GPS radio occultation with CHAMP Jens Wickert, Georg Beyerle, Torsten Schmidt, Christian Marquardt, Rolf K¨onig, Ludwig Grunwaldt, and Christoph Reigber GeoForschungsZentrum Potsdam (GFZ), Division 1, Kinematics & Dynamics of the Earth, Potsdam, Germany,
[email protected] Abstract. The GPS (Global Positioning System) radio occultation experiment onboard the German CHAMP (CHAllenging Minisatellite Payload) satellite was activated in February 2001. By the end of 2001 about 36,500 occultations were recorded. We review the occultation data processing at the GFZ and discuss selected first results. 1) CHAMP allows for atmospheric sounding with high accuracy despite of the Anti-Spoofing (A/S) mode of the GPS. 2) There are advantageous consequences for the GPS data processing due to the termination of the Selective Availability (SA) mode of the GPS. It is possible to reduce the GPS ground station acquisition rate for double difference processing. The application of space-based single differencing technique for precise occultation data processing became feasible. 3) The state-of-the-art GPS flight receiver onboard CHAMP combined with high-gain occultation antenna allows for atmospheric sounding deep into the lower troposphere. Possible applications and improvements of occultation data analysis in the lower troposphere are discussed. Key words: CHAMP, Remote sensing of Earth’s atmosphere, GPS radio occultation
1 Introduction The CHAMP GPS radio occultation experiment is an important step towards establishing an innovative remote sensing method for sounding of the Earth’s atmosphere and ionosphere. The GPS radio occultation technique is based on precise dual-frequency phase measurements (L-band) of a GPS receiver in a Low-Earth-Orbit tracking a setting or rising GPS satellite. Combining these measurements with the satellites’ position and velocity information the small phase path increase due to the atmosphere during the occultation event can be derived. It is converted to atmospheric bending angles. Assuming spherical symmetric atmosphere vertical profiles of refractive index can be determined and converted to atmospheric parameters as pressure, temperature and, using independent knowledge of temperature, also to water vapor within the lower troposphere. The main advantages of the calibration-free GPS occultation technique are global coverage, high vertical resolution and all-weather-capability combined with high accuracy. These properties allow various applications in atmospheric/ionospheric research, weather forecast and climate change detec-
2
Wickert et al.
tion [1, 2]. For the first time the promising potential of the GPS radio occultation technique was demonstrated by the GPS/MET-Experiment [3]. In relation to GPS/MET, the CHAMP satellite possesses enhanced GPS flight receiver (the state-of-the-art GPS receiver ’BlackJack’ provided by JPL, Jet Propulsion Laboratory) and high-gain occultation antenna. Currently, beside CHAMP also the Argentinean SAC-C satellite (launched on November 21, 2000) performs GPS radio occultation measurements. Furthermore, on March 17, 2002 the American-German GRACE satellites (GRAvity and Climate Experiment, twin satellite mission) were successfully launched. Both satellites possess the capability to perform GPS occultation measurements. Further missions flown with GPS occultation receivers onboard are the Danish Ørsted [4] and the South African SUNSAT [5]. The Ionosphere Occultation Experiment (IOX) onboard the PICOSAT satellite (launched on September 30, 2001) provides GPS occultation measurements of the ionosphere. IOX is part of the U.S. American Department of Defense Space Test Program executed by the Air Force [6]. We review the occultation data processing at the GFZ and present selected first results. A detailed evaluation of the quality of CHAMP atmospheric data products is performed in another paper of this issue [7].
Fig. 1. Vertical profiles of (a) dry temperature and (b) specific humidity, derived from one of CHAMP’s first occultation measurements (No. 5; 0.5◦ W, 53.2◦ S; 19:43 UTC) on February 11, 2001, compared to analyses data from ECMWF (European Centre for Medium-Range Weather Forecasts and NCEP (National Centers for Environmental Prediction).(from [9])
2 First Measurements The first 7 occultation measurements with CHAMP were performed on February 11, 2001 between 19:04 and 20:04 UTC. Vertical profiles of dry tempera-
GPS radio occultation with CHAMP
3
ture and specific humidity were derived (an example is shown in Fig. 1). Already these measurements showed that, in spite of the activated anti-spoofing (A/S) mode, CHAMP allows for atmospheric sounding with high accuracy and vertical resolution [8, 9].
3 Status of the experiment Occultation measurements were performed during 217 days in 2001 giving a total of 36,514 events. Vertical profiles of atmospheric parameters were derived for 24,658 occultations (67.5 %). For the remaining profiles (32.5 %) the atmospheric excess phase calibration failed. One reason for that relatively large percentage is the reduced data quality due to on-board-software modifications of the GPS flight receiver for parallel performing of occultation measurements in the atmosphere and ionosphere [10]. These modifications were implemented at the end of July 2001. As a result the number of daily atmospheric occultations was significantly reduced from ∼220-240 daily events temporary down to ∼100 (see Fig. 2). Furthermore, the data analysis software for the automatic calibration of the atmospheric excess phase needs to be improved, to detect and correct for for all types of irregularities in the GPS occultation and ground station data. Furthermore, the influence of the ionosphere during solar maximum as an additional factor for failure has to be investigated in more detail.
Fig. 2. Number of daily CHAMP occultations (duration > 20 s) for 2001. The total height of the columns corresponds to the number of daily measurements. The black and light grey color indicate occultations with calibration failure and insufficient data quality, respectively. The height of the dark grey columns corresponds to the number of vertical atmospheric profiles provide to the CHAMP data center at GFZ.
4
Wickert et al.
4 Occultation data processing and infrastructure at GFZ The main components of the GFZ infrastructure for data generation, transfer, analysis and archiving are: the GPS receiver onboard the CHAMP satellite (provided by JPL) and the ground segment (see Fig. 3). It consists of the downlink station at Neustrelitz, Germany (53.2◦ N, 13.0◦ E) including the Raw Data Center (RDC), the fiducial GPS ground network (Fiducial Network), the Precise Orbit Determination (POD) facility, the Occultation Processing System [Atmospheric Profiling (AP)-Processor] and for archiving and distribution the CHAMP Information System and Data Center (ISDC). The downlink station and the Raw Data Center are operated by the German Aerospace Center (DLR/DFD). The GPS ground network is operated in cooperation between GFZ and JPL, the other components are maintained by GFZ.
Fig. 3. Infrastructure of the CHAMP GPS radio occultation experiment (Overview).
GPS radio occultation with CHAMP
5
Within the framework of the German GASP (GPS Atmosphere Sounding Project, see e.g. [11]) a pre-operational atmospheric data product delivery is aimed at. This requires a maximum time delay of not more than 3 hours between measurement and product delivery to weather services. To reach this objective, every component of the occultation infrastructure (Fig. 3) has to satisfy this 3-hour requirement. We characterize in more detail the Precise Orbit Determination, the Fiducial Network and the occultation data analysis and processing system at GFZ and focus to operational aspects. 4.1 Precise Orbit Determination The GPS and CHAMP orbit ephemerides (Rapid Science Orbits, RSO) are provided by the orbit and gravity component within GFZ’s Science Data System for CHAMP [12]. The orbits are available in the afternoon of the following day. The arc lengths of the CHAMP orbits are chosen at 14 hours. Thus, one day is covered by two arcs with 2 hours overlap. Until November 2001, 3-D position accuracy of the RSO sized at ∼20 cm which was already sufficient for precise occultation processing [13]. Improved POD modelling increased then the accuracy further to ∼10 cm [14]. A current project is the demonstration of near-real-time orbit delivery with 3 hour rhythm (Ultra rapid Science Orbits, USO). 4.2 Fiducial Ground Network To compensate for CHAMP and GPS clock errors, the precise phase measurements of the onboard GPS receiver are differenced with those from a fiducial ground network receiver [9, 15]. To reach global coverage of occultation measurements a global ground network is necessary (High Rate Low Latency GPS Ground Tracking network [16]). This network consists of 38 globally distributed stations (as of January 2002) and is operated jointly by JPL and GFZ with cooperating partners. Redundancy for occultation processing Simulation studies in preparation of the CHAMP occultation experiment using a planned configuration of 28 stations have shown that an average redundancy of ∼4 stations for occultation processing is reached [15, 16]. To confirm this with real data we have processed ∼5,500 CHAMP occultations during day of year 135-160, 2001 (May 15 - June 9) applying the double difference technique using data of 24 ground stations. The average data availability of these ground stations was 91.2 %. The main reasons for loss of data are hardware problems at the ground station sites or data transmission failures from the ground stations to the Level-1 processing center at GFZ.
6
Wickert et al.
Fig. 4. Global distribution of ground station redundancy for CHAMP occultation processing during May 15 - June 9, 2001 (from [17]).
Fig. 4 shows the resulting global distribution of ground station redundancy for occultation processing. On average ∼4 ground stations were available for processing, e.g. over North Atlantic or Europe. Areas with lower redundancy were found as well, e.g. India or South Africa with ∼2 ground stations per occultation. Areas with higher-than-average redundancy with up to 7 ground stations per occultation exist within the Equatorial Pacific region. We conclude that the established network configuration allows for processing of globally distributed occultation events with average redundancy of ∼4 ground stations per event. Data flow The GPS ground data are recorded with an acquisition rate of 1 Hz (Pseudoranges and phases on both GPS frequencies). The data (15 min binary files, ∼540 MByte daily) are transmitted via FTP (File Transfer Protocol) to the Level-1 processing center at GFZ. Here, 1 hourly RINEX (Receiver INdependent EXchange format) files are generated and transferred to the CHAMP Information System and Data Center. The time delay between measurement and data provision is currently between ∼30 min and ∼70 min. Due to the use of 15 min intervals for the data transmission from the ground stations to the Level-1 processing center, these time delays can be further minimized down to about 15-30 min. 4.3 Occultation data analysis and processing system The basics of the standard analysis of CHAMP occultation data are described in [9], a more detailed description can be found in [13]. The standard
GPS radio occultation with CHAMP
7
data analysis is based on the double difference method for deriving the atmospheric excess phase, the geometric optics approximation and the assumption of spherical symmetry for the derivation of the atmospheric refractivity, respectively. For details on the GPS radio occultation method in general see e.g. [2]. The analysis software is embedded in a dynamically configurable and extendable system for operational data processing and product generation [18], which also provides an interface to GFZ’s data archive (ISDC). The analysis and processing software is implemented in FORTRAN, C++ and Perl programming languages. The main processing computer is a SUNFIRE 280R server, 2 UltraSparc III processors (750 MHz), 4 GB RAM.
5 Consequences of the termination of Selective Availability Selective Availability was part of the intentional degradation of the user’s navigation solution. SA implied that GPS transmitter clocks varied by ∼10−7 s (tens of meters in units of length) over timescales of 100 s, with rates on the order of 1 m/s. To correct for these clock rates, which are unacceptable for precise atmospheric profiling [15], the double difference technique is used to process GPS occultation data [9, 19]. Due to the termination of SA on May 2, 2000 the amplitude of the described errors in the transmitter clocks was reduced by orders of magnitude [20]. This made the clocks more predictable and has advantageous consequences for occultation data processing, which are briefly described here. 5.1 Acquisition rate of the ground station data To correct for the clock variations (induced by SA) a ground station acquisition rate of 1 Hz was required. This rate is currently used for standard occultation processing (double differencing) at GFZ. However, already during the preparation of the CHAMP occultation experiment it was shown, that after the termination of SA a reduction of the ground station acquisition rate to 1/10 Hz would not influence the accuracy of the calibration of the atmospheric excess phase for occultation measurements [15]. Here we investigate the influence of reduced (in relation to the standard 1 Hz rate) ground station acquisition rates (1/5, 1/10 and 1/30 Hz) on the quality of vertical dry temperature profiles derived from CHAMP measurements. A set of 436 occultations during April 19-21, 2001 was processed with average redundancy of ∼3.2 ground stations per event using acquisition rates of 1, 1/5, 1/10 and 1/30 Hz, respectively. The resulting 1,400 dry temperature profiles were compared with corresponding ECMWF profiles.
8
Wickert et al.
The 1/5 Hz profiles exhibit identical deviations (mean and standard) from ECMWF as the reference profiles, which were processed using 1 Hz. Even a reduction of the acquisition rate down to 1/30 Hz results in statistically identical profiles compared to ECMWF as the reference set [20]. Our study shows that a reduction of the ground station acquisition rate down to 1/30 Hz will be possible. However, further investigation of larger data sets is required to confirm this result. 5.2 Space-based single differencing As an alternative to double difference processing the application of lessdifferencing techniques became feasible with termination of SA. First results of GPS occultation data processing utilizing a single difference technique, i.e., clock correction using 5 min GPS clock solutions without directly involving GPS ground station data, were presented in [20]. Fig. 5 shows one of the results, the statistical comparison of two sets of 436 vertical profiles of refractivity and dry temperature, generated by the double and single difference method, respectively. The mean deviation in refractivity is
