A Tomographic Study of Seasonal Variations of the Equatorial Anomaly in the Asian Sector

The Low-latitude Ionospheric Tomography Network (LITN) is used to investigate the seasonal behavior of the equatorial anomaly in the Asian Sector during a period of solar minimum. Seasonal variations of the loca­ tion, the time of occurrence, and the height and the electron density of the maximum peak of the anomaly are studied. The starting time for the devel­ opment of the anomaly and the lifetime of the whole phenomenon are also investigated. Extension of the work for possible 4D monitoring of the iono­ sphere using COSMIC data is discussed. (

Figure 1 shows the geometry of the network and the radio rays emitted from the satellite and received at each station for one satellite path.The network was used successfully to image the ionosphere during the 24, October 1995 total eclipse occurred in this region (Huang et al., 1999).Between 1994 and 1997 when the NNSS signal was turned off, LITN was in routine operation.The data has been used to study a variety of ionospheric phenomena.In this paper, we shall concentrate on the seasonal variations of the ionosphere in the anomaly region.Cheng et al. (1992) and Huang and Cheng (1993) used the TEC data from a single station to investi gate this problem.However, because of the sharp latitudinal variation of the ionosphere in this region, a multi-station tomographic technique such as LITN is more suitable for following the details of the variations.TEC data from LITN for the two-year period 1995-96 are used in the paper to reconstruct the ionosphere in the anomaly region in different seasons.From the re constructed ionospheres, the behavior of the anomaly is investigated.In section 2, the comput erized ionospheric tomography (CIT) technique is reviewed and some typical reconstructed images using LITN data are presented.The reconstructed ionospheres for different months are examined in section 3.These images are used to investigate the seasonal variations of the anomaly in section 4. In section 5, some conclusions are given and the possibility of 4D monitor ing of the ionosphere using data from COSMIC is discussed.

COMPUTERIZED IONOSPHERE TOMOGRAPHY (CIT)
Figurel shows a typical two-dimensional CIT geometry.Along any of the ray path con necting the satellite and the receiving station, the TEC is given by (1) where C is the TEC, N" is the electron density and pis the propagation path.For tomographic applications, Eq.( 1) is approximated by the discrete sum where C. represents the TEC along the ith path, a .. represents the length of the ith ray path in the jth pixel, and x. is the election densi ty in the jth pixel.Different algorithms have been developed to solve Eq.( 2) to obtain the electron density distribution.Raymund et al. (1990) have shown the MART algorithm in de t ail, including the inversion procedure, initial guess and em pirical model employed for the reconstruction in derail.For the LITN data reported in this paper the modified MART is used.Combining the estimated lat itudinal distribu tion of the vertical TEC with ionosonde data, model inos phere, and Chapman ionospheric height distri bution generates the initial guess required to start the iteration of the MART algorithm.This a priori information puts constraints in the reconstruction procedure on the peak electron den sity, th e topside ionosphere distribution, and ionospheric heights.The details of the technique have been described by Huang et al. ( 1997Huang et al. ( , 1999)).Figure 2 shows an example of the recon struction using LITN geometry.We note that the technique is capable of reconstruc ting the major features of the ionosphere in the anomaly region.

OBSERVATION AND RECONSTRUCTED N IMAGES
" TEC data from NNSS at the LITN stations were analyzed for the two-year period 94-96.
Reconstructions were carried out for cases where at least three stations recorded data of good quality.Figure 3 shows a typical example of vertical TEC distributions derived from LITN data.The passes at different times of the day provided us with a picture of the anomaly as it developed during the day on 8, March 1995.The anomaly started to develop before 09:26 LT.
At that time, the peak was at 10°N.It began to move northward and grew in magnitude until it reached the highest peak value of 5.52 X 1017 electron/m2 at 13 :26 LT with the peak located at 24 °N.After that, the peak moved south with decreasing magnitude until the anomaly disap peared altogether at 21: 17 LT.Figure 4 shows examples of the tomographic images of the electron density reconstructed from the LITN data for several sample days at the time when the maximum peak of the anomaly occurred.From images like these, we can derive the posi tion, the height and the electron density of the maximum peak of the anomaly.The informa tion can then be used to study the seasonal variation of the anomaly when data from different months were used for reconstruction.

SEASONAL VARIATION OF THE EQUATORIAL ANO MALLY IN THE ASIAN SECTOR
The data were taken from August 1994 to July 1996.During this period, the 12-month running average sunspot number ranged between 30 in 1994 to 10 in 1996.Therefore, it was a period when solar activity moved into its minimum.To investigate the seasonal behavior, we took five quiet days with the smallest k indices ( <3) in each month and used them to represent p the average behavior of the month.The appearance of the anomaly is governed by the F-region dynamic proce sses, which are related to the mo tions and energetic s of both the ionosphere and the thermosphere.It is ex pected that the position where the maximum peak electron density occur s and the time when the den sity grow s to its peak value be clo sely related to these processes.Thus the beha vior of the maximum peak may carry important information that can be used to study these dynamic proce s ses.Our data provided a good opportunity to inve stigate the monthly behavior of th e maximum peak.Figures Sa and Sb show the monthly distribution of the po sition of the maxi mum peak and its appearance time for the period from August 1994 to July 1996.The horizon tal and vertical lines in Figs.Sa and Sb show the median values of latitude and the time of occurrence of the maximum crest in the five quiet days for each month.Figure 6a shows that for the two-year period studied, the peak appeared earlier during winter and summer as com pared to the cor res ponding times du ring spring and autumn.The difference is even more pro nounced during the solar minimum period 1995-96, as shown in Fig. 6b.There wa s also dis-cernible asymmetry as the appearance time was pushed back further to around 18L T during autumnal equinox as compared to around 16LT during vernal equinox.A similar type of asym metry was observed by Cheng (l 992).We next consider the position of the peak shown in Figs.7a and 7b.The peak moved to a higher latitude during spring and autumn.There was also an asymmetry as the peak moved further north during the vernal equinoctial month.It is also interesting to note the monthly variations of the magnitude of the TEC value shown in Fig. 8. Similar re sults and some possible mechanisms responsible for this asymmetry have been re ported by Huang et al. (1989).We summarize the seasonal behavior of the maximum peak as follows: the maximum peak of the anomaly occurred later at local time at higher latitudes with higher TEC value s during spring and autumn.There was an asymmetry in the behavior of the time and location of the peak between the autumn and vernal equinoxes.However, the no asym metry seems for the magnitude of the TEC to exist.

Seasonal Variations of other Phases of the Anomaly
In addition to the maximum peak of the anomaly, other phases of the phenomenon as it de velops during its whole life cyc le are also of interest.Figures 9a and 9b show the monthly average TEC contours for the periods 1994-95 and 1995-96, respectively.We note that there  --,, , �,  ?O <-----t--------

U996
]1] was hardly any seasonal variation of TEC between midnight and sunrise.The starting time for the development of the anomaly was around 08:00 LT and did not show any clear seasonal variation either.The anomaly disappeared between1 8:00 and 20:00LT in winter, at around 18:00LT in summer, and between 20:00 and 23:00LT in spring and autumn.Thus the lifetime of the anomaly is longer in spring and autumn, but shortest during summer.This may be caused mainly by the neutral wind blowing from the summer hemisphere to the winter hemi sphere, which could inhibit the diffusion of plasma along field lines in the summer hemisphere (Rishbeth, 1972).It seems that seasonal anomaly of ionospheric electron density and seasonal variations of the electronic field strength have minor effects.

The Height and the Electron Density of the Maximum Peak
The reconstructed 2D images of the ionosphere for the time when the maximum peak occurred as shown in Fig. 4 can be used to study the vertical position of the maximum anomaly peak and the electron density at the peak.The results are shown in Fig. 10.The height of the maximum peak was found to be highest in summer, at around 410km, and lowest in winter, at around 300km.The electron den sity for the maximum peak was found to be largest during the equinox months and smallest during the month of the summer solstice.There seems to be a      reverse correlation between the height and the electron density of the maximum peak.

DISCUSSIONS AND CONCLUSION
In this paper, we have used the data from LITN to study the seasonal variations of the equatorial anomaly in the Asian Sector.We have shown that the 20 tomographic images of the reconstructed ionosphere allow us to study seasonal variations in the time of occurrence, the height, the latitudinal location and the strength of the maximum peak, as well as the life time of the anomaly.These results are very useful in the effort to model more accurately this 1 � i;; :a: rather active region of the ionosphere, a task to be carried out in the near future.However, because of the limitation in the number of usable passes per day and the 2D nature in the reconstruction geometry, our results did not have very satisfactory temporal resolution and totally neglected the longitudinal variations.This situation may be significantly improved, however, with the launching of COSMIC.Data from the on-board GPS receivers and the network of ground-based stations for the Tri-Band Beacon experiment can be used for 40 ( r, t) tomographic reconstruction of the ionosphere.With these facilities in place, and the tech nique developed and experience gained in our current investigation, a real time monitoring of the equatorial anomaly and even the whole ionosphere, may become a reality.
Fig.I.The tomography geometry for LITN shows ray paths from satellite to ground sation as a NNSS satellite transverses the image region over re ceiving station.

Fig. 2 .
Fig. 2. (a) Ionospheric structure of IRl90 model used for the simulation.(b)Reconstructed image using the modified MART algorithm.
Fig.3.A typical example of vertical TEC distributions at different local time derived from LJTN data.

Fig. 4 .
Fig.4.The tomographic image s of the el ectron desity recon structed from LITN data for several sample days at the time when the maximum peak of the anomaly occurred.

Fig. 5b .
Fig. 5b.The monthly distribution of the maximum peak appearance time and position for the 12 months period (95 -96).

Fig. 6b .
Fig. 6b.The monthly variation of the maximum peak with local time during the 95-96 period.

Fig. 7a .
Fig. 7a.The two-year average monthly variation o f the peak with latutude on LITN.