Sea-State Determination Using GNSS-R Data

IEEE GEOSCIENCE AND REMOTE SENSING LETTERS, VOL. 7, NO. 4, OCTOBER 2010 621 Sea-State Determination Using GNSS-R Data J. F. Marchan-Hernandez, E. Valencia, N. Rodriguez-Alvarez, I. Ramos-Perez, X. Bosch-Lluis, A. Camps, Francisco Eugenio, and Javier Marcello Abstract—Global Navigation Satellite Systems (GNSS) signals can be used to infer geophysical data related to the surface where they scatter. When dealing with the sea surface, its state influences the GNSS scattered signals and, therefore, the GNSS reflectometry (GNSS-R) observables. The aim of the Advanced L-band Emissivity and Reflectivity Observations of the Sea Surface 2008 field experiment was to gather experimental data to study the relationship of the GNSS-R delay-Doppler maps (DDMs) and the sea state. This work describes the field campaign and the main results obtained, where among them is the use of the DDM volume as a roughness descriptor weakly affected by the GPS satellite geometry. Index Terms—Delay-Doppler map (DDM), Global Navigation Satellite Systems reflectometry (GNSS-R), sea state. I. I NTRODUCTION EA SURFACE salinity is a key parameter in both climatology and oceanography fields, and its retrieval can be accomplished by means of L-band radiometry [1]. The sea state introduces a change in the observed brightness temperatures that must be corrected for. This can be achieved using active illumination systems, as planned for the National Aeronautics and Space Administration’s Aquarius salinity mission, or by means of reflected signals from Global Navigation Satellite Systems reflectometry (GNSS-R). The use of GNSS-R has been studied through the last 15 years, which yielded promising results: from altimetry applications [2], [3] to soil moisture determination [4], ice characterization [5], or sea-state retrieval [6], [7]. For the GNSS-R sea-state determination approach to perform the roughnessinduced radiometric correction, the power and mass requirements are significantly lowered since there is no receiver, and a small antenna can be used. Moreover, the bistatic geometry S Manuscript received October 2, 2009; revised December 29, 2009. Date of publication April 1, 2010; date of current version October 13, 2010. This work, conducted as part of the award “Passive Advanced Unit (PAU): A Hybrid L-band Radiometer, GNSS-Reflectometer and IR-Radiometer for Passive Remote Sensing of the Ocean” made under the European Heads of Research Councils and European Science Foundation European Young Investigator (EURYI) Awards scheme in 2004, was supported by funds from the Participating Organizations of EURYI and the EC Sixth Framework Program and was also funded by Research Projects ESP2007-65667-C04 and AYA2008-05906C02-01/ESP. J. F. Marchan-Hernandez, E. Valencia, N. Rodriguez-Alvarez, I. Ramos-Perez, X. Bosch-Lluis, and A. Camps are with the Remote Sensing Laboratory, Departament de Teoria del Senyal i Comunicacions, and the Aeronautics and Space Research Center (CRAE), Institut d’Estudis Espacials de Catalunya, Universitat Politècnica de Catalunya, 08034 Barcelona, Spain (e-mail: jfmarchan@tsc.upc.edu). F. Eugenio and J. Marcello are with the Departamento de Señales y Sistemas, Universidad de Las Palmas de Gran Canaria, 35017 Las Palmas de Gran Canaria, Spain. Color versions of one or more of the figures in this paper are available online at http://ieeexplore.ieee.org. Digital Object Identifier 10.1109/LGRS.2010.2043213 ensures a strong signal return in the specular direction, in opposition to the weak return obtained with monostatic radar off-nadir configurations. Additionally, the GNSS-R approach is less sensitive to power calibration since both direct (opportunity transmitter–receiver link) and scattered signals are processed. The existence of several deployed or planned GNSS constellations (GPS, GLONASS, Galileo, Compass, IRNSS, . . .) increases the availability of temporal and spatial colocated GNSS-R and radiometric measurements. In [8], a model of the reflected waveform is fitted to the measured waveform using the wind speed (WS) as the tuning parameter. Another approach calls for using the so-called interferometric complex field, resulting from the coherent processing of both direct and reflected signals, to obtain a direct estimate of the significant wave height (SWH) [9]. These two approaches require the use of both electromagnetic and sea surface models. To avoid modeling errors in the brightness temperature correction for salinity retrieval, a direct relationship between the sea state and a GNSS-R observable was put forward in [10]. The proposed GNSS-R descriptor is the volume of the normalized delayDoppler map [(DDM); maximum amplitude equal to one] above a threshold to eliminate the noise dependence. Such a descriptor is related to the extent of the glistening zone (the area from which scattered signals are observed), which depends on the sea state. To test this assumption, the Advanced L-band Emissivity and Reflectivity Observations of the Sea Surface 2008 (ALBATROSS 2008) campaign was conducted in the north coast of the Gran Canaria island (Mirador del Balcón, La Aldea de San Nicolás). II. C AMPAIGN D ESCRIPTION The campaign aimed to acquire an extensive data set of collocated GNSS-R and L-band radiometric data over the sea surface under different sea-state conditions using ground-based sensors. The three main requirements of the test site were: 1) its geometry, the highest possible height over the sea surface, the better, to cover the largest possible range of delays; 2) seastate variability; and 3) its orientation to north, to minimize Sun and galaxy contamination in the radiometric measurements. The selected site (Mirador del Balcón) is a scenic viewpoint located in the steep northwest coast of the Gran Canaria island (Canary Islands, Spain). It is located about 400 m over the mean sea level, with a slope of ∼45◦ down to the sea. The area is driven by strong and moist north-component winds (trade winds) mainly in the summer, with minimum influence of the land. The Passive Advanced Unit (PAU)/GNSS-R [11] and the L-band Automatic Radiometer [12] instruments were located at the viewpoint aiming toward the sea (Fig. 1). Oceanographic buoys gathering ground truth data were moored near the observation site within the field of view of the instruments: Two salinity buoys from Universidad de las 1545-598X/$26.00 © 2010 IEEE 622 IEEE GEOSCIENCE AND REMOTE SENSING LETTERS, VOL. 7, NO. 4, OCTOBER 2010 Fig. 2. Buoy ground truth for ALBATROSS 2008. (a) Scatter plot between U10 and SWH for the Agaete buoy (R = 0.26). (b) Evolution in time of the SWH measured by both the Agaete and the TRIAXYS buoys. Fig. 1. Experimental setup for ALBATROSS 2008. (a) Overview of the measurement site. (b) Reflectometer aiming to the sea. Palmas de Gran Canaria and from Universitat Politècnica de Catalunya (UPC) were equipped with SBE-37 thermosalinometers, and the third one was a TRIAXYS directional sea surface spectrum buoy from UPC. In addition to that, data from a SeaWatch buoy 18 km offshore of the deep-water buoy network of the Spanish Harbor Authority were available. The field experiment lasted from May 27 to July 4, 2008, in an attempt to catch different sea-state conditions. PAU/GNSS-R acquired the GPS reflected signals and computed real-time DDMs in two operating modes: 1) 1-min bursts of complex DDMs every 1 ms without averaging; 2) 50-min continuous acquisition, with 1-s incoherent averaging of 1000 1-ms DDMs. Every hour, two consecutive 1-min bursts were acquired followed by a 50-min capture. The site topography imposed the field-of-view mask from where GPS satellite reflections were found: elevations below 45◦ and azimuths between 270◦ and 30◦ north (0◦ : north direction, angles measured clockwise). The down-looking antenna was a hexagonal seven-element array with a 25◦ half-power beamwidth, whereas the antenna to track the direct signal was a single patch, thus having a wider field of view. III. DATA P ROCESSING The real-time computed DDMs had a resolution of 0.36 chips in delay (two samples at the sampling frequency of the system) and 200 Hz in the Doppler domain. Therefore, the minimum sampling values of one chip in delay and 1 kHz in Doppler for a coherent integration time of 1 ms [8] were met. The overall size was 16 × 16 points. Each DDM was stored with additional parameters such as the estimated delay and Doppler used for its computation, the satellite identifier, and the elevation angle. These DDMs were straightforwardly processed to extract the peak value, the peak-to-floor ratio (equivalent to the SNR), and the volume of the normalized DDM (maximum equal to one) that exceeds a predetermined threshold above the noise level, expressed in arbitrary measurement units. IV. D EPENDENCE OF THE DDM VOLUME ON THE S EA S TATE Candidate parameters to act as sea-state descriptors are the SWH, the mean-squared slope (MSS), and the surface WS. Usually, the WS at 10-m height (U10) is used to parameterize the sea state, even though it is known that such a relationship is incompatible at L-band [12] and that other parameters such as swell do play an important role. The Agaete buoy does provide WS, SWH, and time spectrum. The scatter plot between U10 and SWH for this buoy is shown in Fig. 2(a), showing a weak dependence (R = 0.26). On the other hand, the TRIAXYS buoy records SWH and directional spectrum but keeps no record on the WS. Fig. 2(b) shows a pretty good correlation (R = 0.72) between the SWH recorded by the two buoys some 18 km apart. This fact makes possible to work only with the data from the Agaete buoy, to fill the gap existing on the TRIAXYS data, associated to the highest recorded SWH values. The oceanographic data are the result of 20-min integration every hour. Therefore, it is necessary to resample it to match the resolution of the PAU/GNSS-R data. From Fig. 3, it is clear that the volume dependence on the WS is much weaker than the volume dependence on either the SWH MARCHAN-HERNANDEZ et al.: SEA-STATE DETERMINATION USING GNSS-R DATA 623 Fig. 4. Volume dependence on the incidence angle for the ALBATROSS 2008 geometry (h = 360 m, elevation = 30◦ ). (a) Simulated volume. (b) Linear fitting coefficients for the four elevation ranges [(1) 15◦ −20◦ , (2) 20◦ −25◦ , (3) 25◦ −30◦ , and (4) 30◦ −35◦ ]. There is virtually no dependence on the elevation angle. ranges, from 15◦ to 35◦ . Then, each V ol(SW H) dependence was fitted to a first-order polynomial V OL = a · SW H + b Fig. 3. Volume dependence on several sea-state descriptors: (a) WS (R = 0.37), (b) SWH (R = 0.71), and (c) MSS (R = 0.68). The WS has the weakest correlation with respect to the DDM volume. Each color represents data points from the same acquisition. or the MSS. In the ALBATROSS 2008 campaign, these two parameters seem to be quite equivalent, which is reasonable, since both are numerically computed from the same sea surface spectra, even though the integrations cover different ranges of wavenumbers. Therefore, the SWH was chosen as the sea-state descriptor because of it having a slightly higher correlation with the normalized DDM volume (R = 0.71), even though MSS (R = 0.68) is acknowledged to be the parameter actually sensed by GNSS-R. V. I NFLUENCE OF THE E LEVATION A NGLE As already stated, the 1-s incoherently averaged measurements were acquired continuously during 50 min every hour. During this time interval, the GPS satellite significantly changes its position, and so, the elevation could change up to 21◦ . Therefore, the time dependence of the volume can be straightforwardly translated into elevation dependence. For the geometry of the ALBATROSS campaign, the simulations performed show nearly no relationship between these two parameters [Fig. 4(a)]. To verify this fact with the collected data, the volume measurements were assigned to four 5◦ -wide elevation (1) where a and b are the linear fit coefficients. As inferred from the retrieved linear polynomial coefficients [Fig. 4(b)], no significant differences between the different elevation ranges are found. That means that, for the groundbased scenario of the ALBATROSS campaign, the volume observable is approximately independent with respect to the satellite elevation. This result was confirmed by numerical simulations [13]. As explained in the following section, since the volume is a multiangular measurement, a changing geometry has a limited impact on the retrieved measurements. VI. A NTENNA PATTERN I MPACT The DDM peak value exhibits a strong variation with varying elevation angle [Fig. 5(a)], which is largely due to the antenna pattern modulation (25◦ half-power beamwidth). The DDM peak is related to the scattered power over the ocean surface at the delay, Doppler, and incidence angle (complementary of the satellite elevation angle) corresponding to the specular reflection point. Other DDM bins are associated to surface points with different incidence angles. This explains that the antenna pattern affects much more the peak than the DDM volume observable [Fig. 5(b)]. This volume is computed after normalizing the power DDM (squared amplitude DDM) by defining a threshold above the noise level. Lower thresholds imply a larger section of the DDM being actually used to compute the volume and thus result 624 IEEE GEOSCIENCE AND REMOTE SENSING LETTERS, VOL. 7, NO. 4, OCTOBER 2010 TABLE I F ITTING PARAMETERS FOR THE VOLUME D EPENDENCE ON THE SWH the volume dependence on the WS predicted in [10]. The following fitting function is proposed: V OL = α · SW H (1/β) + γ. (2) The obtained fitting coefficients (a, β, and γ) are given in Table I. VII. C ONCLUSION Fig. 5. (a) Sample DDM peak evolution as a function of the elevation angle. (b) Sample normalized DDM volume as a function of the elevation angle. Each DDM volume value is computed after normalizing the measured DDM by its peak value. The ALBATROSS 2008 campaign has provided an extensive set of GNSS-R DDM observables from experimental data. A new observable, the normalized volume of the DDM, has been applied for the first time to that data to experimentally relate it with the L-band sea surface roughness. It has been shown that there is a strong correlation (∼0.7) between the DDM volume and the SWH, and it exhibits a weak dependence on the elevation angle or the antenna pattern. The follow-up ALBATROSS 2009 campaign, whose data are currently under processing, was focused on the direct relationship between the normalized DDM volume and the observed brightness temperature so as to be able to make the brightness temperature correction without scattering and sea surface spectrum models. R EFERENCES Fig. 6. Volume dependence on the SWH for three different volume thresholds (0.01, 0.2, and 0.5), plotted along with the proposed data fitting function. in higher volume values and larger sensitivity. However, the resulting standard deviation is also higher due to the presence of noise peaks and a multipath from points that are farther away from the specular direction (Fig. 6). Higher thresholds lead to lower values of the DDM volume, smaller sensitivity to SWH, and smaller standard deviation. Throughout ALBATROSS, the SWH ranged between 1 and 2.6 m. In this range, the relationship between SWH and volume is pretty linear. An additional point can be obtained analytically without modeling errors for SW H = 0 (perfectly flat sea surface). This is equivalent to using a direct signal, which has not experienced any kind of scattering or multipath. In doing so, it is seen how the dependence can no longer be considered linear throughout the whole SWH range (Fig. 6). This behavior is consistent with [1] J. Font, G. Lagerloef, D. LeVine, A. Camps, and O. Z. Zanife, “The determination of surface salinity with the European SMOS space mission,” IEEE Trans. Geosci. Remote Sens., vol. 42, no. 10, pp. 2196–2205, Oct. 2004. [2] M. Martín-Neira, “A passive reflectometry and interferometry system (PARIS): Application to ocean altimetry,” ESA J., vol. 17, no. 4, pp. 331– 355, 1993. [3] S. Lowe, C. Zuffada, Y. Chao, P. Kroger, L. Young, and J. LaBrecque, “Five-cm precision aircraft ocean altimetry using GPS reflections,” Geophys. Res. Lett., vol. 29, no. 10, p. 1375, May 2002. [4] V. U. Zavorotny and A. G. Voronovich, “Bistatic GPS signal reflections at various polarizations from rough land surface with moisture content,” in Proc. Int. Geosci. Remote Sens. Symp., Honolulu, HI, Jul. 24–28, 2000, pp. 2852–2854. [5] A. Komjathy, J. Maslanik, V. U. Zavorotny, P. Axelrad, and S. J. Katzberg, “Sea ice remote sensing using surface reflected GPS signals,” in Proc. Int. Geosci. Remote Sens. Symp., Honolulu, HI, Jul. 24–28, 2000, pp. 2855–2857. [6] A. Rius, J. M. Aparicio, E. Cardellach, M. Martín-Neira, and B. Chapron, “Sea surface state measured using GPS reflected signals,” Geophys. Res. Lett., vol. 29, no. 23, p. 2122, Dec. 2002. [7] J. L. Garrison, A. Komjathy, V. U. Zavorotny, and S. J. Katzberg, “Wind speed measurements using forward scattered GPS signals,” IEEE Trans. Geosci. Remote Sens., vol. 40, no. 1, pp. 50–65, Jan. 2002. [8] V. U. Zavorotny and A. G. Voronovich, “Scattering of GPS signals from the ocean with wind remote sensing application,” IEEE Trans. Geosci. Remote Sens., vol. 35, no. 3, pp. 951–964, Mar. 2000. [9] F. Soulat, M. Caparrini, O. Germain, P. Lopez-Dekker, M. Taani, and G. Ruffini, “Sea state monitoring using coastal GNSS-R,” Geophys. Res. Lett., vol. 31, no. 21, p. L21 303, 2004. [10] J. F. Marchan-Hernandez, N. Rodríguez-Álvarez, A. Camps, X. Bosch-Lluis, I. Ramos-Perez, and E. Valencia, “Correction of the sea state impact in the L-band brightness temperature by means of delayDoppler maps of global navigation satellite signals reflected over the sea surface,” IEEE Trans. Geosci. Remote Sens., vol. 46, no. 10, pp. 2914– 2923, Oct. 2008. MARCHAN-HERNANDEZ et al.: SEA-STATE DETERMINATION USING GNSS-R DATA [11] J. F. Marchan-Hernandez, A. Camps, N. Rodríguez-Álvarez, X. Bosch-Lluis, I. Ramos-Perez, and E. Valencia, “PAU/GNSS-R: Implementation, performance and first results of a real-time delayDoppler map reflectometer using Global Navigation Satellite System signals,” Sensors J.—Special Issue Remote Sens. Natural Resources Environ., vol. 8, no. 5, pp. 3005–3019, 2008. [12] A. Camps, J. Font, M. Vall-llossera, C. Gabarró, I. Corbella, N. Duffo, F. Torres, S. Blanch, A. Aguasca, R. Villarino, L. Enrique, J. Miranda, J. Arenas, A. Julià, J. Etcheto, V. Caselles, A. Weill, J. Boutin, S. Contardo, R. Niclós, R. Rivas, S. C. Reising, P. Wursteisen, M. Berger, 625 and M. Martín-Neira, “The WISE 2000 and 2001 field experiments in support of the SMOS mission: Sea surface L-band brightness temperature observations and their application to multi-angular salinity retrieval,” IEEE Trans. Geosci. Remote Sens., vol. 42, no. 4, pp. 804–823, Apr. 2004. [13] J. F. Marchan-Hernandez, A. Camps, N. Rodríguez-Álvarez, E. Valencia, X. Bosch-Lluis, and I. Ramos-Perez, “An efficient algorithm to the simulation of delay-Doppler maps of reflected Global Navigation Satellite System signals,” IEEE Trans. Geosci. Remote Sens., vol. 47, pt. 2, no. 8, pp. 2733–2740, Aug. 2009.