1. Introduction

This report describes progress on the analysis of data from an experiment in the N. E. Atlantic, during the first mission of the Shuttle Imaging Radar, SIR-C/X-SAR, in April 1994. The subjects being investigated are:

(i) the behaviour of the mean and distribution of the backscatter cross-section s₀ from the sea surface;

(ii) synthetic-aperture radar (SAR) imaging of ocean waves; and,

(iii) SAR imaging of oil slicks.

The SAR data obtained in this experiment consist of simultaneous, co-registered observations in horizontal (HH) and vertical (VV) polarisations, at both L (1.2 GHz) and C band (5.3 GHz). Observations were also obtained in VV polarisation at X band (9.5 GHz), but these were not precisely co-registered with those at L and C bands. The site was imaged on sixteen occasions, through a combination of ascending and almost orthogonal descending passes. The data were calibrated, typically to within ± 2 dB at L and C bands and within ± 1 dB at X band. Simultaneous wave measurements were obtained from a directional wavebuoy, deployed from an Ocean Weather Ship which also monitored the meteorological conditions. Low to moderate wind speeds (from about 5 to 10 m s^-1) and mild sea states (significant wave heights mostly below 3 m) were encountered. Artificial oil slicks, consisting of a surface-active material known as ICI Emkarox, were deployed from the ship before most of the SAR data-takes.

2. Backscatter Properties

2.1 Mean Backscatter

We have compared the observed mean s₀ values with predictions using the wind speed and direction and the range of incidence angles at each time of imaging. Empirical and theoretical models have been assessed as follows.

(i) Empirical models We have tested two alternative scatterometer model functions, CMOD-3 and CMOD-4, developed for the ERS-1 wind scatterometer at C band, VV polarisation, as well as models developed at L and C band by Snoeij et al. (Delft University) from aircraft data obtained during the TOSCANE-2 campaign off Brittany. CMOD-3 and CMOD-4 fit the data within 1 - 2 dB, hence giving confidence in the calibration of the data at C band. The model of Snoeij et al. shows a small but significant discrepancy with incidence angle; predicted cross-sections are 2 dB or more too high at incidence angles below about 25º. We therefore cannot rule out the possibility of residual calibration errors at L band in the SIR-C or TOSCANE-2 data.

(ii) Theoretical models We have tested the predictions of composite-surface and Kirchhoff-based (Holliday et al.) scattering theories, applied to assumed descriptions of surface wave spectra developed by Donelan & Pierson (DP) and by Donelan, Banner & Jähne (DBJ, as described by Apel). The DP spectrum overestimates the cross-section by 5 -7 dB. The composite-surface model with the DBJ spectrum fits well at C band, but shows systematic misfits with incidence angle at L band and with wind speed at X band. The Kirchhoff-based scattering model shows the same systematic misfits, but gives poorer fits in HH than in VV polarisation. We conclude that, while the results at L band may be caused by a residual calibration error in the data, those at X band reflect a need to modify the short-wave end of the DBJ spectrum. Specifically, its dependence on wind speed is too strong to be consistent with the X-SAR data.

2.2 Backscatter Distribution

In order to quantify the observed backscatter distributions, we have plotted normalised moments of the image intensity as a function of their order. Figures 1 and 2 compare the results for single-look and multi-look images of the same data-take at L band, HH polarisation. In this case, and for all the available data-takes, we find that the single-look data fit closely to a K distribution, whereas the multi-look data fit closely to a lognormal distribution. This distinction has a significant impact on the setting of thresholds to minimise 'false alarms' in marine-target detection.

We find that the observed second moments of image intensity in the SIR-C/X-SAR data can be explained by the modulations of ocean-surface waves. This result comes from applying 'quasi-linear' wave-imaging theory (a linear wave-imaging transfer function plus the 'azimuth fall-off' effect of random wave motions) to the simultaneous buoy measurements collected at the times of imaging. However, relatively large hydrodynamic modulations have to be assumed. Our values are generally consistent with those inferred from tower-radar data.

3. Ocean-Wave Imaging

Initial results on two orbits were reported at IGARSS'95. Good agreement was found between the SAR and simultaneous buoy measurements of the wavelengths and directions of the dominant swell waves, although the SAR data showed a small but significant rotation of the position of the spectral peak with changing radar frequency. Subsequent analysis of the other orbits has confirmed the imaging of spectral peaks due to wind seas as well as swell waves. In some cases multiple peaks are imaged by the SAR but are not resolved by the buoy. Sometimes the relative magnitude of the peaks changes with radar frequency or polarisation.

The simultaneously available radar frequencies and polarisations provide a sensitive test of the wave-imaging mechanisms. For example, we find that the polarisation dependence of the imaged wave modulations is correlated with the swell direction relative to range (Figure 3). This may be explained in terms of the influence of the tilting imaging mechanism, which is strongly dependent on polarisation and which is strongest in the range direction.

Comparison is currently in progress between the SAR image power spectra and the predicted spectra, obtained by applying wave-imaging theory to the simultaneous buoy data. Our approach is more thorough than those of other investigators, as we are taking account of agreement not only in the wavelength and direction of the spectral peaks, but also in their magnitude and in the total power in the spectrum. (The test of the total power is the same as that summarised in Section 2.2 to model the observed second moment of image intensity.) Figure 4 shows an example comparison for Orbit 47, which is dominated by near-range travelling swell. This case tests predominantly the tilting and hydrodynamic wave-imaging mechanisms. The latter in particular is not well understood. The SIR-C/X-SAR observations are demonstrating an ability to constrain its magnitude, phase and directional distribution. Figure 5 shows example results from combined HH and VV polarisations at L band. On Orbit 47, the magnitude and phase of the hydrodynamic modulation are relatively well constrained by the tests, but the directional dependence is not. On Orbit 51 (near-azimuth travelling waves) the reverse is true: consistent fits are only obtained with a weak directional dependence, but a wide range of magnitudes and phases is allowed. Similar modelling for the other data-takes is currently in progress.

Most of the analysis has been undertaken with multi-look images, but some complex single-look images have also been investigated. These provide a more accurate filtering of the speckle component in the image power spectra, leading to a more accurate definition of the imaged wave structure. The analysis procedure has been described by Cordey & Macklin (published in IEEE Transactions on Geoscience & Remote Sensing, 1989). Our initial analysis here has demonstrated that the SIR-C and the X-SAR complex data are both free of processing artifacts and hence suitable for the technique. A demonstration of the impact on the assessment of ocean-wave imaging theory is currently in progress.

4. Oil-Slick Imaging

An experiment was performed to investigate the effects of surface-damping slicks on radar imaging of the sea surface. Films consisting of the surface-active material ICI Emkarox were deployed at various times before most of the SAR data-takes. This chemical has potential to become an environmentally acceptable tracer for ship detection and the monitoring of surface shear flow, with implications for pollution monitoring.

The experiment itself was completed successfully. Initial inspection of the images shows slicks most clearly imaged at the lowest wind speeds encountered. A more thorough analysis will be undertaken later this year. This will investigate how the slick contrast in the image depends on the radar parameters (frequency, polarisation and incidence angle), wind speed, and exposure time before the Shuttle overpasses.

5. Conclusions and Future Plans

The studies already undertaken have demonstrated the significance of this data-set in improving the understanding of radar backscattering from the sea surface and the imaging of ocean waves. It has been possible to test empirical and theoretical models of the mean backscatter cross-section and to identify where modifications are required. The accurate calibration achieved by SIR-C/X-SAR has been crucial to this aspect.

The behaviour of the backscatter distribution has been studied. The image variance can be explained in terms of the influences of speckle together with ocean-wave modulations. However, differences in the behaviour of single-look and multi-look images are not yet fully understood.

The multi-parameter SAR data and simultaneous buoy measurements have demonstrated an ability to test ocean-wave imaging mechanisms. This analysis, along with the interpretation of the oil-slick imaging activity, remains to be completed later this year.

We anticipate that this data-set will lead to further investigations. The possibilities are still under discussion, but they include study of the benefits of fusing the different radar channels to improve maritime target detection or to recover more accurate estimates of directional wave spectra.

6. Publications

The following publications have been produced on the work described here.

6.1 Journal and Conference Papers

'First Results from the SIR-C/X-SAR Experiment on Ocean-Wave Imaging in the N.E. Atlantic', G.E. Keyte, R.A. Cordey, R. Larsen & J.T. Macklin, Proc. IGARSS '95, Firenze, Italy, 10-14 July 1995, pp 1320-1322.

'Radar Backscatter Statistics from the Sea Surface: Implications of SIR-C/X-SAR Observations for Maritime Surveillance', J.T. Macklin, N.R. Stapleton, N.A. Robertson & R. Ringrose, Proc. NATO-AGARD Symposium, 'Space Systems as Contributors to the NATO Defence Mission', Cannes, France, 3-6 June 1996, in press.

'Radar Backscatter Statistics from the Sea Surface: Implications of SIR-C/X-SAR Observations from the N.E. Atlantic', J.T. Macklin & N.R. Stapleton, J. Geophys. Res. (Oceans), submitted November 1996.

'Ocean Radar Backscatter Statistics from Shuttle Imaging Radar SIR-C/X-SAR Observations', C. Anderson & J.T. Macklin, submitted to RADAR-97, Edinburgh, Scotland, 14-16 October 1997.

6.2 Internal Reports

'SIR-C Trials in N.E. Atlantic', R. Larsen & P.R. Dovey, GEC-Marconi Research Centre Report MTR 94/32A on DRA Contract RAE1B/89, August 1994.

'A Study of Ocean Clutter Characteristics Using Multi-Parameter Radar', N.R. Stapleton, J.T. Macklin & P.J. Saich, GEC-Marconi Research Centre Report MTR 96/02A on DRA Contract CSM2/155, January 1996.

'Polarimetric Radar Clutter Characteristics of Ocean and Land Scenes', C. Anderson, P.A. Wright, J.T. Macklin & R.A. Cordey, GEC-Marconi Research Centre Report on DRA Contract CSM/494, in preparation 1997.

Acknowledgements

GEC-Marconi Research Centre was funded for its contribution to this work by the Defence and Evaluation Research Agency, Farnborough, under Contract Nos CSM2/155 and CSM/494. We thank the Jet Propulsion Laboratory and the German Aerospace Research Establishment (DLR) for providing the calibrated SIR-C and X-SAR data, respectively, and the participants in the N. E. Atlantic experiment for collecting the in situ data.

Figure Captions

Figure 1. Comparison of the normalised moments of order n = 2 to 6 observed on a single-look SIR-C image with the predicted dependences of lognormal, K and Weibull distributions which match the observed second-order moment.

Figure 2. Comparison of the normalised moments of order n = 2 to 6 observed on a multi-look SIR-C image (the same case as in Figure 1) with the predicted dependences of lognormal and K distributions which match the observed second-order moment.

Figure 3. Ratio of swell-wave modulations in HH to VV polarisation in the image power spectra at L band, plotted against the swell-wave direction relative to range.

Figure 4. SAR-image power spectrum, directional wavebuoy spectrum, and predicted image power spectrum for Orbit 47 (near-range travelling swell waves) at L band, HH polarisation.

Figure 5. Model fits to magnitude m_h and phase f_h of hydrodynamic modulation, from comparison of SAR spectra with simultaneous buoy data. Top: Orbit 47 (near-range travelling waves), bottom: Orbit 51 (near-azimuth travelling waves), both at L band, combined HH and VV polarisations. Each plot shows the constraints from (left to right) total power, swell direction, magnitude of the spectral peak, and the combined result of all three constraints. Filled squares indicate cases which agree. m_h runs from 0 to 24, and f_h from -90° to +90°.