RAIES: ENVISAT RA2 INDIVIDUAL ECHOES AND S-BAND DATA FOR NEW SCIENTIFIC APPLICATIONS FOR OCEAN, COASTAL, LAND AND ICE REMOTE SENSING

Transcription

1 RAIES: ENVISAT RA2 INDIVIDUAL ECHOES AND S-BAND DATA FOR NEW SCIENTIFIC APPLICATIONS FOR OCEAN, COASTAL, LAND AND ICE REMOTE SENSING C. Gommenginger (1), P. Challenor (1), G. Quartly (1), J. Gomez-Enri (1), M. Srokosz (1), A. Caltabiano (1), P. Berry (2), L. Mathers (2), J. Bennett (2), D. Cotton (3), D. Carter (3), I. LeDuc (4), C. Rogers (4), J. Benveniste (5) (1) Southampton Oceanography Centre, European Way, Southampton, SO14 3ZH, UK (2) De Montfort University, EAPRS, The Gateway, Leicester, LE1 9BH, UK (3) Satellite Observing Systems, 15 Church Street, Godalming, GU7 1EL, UK (4) SciSys (Space & Defence) Ltd, Methuen Park, Chippenham, Wiltshire, UK, SN14 0GB (5) ESRIN, Frascati, IT ABSTRACT RAIES is an ESA-funded study for the scientific exploitation of ENVISAT RA2 Individual Echo and S-band data for ocean, coastal, land and ice remote sensing. This paper will introduce some of the the new scientific applications of RA2 data for the ocean and the coastal zone, including improved rain, wind and wave products using Ku and S-band data. It will also introduce new science applications that will be made possible with RA2 individual echoes or averaged waveforms for ocean, the coastal zone, land and ice remote sensing. The paper will summarise the achievements and scientific findings of the study to date, highlight existing difficulties and provide recommendations to facilitate the exploitation of RA2 data. 1. INTRODUCTION The accurate determination of the height of the satellite in reference to the sea surface is one of the basic requirements of satellite altimetry. However, this precise measurement has to be corrected for atmospheric delays caused by wet and dry tropospheric effects and ionospheric effects. The dispersive nature of the ionosphere means these delays are frequency dependent. Ionospheric delays are thus corrected with a first-order linear combination of altimeter measurements taken simultaneously at two separate microwave frequencies. In principle, S-band should provide improved correction due to the larger frequency separation from Ku-band. RA2 altimeter on ENVISAT operates a dualfrequency system at Ku (13.575GHz) and S-band (3.2 GHz), thus providing altimeter measurements at a lower frequency than C-band. The aims of this study are to examine new and improved scientific applications using S-band or dual frequency products that incorporate S-band. One of the specific objectives of this research is to test the notion that lower microwave frequencies (S-band) may be less sensitive to high frequency variability, and provides more robust estimates of some geophysical parameters, especially in extreme conditions. RA2 is also the first satellite-borne altimeter to allow access to up to 2000 unaveraged individual echoes (IE) samples at Kuband. For the first time it will be possible to investigate the waveform phase, as well as perform pulse-to-pulse correlation in order to study the evolution of waveform at boundaries (sea-ice, landsea) and over rapidly changing surface conditions. This paper is structured as follow: Section 2 starts by reviewing the datasets used in this study. Section 3 presents our results for the scientific application of S- band data for the ocean and the coastal zone. Section 4 will present the scientific applications of S-band for land and ice surfaces. Section 5 will review IE scientific applications for oceans, land and ice surfaces. And Section 6 summarises our conclusions and recommendations. 2. LEVEL 2 RA2 DATASETS Various types of Level 2 RA2 datasets were available and used in different areas of the project. Unless otherwise stated, analyses linked with rain studies were based on data extracted from DVDs, which corresponds to post-calval delivered products. Wind and wave studies were performed with IMAR datasets extracted from the Radar Altimeter Database System (RADS) developed by Delft Institute for Earth- Oriented Space Research (DEOS). Analyses were carried out on a global scale for cycles 15 to 26, which corresponds approximately to the period of May 2003 to May Some datasets displayed marked differences with data in other L2 products from the Proc. of the 2004 Envisat & ERS Symposium, Salzburg, Austria 6-10 September 2004 (ESA SP-572, April 2005)

2 same cycles, while some products displayed changes in the characteristic of the geophysical parameters during the study (Fig.1). As can be noticed, there are clear discrepancies in the distribution of σ 0 at Ku-band in IMAR and GDR products. At the time of writing, reprocessed SGDR products were not available. of difference, values of difference greater than 5 m (Ku-band overestimation) or lower than 5 m (S-band overestimation) were mapped for Cycle 25 (Fig. 3). As expected, it is possible to observe large S-band overestimation, with difference values up to 15 m. In general, large differences are located at the sea-ice boundary, with clear indication of the ice cap in the Southern Ocean. Other larger differences can be observed at Southeastern Asia and over the Pacific warm pool. This could suggest a possible contamination by rain or a difficulty for S-band pulses to deal with an area of several interfaces sea-land-sea. For all the other cycles (not shown), this regional tendency can also be observed. Fig. 1. Distribution of Ku and S-band σ 0 for some of L2 products available 3. SCIENTIFIC RESULTS WITH S-BAND OVER OCEAN AND COASTAL ZONE This section will present the results of the scientific applications of S-band over oceans: improved wind / wave products, rain cell detection and improved ionospheric correction. Fig. 2. Comparison between RA2 SWH Ku and S-band for cycle 25. Solid line shows the best fit. 3.1 Improved wind and wave products Global comparison of Ku and S band SWH For the global comparison between Ku and S-band SWH, RA2 data were extracted from the RADS database for Cycles 15 to 26. Data quality control was performed with standard flags and by discarding data flagged over land by altimeter measurements and retaining only data with peakiness (defined as the ratio between the peak power received and the mean power within the measurement) values between 1.5 and 1.8. Overall, for Cycles 15 to 26 the analyses showed a relatively good agreement between variables measured by Ku and S-band. Global direct comparisons of Kuband and S-band SWH resulted in a mean difference (Ku S band) of 0.08m, rms difference of 0.64m and a r 2 of When looking at data in individual cycles, these results did vary from cycle to cycle. Fig. 2 shows the scatterplot for Ku versus S-band SWH for Cycle 25 (March 2004). S-band retrievals seems to present more scatter, with larger values compared to Ku-band retrievals. In order to consider if there is any regional tendency for the observed values Fig. 3. Map of distribution of SWH difference (Ku-S) for cycle Comparison with buoys With the aim to validate the RA2 dataset, comparisons were performed with in situ buoy data from US National Buoy Data Centre (NDBC), UK Met Office (UKMO), Canada Marine Environment Data Service (CMEDS) and Météo France (MF). The buoys are located in open ocean and are all located in the Northern Hemisphere. RA2 data from Cycle 15 to 26 have been collocated within 50 km of the altimeter tracks and within 30 minutes of the satellite overpass.

3 Comparisons between RA2 1-Hz parameters from the record closest to the buoy location are shown in Fig. 5. In total, 883 records have been compared. For Kuband, SWH is slightly lower than buoys (gradient=0.96, intercept=0.16) with a rms of 0.27 m. However, for S-band retrievals, SWH appeared to be very noisy with a rms of 0.6 m (gradient=1.04, intercept=0.05) Origin of noise in 1Hz S band data numerical simulation with the SOC retracker From the comparison of SWH data retrieved from RA2 (Ku and S-bands) with SWH measured by buoys, it has been shown that S-band SWH retrievals appear to be noisier than Ku-band. Two possible causes for increased noise at S-band can be hypothesized: I) due to the smaller number of waveforms being averaged at S-Band, and II) due to broader bins used at S-band, which could result in loss of sensitivity to changes in the slope of the leading edge. The origin of this poorer performance at S-band has been investigated numerically by simulating waveforms with various bin widths and number of averaged pulses. In order to investigate the relative importance of those two effects, a set of experiments were carried out using the waveform simulator in operation at SOC. A total of four simulations were executed, considering four different cases: Case a: 2000 Ku IE were simulated with a gate spacing of ns and the tracking point at gate 46. And 500 S IE were simulated with a gate spacing of 6.25 ns and the tracking point at gate 23. These represent the parameters currently used for RA2 Case b: same conditions for the Ku-band and 500 S IE with a gate spacing of and the tracking point at gate 23. Case c: same conditions for the Ku-band and 2000 S IE with a gate spacing of 6.25 and the tracking point at gate 23. Case d: same conditions for the Ku-band and 2000 S IE with a gate spacing of and the tracking point at gate 23. In all the cases, two different approaches have been followed. In the first one, different values of SWH and a constant value of σ 0 were used, and in the second one a constant SWH for different σ 0, as detailed: Approach 1: SWH=[10,9,8,7,6,5,4,3,2] (m) and σ 0 =[10] (db). Approach 2: SWH=[4] (m) and σ 0 =[20,18,16,14,12,10,8,6,4] (db). Fig. 5. Comparison between buoy parameters and RA2 1-Hz data. SWH Ku-band (top panel), SWH S-band (bottom panel). The procedure used to obtain the estimates of the two parameters is the following: once the IE are generated, the mean waveforms are estimated at 20-Hz (Ku) and 5-Hz (S) for cases a and b, and 20-Hz (Ku) and 20-Hz (S) for cases c and d. The averaged waveforms were introduced in the SOC simulator to obtain the estimates. The noise used in the IE is different for each (SWH, σ 0 ) realization but remains the same for each case. Therefore, it will be the same noise for SWH=10 m and σ 0 =10 db in cases a, b, c and d. This will allow the comparison of the results in the different cases for every value of the parameters. The analysis shows that for approach 1 (varying SWH values and constant σ 0 ), the simulations with parameters specified for case a demonstrate that the S-band results are much noisier than Ku-band. For lower initial SWH, the Ku-band results seem to be less noisy in respect to higher SWH, which is not the case for the S-band results (Fig. 6). The reduction of the gate spacing from 6.25 ns to ns in the S- band (case b), demonstrate that the estimates are also noisy (not shown). This indicates that the reduction of the gate spacing is not strongly affecting the estimation of SWH. The increase in the number of S- band waveforms used to produce the averaged waveforms from 500 to 2000 waveforms shows that the results are less noisy in all the realizations (Fig. 7). It shows that the strong reduction of the standard

4 deviation associated to the S-band in case c and d, compared to cases a and b. This suggests that the number of averaged waveforms is the key factor for the noise obtained when we estimate the SWH. For approach 2 (constant SWH values and varying s 0 ), simulations with parameters specified in case a, shows that the S-band results are noisier compared to Ku-band (Fig. 8), but it seems that the estimation of s 0 is less affected by the differences between both band as far as the gate spacing and the number of IE averaged is concerned. The increase in the number of waveforms used to calculate the averaged waveforms (case c), slightly reduce the noise in the results of the S-band for higher SWH (Fig. 9). Therefore, as for approach 1, the increase in the number of IE used to produce the averaged waveforms is the key point in the reduction of the noise associated to the estimated σ 0. Fig. 7. Standard deviation of the mean values of the estimated SWH for the true values used in approach 1 for cases c. Black line 2000 Ku-band IE, ns. Red line 2000 S-band IE, 6.25 ns Along-track averaging for noise reduction in S-band The above investigation with the simulator suggested that an increase in the number of averaged IE should lead to reduction of noise in S-band. In order to test this assumption with real data, data have been averaged along track over 5 seconds for SWH Ku and S-band (Fig. 10). Fig. 8. Standard deviation of the mean values of the estimated σ 0 for the true values used in approach 2 for cases a. Black line 2000 Ku-band IE, ns. Red line 500 S-band IE, 6.25 ns. Fig. 6. Standard deviation of the mean values of the estimated SWH for the true values used in approach 1 for cases a. Black line 2000 Ku-band IE, ns. Red line 500 S-band IE, 6.25 ns. Fig. 9. Standard deviation of the mean values of the estimated σ 0 for the true values used in approach 2 for cases c. Black line 2000 Ku-band IE, ns. Red line 2000 S-band IE, 6.25 ns.

5 These results show that along track averaging (here over 5 seconds) reduces rrms for S-band SWH to 0.40m. However, it also has a large effect on regression parameters, compared with results presented in Fig. 5 (gradient from 1.04 to 0.80, intercept from 0.05, 0.53). Averaging seems not to have a large impact on Ku-band results Comparison of ENVISAT and ERS-2 altimeters In order to assess the continuity of the successful retrieval of altimetric data obtained by ERS-2, 1 Hz values of significant wave height (SWH) from RA2 Ku-band and S-band were compared with values from ERS-2 when the two altimeters are on the same ground track, with ERS-2 approximately 29 minutes astern of Envisat. The primary aim of this investigation was to obtain a calibration for the RA2 S-band SWH. This parameter is considerably noisier than the Ku-band SWH, so the usual method of calibrating against buoy data is less satisfactory - the noise produces wide confidence limits on the regression results obtained from the few hundred buoy/altimeter pairs. So the approach here is to relate S-band to RA2 Ku-band SWH, then use the already determined calibration for Ku-band to compute S-band calibration. Data from the North Atlantic (50-60 N E) during October 2003 have been analysed. Data were validated using standard checks; in particular, all data with rain flags set were discarded. But no calibrations were applied to the SWH values. For each pair of RA2 SWH (Ku- and S-band), an ERS-2 SWH value was obtained by linearly interpolation to the Envisat latitude along each pass - if the two validated ERS-2 values about this latitude were more than 1 second apart then all data at that latitude from that pass were discarded. Fig. 10. Comparison between buoy parameters and RA2 data averaged over 5 seconds. SWH Ku-band (upper panel) and S-band (lower panel) SWH = ERS (2) Combining Eq. 1 and 2 gives: SWH = EnvKu (3) This compare to the calibration for RA2 from [1] of: SWH = EnvKu (4) This resulted in 2362 triplets obtained from 37 passes through the area- but with considerable correlation since many adjacent values along a pass were within 7 km. There is a high correlation, of 0.969, between the RA2 Ku SWH and the ERS-2 SWH, shown in Fig. 11. The minimum distance regression (or first principal component) is given by: EnvKu = ERS (1) From a comparison of ERS-2 OPR data and 2830 US NDBC measurements from 1995 to 2000, David Cotton (pers. comm) obtained the following calibrations for ERS-2 SWH: Fig. 11. Linear regression between SWH ERS-2 and SWH RA2 Ku-band. The middle line is the MDR fit Fig. 12 and 13 show the much poorer correlations between RA2 Ku-band and S-band (r 2 =0.830) and between ERS-2 and RA2 S-band (r 2 =0.814). Clearly the S-band SWH values include a considerably higher

6 level of random noise than the Ku-band. So fitting the Minimum Distance Regression (MDR) is inappropriate. This can be illustrated by reducing the noise by averaging over consecutive data. Fig. 14 shows the results from averaging over N=2,3, 5, 10 and 20 values - using only sets of N at 1 second intervals and no gaps. As N increases the MDR fit moves towards the regression of S-band on Ku-band (e.g, for N=20, intercept=0.215, slope=0.894 and r 2 =0.98). One approach is to use the regression of S- band on Ku-band to estimate from the S-band SWH, what the Ku-band would measure, and then apply the Ku-band calibration. Using the regression of S-band on Ku-band: with Eq. 4 gives: EnvS = EnvKu (5) SWH = EnvS (6) A more consistent approach is to compute the MDR allowing for the difference in the variance of noise for S-band and Ku-band. The variance of the noise on the three SWH values - RA2 Ku and S-band and ERS-2 Ku-band - can be estimated using a triple regression analysis technique developed by Challenor (pers. comm.). Taking the RA2 Ku-band as the 'correct' values, this procedure shows that the residual standard deviation of S-band SWH (0.4830) is 4.2 times that of Ku-band SWH (0.1146), i.e. a variance ratio of 18:1. The 1 Hz S-band SWH is estimated from 1/4 the number of values used for the 1 Hz Ku-band SWH (from 450 Hz compared to 1800 Hz for Ku-band), which would account for a factor of 4 in the variance ratio. The S-band bin width is twice that of Ku-band, which presumably accounts for the remaining increase in variance. The larger variance of the ERS-2 Ku-band (0.1454) data times that of RA2 Ku-band - might well be due to these data being interpolated and obtained half an hour later. Fig. 13. Linear regression between SWH RA2 Ku and S-band. The middle line is the MDR fit These regression results - while unable to provide any absolute calibration - indicate that and EnvS = EnvKu (7) ERS2 = EnvKu (8) in reasonable agreement with Eq. 2 and 3, which give: ERS2 = EnvKu (9) Eq. 7 is the MDR allowing for the different variances. Combining this with Eq. 4 gives SWH = EnvS (10) However, Eq. 10 (like Eq. 6) is clearly unsatisfactory for low S-band values (giving negative SWH if EnvS<0.3 m). The data set analysed here has a minimum S-band SWH of 1.06 m with Ku-band SWH of 1.06 m, so a larger data set, with some small SWH measurements is needed to determine whether Eq. 10 needs improving or the linear calibration fails at low SWH for S-band values. Note that regressing S-band averages from 18 consecutive 1 second values against the individual (median) Ku-band values gives - from 73 pairs - the MDR fit in agreement with the tripleregression result in Eq. 7 EnvS = EnvKu (11) Measuring waves in strong rain conditions Fig. 12. Linear regression between SWH ERS-2 and SWH RA2 S-band. The middle line is the MDR fit A separate analysis for SWH was made for the Southern Ocean (50 o S 60 o S, 120 o E 280 o E) for the period of 9 th April to 15 th September This allowed testing the effect of the RA2 rain flag in the SWH measurements. It showed that Ku-band distribution (Fig. 15) is affected to a much greater extent than S-band distribution (Fig. 16), and becomes more skewed and with a lower mode.

7 Fig. 17. Along-track SWH Ku-band (red) and S-band (blue). Black crosses shows RA2 rain flag set Fig. 14. Linear regression between SWH RA2 Ku and S-band. The middle line is the MDR fit Fig. 15. Distribution of RA2 SWH Ku-band with no rain flag applied (left panel) and with rain flag applied (right panel). Fig. 16. Distribution of RA2 SWH S-band with no rain flag applied (left panel) and with rain flag applied (right panel). Fig. 18. Along-track σ 0 Ku-band (red) and S-band (blue). Black crosses shows RA2 rain flag set In addition, analysis of along-track SWH values in strong rain conditions shows that SWH S is very noisy compared with SWH Ku (Fig. 17). However, as expected, σ 0 S is not as badly affected by strong rain than σ 0 Ku (Fig. 18). With these results, it is suggested that σ 0 S may therefore be used with SWH Ku to retrieve wave period information 3.2. Rain studies Over the sea surface the radar altimeter echo has a well-defined shape, enabling sea surface height (ssh) and wave height (SWH) to be determined accurately by fitting to an expected shape for the return. If there are rain cells in part of the altimetric footprint then the return waveform is not ocean-like and the quality of the derived geophysical information is degraded. Thus a key aspect of quality control is the reliable detection of rain-affected data. A dual-frequency detection technique has been already validated for TOPEX. It relies on the normally close relationship in normalised backscatter (s 0 ) at the two frequencies, with

8 significant departures (σ 0 Ku lower than expected) being attributed to rain. There is very good agreement between the backscatter at the two frequencies, especially for moderate winds ( > 3 ms -1 ). The mean difference between σ 0 Ku and s 0 S shows a peak at σ 0 S = 10.7 db (σ 0 Ku = 11.5 db), corresponding to a wind speed of 7 ms 1. The scatter about this mean relationship is very small (std. dev. of less than 0.25 db) for σ 0 S < 12.3 db, but in low wind conditions there is much poorer agreement between the values at the two frequencies. Rain attenuation of the RA2 Ku-band seems to be as expected from models and TOPEX analysis, with intense rain affecting all geophysical parameters. S- band is not affected by rain; but the choice of an attenuation threshold for reliable rain flagging remains to be optimised; because of the variation in sensitivity with wind speed. Of particular concern is the selection of a suitable threshold to minimise the quantity of good data inadvertently discarded. The relationship of s 0 Ku to σ 0 S is affected by SWH as shown in Fig. 19. A constant threshold of, say, -0.5 db is therefore not appropriate, as a simple multiple of the standard deviation does not entirely reflect the change in the underlying distributions. A particular bonus with ENVISAT is the collocation of many sensors providing information on clouds and rain. Reference [2] have demonstrated the complementary information obtained from RA2, MWR and AATSR concerning tropical storms brewing off West Africa Ionospheric correction The approach chosen for the ionospheric correction analysis was based on global comparisons of collocated RA2 dual-frequency and JPL Global Ionospheric Maps (GIM). Ionospheric correction measurements from RA2 showed a good agreement with GIM, with mean difference (GIM RA2) around 0.01m, rms difference around 0.01m and r 2 of Fig. 20 shows the comparison between GIM and RA2 dual frequency ionospheric correction for Cycle 25. Nevertheless, it is essential to remind that the use of peakiness during data quality control played an important role for this good agreement, as an effective editor for sea-ice contaminated data. Preliminary results of ionospheric correction without the use of peakiness showed a large scatter of dual-frequency ionospheric correction data. The largest differences between GIM and RA2 were associated with measurements over ice (RA2 data are underestimated) and sea-ice boundary (RA2 data are overestimated). Fig. 19. SWH effect on mean relationship between s 0 Ku and σ 0 S Fig. 20. JPL GIM against RA2 dual-frequency Fig. 20. Ionospheric correction for Cycle 25. Solid line is the regression best fit S-band anomaly Anomalies associated with in the S-band σ 0 have been observed, which can affect rain detection derived from RA2. In addition, it raises the issue of the necessity of validating any product that utilises dual-frequency algorithms. Fig. 21. Global distribution of S-band anomalies. Red dots: start of anomaly. Blue dot: end of anomaly. Global analysis based on large deviations from the expected Ku/S ratio for cycles 18 to 26 indicated that S-band anomalies occur globally, with slight preference for ascent over Indonesia and descent over South America, plus poles (Fig. 21). No periodicity is evident (Fig. 22). Reference [3] show a particular S-

9 band anomaly event with a rapid increase in σ 0 S followed by damped oscillations in the values. obtained in Ku band. This unexpected finding is tentatively attributed to the wider range window in S band; the Ku band is forced into ocean mode over the majority of the earth s land surface, which significantly de-scopes the ability to recover meaningful data using RA2 Ku band over drainage basins. This is a matter of concern, and is being investigated. Comparison with ERS-2 Ku band data indicates that this result is a consequence of anomalously poor performance in Ku band rather than unexpectedly good performance in S band. Fig. 22. Number of occurrences of S-band anomalies for each of the analysed cycles 4. SCIENTIFIC RESULTS WITH S-BAND OVER LAND AND ICE In the first part of this section, a global analysis was undertaken to evaluate the performance of the RA2 in S band. Data from 7 cycles were analysed using a rule-based expert system, which allows retracking of all echoes for which a leading edge is present within the range window. As the first part of this analysis, the number of echoes successfully processed was examined. To alleviate sampling effects from missing orbits of data or other transient events the data were summed over 45 arc minute cells globally; the best performance for each cell was then taken. The results are shown in Fig. 23. The S band data show high recovery in relatively benign terrain, with lower numbers of echoes retrieved over mountainous areas. Generally the performance is good. Echo shape analysis confirms that the echoes are well behaved, much wider than those obtained in Ku band, an expected finding consistent with the relative behaviours of Topex Ku and C band echoes over land. An example of the waveform shape analysis is given in Fig. 24, showing the relative percentage of flat patch echoes obtained globally over land and ice from RA2 cycle 17. A global analysis over land and ice was then undertaken comparing echo recovery from RA2 Ku band and S band. The results, again using the best performance in every 45 arc minute block over 7 cycles, are shown in Fig. 25. The results show good performance of the S band against Ku over moderate and flat terrain. Ku band, as expected, shows better recovery over mountainous terrain, where the dynamic mode switching allows lock to be maintained. However, S band shows significantly better performance over major drainage basins. Even on this large-scale comparison, the Amazon and Congo basins show significantly better echo capture in S band than that Fig. 23. Global number of points in S band best performance over 7 cycles Fig. 24. Flat patch global echoes from RA2 C band cycle 17 Fig. 25. Comparison of RA2 Ku band and S band (Ku S) best performance over 7 cycles

10 5. SCIENTIFIC APPLICATIONS OF IE OVER OCEANS, COASTAL ZONES, LAND AND ICE SURFACE RA2 IE are currently acquired routinely for 1 second every 3 minutes, corresponding to 2000 successive individual (un-averaged) echoes. The unprocessed IE data are currently available in L1B and SGDR products. One of the aims of the project is to develop the IE processor, which will reconstruct the waveforms and retrieve amplitude and phase of the IE. The processor is currently being finalised Individual Echo Applications over Oceans and coastal zones Using individual waveforms from the RA2 burst sampling mode over the ocean and the coastal zone, it will be possible to investigate: a) pulse-to-pulse correlation as a function of time separation of pulses [4]. From this it will be possible to consider whether SWH can be reliably measured closer to the coast using averages at 18 Hz or higher, or using 1Hz running windows. This will contribute towards examining the blurring effect in the coastal zone due to averaging. b) bin-to-bin correlations. This work will indicate how well the waveforms correspond to the theoretical waveform models, which assume that data in adjacent bins are independent. Access to the individual waveforms enables us to consider the impact of averaging on bin to bin correlation by varying the number of pulse averaged. c) anomalies in the waveform shape, and whether the waveforms conform to the theoretical models. By varying the number of waveforms being averaged, and examining the resulting averaged waveforms, it will be possible to determine the optimum averaging required for conformity to the theoretical model. In addition it will be possible to detect anomalies in the waveforms shapes and investigate the possible effects of these retrievals of geophysical parameters Individual Echo Applications over Land/Ice Individual echoes will be examined over a range of terrain types from mountainous to flat. RA2 Ku band data are available over even mountainous; an example of RA2 s ability to acquire data is given in Fig. 26. Here, an SRTM DEM profile (green) is compared with heights derived from RA2 Ku band (purple) and ERS-2 (red) over the Andes. Another key topic of interest is the behaviour of echo sequences over water/land boundaries, both for inland water analysis and coastal studies. Fig. 26. Envisat RA2 data over Andes 6. CONCLUSIONS New RA2 measurements at S-band and at unaveraged individual echoes are being investigated for scientific applications over the oceans, in the coastal zone, and over land/ice surfaces. Investigations with S-band data over ocean indicate that there is a problem with recurring S-band anomalies, probably of instrumental origin, which impact any products based on S-band parameters, including rain flagging, ionosphere correction, sea state bias estimates and therefore SSH at Ku-band. S- band SWH is intrinsically more noisy, but early results suggest σ 0 S gives more robust results in case of rain. The analysis of RA2 performance over land/ice surfaces shows that S-band data present high recovery in relatively benign terrain, with lower numbers of echoes retrieved over mountainous areas. S band also show good performance against Ku over moderate and flat terrain. Ku band, as expected, shows better recovery over mountainous terrain. However, S band shows significantly better performance over major drainage basins. 7. REFERENCES 1. Cotton, P.D. et al., Geophysical validation and cross calibration of ENVISAT RA-2 wind/wave products, ESA Contract Report, 51 pp, Quartly, G.D and C. A. Poulsen, Coincident cloud observations by altimetry and radiometry, ENVISAT Symposium Proceedings, Salzburg, Lillibridge, J. et al., rain and ice flagging of ENVISAT altimeter and MWR data, ENVISAT Symposium Proceedings, Salzburg, Walsh E.J., Pulse-to-pulse correlation in satellite radar altimeters, Radio Sci., 17, , 1982.