Data CalVal |
Annual Synthesis | T/P Status |
Side-B TOPEX Altimeter Evaluation
CLS/DOS/NT/99.175 |
|
Contents:
A low degradation of TOPEX Side-A altimeter performances had been noticed since approximately cycle 130. SWH estimation was mainly impacted, but no significant influence for most applications could be evidenced. Only studies concerned with signals of less than 1 cm (as Mean Sea Level (MSL) precise monitoring) could be affected (AVISO/CALVAL yearly report, 1999).
The objective of this document is to provide users with a first evaluation of Alt-B data in terms of altimeter parameters and Sea Surface Height (SSH) estimation. The results have been obtained at CLS from the CNES AVISO/CALVAL contribution to the Alt-B CAL/VAL plan. They are based on the Alt-B data processed by AVISO (M-GDRs) in the version that will be disseminated to users.
In the first days of Alt-B, all available types of data were used for first evaluation. In particular, Near Real Time U.S. Navy products were processed and lead to preliminary good news in terms of data quality. Then "official" IGDRs and GDRs were disseminated to the CAL/VAL group in order to get results for the first Alt-B CAL/VAL meeting which took place at GSFC, Washington, on April 22. Taking the meeting conclusions into account, an other version of GDRs was produced, including some parameter adjustments (Callahan, 1999). This version has been processed by AVISO to produce the M-GDRs that are evaluated here.
Cycle by cycle SSH crossovers are routinely computed, so it is easy to monitor the standard deviation of SSH differences, which gives a rough estimate of the data quality. Figure 1 shows the cycle standard deviation as a function of the cycle number. Side-B cycles are represented by red symbols and POSEIDON cycles by dark symbols. From this figure, no difference between Alt-A and Alt-B results can be evidenced. Alt-B cycle standard deviation values are consistent with previous Alt-A values. The result of cycle 236 only differs because of a low number of crossover points (cycle 236 has very important gaps).1. Background
A forecast of the degradation of altimeter parameters was estimated extrapolating the Point Target Response (PTR) change (Hayne et al., 1998). The conclusion was that the signal would be unrecoverable around cycle 250. So it was decided to switch on the Side-B altimeter in order to carry on with the mission. A calibration phase was thus necessary to assess data precision and to compare Alt-B performances relative to Alt-A.2. Objectives
After data quality assessment, the main altimeter parameters are monitored and compared to Alt-A. Then the relative bias between the two TOPEX altimeters is estimated in terms of SSH, and a Sea State Bias (SSB) evaluation is performed. As ERS-2 provides an external source of data for comparison, the main results from TOPEX/ERS-2 cross-calibration are also presented.
A number of figures included in this document involve results from the beginning of the TOPEX/POSEIDON mission. As the objective, here, is mainly to compare Alt-A and Alt-B data, the general comments are not necessarily mentioned and can be found in the AVISO/CALVAL annual report (Dorandeu et al., 1999).3. Results from Alt-B M-GDRs
3.1. Data description
No modification of the AVISO processing was done. In particular the Wallops calibration corrections (Hayne et al., 1999) are actually applied since they are also available for the Side-B altimeter. The altimeter range offsets applied for both TOPEX and POSEIDON data remain unchanged. Since no software modification has been done, the Alt-B data can be directly compared to Alt-A and to POSEIDON.
The first 6 Alt-B cycles (236 – 241) have been used for this study. All the previous Alt-A data serve as a reference and the last POSEIDON cycle (234) is also useful to cross-calibrate the two TOPEX altimeters.3.2. First Alt-B data quality evaluation
The standard deviation of Sea Level Anomalies (SLA) relative to a 4 year mean has also been plotted on figure 2 as a function of time. A good consistency between Alt-B results and last Alt-A values is obtained.
From these two analyses, no anomaly can be detected in terms of Alt-B SSH variance as compared to Alt-A variance. It leads to the conclusion of good quality for the first Alt-B data.
) |
) |
|||
| Figure 1: TOPEX/POSEIDON crossover standard deviation (cm) from cycle 11 to cycle 241. Dark symbols: POSEIDON cycles, red symbols: Alt-B cycles (236 - 241). | Figure 2: Sea Level Anomaly cycle standard deviation (cm) relative to a 4-year mean. Dark symbols: POSEIDON cycles, red symbols: Alt-B cycles (236 - 241). |
3.3.1. RMS of elementary measurements
) |
This parameter is given for each M-GDR measurement and computed among the 10 Hz high rate ranges. It has been averaged over one cycle and plotted as a function of the cycle number on figure 3 (mean and standard deviation). Alt-B cycles are represented by red dots while dark symbols represent Alt-A cycles at the same season and before the PTR change becomes noticeable. All this section will conform to the same convention. |
| Figure 3: Cycle mean and standard deviation of the RMS of elementary measurements (mm). Red symbols: Alt-B, Dark symbols: Alt-A at the same season as Alt-B cycles. |
3.3.2. SWH
) |
This parameter was the most impacted by the PTR change. The Alt-A SWH cycle mean experienced a large increase (of about 30 cm) after approximately cycle 130 (figure 4). Alt-B values are back to usual values of Alt-A before the change. The SWH overestimation is thus solved by switching to Side-B. In terms of standard deviation, the SWH parameter is consistent with Alt-A given the seasonal variations. |
| Figure 4: Cycle mean and standard deviation of Ku Significant Wave Height (cm). Red symbols: Alt-B, Dark symbols: Alt-A at the same season as Alt-B cycles. |
) |
The cycle mean and standard deviation of the difference between the two bands (Ku and C) are plotted on figure 5. It shows an evolution of the Side-A altimeter from the very beginning of the mission, though this mean difference only varies by about 2.5 cm. Alt-B values are very close to those of last Alt-A cycles. The cycle standard deviation again shows a continuous evolution probably due to some instrument degradation. It is impressive to see that with Alt-B, the standard deviation of the (Ku – C) SWH difference is back to low values, as it was the case in the first days of Alt-A. |
| Figure 5: Cycle mean and standard deviation of (Ku – C) SWH differences (cm). Red symbols: Alt-B, Dark symbols: Alt-A at the same season as Alt-B cycles. |
In order to better compare the SWH distributions of Alt-A and Alt-B, 3 scatter diagrams of C SWH as a function Ku SWH have been built for 3 different periods: for Alt-A cycle 51 (before the PTR change and at the same season as first Alt-B cycles), for Alt-A cycle 235 (the last Alt-A cycle) and for cycle 237 (Alt-B).
These diagrams are respectively presented on figures 6, 7 and 8. Figures 6 and 8 are very similar: the shape of the distributions are the same, with a maximum density of points around 2 m in the two cases. Equivalent dispersions are also obtained for both cycles 51 and 237: the standard deviation of (Ku – C) SWH difference is about 17 cm. On the opposite, the last Alt-A cycle 235 differs from the other two cases. The dispersion is higher (standard deviation of about 20 cm), and the distribution is splitted into two distinct spikes with a separation around SWH values of 3 m. The SWH estimation was thus degraded for the last Alt-A cycles. With the Side-B altimeter, SWH estimations are in accordance with Alt-A before the PTR changes.
) |
) |
) |
| Figure 6 | Figure 7 | Figure 8 |
| Figure 6: | Scatter diagram of C SWH as a function of Ku SWH for Alt-A cycle 051 (before PTR change becomes perceptible). |
| Figure 7: | Scatter diagram of C SWH as a function of Ku SWH for Alt-A cycle 235 (last Alt-A cycle). |
| Figure 8: | Scatter diagram of C SWH as a function of Ku SWH for Alt-A cycle 237 (Alt-B). |
3.3.3 Backscatter coefficient
) |
The Ku band Sigma0 cycle mean and standard deviation are plotted on figure 9. The mean is nearly constant during the first 3 years since correction tables have been accounted for until cycle 132 in the M-GDR reprocessing. In the second part of the period, variations are noticed but Alt-B values are consistent with previous Alt-A results, considering the exhibited scattering. In terms of standard deviation, comparable values are obtained with both altimeters at the same season. |
| Figure 9: Cycle mean and standard deviation of Ku backscatter coefficient (10-2 dB). Red symbols: Alt-B, Dark symbols: Alt-A at the same season as Alt-B cycles. |
) |
As far as the difference between the backscatter coefficients in the two bands is considered, figure 10 shows large variations during the second part of the mission. Alt-B cycle means are approximately of the same order as those of previous Alt-A cycles, but with high dispersion among these estimates. In terms of standard deviation of the difference, Alt-B is now consistent with first values of Alt-A. Notice that the Alt-A cycle standard deviation had continuously decreased as a function of time. |
| Figure 10: Cycle mean and standard deviation of (Ku – C) backscatter differences (10-2 dB). Red symbols: Alt-B, Dark symbols: Alt-A at the same season as Alt-B cycles. |
3.3.4 Ionosphere correction
) |
Seasonal comparisons with past Alt-A estimates are difficult since the solar activity has evolved, increasing much in the last cycles. Figure 11 shows that the goal to obtain no bias between Alt-A and Alt-B ionosphere corrections is achieved. |
| Figure 11: Cycle mean and standard deviation of dual-frequency ionosphere correction (mm). Red symbols: Alt-B, Dark symbols: Alt-A at the same season as Alt-B cycles. |
) |
In the AVISO/CALVAL processing, the TOPEX ionosphere correction is routinely filtered using a low-pass filter with a cut-off wavelength of 300 km. It is thus easy to plot the mean and the standard deviation of the difference between before and after filtering. Figure 12 shows variations in this difference for the Side-A altimeter. The standard deviation increase is probably due to more noise in the dual-frequency correction at the end. Now with Alt-B, both mean and standard deviation recover the same level as Alt-A before changes become perceptible. |
| Figure 12: Cycle mean and standard deviation of the difference between raw and filtered TOPEX ionosphere correction. Red symbols: Alt-B, Dark symbols: Alt-A at the same season as Alt-B cycles. |
This section is devoted to the estimation of the relative bias between the two TOPEX altimeters in term of SSH. Repeat-track analysis is used to compare mean SSH estimates at the moment of the switch and to analyse the consistency of Alt-B results with the full Alt-A MSL time series.
3.4.1. Estimation relative to the last Alt-A cycle (235)
This comparison obviously accounts for cycle 235 characteristics but provides an estimate of the potential gap between Alt-A and Alt-B in terms of SSH estimation. Sea Level Anomalies (SLA) relative to cycle 235 are averaged over one cycle and are plotted on figure 13 from cycle 236 to cycle 241. The cycle standard deviation increases due to more and more oceanic signal. The global mean over the 6 estimates of (Alt-B – Alt-A) SSH difference is about -1.3 cm (Alt-A measuring shorter). This figure has been obtained using the NASA orbit. Using the CNES orbit also leads to the same results, as shown on figure 14.
) |
) |
|||
| Figure 13: (Alt-B – Alt-A) mean SSH differences relative to cycle 235. NASA orbit. | Figure 14: (Alt-B – Alt-A) mean SSH differences relative to cycle 235. CNES orbit. |
3.4.2 Estimation relative to the last POSEIDON cycle (234)
The vicinity of a POSEIDON cycle provides an external means of comparison between Alt-A and Alt-B. The results of SLA relative to POSEIDON cycle 234 are reported on figure 15. While the relative bias between TOPEX and POSEIDON was about 1.7 cm in the last Alt-A cycles, it is now much closer to zero with Alt-B (about 0.3 cm).
) |
Figure 15: (TOPEX - POSEIDON) mean SSH differences relative to cycle 234. |
3.4.3. Comparison to the Alt-A MSL time series
To avoid any seasonal signal due to mean atmospheric pressure variations, no inverse barometer correction has been applied when computing the cycle by cycle MSL relative to a 4-year mean. This leads to the figure 16 on which dark symbols are POSEIDON cycles and Alt-B estimates are represented in red. The MSL remains steady until the end of 1996 (around cycle 150). During the large 1997 El-Niño event, MSL rise and fall are observed. In the last Alt-A cycles, the high level of the mean Sea Surface Height can be considered as suspicious: due to altimeter degradation, the SSH estimations are impacted by the range measurement modification and also via the SSB correction due to SWH increase.
) |
Figure 16: Comparison of MSL estimations from TOPEX-A, TOPEX-B and POSEIDON data. Dark symbols: Poseidon cycles, red symbols: Alt-B cycles (236-241) |
Over the 6 estimates from cycle 236 to cycle 241, the mean Alt-B MSL is about 0. This means that it is consistent with the mean computed over the first 4 years of TOPEX/POSEIDON data. The relative bias between Alt-B and POSEIDON is thus also reduced to a few millimetres.
Significant MSL variations can locally occur, at the 1 cm level over a few cycles (for instance, higher values are noticed for cycle 151, 152 and 153). So MSL monitoring over longer time series will be necessary to definitely conclude about the consistency between Alt-A and Alt-B MSL estimations.
The one-parameter SSB model has been adjusted at crossovers for both Alt-A and Alt-B. The SWH coefficient, expressed in percentage of SWH, is plotted as a function of the cycle number on figure 17. A decreasing trend is observed over the full time series, the coefficient increasing in absolute value. This is conflicting with the SWH increase observed for two years and therefore suggests that the range measurement has also altered. The estimates deduced from Alt-B data do not move away from the Alt-A trend, showing no major difference between the two TOPEX altimeters.
) |
Figure 17: BM1 one-parameter model coefficient (% of SWH) from cycle 11 to cycle 241. Alt-B cycles are in red. |
In order to better characterise the Alt-B SSB versus Alt-A, the 3-parameter model (BM3) has been estimated over the 6 alt-B cycles and compared to the NASA-BM3 model. This model was derived from Alt-A data and used in the M-GDR (AVISO User Handbook). The relative bias, expressed as a percentage of SWH is plotted as a function of the wind speed on figure 18. No significant difference between Alt-A and Alt-B can be inferred: the shape and the values of the new BM3 estimations are consistent with the NASA-BM3 model.
) |
Figure 18: BM3 Alt-B relative bias estimations (% of SWH) compared to the NASA BM3 model. |
From this preliminary study (given the low number of Alt-B cycles), no major difference between A and B altimeters can be noticed in terms of SSB. However, this parameter will have to be monitored to compare Alt-A and Alt-B in the long term.
ERS-2 OPR data from cycle 31 to cycle 41 are used to compare ERS-2 and TOPEX SSH, SWH and sigma0 estimations using dual crossover analysis. Averaging crossover differences for each TOPEX cycle allows comparison of Alt-A and Alt-B results relative to ERS-2.
3.6.1. SSH cross-calibration
ERS-2 data have been updated with the CSR3.0 tidal model to get the best homogeneity as possible with TOPEX/POSEIDON data. Both "SPTR" and "USO" corrections have also been applied to ERS data as recommended by ESA/ESRIN. Mean (ERS-2 – TOPEX) differences are plotted on figure 19 as a function of TOPEX cycle numbers. Only time lags lower than 10 days are selected to avoid too much contamination by oceanic signal.
) |
Figure 19: (ERS-2 – TOPEX) SSH crossover differences (cm). (ERS-2 – Alt-B) differences are in red. |
A rough 1 cm bias can be estimated between Alt-A and Alt-B mean SSH around the switch. But large variations are observed in the (ERS-2 – TOPEX) differences because of ERS-2 orbit errors and range bias jumps. Therefore, it is difficult to infer conclusions in terms of Alt-B / Alt-A relative bias at the 1 cm level. This method could however be useful to monitor the ERS-2 and TOPEX/POSEIDON relative bias in the long term.
3.6.2. SWH cross-calibration
The same method is used to compute SWH differences between ERS-2 and TOPEX with time lags lower than 1 hour. The results are plotted on figure 20. A decreasing trend is noticed in the (ERS-2 – TOPEX Alt-A) SWH difference. It corresponds to the Alt-A SWH increase.
) |
Figure 20: (ERS-2 – TOPEX) SWH crossover differences (cm). (ERS-2 – Alt-B) differences are in red. |
A gap of about 30 cm is clearly evidenced between Alt-A and Alt-B estimations. So the results from TOPEX SWH monitoring are fully confirmed. This method thus proves particularly effective to cross-calibrate two different altimeters in terms of SWH.
3.6.3. Sigma0 cross-calibration
Both Ku and C Sigma0 offsets have been applied in the TOPEX data processing in order to get similar backscatter coefficients for Alt-A and Alt-B. Consequently no marked difference was detected in the altimeter parameter monitoring (section 3.3.3). The cross-calibration relative to ERS-2 (figure 21) confirms that, for cycles around the TOPEX switch, there is no significant bias between Alt-A and Alt-B in terms of Sigma0. However, the large Alt-A Sigma0 variations observed in the past suggest that cross-calibration will have to be maintained to detect potential drifts.
) |
Figure 21: (ERS-2 – TOPEX) Sigam0 crossover differences (10-2 dB). (ERS-2 – Alt-B) differences are in red. |
Based on first Side-B TOPEX altimeter cycles, a conclusion of good data quality can be inferred. The precision in terms of SSH estimation is at the expected level in the light of what was obtained with Alt-A.4. Conclusion
The problems linked to Alt-A PTR anomalies are solved using Alt-B. In particular, the SWH estimation is now consistent with the beginning of the mission. Both backscatter coefficient and TOPEX ionosphere correction have been adjusted on Alt-A using offsets in the TOPEX Ground Segment software, to get no gap between the two altimeters. In terms of SSH estimation, a gap of about 1 cm can be noticed between Alt-B and the last Alt-A cycles. However, the MSL deduced from Alt-B data is now more consistent with what was obtained in the first 4 years, and also with POSEIDON estimates.
Altimeter monitoring and cross-calibration with ERS-2 prove to be necessary to link new altimeter data to historical data. More analyses using these methods will provide means to detect potential drifts and so ensure the TOPEX/POSEIDON mission continuity.
