Evaluation of the impact of atmospheric pressure loading modeling on GNSS data analysis

Transcription

1 Evaluation of the impact of atmospheric pressure loading modeling on GNSS data analysis R. Dach a, J. Böhm b, S. Lutz a, and P. Steigenberger c a Astronomical Institute, University of Bern, Bern, Switzerland b Institute of Geodesy and Geophysics, Vienna University of Technology, Vienna, Austria c Institut für Astronomische und Physikalische Geodäsie, TU München, Munich, Germany COST Action: ES71 improved constraints on models of glacial isostatic adjustment Geodetic observation-level modelling and systematic biases (WG1) Velocity determination/reference frame realization (WG2) Vienna University of Technology, Vienna, Austria; 16 to 17 November 21

2 Outline Introduction: CODE contribution to the IGS reprocessing effort (repro1). Atmospheric pressure loading model: e.g., Petrov and Boy (24) How to apply atmospheric pressure loading corrections? based on weekly solutions, on observation level, or with a scaling factor? Validation of the model by estimating scaling factors. Influence of atmospheric pressure loading on other GNSS related parameters. Conclusions and outlook Dach et al.: APL in GNSS analysis 2 / 27

3 Generation of the GNSS solution starting with observation files from CO1 repro. (GPS only) CODE standard processing is solving for CRD, TRP, ORB, ERP modeling: latest hardisp and troposphere VMF1/ECMWF Number of stations daily solution weekly NEQs cumulative solution from NEQs significant outliers and discont. using the FODITS tool of BSW NNR condition for coordinates and linear velocities on IGS5 reference frame sites Degree of freedom x1^ Year Year Dach et al.: APL in GNSS analysis 3 / 27

4 Coordinate time series Repeatability of the weekly solutions Example: Zimmerwald (ZIMM), Switzerland Height variations in mm No atm. loading corrections Year Dach et al.: APL in GNSS analysis 4 / 27

5 Pressure loading model: evaluation 1 Can atmospheric pressure loading explain these seasonal variations? Example: Atmospheric pressure loading model from Petrov and Boy, 24 consists of two components S1/S2 tidal pressure loading coefficients grids for the non tidal component every 6 hours Dach et al.: APL in GNSS analysis 5 / 27

6 Pressure loading model: evaluation 1 Can atmospheric pressure loading explain these seasonal variations? Example: Atmospheric pressure loading model from Petrov and Boy, 24 consists of two components S1/S2 tidal pressure loading coefficients grids for the non tidal component every 6 hours Evaluation 1: What impact on a GNSS solution can be expected from the model? Dach et al.: APL in GNSS analysis 5 / 27

7 Atmospheric loading model Atmospheric pressure loading model from Petrov and Boy, 24 Mean non tidal correction over 15 years Mean value in mm Dach et al.: APL in GNSS analysis 6 / 27

8 Atmospheric loading model Atmospheric pressure loading model from Petrov and Boy, 24 RMS of the non tidal correction over 15 years Standard deviation in mm Dach et al.: APL in GNSS analysis 6 / 27

9 Atmospheric loading model Atmospheric pressure loading model from Petrov and Boy, 24 RMS of the non tidal correction over 15 years 1 2 Standard deviation in mm Dach et al.: APL in GNSS analysis 6 / 27

10 Atmospheric loading model Atmospheric pressure loading model from Petrov and Boy, 24 How does the pressure loading model translate into the geocenter? Variation of the GCC in mm GCC X from grid GCC Y from grid GCC Z from grid Year Dach et al.: APL in GNSS analysis 7 / 27

11 Atmospheric loading model Atmospheric pressure loading model from Petrov and Boy, 24 How does the pressure loading model translate into the geocenter? Variation of the GCC in mm GCC X from grid GCC Y from grid GCC Z from grid GCC X from network GCC Y from network GCC Z from network Year Dach et al.: APL in GNSS analysis 7 / 27

12 Pressure loading model: evaluation 2 Can atmospheric pressure loading explain these seasonal variations? Example: Atmospheric pressure loading model from Petrov and Boy, 24 consists of two components S1/S2 tidal pressure loading coefficients grids for the non tidal component every 6 hours Evaluation 2: The non tidal part is averaged for each station over one week if the station was available for this week. The regression factors between the weekly mean effect from the model and the coordinate time series is evaluated. Dach et al.: APL in GNSS analysis 8 / 27

13 Pressure loading model: evaluation 2 Repeatability of the weekly solutions Example: Zimmerwald (ZIMM), Switzerland Height variations in mm No atm. loading corrections Repeatability: 3.76 mm Year Dach et al.: APL in GNSS analysis 9 / 27

14 Pressure loading model: evaluation 2 Repeatability of the weekly solutions considering atm. loading Example: Zimmerwald (ZIMM), Switzerland Height variations in mm No atm. loading corrections Repeatability: 3.76 mm Corrections from model, weekly mean Repeatability: 3.42 mm Year Dach et al.: APL in GNSS analysis 9 / 27

15 Pressure loading model: evaluation 2 Repeatability of the weekly solutions Example: Arti (ARTU), Russia Height variations in mm No atm. loading corrections Repeatability: 6.65 mm Year Dach et al.: APL in GNSS analysis 1 / 27

16 Pressure loading model: evaluation 2 Repeatability of the weekly solutions considering atm. loading Example: Arti (ARTU), Russia Height variations in mm No atm. loading corrections Repeatability: 6.65 mm Corrections from model, weekly mean Repeatability: 4.78 mm Year Dach et al.: APL in GNSS analysis 1 / 27

17 Pressure loading model: evaluation 2 Correlatogram between height variations and pressure loading Zimmerwald (ZIMM), Switzerland Arti (ARTU), Russia 3 3 Height variations in mm Height variations in mm Atmosphere loading in mm Regression coefficient: 1.3± Atmosphere loading in mm Regression coefficient: 1.27±.3 Dach et al.: APL in GNSS analysis 11 / 27

18 Pressure loading model: evaluation 3 Can atmospheric pressure loading explain these seasonal variations? Example: Atmospheric pressure loading model from Petrov and Boy, 24 consists of two components S1/S2 tidal pressure loading coefficients grids for the non tidal component every 6 hours Evaluation 3: Tidal component is directly applied to the observations Evaluation of the non tidal loading model by estimating scaling factors for each component and station scaling factor of one: model is fully confirmed These scaling factors are introduced as usual parameters in the analysis process and stacked on NEQ level in the cumulative solution. Dach et al.: APL in GNSS analysis 12 / 27

19 Estimated scaling factors for the model Estimated scaling factors for the atmospheric loading model Mean scaling factors over 15 years Scaling factor for model Dach et al.: APL in GNSS analysis 13 / 27

20 Scaling factors for the model Estimated scaling factors for the atmospheric loading model Mean scaling factors over 15 years 4 Number of the stations Scaling factor for model Dach et al.: APL in GNSS analysis 14 / 27

21 Scaling factors for the model Estimated scaling factors for the atmospheric loading model Mean scaling factors over 15 years RMS of the corrections over 15 years 4 1 Number of the stations Percent of the stations Scaling factor for model RMS of model in mm Dach et al.: APL in GNSS analysis 14 / 27

22 Scaling factors for the model Estimated scaling factors for the atmospheric loading model Mean scaling factors over 15 years Dev. from one, norm. with RMS 4 3 Number of the stations Number of the stations Scaling factor for model Dev. from one in mm Dach et al.: APL in GNSS analysis 14 / 27

23 Estim. scaling factors for the model, norm. by RMS Estimated scaling factors for the atmospheric loading model Deviation from one over 15 years, norm. with the RMS: Up Dev. from scaling factor one in mm Dach et al.: APL in GNSS analysis 15 / 27

24 Estim. scaling factors for the model, norm. by RMS Estimated scaling factors for the atmospheric loading model Deviation from one over 15 years, norm. with the RMS: North Dev. from scaling factor one in mm Dach et al.: APL in GNSS analysis 15 / 27

25 Estim. scaling factors for the model, norm. by RMS Estimated scaling factors for the atmospheric loading model Deviation from one over 15 years, norm. with the RMS: East Dev. from scaling factor one in mm Dach et al.: APL in GNSS analysis 15 / 27

26 Scaling factors for the model Repeatability of the weekly solutions considering atm. loading Example: Zimmerwald (ZIMM), Switzerland Height variations in mm Corrections from model, weekly mean Repeatability: 3.42 mm Corrections from model, obs. level Repeatability: 3.26 mm Year Dach et al.: APL in GNSS analysis 16 / 27

27 Scaling factors for the model Evaluation of the scaling factors for the atm. loading model Example: Zimmerwald (ZIMM), Switzerland Scaled loading effect in mm Scaling factors stacked per year Scaling factor over 15 years: 1.95 RMS of the loading model: 4. mm Year Dach et al.: APL in GNSS analysis 16 / 27

28 Scaling factors for the model Evaluation of the scaling factors for the atm. loading model Example: Zimmerwald (ZIMM), Switzerland Scaled loading effect in mm Scaling factors stacked per month Scaling factor over 15 years: 1.95 RMS of the loading model: 4. mm JAN FEB MAR APR MAY JUN JUL AUG SEP OCT NOV DEC Month Dach et al.: APL in GNSS analysis 16 / 27

29 Scaling factors for the model Repeatability of the weekly solutions considering atm. loading Example: Arti (ARTU), Russia Height variations in mm Corrections from model, weekly mean Repeatability: 4.78 mm Corrections from model, obs. level Repeatability: 4.37 mm Year Dach et al.: APL in GNSS analysis 17 / 27

30 Scaling factors for the model Evaluation of the scaling factors for the atm. loading model Example: Arti (ARTU), Russia Scaled loading effect in mm Scaling factors stacked per month Scaling factor over 15 years: 1.9 RMS of the loading model: 7.2 mm JAN FEB MAR APR MAY JUN JUL AUG SEP OCT NOV DEC Month Dach et al.: APL in GNSS analysis 17 / 27

31 Scaling factors for the model Evaluation of the scaling factors for the atm. loading model Example: Kourou (KOUR), French Guyana Scaled loading effect in mm Scaling factors stacked per month Scaling factor over 15 years: RMS of the loading model: 1.3 mm JAN FEB MAR APR MAY JUN JUL AUG SEP OCT NOV DEC Month Dach et al.: APL in GNSS analysis 18 / 27

32 Scaling factors for the model Evaluation of the scaling factors for the atm. loading model Example: La Misere (SEY1), Seychelles Scaled loading effect in mm Scaling factors stacked per month Scaling factor over 15 years:.421 RMS of the loading model:.9 mm JAN FEB MAR APR MAY JUN JUL AUG SEP OCT NOV DEC Month Dach et al.: APL in GNSS analysis 18 / 27

33 Repeatability of the coordinate solution Repeatability of the weekly solutions No atm. loading corrections 1 Repeatability (up) in mm 5 Stations, sorted by magnitude of the pressure loading Dach et al.: APL in GNSS analysis 19 / 27

34 Repeatability of the coordinate solution Repeatability of the weekly solutions considering atm. loading No atm. loading corrections Corrections from model, weekly mean Repeatability (up) in mm 1 5 Stations, sorted by magnitude of the pressure loading Dach et al.: APL in GNSS analysis 19 / 27

35 Repeatability of the coordinate solution Repeatability of the weekly solutions considering atm. loading 1 No atm. loading corrections Corrections from model, obs. level Corrections from model, weekly mean Repeatability (up) in mm 5 Stations, sorted by magnitude of the pressure loading Dach et al.: APL in GNSS analysis 19 / 27

36 Repeatability of the coordinate solution Repeatability of the weekly solutions considering atm. loading 1 No atm. loading corrections Corrections from model, obs. level Corrections from model, weekly mean Corrections from model, scaling factor Repeatability (up) in mm 5 Stations, sorted by magnitude of the pressure loading Dach et al.: APL in GNSS analysis 19 / 27

37 Fictive Variation of the Geocenter Comparison between the model and the solution GNSS: difference of the translation parameters with/without atm-load model Variation of the GCC in mm GCC X from model GCC Y from model GCC Z from model GCC X from GNSS GCC Y from GNSS GCC Z from GNSS Year Dach et al.: APL in GNSS analysis 2 / 27

38 Pressure loading model: results for far Can atmospheric pressure loading explain these seasonal variations? Example: Atmospheric pressure loading model from Petrov and Boy, 24 consists of two components S1/S2 tidal pressure loading coefficients grids for the non tidal component every 6 hours Results so far: Atmospheric pressure loading does not explain the seasonal variations. Nevertheless, the effect was identified in the GNSS coordinate time series: confirmed by estimated scaling factors improvement of the repeatability of the weekly station coordinate time series a systematic impact on the estimated geocenter coordinates was found Dach et al.: APL in GNSS analysis 21 / 27

39 Pressure loading model: evaluation 4 There are influences on other GNSS related parameters? Example: Atmospheric pressure loading model from Petrov and Boy, 24 consists of two components S1/S2 tidal pressure loading coefficients grids for the non tidal component every 6 hours Evaluation 4: GNSS Orbits compensation for the variation of the GCC (daily independent GCC) Estimation of the GNSS orbits by 15 parameters: initial conditions constant and once per revolution in a DYX satellite Sun oriented system no constraints to the orbit parameters Dach et al.: APL in GNSS analysis 22 / 27

40 Pressure loading in the GNSS orbits Comparison between the model and the solution GNSS: translations between the orbits with/without atm-load model Variation of the GCC in mm GCC X from grid GCC Y from grid GCC Z from grid GCC X from orbits GCC Y from orbits GCC Z from orbits Year Dach et al.: APL in GNSS analysis 23 / 27

41 Pressure loading in the GNSS orbits Comparison between the model and the solution 5 GCC X from grid GCC Y from grid GCC Z from grid GCC X from orbits GCC Y from orbits GCC Z from orbits amplitude(gcc) in mm Days Dach et al.: APL in GNSS analysis 23 / 27

42 Pressure loading model: evaluation 4 There are influences on other GNSS related parameters? Example: Atmospheric pressure loading model from Petrov and Boy, 24 consists of two components S1/S2 tidal pressure loading coefficients grids for the non tidal component every 6 hours Evaluation 4: GNSS Orbits compensation for the variation of the GCC (daily independent GCC) Estimation of the GNSS orbits by 15 parameters: initial conditions constant and once per revolution in a DYX satellite Sun oriented system with constraints to once per revolution parameters, as CODE Dach et al.: APL in GNSS analysis 24 / 27

43 Pressure loading in the GNSS orbits Comparison between the model and the solution GNSS: translations between the orbits with/without atm-load model Variation of the GCC in mm GCC X from grid GCC Y from grid GCC Z from grid GCC X from orbits GCC Y from orbits GCC Z from orbits Year Dach et al.: APL in GNSS analysis 25 / 27

44 Pressure loading in the GNSS orbits Comparison between the model and the solution 5 GCC X from grid GCC Y from grid GCC Z from grid GCC X from orbits GCC Y from orbits GCC Z from orbits amplitude(gcc) in mm Days Dach et al.: APL in GNSS analysis 25 / 27

45 Conclusions and outlook The effect of atmospheric loading can be clearly seen in GNSS derived coordinate time series (weekly solutions) and need to be corrected for to generate a reference frame. (compatibility between the solutions/techniques) Atmospheric loading models can be used to correct for this effect an improvement of the repeatability of up to 2% can be achieved. The correction has to be preferably done at the observation level (at least a weekly coordinate solution is a too long interval). The GCC component of the atmospheric loading translates into the satellite orbits if no proper a priori constraining of the once per revolution parameters is introduced. Dach et al.: APL in GNSS analysis 26 / 27

46 Conclusions and outlook The atmospheric loading model from Petrov and Boy (24) has been confirmed by the estimation of station wise scaling factors within the expected uncertainty range. Dach et al.: APL in GNSS analysis 27 / 27

47 Conclusions and outlook The atmospheric loading model from Petrov and Boy (24) has been confirmed by the estimation of station wise scaling factors within the expected uncertainty range. The latency of the model from Petrov and Boy (24) is insufficient for the IGS final processing. Dach et al.: APL in GNSS analysis 27 / 27

48 Conclusions and outlook The atmospheric loading model from Petrov and Boy (24) has been confirmed by the estimation of station wise scaling factors within the expected uncertainty range. The latency of the model from Petrov and Boy (24) is insufficient for the IGS final processing. Technical requirements for an atmospheric loading model from the perspective of an IGS analysis center: consistent time series from 1994 to present for other techniques even earlied latency less than 3 days for the finals (with exceptions up to 7 days) three to four hours after midnight for the rapid Dach et al.: APL in GNSS analysis 27 / 27