Benefit of Multi GNSS Processing with GPS, GLONASS, and QZSS for Tropospheric Monitoring

T. Iwabuchi, C. Rocken, A. Wada, M. Kanzaki

Abstract:	Ground-based GPS precipitable water vapor (PWV) converted from zenith tropospheric delay (ZTD) is currently used in multiple leading weather forecast centers such as NOAA (USA), ECMWF (EU), and JMA (Japan) and in many more counties. These centers have shown the benefit of GPS PWV not only for short-term severe precipitation forecast, but also for the improvement of humidity profile in longer term forecasts. GPS PWV is considered to have observation noise of about 1~2 mm, but GPS PWV sometimes shows larger noise and jumps for some stations and/or at certain times, suggesting that a lower number of GPS satellites and poor line-of-sight condition limits the quality of ZTD estimates in such stations and at such times. Many GPS networks are now being upgraded to multi GNSS observation networks with real-time data streaming, and this upgrade is expected to be beneficial for GNSS tropospheric monitoring. GEONET (GPS Earth Observation Network) deployed by GSI (Geospatial Information Authority of Japan) in Japan consists of about 1,200 station, and GEONET is also being upgraded to be capable to observe GPS, GLONASS, and QZSS (hereafter, GGQ) signals. GSI has started to provide GGQ observation data for about 180 stations which are widely distributed all over the Japanese Islands since July 11, 2012. We processed all available GGQ data to investigate the potential benefit of multi GNSS processing for tropospheric monitoring. We used RTNet GNSS processing software developed by GPS Solutions for GGQ processing. The processing period is from July, 2012 to February, 2013. The length of this period is considered to be enough to investigate the behavior of tropospheric estimates in active weather condition and also to see statistics and seasonal variations of ZTD, PWV, and slant tropospheric delay based on post-fit residuals. GPS and GLONASS satellite products from the IGS, and QZSS satellite product from JAXA (Japan Aerospace Exploration Agency) are used for PPP processing with the ionospheric free phase linear combination. Seamless processing without day boundary offset was performed to handle multiple RINEX files with a forwards running Kalman filter. We first tried kinematic coordinate estimations with GPS (1G), GPS+GLONASS (2G), and GGQ, and confirmed that 2G solutions shows better repeatability than 1G, and GGQ solutions showed the best repeatability. This fact suggests that the quality of satellite clocks used in the study is similar, and thus the increase in the number of observations is helpful. ZTD estimates with multi GNSS processing are more stable than those based only on GPS. Sudden jumps observed in GPS only ZTD are significantly reduced with multi GNSS processing. An interesting fact is that the addition of only one QZSS satellite is noticeably helpful to reduce the sudden jumps. Because the number of satellites with GGQ is about twice that of only GPS observation, the noise due to rising and setting satellites is mitigated thus reducing the size of sudden jumps in ZTD. The fact suggests that ZTD and PWV estimated with multi GNSS is beneficial for its data assimilation into numerical weather models with standard observational operators that assume that the GNSS PWV is just an integration of the wet refractivity above the GNSS station. This assumption is known to be wrong because the real observation of GNSS PWV is averaged information of the wet refractivity with a limited number of GNSS ray paths inside of a cone with a vertex that depends on the cut-off elevation angle. A more realistic observational operator should model the GNSS satellite positions used for ZTD estimation at each epoch, but its computation cost is generally considered too expensive. We can also expect smaller observation noise in GNSS ZTD and PWV based on multi GNSS processing. It means that the moisture field in the background condition is influenced more strongly by the observations. The characteristics of slant delay residuals (post-fit residual) of GLONASS and QZSS are closer to that of GPS though QZSS (currently only one satellite which is usually in higher elevation angles) has much a smaller number of observations at lower elevation angles. Dependency of slant delay residuals on elevation angle and behavior of slant delay in active weather conditions such as weather fronts are similar among GPS, GLONASS, and QZSS. To reduce noise from site-dependent multipath and any systematic errors in the assumed phase center variation (PCV) pattern of each GPS antenna with radome (including multipath from the monument) the post-fit residuals are filtered with a multipath stacking map with 1 x 1 deg. resolution in azimuth and elevation for each of the 180 processed Geonet GGQ stations in the same way as described in Iwabuchi et. al. (ION GNSS 2011). We compared the stacking maps based on GPS, GLONASS, and QZSS, and found that all maps based on different satellite system shows similar multipath pattern due to reflection from rod shaped poles. These facts suggests that slant delays from GPS, GLONASS, and QZSS can be handled in the same way, and that those slant delays from dense multi GNSS networks are helpful for high resolution numerical weather prediction, water vapor tomography, and tropospheric delay corrections for other ground-based sensors such as very long baseline interferometry (VLBI) or synthetic aperture radar interferometry (In-SAR) which uses microwaves passing through the atmosphere. With the higher number of multi GNSS satellite data, the approach to reduce the representativeness error of PWV observation by setting observation cut-off to higher elevation angles (as suggested in Realini et. al., ION GNSS 2012) is expected to work better than with GPS only. Also, the higher number of GNSS satellites can improve other tropospheric estimate in GNSS processing, especially for tropospheric delay gradient (two parameters in addition to ZTD), especially during periods of fewer observed satellites. Improvement of the estimated gradient will help with nowcasting of severe rainfall because the tropospheric delay gradient in GNSS processing is more sensitive at the scale height of water vapor (ground humidity observation is only sensitive near the ground) which is valuable information for the prediction of small-scale heavy rainfall (Shoji et al., 2013).
Published in:	Proceedings of the 26th International Technical Meeting of the Satellite Division of The Institute of Navigation (ION GNSS+ 2013) September 16 - 20, 2013 Nashville Convention Center, Nashville, Tennessee Nashville, TN
Pages:	2496 - 2507
Cite this article:	Iwabuchi, T., Rocken, C., Wada, A., Kanzaki, M., "Benefit of Multi GNSS Processing with GPS, GLONASS, and QZSS for Tropospheric Monitoring," Proceedings of the 26th International Technical Meeting of the Satellite Division of The Institute of Navigation (ION GNSS+ 2013), Nashville, TN, September 2013, pp. 2496-2507.
Full Paper:	ION Members/Non-Members: 1 Download Credit Sign In