a nominal state. The flare irradiance spectral model (FISM) is an empirical model that estimates the solar irradiance at wavelengths from 0.1 to 190 nm at 1 nm resolution. It has a high temporal resolution of 60 s to model variations due to solar flares (for details please see Chamberlin et al. [2007, 2008]). Using the solar spectra data from the FISM model as solar input to the National Center for Atmospheric Research (NCAR) thermosphere-ionosphere-mesosphere electrodynamics general circulation model (TIME-GCM), Qian et al. [2010] investigated the ionosphere/thermosphere response to an X17 solar flare. The results also show the significant thermospheric disturbances including about a maximum increase of 85 K in neutral temperature and 20% in neutral density at height of 400 km. [4] The studies mentioned above mainly focus on cases study for the thermospheric responses to the great solar flares on 28 October 2003 and 04 November 2003. It is not clear how the thermosphere responds to other weaker flares. 1of8

In this study, we analyze the neutral density variation for all X-class solar flares during 2001 2006 based on the density data derived from accelerometers on the Challenging Minisatellite Payload (CHAMP) and Gravity Recovery and Climate Experiment (GRACE) satellites. 2. Data Analysis [5] In present study, we use total neutral density derived measurements of neutral density obtained from the accelerometers aboard CHAMP and GRACE satellites. With a near-polar inclination, the satellite provides near-global coverage at an approximate altitude of 410 km within two local time sectors at most latitudes. It takes about two months to sweep through 12 h of local time. The satellite GRACE is similar to the satellite CHAMP but has a higher orbit. Acceleration, attitude, and orbit ephemeris data files were used in the calculations. The derivation of thermosphere densities from these data is complex, and involves consideration of such effects as radiation pressure, satellite shape and orientation, thruster firings, etc. For detailed documentation about the calculation of density data from satellite accelerometers, please see Sutton et al. [2005]. All density data has been averaged into 3-degree latitudinal bins to reduce any random errors. In addition, neutral density has also been normalized to 2 heights (400 km and 410 km) using the NRL-MSISe-00 empirical density model. [6] In order to examine the neutral density responses to solar flares, the changes between the density during a solar flare and during quiet conditions need to be calculated. As mentioned above, all density data have been averaged into 3 latitudinal bins from 90 to 90 with an interval of 3. For each latitude bin, the mean of three orbits before the flare onset is calculated as the reference value. Thus we can get the reference value at each latitude bin. The period of orbit is about 94 min. There are about 15.4 orbits in a day. In a short time of no more than one day, the satellite covers the different longitude region in different orbit but the local time of each orbit is almost the same. The neutral density has a significant local time variation but has small longitudinal variation. On the basis of neutral density data from CHAMP, Liu et al. [2009] reported the neutral density exhibits a pronounced wave-4 longitudinal pattern in low latitudes, but its peak-trough amplitude is 2 0.2% of the background neutral density, which is much smaller than the enhancement induced by solar flares as mentioned above. Thus, the longitudinal difference is ignored and we can obtain the solar flare effect on the neutral density by comparing the observed data to the reference data (the mean of the previous 3 orbits). 3. Results and Discussion [7] To investigate the neutral density responses to solar flares, the responses of first three orbits (corresponding to the time within about 4 h) after solar flare onset are calculated and their mean is taken as the response for the solar flare. For each solar flare, the percentage variations (DNn) in different latitude bins are calculated. All solar flares are separated into three species: X1 2, X2 5, and X5 30. The neutral density can be significantly affected by geomagnetic storm [e.g., Forbes et al., 2005; Liu and Lühr, 2005; Sutton et al., 2005; Bruinsma et al., 2006]. In addition, on the basis of the analysis of thermosphere densities from the CHAMP and GRACE satellites, Lei et al. [2011] also reported the rapid recovery of 6 8 h. Thus, to reduce the geomagnetic activity effects, a solar flare event is excluded if the Kp index exceeds 4 within 12 h before the solar flare onset or 4 h after the solar flare onset (this 4 h correspond to the time of first three orbits), or the relative change in Kp index exceed 3 in the corresponding period. Although there are some geomagnetic disturbances, the great solar flare events on 28 October 2003 and 4 November 2003 are included for investigation because several studies [e.g., Sutton et al., 2006; Liu et al., 2007] have reported the significant thermosphere responses to the two solar flares. According this condition, for the CHAMP satellite, the number of solar flares having measurements is 20, 11, and 10 for the three species, respectively. For the GRACE satellite, the number of solar flares having measurements is 15, 5, and 8, respectively. In addition, the solar flare on 20 January 2005 is also excluded due to the early observation time (LT = 6.5) which is around the sunrise time. [8] The range of background fluctuations is an important reference to scale the flare effect on the neutral density. For the measurements of CHAMP, there are about 15 orbits each day. We calculated the percentage change of standard deviation comparing to the median for each magnetic quiet day (Kp < 4) during 2001 2006. Then the mean percentage change at each latitude bin can be calculated. The statistical results show the mean range of fluctuation for middle-low latitudes (within 50 degree) is about 6.5% of the background median. [9] The corresponding results are plotted in Figure 1 (left plots for CHAMP and right plots for GRACE). The mean response for each species is also calculated and plotted in the panel. As illustrated in Figure 1, the measurements from CHAMP (GRACE) show that the mean response for latitudes within 50 is about 0.75% (1.4%), 3.01% (4.86%), and 10.86% (13.63%) for solar flares of X1 2, X2 5, and X5 30, respectively. This mean response at a latitudinal bin refers to the average of first three orbits after solar flare onset. As mentioned above, the range of background fluctuation is about 6.5%. Therefore, the results (as shown in Figure 1) suggest that there is almost not any solar flare effect on the neutral density for X1 2, very slight solar flare effect for X2 5, and much stronger solar flare effect for X5 30. That is, the neutral density response increases with increasing flare strength and there is significant response for flares larger than X5. [10] As mentioned above, the observed results show a significant neutral density response to solar flares with class larger than X5. Thus to clearly see the neutral density response for these flares, we illustrate the raw observations for the 15 X5 30 flares since 2001 in Figure 2. The Kp index is also marked in the figure. As shown in Figure 2, there are significant neutral density enhancements just shortly after the flare onset for solar flares on 4/2/2001, 4/6/2001, 10/28/2003, 11/2/2003, and 11/4/2003. These enhancements occurred mainly at low and middle latitudes and there is much less variation at high latitudes. Furthermore, the variations of Kp index show that there is no obvious geomagnetic activity during these solar flare events. On the basis of model simulations, Qian et al. [2011] also showed that the latitude dependence of thermosphere responses to solar flares 2of8

Figure 1. The mean percentage changes in neutral density at three orbits after the solar onset for each flare (dotted line). Solar flares are separated into three species: X1 X2, X2 X5, and X5 X30. For each species, the mean response for the corresponding solar flares is also shown (solid line) in each panel. (left) CHAMP measurements and (right) GRACE measurements. well correlated with the solar zenith angle effect: the large thermosphere responses occurring at low and middle latitudes due to the small solar zenith angle. However, the neutral gas disturbances related with the geomagnetic activity often started at high latitudes and then propagated to lower latitudes [Bruinsma et al., 2006; Lei et al., 2011]. Thus, the significant enhancements mainly occurring at low and middle latitudes strongly suggest that these responses should be caused by an increase in solar radiation flux. [11] Figure 2 also shows that there are some enhancements in neutral density during solar flares on 8/25/2001, 10/29/2003, 9/9/2005, and 12/6/2006. However, the enhancements during these events not only occurred in low and middle latitudes, but also occurred in high latitudes and pole region. Furthermore, the enhancements first appeared in the high latitudes and then propagated to the lower latitudes. In addition, the enhancements do not just occur after the flare onset; in contrast, some of them have happened before the solar flare onset. As illustrated in Figure 2, the Kp index increased significantly during the four solar flare event. Thus, judging from the results as mentioned above, one can come to a conclusion that the enhancements in neutral density during solar flares on 8/25/2001, 10/29/2003, 9/9/2005, and 12/6/2006, are mainly contributed to by geomagnetic activity increases, not induced by solar flare effect. 3of8

Figure 2. Latitude versus time depictions of total mass density measured by CHAMP satellite during 15 larger than X5 solar flares. The density is normalized to 400 km. The Kp index is superimposed in each panel. The vertical solid line indicates the flare onset. The observation time (local time) is also noted in each panel. [12] In addition, as illustrated in Figure 2, there is no obvious neutral density response to solar flares on 12/13/2001, 10/23/2003, 1/20/2005, 9/7/2005, 9/8/2005, and 12/5/2006. With the exception of the solar flare on 12/13/2001, these solar flare events have a common feature: limb flare with central meridian distance (CMD) larger than 60 degree. It is known that the ionospheric responses to solar flares have a significant dependence on solar location (CMD value) [Zhang et al., 2002; Tsurutani et al., 2005; Mahajan et al., 2010]. Then a question arises as to whether there is the same solar location effect for the thermospheric response and what factors can affect or control the thermospheric response. Qian et al. [2010] investigated the flare location effect on ionosphere/thermosphere responses by the TIME-GCM model with two different solar irradiance inputs derived from FISM model: one is for center flare and the other is for limb flare. The results show that flare-driven changes in the F region ionosphere, total electron content, and neutral density in the upper thermosphere, are 2 3 times stronger for a disk-center flare than for a limb flare, due to larger EUV enhancements for a disk-center flare. [13] As is known, the most important source of external forcing to the thermosphere is solar extreme ultraviolet (EUV) and soft X-ray (XUV) irradiation which results in the dissociation and ionization of background atmosphere from roughly 90 km to 200 km. During solar flares, the incident solar radiations carry much more energy than the requirement of dissociation and ionization of the gas. The leftover energy can further heat the neutral gas. The much enhanced radiations produce much more ionizations and high energy photoelectron, which also further transfer the energy to the neutrals by collisions. The heating causes the expansion of 4of8

Figure 3. Dependence of percentage enhancement in neutral density measured by CHAMP satellite during 11 solar flares on (a) Peak X-ray, (b) modified Peak X-ray with CMD value (peak X-ray*Cos(CMD)), (c) Time-integrated X-ray, (d) Peak EUV, and (e) Time-integrated EUV, respectively. The solid line represents the linear fitting. Correlation coefficients are also marked in the figure. The data of X-ray flux is from the GOES satellite and data of EUV (26 34 nm) is from the SOHO satellite. the thermosphere, which results in the enhancements of neutral density as observed by the CHAMP and GRACE. During a solar flare, the soft X-ray and parts of the extreme ultraviolet (EUV) irradiance can increase by as much as a factor of 50 in minutes [Woods et al., 2004]. [14] As mentioned above, CMD value might also be one of the important factors for the response of upper atmosphere to a solar flare. Thus, to further analyze the relationship between neutral density responses and X-ray and EUV flux, as well as the flare location effect, we plotted percentage enhancement in neutral density measured by CHAMP satellite versus peak X-ray flux and peak X-ray flux*cos(cmd), time-integrated X-ray during 10 X-class solar flares, peak EUV and time-integrated EUV during 7 X-class solar flares in Figure 3. The data of X-ray flux (0.1 0.8 nm) from GOES satellite and the data of EUV (26 34 nm) from the SOHO satellite are used in this study. The 10 solar flares are selected from 15 X-class flares; the four flares on 8/25/2001, 10/29/2003, 9/9/2005, and 12/6/2006 are excluded due to the significant geomagnetic activity and CHAMP satellite has no measurement during the flare on 1/20/2005. In addition, only 7 flares are left to be used in the analysis for the relation between EUV flux and DNn due to no EUV measurement during the three solar flares on 9/7/2005, 9/8/2005, and 12/5/2006. For each solar flare, the value of DNn averaged over 50 latitude represents the neutral density response to the solar flare. [15] As shown in Figure 3a, the correlation between the DNn and peak X-ray flux is very poor; its correlation coefficient is only 0.30. To consider flare location effect, we calculated the values of peak X-ray*Cos(CMD) and illustrated the dependence of DNn on peak X-ray*Cos(CMD). Figure 3b shows a much higher correlation coefficient (0.85) between them. The large difference between the results in Figures 3a and 3b suggests that the thermosphere responds differently depending on flare location on the disk. It is known that X-ray flux intensity is independent of solar origin location [e.g., Mosher, 1979; Samain, 1979]. Their results indicated that the EUV and UV radiation intensity decreases when a radiation source moves from the central solar meridian to the limb, whereas the intensity of soft X rays remains almost unchanged. Thus, we use peak X-ray*Cos(CMD) to reflect the flare location effect on the EUV flux. That is, for the same class solar flare (the same value of peak X-ray flux), the larger value of CMD means the smaller value of peak X-ray*Cos(CMD), which represents the smaller EUV flux. The higher correlation between neutral density response with peak X-ray*Cos(CMD) than with peak X-ray flux indirectly reflects the flare location effect on the EUV flux, as neutral density response mainly comes from the enhancements in EUV flux. Mahajan et al. [2010] and Zhang et al. [2011] also used the same method to investigate the flare location effect on the ionospheric response 5of8

Table 1. The Basic Information During 15 X5+ Solar Flares a Date Peak X-ray CMD (degree) Peak X-ray *Cos(CMD) Peak EUV Time-Integrated EUV Kp DNn (%) Observation Time (LT) 04/02/2001 20.0 72 6.18 4.12 6.57 14.71 14.8 04/06/2001 5.6 31 4.8 3.43 4.08 11.86 14.4 08/25/2001 5.3 34 4.4 3.60 7.50 12.50 13.6 12/13/2001 6.2 9 6.12 4.73 3.35 6.12 15.6 10/23/2003 5.4 88 0.19 2.23 3.65 1.56 13.6 10/28/2003 17.2 8 17.03 5.49 10.83 27.14 13.1 10/29/2003 10.0 2 9.99 4.82 6.27 17.47 13.0 11/02/2003 8.3 56 4.64 2.93 5.59 17.40 12.6 11/04/2003 28.0 83 3.41 3.50 6.01 9.61 12.5 01/20/2005 7.1 61 3.44 7.8 09/07/2005 17.0 77 3.82 6.54 10.6 09/08/2005 5.4 75 1.39 9.58 10.5 09/09/2005 6.2 67 2.42 22.2 10.5 12/05/2006 9.0 68 3.37 6.10 16.8 12/06/2006 6.5 64 2.85 2.08 2.62 13.75 16.7 a CMD value, X-class*Cos(CMD), peak and time-integrated EUV, change in Kp, percentage change (DNn) and corresponding observation time of CHAMP, respectively. It is in unit of 10 10 photons/cm 2 s for peak EUV and 10 13 photons/cm 2 for Time-integrated EUV. It is in unit of 10 4 W/m 2 for peak X-ray and peak X-ray*Cos(CMD). The symbol represents no data for the corresponding item. The symbol represents the significant geomagnetic activity. to solar flares, which also represents the location effect on the EUV flux. [16] The dependence of DNn on peak EUV flux is illustrated in Figure 3d, the results show a relative low correlation coefficient (0.65) between the DNn and peak EUV flux, but it is still much better than that between the DNn and peak X-ray flux. Furthermore, Figure 3e shows a much higher correlation coefficient (0.91) between the DNn and time-integrated EUV flux. The time-integrated EUV flux is integrated between the onset and end of a solar flare. Because the large mass and high heat capacity in the thermosphere, the neutral gas response to a solar flare would be sluggish with regard to transient enhancement in EUV and X-ray flux, which typically lasts less than one hour. In other words, it needs longer time for neutral gas to react to the increase in EUV and soft X-ray flux, and the neutral gas response also can last a longer time. Such a character causes that neutral density would increase gradually until to the end of a solar flare. Thus neutral gas responses are related to the total solar EUV coming into the thermosphere. For the comparison, we also show the relation of DNn with the time-integrated X-ray flux in Figure 3c. The correlation coefficient between them is only 0.27, which again suggest that neutral density response mainly comes from the enhancements in EUV flux not that in X-ray flux. [17] Comparing to the slow responses in the thermosphere, the ionosphere responds much faster due to the small time constant of electron and ions. The time and amplitude of peak response is determined by peak solar irradiance [e.g., Afraimovich, 2000; Zhang et al., 2002; Le et al., 2007]. In addition, the simulations based on Global Ionosphere- Thermosphere Model (GITM) and synthetic solar irradiance data during a solar flare [Pawlowski and Ridley, 2011] also suggest that the neutral density response at 400 km altitude is linearly dependent on the total integrated energy throughout the duration of a solar flare above the background level being deposited into the atmosphere. Therefore, the correlation of DNn and time-integrated EUV is better than that of DNn and peak EUV, as shown in Figures 3d 3e. [18] For better analysis of the neutral density response to the 15 solar flares, we list the detailed information for each solar flare in Table 1, including peak X-ray flux, CMD value, modified peak X-ray flux, peak EUV flux, time-integrated EUV flux, DNn, and variation of Kp index. As shown in Table 1, although the X6.2 flare on 12/13/2001 is a central flare with CMD = 9, it has a much smaller time-integrated EUV of 3.35 10 13 photons/cm 2 compared to other flares, which results in a small neutral density response of DNn = 6.12%. Table 1 also shows solar flares on 10/23/2003 (X5.4), 9/7/2005 (X17), 9/8/2005 (X5.4), and 12/5/2006 (X9) have small neutral density responses with DNn < 10%, which is due to the large values of CMD. Even the great X28 solar flare on 11/4/2003 only has a small value of DNn = 9.6%. Actually, because the neutral density response is highly controlled by EUV flux variation, one can come to a conclusion that the dependence of neutral density response on solar flare location is mainly due to the solar flare location effect on solar EUV enhancements. Le et al. [2011] statistically analyzed the relationship between enhancements in X-ray flux and in EUV flux for solar flares during 1999 2006. The results show a significant flare location effect on EUV enhancements for X-class solar flares. One of the reasons for the flare location effect on solar EUV variation is that solar EUV is produced lower in the solar atmosphere than soft X-ray [Donnelly, 1976]. The more important factor is that many bright chromospheric features in the EUV (e.g., the He II 30.4 nm line) are optically thick and self-absorbed due to the abundance of the lower state of the transition. By contrast, most hot coronal features in the soft X-ray region, typically highly ionized metallic lines, are optically thin, and hence are not subject to limb absorption effects [Chamberlin et al., 2008]. The dependence of ionospheric responses on flare location (CMD value) also comes from the flare location effect on solar EUV enhancements [e.g., Zhang et al., 2002; Tsurutani et al., 2005; Mahajan et al., 2010; Qian et al., 2010]. 4. Conclusion and Remarks [19] On the basis of high sensitivity accelerometers on board the CHAMP satellite and GRACE satellite, several studies [e.g., Sutton et al., 2006; Liu et al., 2007] show some 6of8

obvious enhancements in neutral density during the two great solar flares on 10/28/2003 and 11/04/2003. However, other flare events have not been studied in detail. In this study, neutral density data derived from observations made by the CHAMP and GRACE satellites have been used to investigate the neutral density response to solar flares for all the X-class solar flares during 2001 2006. [20] The observed results show the density response for X1 5 flares falls within the noise level in the thermosphere but there is significant neutral density response for X5+ flares. Then we carried out an analysis of the X5+ events to study the dependence of neutral density response on peak X-ray, modified peak X-ray with CMD, peak EUV, and timeintegrated EUV, respectively. The results show a very poor correlation coefficient of 0.3 between DNn and peak X-ray, but a much improved correlation coefficient of 0.85 between DNn and modified peak X-ray with CMD, which means that same as the ionosphere, the thermospheric response to a solar flare also can be affected by solar flare location on the disk. This flare location effect mainly comes from the dependence of solar EUV flux on solar flare location, because the neutral density enhancement is highly related to the EUV variation. The correlation coefficient between DNn and peak EUV is about 0.65. Furthermore, it reaches about as high as 0.91 for time-integrated EUV. The results show the neutral density response is highly related to the variation in EUV flux during a solar flare and its magnitude is mainly determined by time-integrated EUV flux. [21] Our results show that with the exception of the great solar flares like (X17.2 October 28 2003 and X28 November 4 2003), the moderate X-class flares (X5+) also has a nonignorable effect on the thermosphere, with an average enhancement of 10 13% in neutral density at mid-low latitudes within about 4 h after solar flare onset. In addition, changes in the density and composition of the thermosphere can significantly affect the ionospheric density [Fuller-Rowell et al., 1994; Pawlowski and Ridley, 2009]. Tsurutani et al. [2005] reported that there is a long duration ionospheric response with 3 h enhancements in TEC during X17.2 October 28 2003 solar flare, which might partly be due to the long time variation in neutral density. The study on coupling between the ionosphere and thermosphere during a solar flare can be done by more synchronous observation for the ionosphere and thermosphere, as well as by modeling. Pawlowski and Ridley [2009] modeled the ionospheric responses to the solar flare on October 28 2003 by using a coupled ionosphere-thermosphere model; the simulated results show that the ionospheric responses can be affected due to the thermospheric perturbations during the solar flare. [22] Acknowledgments. The authors acknowledge the SEM/SOHO team to providing EUV flux data at their Web site. SOHO is a project of international cooperation between ESA and NASA. The neutral density data are provided by the University of Colorado CHAMP-GRACE Team. The raw data was provided by the CHAMP Information System and Data Center at GeoForschungsZentrum (GFZ), Postdam, Germany. This research was supported by the National Natural Science Foundation of China (41004069, 41074112, 41174137), the National important Basic Research Project (2012CB825604 and 2011CB811405), and the Specialized Research Fund for State Key Laboratories, and by the CMA grant GYHY201106011. [23] Philippa Browning thanks the reviewers for their assistance in evaluating this paper. References Afraimovich, E. L. (2000), GPS global detection of the ionospheric response to solar flares, Radio Sci., 35(6), 1417 1424, doi:10.1029/ 2000RS002340. Bruinsma, S., J. Forbes, R. Nerem, and X. Zhang (2006), Thermosphere density response to the 20 21 November 2003 solar and geomagnetic storm from CHAMP and GRACE accelerometer data, J. Geophys. Res., 111, A06303, doi:10.1029/2005ja011284. Chamberlin, P. C., T. N. Woods, and F. G. Eparvier (2007), Flare Irradiance Spectral Model (FISM): Daily component algorithms and results, Space Weather, 5, S07005, doi:10.1029/2007sw000316. Chamberlin, P. C., T. N. Woods, and F. G. Eparvier (2008), Flare Irradiance Spectral Model (FISM): Flare component algorithms and results, Space Weather, 6, S05001, doi:10.1029/2007sw000372. Donnelly, R. F. (1976), Empirical models of solar flare X-ray and EUV emission for use in studying their E and F region effect, J. Geophys. Res., 81, 4745 4753, doi:10.1029/ja081i025p04745. Forbes, J. M., G. Lu, S. Bruinsma, S. Nerem, and X. Zhang (2005), Thermosphere density variations due to the 15 24 April 2002 solar events from CHAMP/STAR accelerometer measurements, J. Geophys. Res., 110, A12S27, doi:10.1029/2004ja010856. Fuller-Rowell, T. J., M. V. Codrescu, R. J. Moffett, and S. Quegan (1994), Response of the thermosphere and ionosphere to geomagnetic storms, J. Geophys. Res., 99, 3893 3914, doi:10.1029/93ja02015. Le, H., L. Liu, B. Chen, J. Lei, X. Yue, and W. Wan (2007), Modeling the responses of the middle latitude ionosphere to solar flares, J. Atmos. Sol. Terr. Phys., 69, 1587 1598, doi:10.1016/j.jastp.2007.06.005. Le, H., L. Liu, H. He, and W. Wan (2011), Statistical analysis of solar EUV and X-ray flux enhancements induced by solar flares and its implication to upper atmosphere, J. Geophys. Res., 116, A11301, doi:10.1029/ 2011JA016704. Lei, J., J. P. Thayer, G. Lu, A. G. Burns, W. Wang, E. K. Sutton, and B. A. Emery (2011), Rapid recovery of thermosphere density during the October 2003 geomagnetic storms, J. Geophys. Res., 116, A03306, doi:10.1029/2010ja016164. Leonovich, L. A., E. L. Afraimovich, E. B. Romanova, and A. V. Taschilin (2002), Estimating the contribution from different ionospheric regions to the TEC response to the solar flares using data from the international GPS network, Ann. Geophys., 20, 1935 1941, doi:10.5194/angeo-20-1935-2002. Liu, H., and H. Lühr (2005), Strong disturbance of the upper thermospheric density due to magnetic storms: CHAMP observations, J. Geophys. Res., 110, A09S29, doi:10.1029/2004ja010908. Liu, H., H. Lühr, S. Watanabe, W. Köhler, and C. Manoj (2007), Contrasting behavior of the thermosphere and ionosphere in response to the 28 October 2003 solar flare, J. Geophys. Res., 112, A07305, doi:10.1029/ 2007JA012313. Liu, H., M. Yamamoto, and H. Lühr (2009), Wave-4 pattern of the equatorial mass density anomaly: A thermospheric signature of tropical deep convection, Geophys. Res. Lett., 36, L18104, doi:10.1029/2009gl039865. Liu, J. Y., C. H. Lin, H. F. Tsai, and Y. A. Liou (2004), Ionospheric solar flare effects monitored by the ground-based GPS receivers: Theory and observation, J. Geophys. Res., 109, A01307, doi:10.1029/2003ja009931. Liu, J. Y., C. H. Lin, Y. I. Chen, Y. C. Lin, T. W. Fang, C. H. Chen, Y. C. Chen, and J. J. Hwang (2006), Solar flare signatures of the ionospheric GPS total electron content, J. Geophys. Res., 111, A05308, doi:10.1029/ 2005JA011306. Liu, L., W. Wan, Y. Chen, and H. Le (2011), Solar activity effects of the ionosphere: A brief review, Chin. Sci. Bull., 56(12), 1202 1211, doi:10.1007/s11434-010-4226-9. Mahajan, K. K., N. K. Lodhi, and A. K. Upadhayaya (2010), Observations of X-ray and EUV fluxes during X-class solar flares and response of upper ionosphere, J. Geophys. Res., 115, A12330, doi:10.1029/ 2010JA015576. Mosher, J. M. (1979), The height structure of solar active regions at X-Ray wavelengths as deduced from OSO8 limb crossing observations, Sol. Phys., 64(1), 109 119, doi:10.1007/bf00151119. Pawlowski, D. J., and A. J. Ridley (2008), Modeling the thermospheric response to solar flares, J. Geophys. Res., 113, A10309, doi:10.1029/ 2008JA013182. Pawlowski, D. J., and A. J. Ridley (2009), Modeling the ionospheric response to the 28 October 2003 solar flare due to coupling with the thermosphere, Radio Sci., 44, RS0A23, doi:10.1029/2008rs004081. Pawlowski, D. J., and A. J. Ridley (2011), The effects of different solar flare characteristics on the global thermosphere, J. Atmos. Sol. Terr. Phys., 73, 1840 1848, doi:10.1016/j.jastp.2011.04.004. Qian, L., A. G. Burns, P. C. Chamberlin, and S. C. Solomon (2010), Flare location on the solar disk: Modeling the thermosphere and ionosphere response, J. Geophys. Res., 115, A09311, doi:10.1029/2009ja015225. 7of8

Qian, L., A. G. Burns, P. C. Chamberlin, and S. C. Solomon (2011), Variability of thermosphere and ionosphere responses to solar flares, J. Geophys. Res., 116, A10309, doi:10.1029/2011ja016777. Samain, D. (1979), Solar continuum data on absolute intensities, center to limb variations and laplace inversion between 1400 and 2100 Å, Astron. Astrophys., 74(2), 225 230. Sutton, E. K., J. M. Forbes, and R. S. Nerem (2005), Global thermospheric neutral density and wind response to the severe 2003 geomagnetic storms from CHAMP accelerometer data, J. Geophys. Res., 110, A09S40, doi:10.1029/2004ja010985. Sutton, E. K., J. M. Forbes, R. S. Nerem, and T. N. Woods (2006), Neutral density response to the solar flares of October and November, 2003, Geophys. Res. Lett., 33, L22101, doi:10.1029/2006gl027737. Tsurutani, B. T., et al. (2005), The October 28, 2003 extreme EUV solar flare and resultant extreme ionospheric effects: Comparison to other Halloween events and the Bastille Day event, Geophys. Res. Lett., 32, L03S09, doi:10.1029/2004gl021475. Wan, W., L. Liu, H. Yuan, B. Ning, and S. Zhang (2005), The GPS measured SITEC caused by the very intense solar flare on July 14, 2000, Adv. Space Res., 36, 2465 2469, doi:10.1016/j.asr.2004.01.027. Woods, T. N., F. G. Eparvier, J. Fontenla, J. Harder, G. Kopp, W. E. McClintock, G. Rottman, B. Smiley, and M. Snow (2004), Solar irradiance variability during the October 2003 solar storm period, Geophys. Res. Lett., 31, L10802, doi:10.1029/2004gl019571. Zhang, D. H., and Z. Xiao (2005), Study of ionospheric response to the 4B flare on 28 October 2003 using international GPS service network data, J. Geophys. Res., 110,, doi:10.1029/2004ja010738. Zhang, D. H., Z. Xiao, and Q. Chang (2002), The correlation of flare s location on solar disc and the sudden increase of total electron content, Chin. Sci. Bull., 47(1), 82 85, doi:10.1360/02tb9017. Zhang, D. H., X. H. Mo, L. Cai, W. Zhang, M. Feng, Y. Q. Hao, and Z. Xiao (2011), Impact factor for the ionospheric total electron content response to solar flare irradiation, J. Geophys. Res., 116, A04311, doi:10.1029/2010ja016089. H. Le, L. Liu, and W. Wan, Beijing National Observatory of Space Environment, Institute of Geology and Geophysics, Chinese Academy of Sciences, Beijing 100029, China. (lehj@mail.iggcas.ac.cn) 8of8