To Estimate The Regional Ionospheric TEC From GEONET Observation Jinsong Ping(Email: jsping@miz.nao.ac.jp) 1,2, Nobuyuki Kawano 2,3, Mamoru Sekido 4 1. Dept. Astronomy, Beijing Normal University, Haidian, Beijing 100875, China 2. Dept. Astronomy, the Graduate University for Advanced Studies, Japan 3. Earth Rotation Division, National Astronomical Observatory, Japan 4. Communications Research Laboratory, Japan [Abstract] Dual frequency phase observations of Global Positioning System (GPS) prove a possibility to detect the vertical total electron content (TEC) with high precision above a super-dense network of GPS receivers. A method is discussed to estimate a regional TEC distribution over Japanese Islands from the data obtained by the GPS Earth Observation Network (GEONET), which is belong to the Geographical Survey Institute (GSI) of Japan. Using the GPS phase observations of GEONET on August 03, 1999, a daily TEC distribution map has been obtained. It is compared with the global TEC model published by IGS. Keywords: Ionospheric TEC, GPS, GEONET Introduction In the ground based satellite radio tracking observation, the time delay or time advance for the group or phase link signal, respectively, is proportional to the TEC along the path, and to the inverse of the central frequency squared. The primary purpose of the second frequency in GPS is to neutralize the effect of the ionosphere on signal propagation [1]. At the same time, dual frequency phase observations of GPS also provide a possibility to obtain this kind of time delay or time advance. From it, the information of TEC can be retrieved. Lanyi and Roth [2] have assumed that the electronic distribution lies in a single layer thin shell at a fixed altitude above the surface of the Earth, and modeled it by a two-dimensional polynomial function of shell angular position. Wilson et al. [3] have extended this thin spherical shell fitting technique to multi-site GPS data sets using two-dimensional spherical harmonics as a global basis: TEC( l, s) nmax n = P (sin l )( a cos( ms) a nm a nm + n= 0 m= 0 b nm sin( ms)). (1) where n max is the maximum degree of the spherical harmonic expansion; P nm (sinl a ) are the normalized associated Legendre functions of degree n and order m; a nm and b nm are the unknown TEC coefficients of the spherical harmonics, i.e. the global or regional ionospheric TEC model parameters to be estimated; l a is the geocentric latitude of the intersection point of the line receiver-satellite with the ionospheric layer and s is the sun-fixed longitude of the ionospheric pierce point [4], which is related to the local solar time (LT) according to s = LT - 12h UT + l o - 12h. (2) UT is Universal Time and l o denotes the geographical longitude of the sub-ionospheric point. The Equation (1) is well suited for a regional or a global TEC model. It has been adopted to estimate the daily and subdaily TEC model based on the global GPS surveying data set of IGS, and on the regional GPS surveying data sets obtained in North America and in Europe [5-7], respectively.
At here, the possibility is firstly studied to retrieve the regional ionospheric TEC distribution in the mid-latitude area using the data sets obtained by the GPS receivers of GEONET in Japan. In this area, VLBI, satellite tracking and GPS observation have also been done at single frequency. One of the motivations and targets of the research in this article is to estimate a precise TEC model so as to calibrate the effect due to the ionosphere on the single frequency space technique observations in this area. Ionospheric Regional Model Figure 1. The distribution of the GPS receivers in GEONET Project To obtain the ionospheric TEC model defined by Equation (1), the TEC measurements in the direction of all GPS satellites in view from a ground GPS network are scaled to the vertical direction. These vertical TEC observed during a time interval are attributed to the intercept point of the slant path with an thin shell at an altitude of 250~450 km. Then, they are fitted by Equation (1) in a Sun-fixed, i.e. geomagnetic, reference frame, because the ionospheric pattern rotates with the Sun. In the published global standard TEC subdaily model of IGS, a height of 450 km is adopted for this thin ionospheric shell. In Asian area, the GSI of Japan launched a national GPS surveying network on the Japanese Islands from 1994. The total number of dual-frequency receivers is larger than 1000 in 1999. The average separation between the fixed sites is about 25 km. The positions of them available on August 03 of 1999 are shown in Figure 1. Although the primary target of GEONET Project is a monitoring of crustal deformation [8], this dense array has provided useful information on meteorological study [9] and on ionospheric research [10]. The phase observable of two carrier waives obtained by GEONET from August 01-03 have been processed. The sampling rate of each GPS receiver is 30 seconds. The data are separated into 3 sessions, 24 hours for each. Because there is not software of GPS data analysis can handling a network with several hundreds receivers simultaneously, the data from GEONET have
been rearranged into 8 sub-networks, each has 100-200 receivers, which can cover all of the Japanese Islands. As a example, one of the sub-networks has been shown in Figure 1 by using solid squares. After removing the integer periodic ambiguities in phase measurements by using the measured pseudoranges, the coordinates of the GPS sites are resolved in ITRF97 with RMS of 2-15 mm in each direction, where the fixed site is Tsukuba IGS GPS site (TSKB). The new coordinates are fixed and introduced into an iterative process to estimate the daily average regional TEC model by using each sub-network. At last, an average model is obtained from the 8 sub-networks for each session. (a) (b) Figure 2. TEC daily model (a) and errors (b) on August 03, 1999 from GEONET GPS Surveying Figure 3a. TEC subdaily models on August 03, 1999 from GSI GPS Surveying
Figure 3b. Errors of TEC subdaily models on August 03, 1999 from GSI GPS Surveying In the data analysis, different options and parameters have been tested. The key ones are mean altitude of the TEC shell, the maximum degrees and orders of the spherical harmonics, the re-sampling rate for the phase data, and the cut-off elevation angle of the observable. The heights from 250 km through 500 km have been tested, 300 km through 400 km can give almost the same result, the lower or higher value show considerable biases. The result from adopted 350 km is given in this article. Usually, a harmonics of 12 orders and 8 degrees have been used in global TEC model. In our study, due to the limitation of the observations, no reasonable result can be obtained for the regional model from GEONET observation by using a harmonics of orders or degrees larger than 3. A 2x2 model has been adopted. Different re-sampling rates of the data have been tested, the minimum of 30 seconds is recommended because the data will be averaged or smoothed seriously by long sampling interval, and more information will be lost. The cut-off elevation angle is 15 o for the observation. When using this value to estimate the TEC model, an circle area with diameter of about 2000 km can be covered for each site at given epoch. The data of low elevation angle will introduce large error into the estimation. Different cut-off elevation angles from 15 o through 45 o have been tested. Of them, 20 o -30 o are reasonable for introducing small model biases and losing less information. In the Sun-fixed geomagnetic reference frame, the daily average TEC distribution in the zonal range of 8 o ~40 o in latitude has been obtained for each session. The pattern of the daily regional average TEC model of the three sessions are similar. Only the contour map of the model on August 03, 1999 is shown in Figure 2, together with the error gray map. The RMS is smaller than 0.4 TECU.
Comparison and discussion The gray map of global subdaily TEC models and the TEC errors of the same day are shown in Figure 3a and 3b, respectively. In the same zonal range, the average RMS is about 1~2.0 TECU. The maximum difference between the regional model and the global model for August 03, 1999 is ~15 TECU. The average difference between them is about 10~12%. It is similar to the result of differences between a global model and a local model given by Wilson et al. [3]. The TEC distribution has been smoothed seriously in a global model. From this point, it is not easy to say the regional model is more reasonable. A simple way can be used to compare and check them between each other. It is to introduce thesa models into the data analysis processing, and using the double differencing single frequency phase L1 data to re-estimate the site position, separately; then, from the results, compare the baseline vertical/horizontal RMS vs. length with each other. Figure 4. Daily baseline vertical RMS vs. length for the 3 solution types A sub-network of 22 sites have been adopted in the data analysis of comparison. It covers the Japanese Island with unit distribution, TSKB is in the middle area of this sub-network. The baselines are defined by combining each other site with TSKB. The lengths vary from 100 km through 1100 km. Firstly, the ionosphere-free linear combination phase observable, L3, are used to estimate the site positions. Then, the single frequency phase data, L1, are used, where the global and regional TEC model have been introduced separately to calibrate the effect due to ionospheric time advance. In all of the three analysis, the tropospheric delays in the zenith direction of each site have been estimated per three hours. A simple mapping function has been adopted. The CPU time consuming is about 2:10:1 for the data processing by using L3, L1 + global TEC mode and L1 + regional TEC model. The results of RMS in vertical direction are given in Figure 4. In Figure 4 it can be seen that L1+regional TEC model and L1 + global TEC model give almost the same results, however, the result from regional TEC model shows insignificantly better. It is proved that an average daily regional ionospheric TEC distribution model can be obtained from GSI/GEONET observation, which is comparable to the IGS global TEC model. The results from L1 is about three times worse than the results from L3. Usually, if the
baseline length is shorter than 10 km, the baseline vertical RMS is comparable for different solution types of L1 single frequency data and of L3 ionosphere-free combination data in the daily resolution. This conclusion can be kept till several tens kilometer of the baseline lengths if a high resolution ionospheric modeling technique is applied in a local small area[11]. Beyond this baseline length, although a global/regional TEC model is good enough for ionospheric research, the single frequency observation with these models cannot give comparable RMS to the result from dual frequency observation. In order to calibrate the ionospheric effects on the single frequency GPS, VLBI or Space Tracking observation at local area, the global/regional model need to be compared with the local TEC model in some detail so as to test the effectiveness of them. Our next work will focus on developing a local TEC model and compare it with the regional/global models. The effort of estimating a regional TEC model over Japanese Island is limited mainly by the distribution of GSI/GEONET in a small narrow range of Japanese Islands. To estimate an improved regional TEC model for eastern Asia area, data obtained by other GPS net-works from neighbor areas should be included in the future research. Acknowledgement: The RINEX GPS data is obtained by GEONET, GSI of Japan. Without the help from Dr. T. Iwabuchi of Meteorological Research Institute of Japan, this work cannot finished smoothly. The author received the financial supports from National Astronomical Observatory of Japan and from the APSG to take part in this workshop. References [1] Leick, A. in GPS Satellite Surveying 2nd ed., New York, 1995, 294-305. [2] Lanyi, G, Roth, T. A comparison of mapped and measured total ionospheric electron content using global positioning system and beacon satellite observations, Radio Sci., 1988, 23:483-492. [3] Wilson, B D, Mannucci, A J, Edwards, C D. Subdaily northern hemisphere ionospheric maps using an extensive network of GPS receivers. Radio Sci.,1995, 30:639-648. [4] Rothacher, M, Mervart, L. in Bernese GPS Software Version 4.2, Univ. Berne, August 2000, 186-205. [5] Schaer, S, Beutler, G, Mervart, L, Rothacher, M, Wild, U. Global and Regional Ionosphere Models Using the GPS Double Difference Phase Observable, in IGS Workshop Proceedings on Special Topics and New Directions, eds. by G. Gendt and G. Dick, 77-92, GeoForschungsZentrum, Potsdam, Germany, May 15-18, 1995. [6] http://www.aiub.unibe.ch/ionosphere.html. [7] http://iono.jpl.nasa.gov/. [8] Tsuji, H, Hatanaka, Y, Sagiya, T, Hashimoto, M. Coseismic crustal deformation from the 1994 Hokkaido-Toho-Oki earthquake monitored by a national wide continuous GPS array in Japan, Geophys. Res. Let., 1995, 22:1969-1972. [9] Iwabuchi,T, Naito, I. A comparison of Global Positioning System retrieved precipitable water vapor with the numerical weather prediction analysis data over Japanese Islands, J. Geophys. Res., 2000, 105:4573-4585. Saito, A, Fukao, S. High resolution mapping of TEC perturbations with the GSI GPS network over Japan, Geophys. Res. Let., 1998, 25:3079-3082. Rocken, C, Johnson, J M, Braun, J J, Kawawa, H, Hatanaka, Y, Imakiire, T. Improving GPS surveying with modeled ionospheric corrections, Geophys. Res. Let., 2000, 27:3821-3824.