Real-time High Accuracy Retrievals of Precipitable Water Vapour from GNSS Precise Point Positioning

Transcription

1 Real-time High Accuracy Retrievals of Precipitable Water Vapour from GNSS Precise Point Positioning Yubin Yuan, Kefei Zhang, Suelynn Choy RMIT University, Melbourne, AUSTRALIA Witold Rohm Wroclaw University of Environmental and Life Sciences, Wroclaw, POLAND IGNSS, Gold Coast, 16/07/2015

2 Outline Background BNC - modifications ZTD and PWV retrievals GPS-only Error analysis ZTD retrievals GPS+GLONASS Summary 2/46

3 Background Australia suffers from a wide range of severe weather events, e.g. flooding and storms Storms hit Melbourne on Christmas Day 2011: Greensborough from australiasevereweather.com 3/46

4 Atmospheric water vapour (AWV) is strongly related to storms Difficult to determine. One of the least understood and poorly described components of the Earth's atmosphere Atmosphere sounding Traditional means Radiosonde Balloon-borne equipment; A key meteorological sensor in the past seventy years; a limited number of launches and stations due to high cost; a poor spatial-temporal resolution of water vapour Water vapour radiometer (WVR) AWV can be estimated accurately; Ground-based WVRs: good temporal but poor spatial coverage; Space-based WVRs: poor temporal but good spatial coverage; Extremely expensive 4/46

5 Water vapour sounding GPS-Met A novel remote sensing approach of water vapour retrievals Complementary technology to the conventional sounding means Ground-based GPS-Met: using GPS tropospheric delay (PWV, ZTD, slantpass delay) for meteorology Space-based GPS-Met: GPS radio occultation of atmosphere in GPS antenna/receiver on-board satellites in low earth orbit (LEO) for meteorology 5/46

6 Current research GPS-MET was first proposed by Bevis et al. [1992] Post-processing scenarios investigated [Duan et al., 1996; Rocken et al., 1997; Tregoning et al., 1998; Hagemann et al., 2003] Near-real-time GPS-MET products applied in NWP models [Poli et al., 2007; Smith et al., 2007]. The assimilated ZTDs were beneficial for cloud forecasting [Bennitt and Jupp, 2012] and precipitation forecasting [Boniface et al., 2009] Real-time experiments since late 2014 [Li et al., 2014; 2015] 6/46

7 GPS-MET to date based on post-processing or near-real-time processing using differenced GPS observations still challenging at a high temporal resolution and in the real-time mode Modern weather forecasting systems have an ever increasing demand for PWV retrievals short latency or no latency higher spatial resolutions of GPS stations without much increase of computational resources higher temporal resolutions specifically for severe weather nowcasting 7/46

8 PPP vs DD PPP DD need reference stations no yes GPS data processed data processing time ZTD epochs all observations from one station linearly proportional to the number of GPS stations epoch-by-epoch, highresolution simultaneous observations from the same pair of two stations and two satellites Exponentially proportional to the number of GPS stations Batch processing, mean values Advantages of PPP when large amount of GPS data is processed and multi-thread processing is adopted for the applications that require near-real-time or real-time solutions 8/46

9 Modern weather forecasting systems have an ever increasing demand for PWV retrieval short latency or no latency higher spatial resolutions of GPS stations without much increase of computational resources higher temporal resolutions specifically for severe weather nowcasting This study aims to take these three requirements into account and investigate the retrieval of ZTD and PWV using the real-time PPP technique. 9/46

10 The BNC software The BKG Ntrip Client V2.8 BNC Developed by the Federal Agency for Cartography and Geodesy (BKG), Frankfurt, Germany A program for simultaneously retrieving, decoding, converting and processing real-time GNSS data streams Developed within the framework of the IAG sub-commission for Europe (EUREF) and the International GNSS Service (IGS) Written under GNU General Public License (GPL) and Source code is available online 10/46

11 BNC - Modifications the PPP module of the original BNC software: low positioning accuracies (5m accuracy in coordinates) Modifications in BNC Modelling of ZTD l c( dt dt) T p r P l c( dt dt) Tr N Tr zt mft ( original BNC) mft 1/ sin e Tr zh mfh zw mfw z h,z w,mf h,mf w from external models ( modified BNC) 11/46

12 BNC - Modifications Mapping functions Hopfield [1971] Saastamoinen [1973] UNB3M [Leandro et al., 2006] Global Pressure and Temperature 2 / Global Mapping Function (GPT2/GMF) [Boehm et al., 2006a] and [Lagler et al., 2013] Forecast Gridded Vienna Mapping Functions 1 (VMF1- FC) [Boehm et al., 2009] 12/46

13 VMF1 is proposed by Boehm et al. (2006b). Each mapping function depends on different sets of constants a, b and c. Those constants are determined using the 40-year period ECMWF Re-Analysis (ERA-40) data [Uppala et al., 2005] available ( station-specific parameters such as ZHD, ZWD, a h and a w --- for IGS stations Gridded VMF1 Forecast VMF1 (VMF1-FC) --- for real-time applications 13/46

14 Implementation of VMF1-FC T z mf z mf r h h w w mf ( E) a 1 b 1 1 c a sin( E) b sin( E) sin( E) c a, a, z, z h w h w??? b b h w doy 28 c11 ch c0 ((cos( 2 ) 1) c10 ) (1 cos ) c w 14/46

15 Interpolation to the current epoch using consecutive files. A sliding window ensures that no big jumps in values of the interpolated parameters occur. Interpolation to the specific location using data from neighbouring grid points. 15/46

16 Height corrections ([Niell et al., 1996; Kouba, 2008; Fund et al., 2009]) a0 1 b c0 h0 hcorr sin( E) a sin( E) b0 sin( E) sin( E) c 0 mf mf h h h corr gm Pg ( ) zh( s) zh( g) ( h( s) h( g)) R T ( g) z ( s) z ( g) e w w ( h( s) h( g ))/ /46

17 Global Mapping Function (GMF) model is a simplified model of VMF1 [Boehm et al., 2006a] Global Pressure and Temperature 2 (GPT2, [Lagler et al., 2013]) provides pressure, temperature, temperature lapse rate and water vapour pressure at any site on the earth s surface on a global 5 x 5 grid. ZHD from GPT2 is then computed using Saastamoinen formulas [Saastamoinen, 1973] as modified by Davis et al. [1985] z P h h / ( cos(2 ) ) Tr zh mfh zw mfw 17/46

18 Validation of ZHD Using In-situ Meteorological Measurements 71 IGS stations with RINEX Meteorological files 18/46

19 19/46

20 Cumulative distributions of RMS errors of ZHDs Reference data taken as true values are derived using in-situ pressure measurements at 71 globally distributed IGS stations during September 2013 Stations with RMS error <15.3 mm account for 95% of the 71 IGS stations (VMF1-FC). 20/46

21 PPP-ZTDs using GPT2/GMF and VMF1-FC are almost identical So in PPP processing, GPT2/GMF is used In conversion from ZTD to PWV, VMF1-FC is used 21/46

22 BNC - Modifications Modifications in BNC l c( dt dt) T p r P! Lots of errors to be corrected l c( dt dt) Tr N Ocean tide loading, Solid Earth Tides are corrected following the latest IERS 2010 Conventions [Petit and Luzum, 2010] Satellite antenna variations Receiver antenna offsets Receiver antenna variations 22/46

23 Data selection - orbits & clocks Real-time clocks and orbits from BKG for this investigation Caster IP: Port Mount point Message Content products.igsip.net:2101 RTCM3EPH 1019 GPS broadcast ephemerides products.igsip.net:2101 IGS GPS orbit corrections to broadcast ephemerides products.igsip.net:2101 IGS GPS clock corrections to broadcast ephemerides 23/46

24 Data selection GPS observations From 20 stations in September 2013 (DOY ) GPS observations are available dataset covers different climatic regions reference data from CODE and USNO is available to validate the derived ZTDs radiosonde stations within 60 km distance of the IGS stations are available to validate the derived PWVs GPS observations with a sampling rate of 1 Hz PPP-ZTDs are estimated every 30 seconds hence approximately epoch-solutions are available for each station in the period of 30 days 24/46

25 Data selection GPS observations Global distribution of the IGS stations selected for test 25/46

26 PPP-ZTD PIMO is a station where PPP-ZTDs show poorer agreement with the reference data. HERT is a station where real-time GPS data suffers from internet instability in DOY 256. Other stations are presented here for comparison. 09:47 26/46

27 PPP CODE PPP USNO USNO CODE Sites RMSE(mm) <20mm (%) RMSE(mm) <20mm(%) RMSE(mm) <20mm(%) ABPO BOGT CHUR COCO FAIR GODE HERT HOB HRAO MAC MCM MKEA MOBS NRIL ONSA PERT PIMO SHAO RMSE (mm) of the differences in ZTDs between real-time PPP, USNO and CODE. Column USNO-CODE shows the differences between CODE and USNO ZTDs vary from 3.4 mm to 7.5 mm. Column PPP-CODE and PPP-USNO show that the accuracies of PPP-ZTDs are better than 12 mm assuming CODE ZTDs are true values. PPP-ZTDs are accurate enough as input to NWP models (15mm required) 27/46

28 PPP-ZTD -> PPP-PWV PWV ( z z ) z ( PPP ZTD) t t z ( from VMF1 FC) h h 6 10 M w Mw k 3 Rk2 k1 M d T m T T ( T from GPT 2) m /46

29 GPS RS. Height Height IGS Site (m) (m) Distance RMSE (km) (mm) CHUR COCO FAIR GODE HERT HOB MAC MCM MOBS NRIL ONSA PERT SHAO STHL ZIM ABPO BOGT HRAO MKEA PIMO RMSE (mm) of the differences in PWVs between real-time PPP and radiosonde data. Accuracies of PPP-PWVs are at the level of 3 mm, which is the threshold for weather nowcasting. The five stations at the lower part of the table are less conclusive due to the height issues. 29/46

30 Error analysis A number of models and datasets are used. Each step of the process is performed with a certain degree of accuracy and introduces its own instrumental or computational errors. It is therefore important to analyse the errors of the retrieved PWVs from different sources. Is PPP-PWV of 3mm-accuracy achievable? if Accuracy of PPP-ZTD is 12mm Empirical models are used to obtain pressure, ZHD, and weighted mean temperature 30/46

31 Error analysis 6 10 M PWV ( zt zh) Rk, R, M, M, k, k and k w d Assuming the RMSE of w ' 3 2 Tm k are constants z ( ), z ( ) and T ( ) t t h h m T are Gaussian and independent from each other m the RMSE of PWV is PWV M w Mwk3 2 2 t h z ' 2 w T ' k R k 3 R k T 2 2 m k T 3 m 2 m 31/46

32 M w M wk3 2 2 PWV t h z ' 2 w T ' k 3 R k R k2tm k 3 2 T m m is a strictly increasing function of variables z,, and PWV w t h T m z T PWV w m is a strictly increasing function [0, 0.6] m [273 40, ] K [3, 12] mm t [1, 15.3] mm h T m [0, 6] K o f T as m proved by [Yuan et al., 2014] if the retrieval of PPP PWV is more challenging in conditions of higher humidity and weighted mean temperature 32/46

33 Error analysis Theoretical accuracy of PPP- PWV (a) in polar regions like station MCM4 where it is cold and dry (b) in moderate conditions like CHUR, HERT, MOBS, ONSA, PERT, ZIM2 (c) in the tropics or subtropics like station SHAO and COCO where it is hot and humid 3mm accuracy is achievable RMS of PPP-ZTDs are <12mm Even coarse empirical models are used to convert ZTDs to PWVs 33/46

34 GPS + GLONASS 34/46

35 35/46

36 Sites GPS only GPS+GLO weighting 1:5 GPS+GLO weighting 1:1 ALIC COCO DAV GRAZ HERT HOB MOBS NTUS ONSA PERT STHL TOW RMS (mm) of the differences between real-time PPP-ZTDs and CODE-ZTDs during DOY , /46

37 Origins of the Degraded Accuracy 1. Deficiencies in the Functional Model: ISB: two clock offsets per epoch: one for GPS and another for GLONASS IFB for GLONASS 37/46

38 Origins of the Degraded Accuracy 2. the increase of new ambiguity parameters accordingly Station GPS SVs GLONASS SVs Multi-GNSS SVs Increase of Observations (%) Increase of Parameters (%) NYA WUHN NTUS MOBS CAS Increases of observations and parameters in terms of GPS-only and GPS+GLONASS scenarios during DOY , 2014 Consequently no significant improvement in terms of accuracy can be expected from adding GLONASS data 38/46

39 Origins of the Degraded Accuracy 3. Less Accurate GLONASS Orbits and Clocks Performance statistics of GPS and GLONASS corrections [Agrotis et al., 2014] Case MGEX testing 31/8/2014 RT MGEX 21/5/2014 RT MGEX 22/5/2014 GPS Orbit 1-D RMS a (mm) GPS Clock SD a (ns) GLONASS Orbit 1-D RMS b (mm) GLONASS Clock SD b (ns) b Against IGS rapid products. c Against ESA rapid products. 39/46

40 Summary PPP-ZTDs are shown to be accurate (RMSE <12 mm) and sufficient as an input to NWP models while PPP-PWVs are sufficient (RMSE 3 mm) for weather nowcasting Results are validated in different climatic conditions, on a global scale, by IGS products and radiosonde data PPP-PWVs of 3 mm accuracy is theoretically proved achievable if the accuracy of PPP-ZTDs is <12 mm, even empirical models are used for conversion from PPP-ZTDs to PPP-PWVs Adding GLONASS observations degrades the accuracies of PPP-ZTDs 40/46

41 Reference Yuan, Y., K. Zhang, W. Rohm, S. Choy, R. Norman, and C.-S. Wang (2014), Real-time retrieval of precipitable water vapor from GPS precise point positioning, Journal of Geophysical Research: Atmospheres, 119(16), 2014JD021486, doi: /2014jd /46

42 Thank you! 42/46