150 likes | 236 Views
GPS data during SOP1 & SOP2 (D4B2). IODAMED meeting, 16 May 2014. Olivier Bock, LAREG IGN Pierre Bosser, ENSG IGN Contributions from : C. Champollion, E. Doerflinger (LDL) France R. Pacione (E-GEOS/ASI) Italy
E N D
GPS data during SOP1 & SOP2(D4B2) IODAMED meeting, 16 May 2014 Olivier Bock, LAREG IGN Pierre Bosser, ENSG IGN Contributions from: • C. Champollion, E. Doerflinger (LDL) France • R. Pacione (E-GEOS/ASI) Italy • J. A. Sanchez Sobrino (IGN-E), A. Seco (UPN), A. Botas (Logica), C. Sanchez Tugores (Sitibsa), L. García Pellicer (ARAGEA), Spain • M. Vasconcelos (DGT), Portugal • J. Jones, D. Offiler (UKMO), UK IODAMED meeting, 16 May 2014
Outline • The operational GPS network and ZTD data flow (EGVAP -> assimilation in NWP models) • A reprocessed dataset for HYMEX SOP1+2 • Screening of ZTD data • Conversion of ZTD to IWV • Comparison of GPS IWV to AROME WMED Op An IODAMED meeting, 16 May 2014
May 2012 E-GVAPEUMETNET EIG GPS Water Vapour Program • Aim: use NRT GNSS ZTD for operationalmeteorology. • EGVAP is a consortium of 13 National Weather Services working in close collaboration with11 National MappingAgencies. • GNSS network iscomposedof public & private stations and counts ~ 1500 sites. • GNSS rinex data are processed by 15 ACs in Near Real Time (NRT), i.e. latency < 1h45min. • NRT ZTD disseminated by UMKO on the GTS (BUFR files) and assimilatedin NWP models by UKMO, Meteo-France, DMI… IODAMED meeting, 16 May 2014
ZTD GPS data input AROME - France 10 000 values per 3h 15 GPS ACs with different sofware and processing procedures => ZTD dataset is inhomogeneous IODAMED meeting, 16 May 2014
NRT ZTD data archived on hymex DB • NRT_EGVAP_SGN (IGN/SGN, France) • Network: • - RGP (France) + reference stations (321 sites) • Processing: • BerneseGPS software • Ultra-rapidIGS orbits, fixed positions, CO=10°, no gradients, ZTD values at HH+00, 15, 30, 45, 59’ • NRT_EGVAP_IGE (IGN, Spain) • Network: • - Spain + Portugal + reference stations (288 sites) • Processing: • BerneseGPS software • Ultra-rapidIGS orbits, ZTD values: HH+00, 15, 30, 45, 59’ • NRT_EGVAP_ASI + NRT_EGVAP_ASI1 (E-GEOS/ASI, Italy) • Network: • - Italy+ reference stations (164 sites) • Processing: • GIPSY-OASIS II software • Ultra-rapidJPL orbits, fixed positions, ZTD values HH+00, 15, 30, 45’ IODAMED meeting, 16 May 2014
Hymexdatabase sedoo.fr IODAMED meeting, 16 May 2014
Comparison of NRT ZTD data at common sites ASI – SGN, 44 common sites, different software Mean = -3 ± 5 mm Std. = 7 mm IGE – SGN, 50 common sites, same software Std. = 7 mm Mean = 0 ± 2 mm IODAMED meeting, 16 May 2014
Reprocessed network REPRO EGVAP Orphéon SGN IGE HYMEX_FR HYMEX_DE 15 networks ASI Sardinia 773 stations (3 Analysis Centres) 970 stations (2 ACs, 21 networks) IODAMED meeting, 16 May 2014
Reprocessed network Homogenous reprocessing of 970 sites for period 1 Sep 2012 – 31 Mar 2013 • Reprocessing by IGN/LAREG, France • Network: 823 sites • France : RGP (349 sites) + Orphéon (182 sites) • Temporary sites : Cévennes (8 sites, Géosci. Montpellier) and Corsica (5 sites, GFZ) • Spain: 14 regions (231 sites) + IGE national network (37 sites) • Portugal: RENEP (43 sites) • North Africa: UNAVCO (5 sites) • Processing: • GIPSY-OASIS II software • Final JPL orbits & clocks, ZTD + gradients every5’, VMF1, 30 h session • Reprocessing by e-GEOS/ASI, Italy • Network: • Italy: ASI1 + new stations (Sardinia) = 147 sites • Processing: • Idem repro_LAREG IODAMED meeting, 16 May 2014
Comparisonof NRT to post-proc. data IGE Mean= 0.5 ± 3 mm Std. = 8 mm SGN Mean= -0.1 ± 2 mm Std. = 6mm ASI Mean= 0.8 ± 2 mm Std. = 4 mm IODAMED meeting, 16 May 2014
Screening of ZTD data 1st step Analysis of formal error for ZTD, station height, and other processing output information After Before mean 1 – 99 percentiles of ZTD formal error Std.dev. 2nd step Comparison of GPS ZTD data to ZTD data from a NWP model, rejection of data at ± 3 IODAMED meeting, 16 May 2014
Conversion of ZTD to IWV accurate to 0.006% Surface pressure or NWP Analysis accurate to 0.1-1 hPa (0.01-0.1%) accurate to 0.06% accurate to 10-3 m s-2(0.01%) Principle: IWV=K(Tm)*( ZTD – ZHD ) Requires auxiliairy data accurate to 2.4 % accurate to ±3 % but k’2 = 0.047 * k3/Tm Empirical model Tm = f(Ts) or NWP Analysis Tm= 250 ± 20 K (8% variations) IODAMED meeting, 16 May 2014
Uncertainty budget of GPS IWV ZTD uncertainty + ZHD uncertainty + K(Tm) uncertainty 1 mm (0.06% k1 error) + 1-2 mm (0.5-1 hPa Ps measurement error) + 1-2 mm (interpolation error) 1-2% (k2 & k3 error) + 1-2% (Tm error < 5K) 4-6 mm (IGS solution) IWV=40 kg/m2 IWV=5 kg/m2 0.3-0.5 kg/m2 0.1-1.1 kg/m2 0.6-0.9 kg/m2 GPS IWV uncertainty: bias≈ 0.2-0.8 kg/m2 random ≈ 0.2-1.5 kg/m2
Comparison of GPS IWV to AROME WMED Mean= 1.5 kg m-2 Std.= 1.5 kg m-2 Slope = 0.96 Correl = 0.92 16 networks, 1 month (Sep 2012) Auxiliairy data for IWV conversion: AROME WMED (surf + T) + emp. model Tm =f(Ts) IODAMED meeting, 16 May 2014
Next steps • Finalise validation of reprocessed GPS ZTD dataset • ZTD data screening (improve method, 1st step) • ZTD to IWV conversion (assess auxiliairay datasets) • Validate GPS IWV data by comparison with other data (MWR…) • Assimilation of reprocessed GPS ZTD data • AROME WMED reanalysis • Prepare data (BUFR files) and assimilation system • Test impact of assimilation, case studies • Assimilation in research models • Monitoring of radiosonde humidity biases • Validation of simulations (case studies) • Process studies (HPE) IODAMED meeting, 16 May 2014