The paper deals with initial analyzes of radio occultation measurements of the ionosphere carried out on board the CHAMP satellite since 11 April 2001. The accuracy of the operationally retrieved electron density profiles has been estimated by comparing with independent measurements. The derived ionospheric key parameters such as f0F2 and hmF2 agree with a standard deviation of 18 and 13%, respectively. It is shown that the CHAMP data products can essentially contribute to the establishment of operational data sets of the global electron density distribution for developing and improving global ionospheric models and to provide operational space weather information.
 LEO missions such as CHAMP carrying a dual frequency GPS receiver onboard, offer a unique opportunity to improve our knowledge of the ionospheric behavior and to monitor the actual state of the ionosphere on a continuous basis. The GPS limb sounding technique has been demonstrated to be a powerful tool for remote sensing the Earth's neutral atmosphere and ionosphere by analyzing GPS radio occultation data obtained from the GPS/MET instrument, flown on the Microlab-1 LEO satellite (e.g. Ware et al. ). The German CHAMP satellite (Reigber et al. ) was successfully launched from the Russian launch site Plesetsk by a COSMOS rocket on July 15, 2000. The main scientific objectives of this geo-science mission are based on precise gravity and magnetic field measurements and atmosphere/ionosphere sounding data by using GPS radio occultation techniques. The GPS radio occultation measurements are performed by the most recent generation of a space qualified GPS receiver (“Black Jack”) especially developed by the Jet Propulsion Laboratory/USA for such purposes. The GPS signals are received by a RHCP helix antenna (10 dB) mounted on the 21° inclined backward side of CHAMP for a high elevation recording of setting occultations. First neutral gas limb sounding measurements onboard CHAMP starting on February 11, 2001 were reported recently by Wickert et al. . Two months later, on 11 April, the first ionospheric radio occultation measurements have been carried out on board CHAMP. This paper focuses on the preliminary analysis of data measured between April 11 (day 101) and August 12 (day 224), 2001. The ionospheric radio occultation (IRO) technique has a large potential for measuring the vertical electron density structure of the ionosphere with high data coverage on global scale (Hajj and Romans ; Schreiner et al. ). Whereas one limb sounding mode (rising or setting) can provide about 200 IRO measurements per day, future satellite missions such as GRACE planned to be launched in 2002 will allow to measure both rising as well as setting modes in forward and backward directions.
2. Data Processing
 The data processing system for IRO data analysis up to level 3 data products fulfils operational requirements, i.e. the ionosphere data products are available within 3 hours after data dump from the CHAMP satellite. A corresponding dynamically configurable processing system has been developed in DLR in the preparatory phase of the CHAMP mission (Wehrenpfennig et al. ). After data reception in the DLR Remote Sensing Data Center, Neustrelitz the GPS data are automatically checked and preprocessed by the implemented software. If the controlling subsystem indicates the availability of all data needed for a certain data product, the corresponding generation module is started immediately. The modular structure enables high flexibility if retrieval modules shall be modified or replaced in the course of the CHAMP mission or if supplementary data shall be included to improve the products. At present the IRO processing system provides three official data products that will be made available for the scientific community through the CHAMP Information System and Data Center (ISDC) at GFZ after finishing the validation phase whose initial results are presented here. These products can be described in brief as: diurnal GPS data files, relative TEC data along radio occultation links and vertical electron density profiles. The required orbit information for GPS and CHAMP satellites is provided by GFZ's CHAMP orbit processing group (http://op.gfz-potsdam.de/champ).
3. Data Retrieval Method
 Because the widely used Abel inversion technique for retrieving the vertical refractive index profile is fundamentally based on a spherically layered atmosphere/ionosphere, horizontal gradients or structures in the electron density distribution are principally ignored by applying this technique. Since near sunrise/sunset hours, in the course of ionospheric storms and/or near the crest region strong spatial plasma density gradients are expected in particular under high solar activity conditions, this assumption cannot be held in general. So we preferred to establish a tomographic approach dividing the ionosphere and plasmasphere into spherical shells with upward growing thickness and constant electron density inside.
 This has the advantage that additional information from ground based GPS measurements, models and/or other sources such as peak electron densities obtained at vertical sounding stations can easily be included in the reconstruction of the electron density profile (Jakowski et al. ). After measuring the dual frequency GPS signals the corresponding differential L1/L2 phases provide an accurate expression for the integrated electron density along the considered ith ray path. So, taking into account the sample rate of 1 Hz, every second a linear equation according to
can be established. Here neik denotes the electron density within shell k and Δsik denotes the length of the ray path element going through shell k. One occultation event is defined by a series of measurements TECi along ray paths whose tangential heights come closer and closer to the Earth down to the bottom of the ionosphere thus providing a system of linear equations for the electron density in the different shells. Taking into account that the initial orbit altitude of CHAMP was 454 km and that this height will decay due to atmosphere drag to 300 km in the course of the projected 5 years lifetime, it has to be noted that the upper boundary condition at the beginning of the occultation needs special consideration. This is due to the fact that the topside ionosphere above the orbit height may contribute up to 50% of the total signal or even more if the starting ray path comes close to the F2 peak height. To overcome this problem, the inversion is assisted by an adaptive electron density model of the topside ionosphere and plasmasphere. The initial vertical electron density profile is defined by a Chapman layer which is superposed by a simple plasmasphere model characterized by an exponential decay with altitude (scale height Hp = 104km). If the electron density distribution above the top shell traversed by the first ray at the beginning of the occultation event is assumed to be known by the model (first guess), the electron density within this shell can directly be computed. So, starting from the uppermost shell, the electron density of each shell can be successively determined. However, since the first guess is usually not the best one, there appears usually a discontinuity in the profile shape at the transition height between model and preliminary retrieval results indicating an erroneous solution. To get a homogeneous transition from model to inversion data, peak density Nmax, corresponding height hmax and the topside scale height Hp of the Chapman layer are adjusted within 6 iterations. After this initialization the entire profile is computed starting from the highest layer crossed by the first occultation ray down to the bottomside. To define the site of an occultation event, the vertical electron density profile is assumed to be representative for that location where the tangential point of the Earth approaching occultation rays reaches hmax.
4. Measurements and Discussion
 The IRO measurements reported here were carried out on 35 days within the period 11 April – 12 August 2001. Altogether we obtained 1406 vertical electron density profiles during this period, i.e. about 40 profiles per day. When discussing the results of IRO measurements it has to be taken into account that the occultation rays travel a long distance through the ionosphere. So, as a principal limiting factor of this method, only a poor horizontal resolution in the order of the horizontal scale length (1000 – 2000 km) can be expected. However, this averaging methodological effect may be even advantageous if ionosphere models shall be evaluated or developed. In order to fulfil the operational requirements of data product generation there is only a little time to access additional ionospheric data from other facilities. At present such data are not yet included. This is the reason why the initial results discussed in this section are based on a retrieval algorithm that still assumes a spherically layered electron density distribution as Abel inversion technique too. When evaluating the data, it has to be taken into account that strong ionospheric gradients may degrade the quality of the retrieved profiles. As a first order quality information for users, the internal consistency of the retrieved results is evaluated and scaled by a certain quality number.
 To validate the retrieved electron density profiles, vertical sounding, incoherent scatter and radio beacon techniques can be considered as very powerful sensing methods to get key information for comparison. The first approach to the German ionosonde station Juliusruh were achieved on 27 April 2001 during the HIgh RAte GPS ground station measuring Campaign (HIRAC) initiated by the International GPS service (IGS) for the period 23–29 April 2001. In Figure 1 the IRO reconstructed electron density profile is compared with the vertical profile deduced from the Juliusruh digisonde and the corresponding in situ plasma density measured by the Planar Langmuir Probe (PLP) on board CHAMP. It can be seen that both profiles fit quite well and also the PLP value is in good agreement with the retrieved electron density. Since the model assisted retrieval technique allows the computation of the vertical total electron content, a comparison with regional or global TEC maps (e.g. Jakowski , http://www.kn.nz.dlr.de) provides a rough check of the topside model assumption. On the other hand, horizontal gradients derivable from TEC maps shall be used in future to improve the retrieval procedure (Jakowski et al. ). As already indicated in Figure 1, a very effective way of checking the topside electron density profiles is a comparison with the PLP data that are measured along the satellite track. The PLP is part of the Digital Ion Drift Meter and was provided by the US Air Force Research Laboratory, Hanscom, MA. Figure 2 shows a direct scatter plot of IRO data against corresponding Langmuir probe data. To obtain a reasonable number of coincidences, the longitudinal distance between the localization of the IRO event and the satellite track where PLP measurements where carried out amounts up to 6°. As it can be seen, the correlation between IRO and PLP data is quite consistent over a large dynamic range of more than two decades indicating a reasonable estimation of the electron density in the transition region between the adaptive model and IRO measurements.
 In order to come up with significant conclusions on the accuracy and reliability of ionospheric key parameters such as f0F2 and hmF2 retrieved by IRO techniques, statistical studies have to be performed. Here the global network of vertical sounding (VS) stations can effectively be used for validation tasks. About 100 measurements coincidences within a cross section diameter of about 1800 km between IRO location and ionosonde sites were available. Data were mainly provided by the Ionospheric Prediction Service/Australia (P. Wilkinson), SPIDR (O'Loughlin ) and by the European COST 271 community (http://www.cost271.rl.ac.uk).
 Initial results are illustrated in Figure 3. by plotting the distribution function of the percentage deviations of the parameters f0F2 and hmF2. The bias between IRO and VS parameters amounts to −1.7% for the peak plasma frequency f0F2 and to −4.1% for the peak height hmF2 indicating a slight underestimation of IRO data. The derived rms percentage deviations of 17.8% and 13.1% for f0F2 and hmF2, respectively are quite reasonable if the cross section diameter of 16° (≈1900 km) is taken into account. Schreiner et al.  report a percentage error of 13% for f0F2 by applying the Abel inversion technique on GPS/MET data.
 It should be underlined that the data presented here are the direct output of the automatically working IRO processing unit without further reviewing. Additional selection criteria shall be developed to remove a few remaining outliers. Although such validation studies are far from being complete, this letter is intended as a brief demonstration of the potential of IRO measurements for ionospheric modeling and monitoring from the CHAMP mission. Figures 4 and 5 illustrate for example how the peak electron density can be monitored on global scale. It can be concluded that the latitudinal distribution of NmF2 agrees in general with existing models such as the Parameterized Ionospheric Model - PIM (Daniell et al. ). In addition to the value of the F2 plasma frequency f0F2 indicated by colored pixels, Figure 5 provides an impression on the global distribution of IRO measurements. It is evident that large gaps in the global ionosonde network over the oceans and polar regions can easily be filled in. On the other hand we find an under-representation of successful retrievals in the equatorial region which is probably due to the high altitude of the F2 peak density.
 We have reported first experiences with the ionospheric radio occultation experiment on board CHAMP. Whereas we have obtained about 40 profiles per day during the period reported here, the number increased up to about 100 at present due to onboard GPS software upgrades. Early retrieval results for f0F2 and hmF2 are in good agreement with corresponding ionosonde data. Further work is needed to validate the entire profiles from CHAMP orbit down to the E-region height under different geophysical conditions using a more extended data base. Taking into account the upper boundary problem, accuracy and reliability of retrieved electron density profiles can be enhanced by an improved adaptive topside/plasmasphere model and by including supplementary ionospheric data in the operational data processing system. It should be stressed that independent of the absolute accuracy of the IRO profiling technique each GPS limb measurement can contribute to 3-D electron density reconstruction by data assimilation techniques. The combination of ground based GPS and IRO measurements for reconstructing the structure of the ionosphere/plasmasphere systems is promising especially over good conditioned areas with high dense networks of GPS ground stations such as California, Europe or Japan. It can be expected that radio occultation measurements aboard satellite missions such as CHAMP, SAC-C, GRACE, and COSMIC will essentially contribute to a global space weather monitoring system by measuring the electron density distribution permanently on global scales.
 The authors are very grateful to all colleagues of the CHAMP team and from the payload providers JPL and AFRL. We thank also Phil Wilkinson (IPS, Australia), Jens Mielich (IAP, Germany) and numerous colleagues involved in the European COST 271 action who contributed by their data to validate IRO profiles in short time. This study was carried out under grant number 01SF9922/2 of the German Federal Ministry of Education and Research (BMBF).