Local and global calibration/validation P. Bonnefond, S. Desai, B. Haines, S. Nerem and N. Picot Jason-1 - T/P Sea Surface Height Formation Flying Phase (Jason-1 Cycles 1-21) ?
March 12-15, 2007Ocean Surface Topography Science Working Team Meeting Cal/Val Splinter Session Long-Term Sea-Surface Height Bias Estimates Nominal (T/P: MGDR-B+TMR+GSFC TVG; Jason-1: GDR-B) Haines et al.
Recent Results from the Corsica Calibration Site OSTST Meeting, Hobart, March 2007 CONCLUSIONCONCLUSION Jason-1 GDR SSH remains biased (high) –New GDR-B reduces bias ( epoch) from +112 to +81 mm. –New POE in GDR-B improves the standard deviation. No significant drift in Jason-1 bias (GDR-B) –Drift due to JMR wet path delay is removed (Senetosa wasn't affected by GDR-A orbit). No detectable drift in JMR path delay –~14 mm bias in JMR GDR-B remains (compared to GPS). –~19 mm / ECMWF. –Suggests JMR is too short at Senetosa. Explains half of the Harvest/Senetosa difference. Biases (SSH) in T/P altimetric measurement decreased with MGDR + –3 mm due to TMR replacement product (and 6mm more when compared to the old TMR correction). –-10mm due to the new GSFC orbits (included in the Retracked GDR products) Insignificant drift for ALT-B in T/P altimeter measurement systems T/P retracking increases the bias (less negative) –+13 mm for RGDR-1 (LSE) and +6 mm for RGDR-2 (MAP). –MAP (RGDR-2) retracking gives more noisy results than LSE (RGDR-1) one (standard deviation of 42 mm and 32 mm respectively). Numbers should be revised when new SSB and ionospheric corretion will be delivered Corsica Results (Harvest results): ALT-A (aging): -8 ±9 mm (-7 ± 3) ALT-B: -22 ±3 mm (-3 ±4) POSEIDON-2: +84 ±4 mm (+115 ±3) ALT-B (MGDR + ): Bias = -23 ±5 mm (-5 ±5) Slope = -1 ±3 mm/yr (-2 ±3) POSEIDON-2 (GDR-B): Bias = +81 ±9 mm (+114 ±6) Slope = +1 ±3 mm/yr (0 ±2) Bonnefond et al.
March 14, 2007Ocean Surface Topography Science Working Team Meeting Cal/Val Splinter Session Average of Differences w.r.t. Independent Measurements JMR comparisons to SSMI, TMI, and GPS statistically identical. TMR comparisons to GPS are biased by ~ -9 mm versus SSMI and TMI. Separation of GPS comparisons by island and continental sites shows large discrepancy for TMR, but not for JMR. TMR likely dominant source for errors near continents. Desai et al.
Further improvement expected using similar treatment for: Geographically correlated orbit errors, ALT-B waveform leakage, SSB correction… Haines & Bonnefond
Cycles 1 to 185 Jason-1 GDR-B Results Mean J-1 Bias: GDR-B Data: +107 mm Pg 8 of 11 Watson et al.
Jan et al.
Mitchum et al.
Bosch et al.
OSTST Hobart 2007 – Performance assessment Jason-1 data Conclusion Good performances of Jason-1 GDR data : 5.1 cm RMS at crossover, and stable Data performances better with GDR ‘B’: SSH variance at crossovers using GDR “A” [cm²] SSH variance at crossovers using GDR “B” [cm²] 0 cm²100 cm² 0 cm² 100 cm² Some improvements will be provided in next GDR release : –New SSB (Labroue, Venice 2007) –JMR corrections –New Orbits –New geophysical corrections :DAC (MOG2D HR) Variance gain = 21 cm² 35 % Ablain et al.
OSTST Hobart 2007 – Performance assessment Jason-1 data Variance of SLA differences Variance of Jason-1/TOPEX SLA differences computed after filtering out SLA signals smaller than 50 km (in order to remove the SSH high frequency content). Using GDR ‘A’ for Jason-1 and MGDR for T/P: – variance is about 7.4 cm² – larger differences in strong waves areas due to SSB discrepancies Using GDR ‘B’ for Jason-1 and RGDR for T/P: –variance is reduced by 2 cm² –Variance is mainly reduced in strong wave areas showing the better SSB consistency between Jason-1 and T/P. 0 cm²20cm ² GDR ‘A’ / MGDR GDR ‘B’ / RGDR 0 cm² 20cm ² Variance=5.31 cm² ( 2.30 cm RMS) Variance=7.37 cm² ( 2.71 cm RMS) Ablain et al.
OSTST Hobart 2007 – Performance assessment Jason-1 data Jason-1 and Envisat Mean Sea Level trends The EN and J1 MSL trend over the 4 years are not consistent However the EN and J1 MSL trend are consistent over period Selection on Latitude (<66°), seasonal signals removed Model wet tropo used Jason-1 Envisat Consistent Not consistent Faugere et al.
Brown et al. OSTST07-Hobart Summary TMR recalibration is complete T B drifts, gain and offset errors, and instrument temperature dependent errors were removed PD coefficients were reverted to pre-launch values TMR PDs are in good agreement with several validation sources –No drift compared to SSM/I –Low bias and negligible scale error compared to RaOb, GPS, and ECMWF After JMR PD coefficient adjustment, TMR and JMR are in excellent agreement –Although, there is still room for regional improvement –JMR calibration will be updated on version-C GDRs Brown et al.
March 14, 2007Ocean Surface Topography Science Working Team Meeting Cal/Val Splinter Session Validation of JMR Version B and Ongoing Recalibration JMR from GDRB All 4 comparisons agree to within 3.2 mm. Residual (< 3mm) yaw state dependence may remain after cycle 100. JMR-ECMWF has a drift of –0.3 mm/year. Recalibration of JMR –In preparation for Version C GDRs. Will be ~3mm drier than version B GDRs. –Scale error corrected. JMR-ECMWF drift reduced to –0.25 mm/yr –May have contribution from < 2 mm shifts after cycle 136 and 179 safeholds. –Will be corrected in Version C. Desai et al.
OSTST, March 12-15, 2007 – 15 Comparison between measures and simulations along track : Validated for 2 different characteristic configurations 2D maps are simulated: allow the evaluation in any possible configuration 2 – Radiometer simulator Obligis et al.
OSTST, March 12-15, 2007 – 16 Tournadre et al.
Key points discussed 1.What is best approach to aligning TOPEX and Jason for SSH time series? Need integrated approach that reconciles global differences with results (coastal) at calibration sites. Incumbent on Pis for cal sites to reconcile results/error budget for their (coastal) sites with expectations for global (open ocean) comparisons, e.g.: - Use in-situ (GPS) model path delays - Segregate results for low and high sea states (maybe regress against SWH, or eliminate low SWH overflight results) - Use competing orbit products to develop empirical (site specific) corrections for geographically correlated orbit errors. - Could be extended to other effects (e.g., waveform leakages on ALT-B) 2.Will the future new products (Retracked T/P GDR) and GDR-C answer to #1 question? Yes from global statistical approaches 3.Is there any improvement to be done before complete reprocessing? Some are under progress but inducing minor changes (e.g. JMR) 4.Is it possible to give a unique error budget? Global? Geographic distribution? Coastal and or inland versus open ocean (radiometer land contamination)? Priorities is to improve the correction for distance 50 to 250km See #1; also, in-situ calibration sites are able to separate the origin of errors 5.Need to identify/separate the constant and time varying parts of the errors 6.Is Formation Flight (6 months) period sufficient or do we need to increase it for Jason-1 / Jason-2? In balance with meso-scale studies. So 6 MONTHS IS ENOUGH 7.Implications of the change of orbits for Jason-3 on CALVAL activities? Needs to be evaluated in detail but keeping only one calibration site may not be reasonable. Moreover continuity is the cornerstone of the T/P Jason-1 missions