STORM: A NEW AIRBORNE POLARIMETRIC REAL-APERTURE RADAR FOR EARTH OBSERVATIONS

Transcription

1 STORM: A NEW AIRBORNE POLARIMETRIC REAL-APERTURE RADAR FOR EARTH OBSERVATIONS HAUSER Danièle, T. PODVIN, M. DECHAMBRE, R. VALENTIN, G. CAUDAL, J-F DALOZE CETP-IPSL (CNRS/université de Versailles-Saint-Quentin en Yvelines) Avenue de l Europe, Vélizy, France, hauser@cetp.ipsl.fr 1 INTRODUCTION In the context of increasing interest in polarimetric radar measurements for earth observation, airborne radar systems are useful tools to analyze the polarimetric signature of natural surfaces and to develop and test algorithms. Because SAR (synthetic Aperture Radar Systems) systems are quite expensive to develop and not trivial to use, the use of realaperture radar is interesting. Although they cannot provide radar images of the surface, they can have range resolution of the same order of magnitude as SAR systems (in particular when the Frequency-Modulated-Continuous Wave technique is used) and hence can be used to investigate the polarimetric signature of natural surfaces with a high resolution. Their use is usually flexible since the incidence angle and look angle can be chosen or even made variable. At CETP, the first polarimetric radar developed and used was the RENE radar at X and S-Band [1]. It was designed to be installed on board an helicopter and was mainly devoted to the study of continental surfaces. Since 2001, CETP has developed a new polarimetric radar called STORM (Système de Télédétection pour l Observation par Radar de la Mer) derived from RENE but working at C-Band (5.35 GHz), designed to be installed on an airplane, and mainly devoted to the study of the oceanic surface. It can also be used, however, over continental surfaces. In this presentation, we present the main characteristics of the STORM radar, the calibration procedure and results, and first results over natural surfaces. The first scientific campaign performed with STORM is VALPARESO (VALidation with a Polarimetric Airborne Radar of ENVISAT SAR over the Ocean) which took place in October-November An overview of this campaign and first results are presented in a companion paper ( [2], this conference). 2 PRESENTATION OF THE STORM RADAR STORM combines the characteristics of the mono-polarized C-Band radar RESSAC [3] designed for wave measurements and compatible with an installation on airplane, with the characteristics of the polarimetric radar RENE [1]. STORM has been designed and built in 2000 and 2001 with the support of National Space Agency "CNES", and the French " Institut des Sciences de l'univers". It is compatible with an implementation on the MERLIN-IV airplane of Météo-France [4]. It is operational since October 2001 and has been used for test flights in October 2001, and in the experimental campaign VALPARESO in October-November 2002, in the context of the ASAR ocean wave products validation. The main characteristics of STORM are given in Table 1 and Figure 1 shows the geometry. STORM is a FMCW (Frequency-Modulated-Continuous Wave) radar. The modulation is a saw-tooth shape (linearly increasing frequency) with a mean frequency of 5.35 GHz, a 192 MHz beamwidth, and a 8 ms duration. It is synthesized through a direct numerical synthesis system. The corresponding slant-range resolution is 1.53 m. The electromagnetic wave is successively transmitted in two orthogonal linear polarizations (horizontal and vertical). Because the microwave signal is transmitted continuously, a dual antenna system is necessary with one antenna for transmission, and one for reception. This set of two antennae is mounted in a radome placed under the airplane (Figure 2) and has the capability to rotate over 360 in the horizontal plan. The transmitting antenna is a horn with an orthomode system to switch the polarization of the transmitted electromagnetic wave. The receiving antenna is a plane antenna composed of an array of 4 x 10 couples of crossed-dipoles designed for receiving H and V polarizations. simultaneously. The 3 db beamwidth of the transmitting antenna is large (50 x 50 ), so that the 2-ways equivalent beam-width (30 x7.5 ) is not very different from that of the transmitting antenna (33 x7.5 ). A large beamwidth in elevation and azimuth was chosen, because of the geophysical applications (see Hauser et al, this conference). For financial reason, the antenna has not been specifically designed for STORM but we have chosen to use a pre-existing antenna developed for the airborne SAR of CNES. The antenna incidence was fixed to 20 (with respect to nadir). This mean incidence value, combined with the large beam aperture of (33 x7.5 ), and the use of a rotating system to scan over 360 in azimuth, is compatible with the principle used to estimate the directional wave spectrum and the wind (see [2], this conference) from the backscatter coefficient. The rotation system of the antenna has been changed with respect

2 to the RESSAC system. Instead of using a continuous clockwise rotation (as for the RESSAC radar), a system has been installed for a rotation over 360 alternatively clockwise and counterclockwise. This was chosen to avoid the use of double-channel rotary joints which have poor performances for polarimetric applications. The rotation speed is 3 rotations per minute. Isolation between polarization is better than 30 db for the transmitting antenna, and 40 db for the receiving antenna. Isolation between transmitting and receiving in the co-polar condition is better than 70 db. The receiving chain is designed for a simultaneous reception in H and V. The microwave received signal and the transmitted one are supplied to a mixer and low-pass filter to obtain a resulting video signal at Intermediate Frequency (IF). Table 1: Main characteristics of the STORM radar Microwave characteristics Type Mean Frequency FM/CW, bandwidth FM/CW repetition period Range Resolution Transmitted power Emitting alternatively in H and V (each 8 ms) Receiving simultaneously in H and V FM/CW GHz 192 MHz 8 ms 1.53 m 2 Watts Antenna 3 db beamwidth of the transmitting antenna (H or V ) 50 x 50 3 db beamwidth of the receiving antenna (H and V) 33 x 7.5 Cross-polarization isolation better than 30 db Isolation between transmission and reception better than 70 db Geometry 2 ways Beam Aperture 30 x 7.5 Mean incidence 20 Azimuth looking angle: scans over 360 Typical flight altitude 2000 to 3000 m H= 2000 to 3000m x z Φ Scan over m Incidence range from 5 to 35 y Y (azimuth) X(elevation) Look direction Figure 1: Geometry of observations with STORM. The flight level is usually between 2000 and 3000 m. corresponding to footprint sizes of (1200m x 280m) to Processing and Recording Video signal digitzed at a frequency of 2 MHz On 8192 samples Real-time FFT providing Complex spectra (Real + Imaginary parts) if no integration (1 sample each 8 ms) 4096 points Amplitude spectra if integration (8 samples) 4096 points Recording of the complex or amplitude spectra on 1024 points Ancillary data roll, pitch, drift, speed of aircraft, time, (1800 m x 420 m). Figure 2: Photo of the MERLIN-IV, with the radome protecting the STORM antenna visible under the airplane The real-time processing and recording is performed by using a Hewlett Packard-VXI system. The video signal obtained in IF over a 1 MHz bandwidth is digitized over a ms interval with a sampling frequency of 2 MHz, providing time series over 8192 samples. A FFT is then applied in real time to obtain a complex signal defined over 4096 points with a 244 Hz resolution. In order to limit the data recording rate, only part of the spectral components is recorded. So, the raw data recorded by the computer (on disks) is composed of the real and imaginary parts for 1024 frequency bins covering a frequency range of about 250 KHz, compatible with the range in distance to be analyzed for a to 3000 m flight level. The smallest recorded frequency is chosen by the operator, according to the flight level. No real-time integration is performed on the data in order to permit a post-processing in a polarimetric chain including all necessary corrections and adequate temporal average. Data recorded each 8 ms also include ancillary information which are necessary for the post-processing (pitch, roll, drift, speed of the aircraft, clock, ). The overall dynamic range of the received power spectrum signal is 80 db. To use the data in a full polarimetric mode a great care must be taken on the measurement of the phase of the signal. It was checked that the phase noise corresponds

3 to an accuracy of ±2 on the phase (without antenna). It was also checked that the rotation of the cables during the scanning of the antenna does not introduce fluctuations of phase incompatible with a polarimetric processing. 3 CALIBRATION PROCEDURE AN RESULTS 3.1 Principle and Notations Taking into account the power budget equation, and neglecting the effects of cross-coupling between vertical and horizontal receiver channels, the received signal written in a matrix form [Z] is related to the scattering matrix [S] of the natural or artificial scatter (see, [1]) through: ( ) [ Z]= Ae jθ e j ( φ t,v + φ r, v ) G e h G r h S hh e j φ t + φ r G e v G r h f t S hv e jφ r G e h G r v f r S vh e jφ t G e v G r (1) v f t f r S vv where A is a constant, S hh, S vv, S hv, and S vh are the elements of the scattering matrix [S], G ip is the gain where subscript i stands for reception (r) or transmission (t), subscript p for polarization (H or V), θ is the phase shift of the signal due to propagation in the atmosphere, φ i,p is the phase shift due to scattering and to the radar system (same subscript convention as for G ip ), f t and f r are coefficients to account for imbalance between the H and V channels in transmission and reception, respectively. Studying the polarimetric response of a natural surface, requires to use Eq. (1), in which the f t and f r and phases must be fixed. In fact, only relative phases (difference of phases are needed) so that the parameters to be determined by the calibration procedure are f t and f r, φ t and φ r where φ t =φ t,h -φ t,v and φ r =φ r,h -φ r,v. 3.2 Calibration procedure Part of the calibration procedure of STORM is based on laboratory measurements using an optic delay-line integrated in the radar system and a noise signal, but the full calibration (in particular for the polarimetric aspects) requires measurements over artificial targets. A combination of trihedral and a dihedral corner reflectors (see [5], [1]) enables the determination of the phase differences φ t and φ r and imbalance terms f t and f r. Indeed for a trihedral reflector, and assuming a perfect isolation between radar channels, the radar signal is backscattered only in co-polarized channels (terms of matrix [S] are S hh =S vv =1; S hv =S vh =0 ), the measured phase difference φ vv - φ hh is equal to (-φ t -φ r ) and the ratio of Z vv / Z hh is equal to the product f t f r. Similarly, for a dihedral reflector tilted at 45 from the horizontal plane, the signal is backscattered only in the cross-polarized channels (terms of matrix [S] are S hh =S vv =0; S hv =S vh =1), φ vh - φ hv is equal to (φ t -φ r ) and Z vh / Z hv is equal to the ratio f r /f t. In addition to the phase calibration needed for the polarimetric mode, standard calibration in amplitude for co-polarized (HH, VV) and cross-polarized (HV) signal are also performed, using internal calibration (delay line, noise source) and external calibration using the corner reflectors. Antenna pattern measurements were also performed in laboratory for all polarimetric configurations, and taking into account the effects of the radome installed to protect the antenna below the airplane. Measurements over corner reflectors have been performed three times since STORM is operational. The procedure was to fly over 5 trihedral and 1 dihedral corner reflectors installed on a aircraft runway. Figure 3a-c shows these corner reflectors and Fig 3c shows how they were installed on the runway. All the reflectors were installed in such a way that for each radar measurement and for the optimal flight-altitude (450 m) all of them are included in the radar footprint. When positioning the antenna in the forward direction (parallel to the aircraft), this allows to measure the corner reflectors response at different incidence angles at the same time. The use of several reflectors is also needed to ensure that at least one of them is close to the center of the beam in azimuth.

4 T T T 10 m Figure 3a: Trihedral reflector Figure 3b: Dihedral reflector T D T T = trihedral reflector, D= Diedral reflector Figure 3c: Installation scheme of the dihedral and trihedral reflectors 3.3 Calibration results for amplitude analysis Estimation of the receiver gain from the internal calibration showed that the two receiving channels (for H and V polarization) are well-balanced with only a 0.2 db difference in gain. The stability of the gain was checked by performing long-term measurements (using a known noise source). Gain variations due to temperature variation of the system were found to be less than 1 db over a 2 to 3 hours flight. Measurements over the corner reflectors were processed to estimate the radar cross-section of the reflectors an check the overall gain of the radar. In HH polarization, the estimated radar cross-section using the gain values from the internal calibration, ranges from 32.4 to 32.8 dbm 2 in HH, 32.1 to 32.5 in VV and 36.5 to 36.7 in HV. Taking into account the uncertainty of such measurements due to the difficulty of alignment of the radar beam with the reflectors, this is quite consistent with the theoretical expected values (32.5 dbm 2 in HH and VV, 36 dbm 2 in HV). This validates the internal calibration. The antenna gain pattern in elevation for HH and VV could also be partly checked from the flights over the trihedral reflectors by combining the radar cross-section of the reflectors at different elevation angles. Figure 4a shows an example for one of the pass over the reflectors. This pattern is quite consistent with the antenna pattern measured in laboratory. (Figure 4b) Antenna gain pattern from laboratory measurements (combination transmission/reception) 2 0 Figure 4a : Antenna gain pattern in elevation and in VV polarization from the STORM measurements over the trihedral reflectors, Angles along the x-axis are relative to the mean incidence angle and values in the y-axis refer to radar cross-section Gain (db) HH VV HV VH Elevation angle ( ) Figure 4b :Two-way antenna gain pattern in elevation from laboratory measurements (in VV, HH, HV, VH)

5 3.4 Calibration for polarimetric analysis Laboratory measurements have shown that the phase variation rate of each reception channel is less than 0.1 by second, due to variability of the local oscillator that generates the microwave transmission. This variation is quite negligible with respect to the time difference between two chirps (8 ms). Phase noise also estimated from laboratory measurements (with a delay line) indicated an standard deviation value of about 2. Histograms of φ vv - φ hh and φ vh - φ hv have been obtained from the flights over the corner reflectors (respectively trihedral and dihedral). Note that interpolation in time has been applied on one of the channels because transmission in H and V are not simultaneous. Fig 5a shows the results for one pass over the trihedral reflector (on 8 October 2002). In this case, the data come from the signal backscattered from 5 trihedral reflectors and all the range points located within the 3dB beamwidth have been kept to compute this histogram. The histogram shows a double peak (with an ambiguity of 180 ). This ambiguity is attributed to the interpolation process that has been applied to compensate for the time separation between to consecutive transmissions in H and V polarization. The mean and median values of φ vv - φ hh (calculated between 0 and 180 ) are very close (46.0±180 and 45.6±180 ) indicating that the histogram is close to symmetrical. The standard deviation is small (13 ). When plotting φ vv - φ hh as a function of incidence angle (not shown) we could check that φ vv - φ hh decreases slowly by less than 20 with incidence angle within the 3 db beam-width. Fig 5b shows the results obtained for one pass over the dihedral reflector. In this case, the number of data points is less than for the trihedral reflectors, because only one reflector was deployed and because the response of the dihedral reflector is seen only for a very strict conditions (when it is close to the center of the beam). Here again the histogram has a double-peak signature (180 ambiguity) due to the interpolation procedure. The shape of the histogram is less clear than in the case of the trihedral data, due to the small number of points. However, we could establish that the mean of the histogram is 84±180 with a standard deviation of 44. Figure 5a : Histogram of the phase difference between HH and VV polarized signals masure over the triehdral corner reflectors on October 8th, 2002 Figure 5b : Histogram of the phase difference between HV and VH polarized signals measured over the dihedral corner reflector on October 8th, 2002 Results very close to those presented in Fig 5a-b were found for the other passes on the same day performed in optimal conditions (good alignments with the reflectors): for the calibration flight of 8 October 2002 (3 passes in optimal conditions), mean values of φ vv - φ hh range from 46 to 47 with a standard deviation (std) of less than 14, and mean values of φ vh - φ hv range (for 3 passes) from 83 to 86 with a standard deviation of 41 to 44. When the calibration flight was repeated 1.5 month later (28 November 2002) after the VALAPRESO experiment, we found slightly different values: φ vv - φ hh range from 65 to 66 with a standard deviation less than 13 and φ vh - φ hv range from 86 to 95 (std 45 ). But with the delay line and the internal calibration, we did not find any evolution in these phase differences. This shows that phase shifts may appear, due to antenna or wires evolution. It also indicates that calibration flights must be repeated to follow the potential shifts of the phase during an experiment. From the trihedral and dihedral data analysis, we could check that f t and f r are values close to 1 (as expected for well balanced channels). Indeed we found (from the 8 October flight), Z vv / Z hh varying from 0.95 to 0.98 and Z vh / Z hv varying from 1.06 to The associated values of ft and fr are to 1.03 and 0.95, respectively. 4 RESULTS ON DISTRIBUTED SCATTERS

6 During the first test flights of STORM (October 2001) measurements have been performed with flights over various natural surfaces (grass, forest, ciment runway) with the aim to further validate the STORM radar and to provide an additional way of polarimetric calibration. Indeed, as proposed by [6] and also used by [1] with our RENE radar, the properties of some natural surface (in particular grass) can be used to estimate the coefficients needed in Eq.(1) to calculate the scattering matrix. We present hereafter results on phase distributions obtained with flights over a grass terrain. In the second part of this section we will also show preliminary results on phase distributions over ocean surface obtained during the VALPARESO experiment (October-November 2002) and will show that they significantly differ from what is observed over grass. 4.1 Phase distributions over grass Flight were performed at 300 and 450 altitude over a grass airfield. Two flight configurations have been chosen. In the first one the antenna was pointing forward (along the flight track). In the second one, the antenna was pointing on one side (perpendicular to the flight track). In all cases, it was checked that the entire footprint was included within this grass area. Figure 6 shows the distribution of φ vv - φ hh and φ vh - φ hv obtained for a flight at 300 m altitude with the antenna pointing forward, and for range points included in the 3 db beam-width. It is clear that these distributions are well peaked with two peaks separated by 180. The presence of these two peaks is again explained, as in the case of the analysis over corner reflectors by the interpolation procedure to account for non-simultaneous transmission in H and V polarization. It is important to note that the peak of the histograms is very close to the one obtained the same day over corner reflectors (see arrows in Fig. 6.). This shows the possibility to estimate the phase coefficients in Eq.(1) from observations over natural surfaces. At it was shown that f t and f r are very close to 1, this procedure may be sufficient for the complete calibration of the polarimetric radar. Results obtained from other flights on the same day or the day after over the same area show very similar results in the same configuration (same altitude, same antenna pointing). Changing the altitude (450 m instead of 300 m) also gives the same results. On the contrary, for the flights performed with the antenna pointing on the side of the aircraft, the distribution of φ vv - φ hh and φ vh - φ hv are almost uniform (not shown). We do not have a clear explanation of this fact but we just note that similar measurements with the RENE radar on-board an helicopter gave consistent results with a side-looking antenna [2]. The difference with the present case is the slower speed of the platform (30 to 40 m s -1 for helicopter against about 80 to 90 m s -1 for aircraft). In the present case, it is possible that for the side-looking case, the lateral displacement of the antenna beam (perpendicular to the elevation direction) between successive pulses transmitted in orthogonal polarization (separated by 8 s) is too large to combine signals from successive pulses. 4.2 Phase distribution over the ocean During the VALPARESO experiment, flights were performed in 3 different modes over the ocean (see [2]). The results presented here-after correspond to observations performed at a flight level around 2000 m with the antenna rotating over 360 in azimuth (at a rate of 3 rotations per minute) and the aircraft flying in strait line (no imposed roll). The radar data were recorded each 8 ms as complex values (amplitude, phase). Compared to most of the continental surfaces, the sea surface is rather smooth, which leads to a moderate depolarization. Indeed we measured cross-polarized backscattering coefficients about 20 db smaller than the copolarized ones. Due to this fact, the signal to noise ratio is not very high and the polarimetric analysis can only be performed over a limited range of incidence angles. The data are still in processing and we present hereafter only very preliminary results concerning phase histograms. Radar cross-section at different polarization configurations are shown in [2]. Figure 7 shows histograms of φ vv - φ hh, φ vh - φ hv, φ hh - φ hv, φ vv - φ vh for range points close to the mean incidence angle (20 ), where the signal-to noise ratio is larger than 10 db even in HV polarization, and for 1850 samples covering a complete scan of the antenna over 360 in azimuth. As for the processing of the data over corner reflectors, we applied here a phase interpolation to estimate φ vv - φ hh, φ vh - φ hv to account for the time separation (8 ms) between transmissions in H and V polarization. As shows Fig7, the histograms of φ vv - φ hh, φ vh - φ hv are much broader than in the case of corner reflectors (see above) or in the case of observations over grass. Two peaks separated by 180 also seem to exist but this is less clear than in the cases of grass or corner reflectors. In contrast, histograms of φ hh - φ hv, φ vv - φ vh show a clear peaked distribution with mean values of about 90 and 60 respectively. This first analysis did not show a clear variation of these histograms when restricting the data to particular azimuth look angle.

7 Figure 6: Histograms of phase difference (curves in a polar representation) from STORM measurements over a grass runway(airfield at a flight altitude of 300 m and when the antenna was pointing parallel to the flight track.). Data analyzed here correspond to range points located within the 3 db beam-width. The two arrows indicate the position of the peak of the histograms found from corner reflectors measurements during the same experiment Figure 7 : Histograms of phase difference from STORM measurements over the ocean. Data analyzed here correspond to incidence angles from about 19 to 21 and to data collected over a complete rotation of the antenna. To go further in the analysis, we plan to calculate the different terms of the matrix of conjugate products as proposed by Yueh [7, 8]. The terms of these matrix M are given below and can be obtained from the measured signal by calculating mean complex terms after geometric corrections are applied to replace the measurements in a frame associated with the surface and not with the antenna plane; S S * S S * S hh hh vv hh hv S * hh M = S S * S S * S S * hh vv vv vv hv vv S S * S S * S S * hh hv vv hv hv hv where each element is the average value of conjugate products. These conjugate products describe the correlation between co-polarized and cross-polarized responses. Averaging in space and/or time is necessary to decrease the statistical uncertainty due to thermal noise and speckle. The averaging process (indicated by the < > symbol in Eq.(2)) will be applied typically on 16 to 32 successive samples corresponding to time intervals of about 128 to 256 ms. This time interval is a trade-off between the necessity to increase the number of independent samples in the estimate (decrease the statistical uncertainty) and the necessity to use data of mean constant. With the antenna rotation speed of 3 rotations per minute, this corresponds to an azimuth angle variation of 0.4 to 0.8. The three terms of the diagonal of this matrix are real, and related through the radar power budget equation, to the normalized radar cross-section in HH, VV and HV polarization, respectively. The other terms have a real and an imaginary part, and they can be used to study the cross polarization correlations [8]. Other formulations have been proposed to study the polarimetric signature of the ocean surface, using the Mueller matrix or the Graves matrix (see e.g. [9]). In each case, the elements of these matrices can be derived from the coefficients of the S matrix (Eq.2), and an appropriate averaging process. 5 CONCLUSION (2)

8 The first results obtained with the new airborne polarimetric radar STORM have been presented here. We recall here that STORM is a C-Band real-aperture FM/CW radar with an antenna which has the capability to scan in azimuth. The results show that STORM works well. Calibration over corner reflectors give consistent and reproductible results (as far as no change in the wires or radar installation occur). It also shows that imbalance between channels is weak. Measurements over natural surface have provided consistent phase distributions over grass when the antenna was pointing in the direction of flight but inconsistent phase distributions (uniform distribution) when it was pointing in the perpendicular direction. This problem has still to be analyzed. Phase distributions over the ocean seem to contain information (not uniform distributions), but differ significantly from results over grass terrain. This was already mentioned by [1]. To go further in the analysis over the ocean, it is now necessary to analyze the terms of coherence or co-variance matrixes as proposed by several authors. Results on radar-cross-section over the ocean in different polarization configurations are presented in a companion paper [2]. 6 REFERENCES [1] Leloch-Duplex N., D. Vidal- Madjar, and J-P. Hardange, "On the calibration of the helicopterborne polarimetric radar RENE", Annales of Telecommunications, 51, no5-6, , 1996 [2] Hauser D., T. Podvin, M. Dechambre, G. Caudal, A Mouche, J-F Daloze, Polarimetric measurements over the seasurface with the airborne STORM radar in the context of the geophysical validation of the ENVISAT ASAR, this conference, (POLINSAR, ESA, Frascati,(It), 2003) [3] Hauser, D., G. Caudal, G.J. Rijckenberg, D. Vidal-Madjar, G. Laurent, and P. Lancelin, "RESSAC: A new airborne FM/CW radar ocean wave spectrometer", IEEE Trans. Geosci. Remote Sensing, 30 (5), , [4] Chalon J-P., André M., Brenguier J-L., Druilhet A., Gayet J-F., Flamant P., Guillemet B., Hauser D., Kerr Y., Pelon J., Ravaut M., Tanré D., "Les avions Français de recherche atmosphérique et de télédétection: bilan et perspectives," La Météorologie, 8è série, N 22, 14-44, juin 1998 [5] Sarabandi K., F.T. Ulaby, M.A. Tassoudji, Polarimetric SAR calibration radar systems with good polarization isolation. IEEE Trans on GRS (1990) 8 n 1 pp70_ [6]Freeman A., Van Zyl J.J., Klein J.D., H.A. Zebker, Y. Shen, Calibration of Stokes and scattering matrix for polarimetric data, IEEE Trans. On Geosc. And Reomte Sensing, 30, , 1992 [7] Yueh S.H., R. Kwok, and S. V. Nghiem, "Polarimetric scattering and emission properties of targets with reflection symmetry", Radio Science, Vol. 29, n 6, , 1994 [8] Yueh S.H., S.V. Nghiem, and R. Kwok, "Comparison of a polarimetric scattering and emission model with ocean backscatter and brightness temperature", Proceedings of IGARSS'94, CD-ROM, IEEE Editor, [9] Bahar E., and X. Shi, "The identification of sea swell using polarimetric radars: a new full wave approach", Int. J. Remote Sensing, Vol 19 No 11, , 1998.