Wind Field Retrieval from ALOS PALSAR - Comparison to Weather Model and ENVISAT ASAR results
Jochen Horstmann(1), Frank M. Monaldo(2), Donald R. Thompson(2) and Wolfgang Koch(3)
(1) NATO, Undersea Research Center, , Italy
(2) The Johns Hopkins University, Applied Physics Laboratory, , United States
(3) GKSS Research Center, , , Germany
Up to the launch of the PALSAR L-band synthetic aperture radar (SAR) on the Japanese Advanced Land Observing Satellite (ALOS) satellite, there was little impetus to determine a general L-band model function that relates wind speed and direction to normalized radar cross section (NRCS) as function of incident angle and polarization. The Seasat SAR satellite launched in 1978 was not adequately calibrated and operated at only a single incident angle over a narrow 25 km swath. Nonetheless, Gerling (1986) was able to demonstrate the rough correspondence of Seasat SAR image intensity and wind speed measured form Seasat SASS scatterometer. The Japanese launched the L-band JERS-1 SAR in 1992. It, too, was limited in oceanographic application with a less than robust calibration and a limited incident angle. However, the work of Shimada et al. (2003) showed the promise of L-band wind retrievals.
Our first effort to derive a geophysical model function (GMF) was to use the Elfouhaily wave spectrum and a scattering model to estimate an L-band model function. These results agreed well with aircraft L-band SAR measurements. However, the results with the ALOS PALSAR yielded wind speeds that were too high and showed unexpectedly high wind speeds at mid-incident angles (30 degrees). Therefore, for the wind speed retrieval we used a totally empirical GMF, which was recently developed by Isoguchi and Shimada (2007). Concerning wind direction retrieval we follow the approach common for C-band SARs, which is based on detection of linear features in the SAR image that are associated to wind streaks. The orientations of these features we retrieve using the Local Gradient Method.
We make a systematic comparison of wind directions and wind speeds computed from over 50 PALSAR data against wind analyses from an operational weather model. We also compare the results to wind fields retrieved from the advanced SAR aboard the European satellite ENVISAT, which were collocated in space and with a short time difference of ~ 30 min. The standard deviation of the residual difference is very similar to results from C-band SARs on Radarsat-1 and ENVISAT when compared against models, albeit with a systematic and easily removed bias in the wind speeds. At the moment our PALSAR results are limited to HH-polarization data.