Beside the GPS radio occultation measurements of the troposphere and ionosphere, the CHAMP satellite tracks presently up to 8 GPS satellites simultaneously for navigation. The received dual frequency GPS signals provide valuable information on the ionization state of the topside ionosphere and plasmasphere. This paper briefly describes a new developed model assisted technique to improve the three dimensional electron density estimation above the CHAMP orbit from these GPS measurements. Application of this assimilation algorithm provides a 2-D electron density slice of the ionosphere/plasmasphere system within the CHAMP orbit plane from the CHAMP altitude up to GPS orbit heights. We estimate the accuracy of the GPS TEC calibration and perform a first validation by comparing the retrieved electron density of the ionosphere/plasmasphere slice at CHAMP altitude to simultaneous electron density measurements of the Langmuir Probe onboard CHAMP.
 The German small satellite CHAMP (CHAllenging Minisatellite Payload) (Reigber et al. ) has been launched successfully on 15 July 2000 from Plesetsk, Russia on board of a Cosmos rocket (http://op.gfz-potsdam.de/champ). The main scientific objectives of this geo-science mission are the precise recovery of the Earth's gravity and magnetic field as well as the sounding of the Earth's neutral atmosphere and ionosphere using GPS (Global Positioning System) radio occultation measurements. The CHAMP orbit is almost circular and near polar with an inclination of 87.2° and an initial altitude of 454 km that will decrease continuously during the planned 5 years mission lifetime but will be maintained above 300 km.
 GPS ground based observations are well established as a powerful tool to monitor continuously the actual state of the ionosphere (e.g., Jakowski ). The installation of dual frequency GPS receivers on board of LEO (Low Earth Orbiting) satellites such as CHAMP offers new opportunities of ionospheric sounding on global scale. The GPS/MET (Global Positioning System/Meteorology) experiment, flown on the Microlab-1 LEO satellite, demonstrated the GPS radio occultation technique as a powerful tool for limb sounding of the Earth's ionosphere and neutral atmosphere (e.g., Ware et al. ). The GPS measurements on board CHAMP are performed by the most recent generation of a space qualified GPS receiver (“Black Jack”) developed by the Jet Propulsion Laboratory (JPL, USA). For precise orbit determination CHAMP tracks presently up to 8 GPS satellites simultaneously with a sampling rate of 0.1 Hz, using a dedicated zenith looking antenna. Thus the navigation links are sounding the topside ionosphere and plasmasphere at heights above the CHAMP orbit. This paper briefly describes a newly developed model assisted technique to reconstruct the three dimensional electron density structure above the CHAMP orbit from GPS signal measurements. We discuss preliminary results that provide a reconstruction of the ionosphere/plasmasphere electron density in the CHAMP orbit plane from the CHAMP altitude up to GPS orbit heights. First validation checks with in situ electron density measurements by the Langmuir Probe on board CHAMP are presented.
2. Data Processing and Retrieval Method
2.1. Preprocessing of GPS Data
 For the retrieval algorithm two main inputs are required: GPS low rate (0.1 Hz) navigation measurements from CHAMP and orbit data of CHAMP and the involved GPS satellites. The CHAMP GPS data are received in the DLR Remote Sensing Data Center Neustrelitz and GPS low rate data as well as orbit information are provided by the GFZ. During the preprocessing of the GPS data, outliers and cycle slips are detected and removed or corrected, respectively. Here an algorithm described by Blewitt  is applied. Due to the dispersive propagation properties of the ionosphere for radio signals, differential phases provide the integrated electron density along the signals ray path. To calculate this Total Electron Content (TEC) we use the ionospheric combination of GPS pseudoranges (P2 – P1) and carrier phases (L1 – L2) (Blewitt ) and work finally with pseudorange–leveled carrier phase differences. The so derived TEC has to be calibrated for the receiver and satellite differential group delay biases. Details on the estimation of TEC as well as ground receiver and satellite biases are given e.g. by Sardon et al. .
2.2. Calibration of Link Related TEC
 To calibrate the numerous link related TEC measurements for the instrumental biases, a model assisted technique has been developed. For assistance, the Parameterized Ionospheric Model (PIM) (Daniell et al. ) is used, that also includes the Gallagher model (Gallagher et al. ) of the plasmasphere. While the CHAMP receiver bias is not known, GPS differential code biases are estimated in a reliable manner by different GPS processing centers. A similar bias solution is regularly processed at DLR IKN Neustrelitz as well (Sardon et al. ). Therefore the calibration procedure is focused on the estimation of the CHAMP receiver bias. We assume that the assisting model meets the real ionospheric/plasmaspheric electron density distribution on average over several days and various geographic regions. Thus, the CHAMP receiver bias is estimated by the difference of the TEC measurements of CHAMP and the corresponding modeled TEC values. The average difference is calculated for nighttime intervals, considering only TEC measurements recorded in middle and high latitudes under an elevation of at least 50° with respect to the zenith antenna. These restrictions promise small TEC values because of a low ionization level and a comparatively short ray path through the ionosphere and plasmasphere, and therefore small calibration errors. Figure 1 shows the estimated CHAMP bias in TEC units from day 110 till day 230 of 2001, using GPS measurements from 5 consecutive days as a running mean. Because of a stable temperature handling on board CHAMP, which is most important for the avoidance of short term receiver bias fluctuations, it is possible to estimate reliable bias means over several days. As Figure 1 shows, the estimated CHAMP bias is quite stable over a considerable time period varying around −20 TEC Units (1 TECU = 1016m−2) by at most 1 TECU. It has to be mentioned that in particular severe large scale disturbances of the ionospheric and plasmaspheric state could violate the main assumption of this calibration technique resulting in fluctuations of the estimated receiver bias. Furthermore, our approach may hide a bias resulting from a continuous mismodelling of the PIM regarding to the topside ionosphere and plasmasphere. But this will be kept small because of the selection criteria discussed above.
2.3. Assimilation of Calibrated TEC
 In order to derive local electron density information from the integral TEC measurements we have to solve an inverse problem. To do this the calibrated link related TEC data derived for each full CHAMP revolution are assimilated into the PIM. Depending on the data quality up to 4000 radio links are available for such a period. A full revolution takes 93 minutes, so the ionospheric/plasmaspheric system is considered to be stationary over this period. To assimilate the link related TEC data into an ionospheric/plasmaspheric model, a global 3-dimensional voxel structure has been constructed. Due to the particular importance of the Earth's magnetic field for the plasmaspheric electron density distribution, this structure is defined in dipolar magnetic coordinates and its shape follows the natural (undisturbed) electron density distribution in the ionospheric/plasmaspheric system. The vertical resolution amounts to 10 km at lower end near the CHAMP orbit and extends up to 1000 km at GPS orbit height, while the horizontal resolution is 2.5 × 5 degree in latitude and longitude, respectively.
 At the beginning of the assimilation process, PIM is used to initialize the described voxel structure. In the following an iterative process is carried out that modifies the electron density inside the voxels crossed by the CHAMP - GPS radio links to meet the link related TEC measurements. Going to the next iteration step the electron density values inside the crossed voxels are multiplied by the ratio between the TEC measurement and the voxel-TEC (integration of voxel electron density along the radio link). This procedure resembles the well known Multiplicative Algebraic Reconstruction Technique (MART, e.g., Gordon et al. ). After this iteration we have already considered the complete information provided by the input data, but the resulting local electron density information is still restricted only to voxels crossed by radio links. In order to get a coherent 3-dimensional electron density distribution, each crossed voxel obtains an influence on its environment depending on distance and location. Three Gaussian functions are used to determine the amount of this influence. Of course, the result of this procedure will not be a detailed correct reproduction of the real state of the topside ionosphere and plasmasphere. However, depending on data coverage the result is an improved model output. Independent on data coverage the result is always stable and physical reasonable. It should be mentioned that the available integral TEC information is not sufficient to get a unique solution of this inverse problem. Therefore the reconstructed electron density distribution is significantly dependent on the data coverage, the geometrical situation of the link related TEC measurements and the quality of the input model.
3. Results and Discussion
 The result of the assimilation process is a 3-dimensional electron density distribution. But obviously the data coverage and the influence of the assimilation process have their maximum inside and near the CHAMP orbit plane, while other areas of the voxel structure remain completely unaffected and give therefore a pure model output. Thus, the probability of a satisfying assimilation result that meets the real electron density distribution is highest if we consider only a 2-dimensional extract of the assimilated voxel structure along the orbit plane. The left panel of Figure 2 gives an impression of the reconstructed electron density distribution in the orbit plane (from around 00:00 LT in the East to 12:00 LT in the West) obtained after a selected assimilation process. The difference between the day- and nighttime ionosphere and plasmasphere is clearly visible. The right panel of Figure 2 shows the percentage deviation of the assimilation result from the model output used for initialization. A remarkable and quite strong deviation of the assimilation output from the model input occurs in the area of the modeled plasmapause. This could be considered as a correction of the PIM plasmapause position by the assimilation process.
 To get information on the data coverage of the voxel structure it is useful to know how many link related TEC measurements are influencing the assimilation result for a particular voxel. In order to calculate this so called link density it is possible to process the voxel structure initialized by the number of crossing links for every voxel in a similar manner as for the electron density. Figure 3 shows a link density plot corresponding to the assimilation output in the orbit plane shown in Figure 2. The missing data coverage above the polar regions is due to the 55° inclination of the GPS orbit planes. It is interesting to note that Figure 3 shows high link densities not only near the satellite path but also near the GPS orbit height. Furthermore, it should be mentioned that a comparison of Figures 2 and 3 indicates no correlation between the relative changes of the initializing model and the link density.
 Due to the shortcoming of independent ionospheric data concerned with the topside ionosphere and plasmasphere, the validation of the derived electron density distribution is a rather difficult problem. On the other hand, the in situ plasma density measurements of the Planar Langmuir Probe (PLP) on board CHAMP offer a unique opportunity to validate the electron density derived by assimilation along the satellite path. The PLP is part of the Digital Ion Drift Meter and was provided by the US Air Force Research Laboratory (AFRL), Hanscom, MA. Of course, it is not possible to validate the entire assimilation result using the PLP data. However, systematic errors in the derived electron density along the satellite path would clearly indicate deficiencies of the assimilation technique resulting in errors of the derived electron density above the orbit. The left panel of Figure 4 shows a scatter plot of the electron density derived from assimilation versus corresponding PLP measurements from 2 to 9 August of 2001. On the other hand, the right panel shows the corresponding scatter plot where the assimilation data are replaced by the initial model. The number of samples amounts to 64678 for this period. As can be seen, the correlation between assimilation and PLP is quite consistent over a large dynamic range of more than two decades. It is clearly visible how the assimilation process changes and improves the initial model assumption of the electron density along the satellite path. The assimilation data coincide with PLP electron density values within an RMS spread of 20.3 × 1010m−3 (34.8% mean relative deviation) whereas initial model values agree only within 27.4 × 1010m−3 (90.5%). It should be noted that there is no significant bias between the PLP measurements and the assimilation result.
 Before discussing geophysical issues of the derived plasmaspheric electron density distributions, further validations have to be done using extended data sources such as plasmasphere measurements from the IMAGE satellite mission or incoherent scatter radar data.
 We have reported first experiences with a new assimilation technique for the reconstruction of ionospheric/plasmaspheric structures using GPS satellite-to-satellite tracking measurements on board CHAMP. Preliminary results, providing the first time topside ionosphere/plasmasphere slice by means of CHAMP GPS data, have been presented and discussed. The processing algorithm has been described briefly. First validation checks with independent in situ Langmuir Probe measurements on board CHAMP are promising and show the fundamental suitability of the presented technique for the inversion of integrated TEC measurements into adequate electron density information. Nevertheless, the validation of the derived electron density distribution needs further work. Depending on the reliability of the CHAMP topside data a statistical analysis could give for instance the 3-D shape of the plasmasphere in the future. It should be mentioned that additional space- and/or ground based TEC measurements as well as in situ electron density measurements can easily be included in the described assimilation process. It can be expected that the combined assimilation of GPS navigation measurements aboard satellite missions such as CHAMP, SAC-C, GRACE, COSMIC and further planned LEO missions will enable a permanent global monitoring of the topside ionosphere/plasmasphere system. This could essentially contribute to a global space weather monitoring system and would provide a valuable database for the improvement of ionospheric/plasmaspheric models.
 The authors are very grateful to all colleagues of the CHAMP team and from the payload providers JPL and AFRL. This study was carried out under grant number 01SF9922/2 of the German Federal Ministry of Education and Research (BMBF).