OVERVIEW OF IGS PRODUCTS & ANALYSIS CENTER MODELING

Jim Ray, NOAA/NGS

Jake Griffiths, NOAA/NGS

• Status of core products– focus on Ultra-rapid predicted orbits– issues with current products

• Comparisons of AC analysis strategies– evidence for systematic errors, esp. fortnightly harmonics

• Recommendations

IGS 2008 Workshop, Miami Beach, 2 June 2008

SUMMARY OF IGS CORE PRODUCTSSUMMARY OF IGS CORE PRODUCTSPRODUCT SUITES

# ACs

CURRENT

PRECISION

LATENCY UPDATES SAMPLE INTERVAL

QUALITY

ASSESSMENT

Ultra-Rapid (predicted)

• orbits

• SV clocks

• ERPs

7 (2)*

4

7 (2)*

< 5 cm

~5 ns

< ~1 mas

real time 03, 09, 15, 21

UTC

15 min

15 min

6 hr

marginally robust

extremely poor

very weak

Ultra-Rapid (observed)

• orbits

• SV clocks

• ERPs

7 (2)*

4

7 (2)*

~3 cm

~0.2 ns

~0.1 mas

3 - 9 hr 03, 09, 15, 21

UTC

15 min

15 min

6 hr

fairly robust

weak

fairly robust

Rapid • orbits

• SV, stn clocks

• ERPs

8

5

8

~2.5 cm

~0.1 ns

~0.06 mas

17 - 41 hr daily

15 min

5 min

daily

robust

marginally robust

robust

Final • orbits

• GLO orbits

• SV, stn clocks

• ERPs

• terr frame

8

4

6

8

8

~2.5 cm

< ~15 cm ?

~0.1 ns

~0.03 mas

3 (h), 6 (v) mm

13 - 20 d weekly

15 min

15 min

5 min, 30 s

daily

weekly

robust

not robust

robust for 5 min

robust

robust

* indicates AC contributions that are weaker than others

Predicted IGU Orbit WRMS

• WRMS of IGU orbit predictions have improved to <5 cm RMS

IGU Orbits (1st 6 hr of predictions) wrt IGR Orbits

• PRN29 (IIA) decommissioned• GOP solutions improved

Scale & Rotations of Predicted IGU Orbits

• Z rotations (UT1 prediction error) reach 1 mas level• equivalent to equatorial shift of 12.9 cm at GPS altitude

(shifted)

(shifted)

IGU Orbits (24 h of predictions) wrt IGR Orbits

Issues with Current Products• IGU orbit combination only marginally robust

– sometimes AC predictions are better than combo

Ultra-Rapid IGS Orbit Comparison for 1478_6_06 (10 May 2008 06h)CENT STA| DX DY DZ RX RY RZ SCL RMS WRMS MEDI | [mm] [mm] [mm] [uas] [uas] [uas] [ppb] [mm] [mm] [mm]--------+--------------------------------------------------------cou 73| 11 -1 -4 536 -389 254 -.29 64 34 33emu 49| 7 0 0 486 38 -60 .03 84 44 21esu 95| 4 5 -2 -396 687 -72 .13 77 77 29gfu 65| 1 -2 -2 302 -21 127 -.34 77 78 29gou 82| -5 -4 -1 260 334 48 -.35 89 78 31siu 62| 0 17 -33 -221 1068 730 .02 130 131 71usu 33| 19 9 0 297 -394 -20 .14 123 111 56

igu n/a| 5 0 -5 283 103 45 -.08 74 79 18

– would benefit from more high-quality ACs– accuracy now limited by ERP predictions, mostly

• may also apply to IGR orbits (but less severe)

• IGU combined clocks are very poor– clock predictions no better than BRDC– not enough clock ACs

• even IGR clocks sometimes weak when some ACs miss

COMPARISON OF AC DATA USAGECOMPARISON OF AC DATA USAGEANALYSIS CENTER

OBS TYPE ORBIT DATA ARC LENGTH

DATA RATE

ELEVATION CUTOFF

ELEVATION INVERSE WGTS

CODE DbDiff ( weak redundant)

24 + 24 + 24 h 3 min 3 deg 1/cos2(z)

EMR UnDiff 24 h 5 min 15 deg none

ESA UnDiff 24 h 5 min 10 deg 1/sin2(e)

GFZ UnDiff n x 24 h

n = 3 (Rapid = 2)

5 min 7 deg 1/2sin(e)

for e < 30 deg

GRGS (new) UnDiff 48 h 15 min 10 deg 1/cos2(z)

JPL UnDiff 3 + 24 + 3 h 5 min 15 deg → 7 deg none

MIT DbDiff (weak redundant)

24 h

(SRPs over 9d)

2 min 10 deg a2 + (b2/sin2(e))

a,b from site residuals

NGS DbDiff (redundant)

24 h 30 s 10 deg [5 + (2/sin(e)) cm]2

PDR (Repro) DbDiff (weak redundant)

24 + 24 + 24 h 3 min 3 deg 1/cos2(z)

SIO DbDiff (weak redundant)

24 h 2 min 10 deg a2 + (b2/sin2(e))

a,b from site residuals

effect of 15-deg cutoff

COMPARISON OF AC TIDAL MODELSCOMPARISON OF AC TIDAL MODELSANALYSIS CENTER

SOLID EARTH

EARTH POLE

OCEAN LOAD

OCEAN POLE

OCEAN CMC

SUBDAILY EOPs

CODE IERS 2003;

dehanttideinel.f

eqn 23a/b mean pole

FES2004;

hardisp.f

none sites & SP3 IERS 2003;

subd nutation

EMR IERS 2003 eqn 23a/b mean pole

FES2004;

gipsy

none sites & SP3 IERS 1996;

no subd nutation

ESA IERS 2003;

dehanttideinel.f

eqn 23a/b mean pole

FES2004;

hardisp.f

none sites & SP3 IERS 2003 & PMsdnut.for

GFZ IERS 1992 eqn 23a/b mean pole

FES2004;

hardisp.f

none sites & SP3 IERS 2003;

subd nutation

GRGS (new) IERS 2003 nominal mean PM

FES2002 none none IERS 2003 & PMsdnut.for

JPL IERS 2003 eqn 23a/b mean pole

FES2002;

gipsy

none none →

sites & SP3

IERS 1996 →

IERS 2003

MIT IERS 2003 eqn 23a/b mean pole

FES2004 none sites & SP3 IERS 2003 & PMsdnut.for

NGS IERS 2003;

dehanttideinel.f

eqn 23a/b mean pole

FES2004;

hardisp.f

none sites & SP3 IERS 2003 & PMsdnut.for

PDR (Repro) IERS 2003;

dehanttideinel.f

fixed mean pole

GOT00.2 w/ 11 terms

none none IERS 2003;

no subd nutation

SIO IERS 2003 eqn 23a/b mean pole

FES2004 none sites & SP3 IERS 2003 & PMsdnut.for

Aliased Tidal Peaks in PM Discontinuities

• Spectra of polar motion day-boundary discontinuities show signatures of aliased O1, Q1, & N2 tides + unknown 7.2 d line

Peaks in PM Differences

AC 14 d 9 d 7 d EMR PM-x

± 14.20.2

9.350.09

7.180.05

EMR PM-y±

14.10.2

9.6 & 9.00.1

7.160.05

GFZ PM-x±

14.20.2

9.40.1

7.210.05

GFZ PM-y±

14.20.2

9.6 & 8.90.1

7.140.05

JPL PM-x±

14.20.2

9.40.1

7.230.05

JPL PM-y±

14.20.2

9.20.1

7.260.05

• VLBI (1-hr) UT1 VLBI (1-hr) UT1 residuals residuals

– white over full frequency range

• 14.19-d periodic14.19-d periodic– treated as GPS artifact– amplitude varies between

5 & 11 μs – phase varies linearly w/ time

due to changing period

• GPS LOD residualsGPS LOD residuals– approximately white– with small peak at 13.7 d– possible difference in

a priori tidal models wrt VLBI

• Gauss-Markov values Gauss-Markov values for GPS LOD biasesfor GPS LOD biases

– peak-to-peak range = ± 40 μs

– RMS = 9 μs

EMR analysis upgrade

Kalman Filter of VLBI UT1 + GPS LOD(Senior, Kouba, Ray – EGU 2008)

14.12 d signal in IGS & C04

probably due to mix of

GPS errors

LODS – (AAM+OAM) spectra

Fortnightly Band – Spurious IGS LOD(Senior, Kouba, Ray – EGU 2008)

Day-boundary Orbit Discontinuities

• Orbit discontinuities between days show temporally correlated errors & broad fortnightly spectral peak

• From Griffiths & Ray (AGU 2007)

COMPARISON OF AC GRAVITY FORCE MODELSCOMPARISON OF AC GRAVITY FORCE MODELSANALYSIS CENTER

GRAVITY FIELD

EARTH TIDES

EARTH POLE

OCEAN TIDES

OCEAN POLE

RELATIVITY EFFECTS

CODE JGM3; C21/S21 due to PM

IERS 2003 IERS 2003

CSR 3.0 none dynamic corr &

bending applied

EMR JGM3; C21/S21 due to PM

freq-depend. Love #

IERS 2003

CSR none no dynamic corr;

bending applied

ESA EIGEN; C21/S21 due to PM

IERS 2003 IERS 2003

IERS 2003

none dynamic corr &

bending applied

GFZ JGM2; C21/S21 due to PM

Wahr Love # GFZ model

GEM-T1 none dynamic corr &

bending applied

GRGS (new) EIGEN; C21/S21 due to PM

IERS 2003 IERS 2003

FES 2004 none dynamic corr applied; no bending

JPL JGM3; C21/S21 due to PM

IERS 2003 IERS 2003

CSR → FES2004

none dynamic corr &

bending applied

MIT EGM96; C21/S21 due to PM

IERS 1992; Eanes Love #

none none none no dynamic corr;

bending applied

NGS GEM-T3; C21/S21 due to PM

IERS 1992; Eanes Love #

none none none no dynamic corr;

bending applied

PDR (Repro) JGM3; constant C21/S21

IERS 2003 except step 2

IERS96; fixed pole

CSR 3.0 none dynamic corr &

bending applied

SIO EGM96; C21/S21 due to PM

IERS 1992; Eanes Love #

none none none no dynamic corr;

bending applied

COMPARISON OF AC SATELLITE DYNAMICSCOMPARISON OF AC SATELLITE DYNAMICSANALYSIS CENTER

NUTATION & EOPs

SRP PARAMS

VELOCITY BRKs

ATTITUDE SHADOW ZONES

EARTH ALBEDO

CODE IAU 2000; BuA ERPs

D,Y,B scales; B 1/rev

every 12 hr + constraints

none E+M: umbra & penumbra

none

EMR IAU 1980; extrap. past 3d

X,Y,Z scales stochastic

none yaw rates estimated

E: umbra & penumbra

none

ESA IAU 2000; BuA ERPs

D,Y,B scales; B 1/rev

none; Along, Along 1/rev

accelerations

none E+M: umbra & penumbra

applied + IR

GFZ IAU 2000; GFZ ERPs

D,Y scales @ 12:00 + constraints

yaw rates estimated

E: umbra & penumbra

none

GRGS (new) IAU 2000; C04 + BuA ERPs

D,Y,B scales; D,B 1/rev

none none E+M: umbra & penumbra

applied + IR

JPL IAU 1980; BuB ERPs → BuA

X,Y,Z scales stochastic

none nominal yaw rates used

E+M: umbra & penumbra

applied

MIT IAU 2000; BuA ERPs

D,Y,B scales; B(D,Y) 1/rev

none; 1/rev constraints

nominal yaw rates used

E+M: umbra & penumbra

none

NGS IAU 2000; IGS PM; BuA UT1

D,Y,B scales; B 1/rev

@ 12:00 + constraints

none; delete eclipse data

E+M: umbra & penumbra

none

PDR (Repro) IAU 2000; BuA ERPs

D,Y,B scales; B 1/rev

every 12 hr + constraints

none E+M: umbra & penumbra

none

SIO IAU 2000; BuA ERPs

D,Y,B scales; D,Y,B 1/rev

none; 1/rev constraints

nominal yaw rates used

E+M: umbra & penumbra

none

COMPARISON OF AC TROPOSPHERE MODELSCOMPARISON OF AC TROPOSPHERE MODELSANALYSIS CENTER

METEO DATA

ZENITH DELAY

MAPPING FNCT

GRAD MODEL

ZENITH PARAMS

GRAD PARAMS

CODE GPT Saastamoinen dry

GMF

dry

none 2-hr contin. w/ GMF wet

24-hr NS + EW continuous

EMR ECMWF 6-hr grids

ECMWF

dry + wet

NMF

dry + wet

none 5-min stochastic ZTD

5-min stochastic

ESA GPT Saastamoinen dry

GMF

dry

none 2-hr contin. w/ GMF wet

none

GFZ GPT Saastamoinendry + wet?

GMF

dry + wet ?

none 1-hr constants w/ GMF ?

24-hr NS + EW constants

GRGS (new) ECMWF 6-hr grids

ECMWF

dry + wet

Guo

dry + wet

none 2.4-hr contin. w/ Guo dry

none

JPL none → GPT

dry=hgt scale

wet=0.1 m

NMF → GMF none 5-min stochastic ZTD

5-min stochastic

MIT GPT Saastamoinen dry + wet

GMF

dry + wet

none 2-hr contin. w/ GMF wet

NS + EW vary linearly

NGS GPT Saastamoinen dry + wet

GMF

dry + wet

none 1-hr constants w/ GMF wet

NS + EW vary linearly

PDR (Repro) Berg (1948)

Saastamoinen dry

IMF dry w/ ECMWF z200

none 2-hr contin. w/ NMF wet

24-hr NS + EW continuous

SIO GPT Saastamoinen dry + wet

GMF

dry + wet

none 2-hr contin. w/ GMF wet

NS +EW vary linearly

Conclusions

• Despite huge progress by IGS since 1994, numerous small systematic errors remain in products– see EGU 2008 presentation by J. Ray

http://www.ngs.noaa.gov/IGSWorkshop2008/docs/igs-errs_egu08.pdf

• Applications to cutting-edge science are currently limited– need to focus on identifying, understanding, & mitigating errors– should avoid rush to premature science conclusions– must renew basic GNSS research efforts, not just in geophysical

applications– requires accurate knowledge of AC processing strategies

• Improvements will probably require better station installations (to reduce near-field multipath biases) & analysis upgrades– more research into field configuration effects badly needed– need better leadership to popularize lessons learned– need better cooperation & coordination between analysts & network

Recommendations

• For more robust products:– recruit new or improved IGU ACs & more IGR clock ACs– investigate improved near-RT & predicted ERPs– should IGS start (UT1 + LOD) service ? (à la Senior et al., EGU08)

• Reject GGOS UAW actions for:– SINEX parameter & naming extensions– piecewise, continuous segment parameterization as SINEX standard

• Reject rigidly standardized AC procedures & parameterizations– would lead to stagnation & end of progress– would eliminate basis for multi-solution product combinations– but ACs must agree on conventional choices & use of modern models

• Instead, set up inter-service SINEX & combinations WG– investigate technique-specific systematic errors– maintain SINEX format

Recommendations (cont’d)

• Updated AC summaries are required:– EMR 23 Jan 2002– GFZ 27 Feb 2003– JPL 13 Apr 2004– SIO 31 Oct 2005– (USNO 12 Sep 2006)

• Suggest suspending ACs with no updates by 30 Sep 2008– if processing summary is older than 2 years– submissions would be rejected from IGS products after Sep 2008

• Rescind AC status if no updates by 31 Dec 2008– would need to formally rejoin IGS ACs after Dec 2008

• Or ask above ACs for effective alternative proposal

Backup Slides 1

A SURVEY OF SOME SYSTEMATIC ERRORS IN IGS PRODUCTS

Jim Ray, NOAA/National Geodetic Survey

• Clock jumps at day boundaries & near-field multipath

• Position time series show N * 1.04 cpy harmonics

• N/S distortions in IGS frames

• Earth rotation parameters smoothed & filtered

• Spurious tidal lines in EOPs

• Orbit discontinuities have fortnightly variations

EGU 2008 General Assembly, Paper G4 #A-01694, Vienna, 15 April 2008

Context• GPS errors are propagated formally but true data

noise is unknown– highly site-dependent & not white noise– e.g., variances of AC frame solutions differ by > x 100

• dealt with by empirical rescaling of covariance matrix

• Evidence for systematic effects in IGS product covariances is well known– e.g., user velocity errors are routinely inflated to

account for temporal correlation of position errors• but methods are purely empirical

• Objective: Survey systematic errors in some IGS product values– underlying causes mostly unknown or not confirmed

1) Day-boundary Clock Jumps• clock bias accuracy is based on

mean of code data per arc

• for 24-hr arc with code σ = 1 m, clock accuracy should be ~120 ps

• can study local code biases via clock jumps at day boundaries (H-maser stations only)

• observed clock jumps vary hugely among stations:110 ps to >1500 ps

• presumably caused mainly by local code multipath conditions, esp. in near-field of antenna

Near-field Multipath Mechanism• expect largest & longest-period MP

errors when height H of antenna is small [Elósegui et al., 1995]

• may have special problems when H is near multiples of λ/4

• reflected RCP GPS signals enter from behind as LCP

• choke-ring design esp sensitive to L2 reflections from below [Byun et al. 2002]

• most IGS RF antennas mounted over flat surfaces!

Correlated Clock & Position Effects: ALGO

• ALGO day-boundary clock jumps increase in winters

• every winter ALGO also has large position anomalies

– IGS deletes outliers >5 σ

• implies common near-field multipath effect is likely(phase & code)

Probably better to mount antennas away from closereflecting surfaces!

worse better

Other Hardware Choices Also Important

• receiver health, firmware, antenna model, & cables also affect day-boundary clock jumps

Rogue SNR-8000 receiver Ashtech UZ-12 receiver

Ashtech UZ-12 receiver

AOA firmware 3.2.32.8 3.2.32.11

AOA D/M_Tantenna

ASH701945E_Mantenna+new cables

PIE1

2) Stacked/Smoothed Spectra of Site Residuals

• for 167 IGS sites with >200 weekly points in 1996.0 – 2006.0• large annual + semi-annual variations• plus harmonics in all components at N * (1.040 ± 0.008) cpy• flicker noise spectra down to periods of ~few months

(shifted)

(shifted)

Position Harmonics Linked to GPS Year

• 1.040 ± 0.008 cpy fundamental does not match any expected alias or geophysical frequency– also not seen in VLBI, SLR, or fluid load spectra

• Closely matches GPS “draconitic” year– rotation period of Sun w.r.t. GPS nodes (viewed from Earth)– GPS nodal drift is -14.16° per year (due to Earth’s oblateness)– period = 351.4 day or frequency = 1.039 cpy

• Two possible coupling mechanisms suggested:1) direct orbit modeling errors (e.g., related to eclipse periods &

planes)2) alias of site position biases (e.g., near-field phase multipath) due to

beating of 24-hr processing arc against 23.93-hr GPS repeat period

– useful distinguishing tests not yet made

3) N/S Distortions of IGS Frames

N

E

U

• Weekly mean biases of IGS frames compared to long-term frame

IGS Frame Distortions

• N/S mean component of IGS weekly frames shows largest annual variation– after weekly 7-parameter Helmert alignment– also largest dispersion among ACs in N/S direction

• Not likely to be caused by annual inter-hemisphere fluid load cycle– load signal should be largest in heights, not N/S

• Could possibly be related to along-track GPS orbit errors– but no mechanism identified

• Likelier explanation: possible neglected 2nd order ionospheric effect

4) High-frequency Smoothing of EOPs

• Day-boundary continuity constraint by some ACs smoothes & filters EOP estimates near Nyquist limit

Filter/Smoother by Continuity Constraints

• Some ACs estimate EOPs (& others) by continuous linear segments– attenuates power by factor 4 at Nyquist limit– smoothes estimates– filters certain phase components

• To avoid contaminating IGS combination, such EOP solutions rejected since January 2008 (wk 1460)– but effects on other parameters probably still present

• Past high-frequency studies should be reconsidered

• Can use GFZ polar motion to estimate background, non-tidal, sub-daily variance: 13.6 to 20.7 μas2

5) Aliased Tidal Peaks in EOP Discontinuities

• Spectra of polar motion day-boundary discontinuities show signatures of aliased O1, Q1, & N2 tides + unknown 7.2 d line

Peaks in PM Differences

AC 14 d 9 d 7 d EMR PM-x

± 14.20.2

9.350.09

7.180.05

EMR PM-y±

14.10.2

9.6 & 9.00.1

7.160.05

GFZ PM-x±

14.20.2

9.40.1

7.210.05

GFZ PM-y±

14.20.2

9.6 & 8.90.1

7.140.05

JPL PM-x±

14.20.2

9.40.1

7.230.05

JPL PM-y±

14.20.2

9.20.1

7.260.05

6) Day-boundary Orbit Discontinuities

• Orbit discontinuities between days show temporally correlated errors & broad fortnightly spectral peak

Conclusions• Despite huge progress by IGS since 1994, numerous small

systematic errors remain in products

• Applications to cutting-edge science must recognize limitations– need to focus on identifying, understanding, & mitigating errors– must renew basic GNSS research efforts, not just in geophysical

applications– should avoid rush to premature science conclusions

• Improvements will probably require better station installations (to reduce near-field multipath) & analysis upgrades– more research into field configuration effects badly needed– need better leadership to popularize lessons learned– need better cooperation & coordination between analysts & network