On polarimetric characteristics of convection in the marine atmospheric boundary layer

Convection in the marine atmospheric boundary layer (MABL) is an important atmospheric phenomenon. SAR is a powerful sensor to study the vast array of phenomena that inhabit the MABL. These phenomena can modulate the local wind stress vector fields and change surface wave spectra on scales comparable to the wave length of the radar; moreover, they can result in convection in MABL which can produce corresponding characteristic signatures in radar imagery (Sikora and Ufermann, 2005). Sikora et al (1995) provide convincing evidence linking the mottled SAR image patterns with convection in MABL, co-located with concurrent, high resolution boundary observations. Previous studies have been limited by single polarization data, and have therefore focused on the variations in intensity of the return radar wave, and which are hampered by the existence of the same image patterns from other atmospheric and oceanic phenomena. In the paper, we first study the polarimetric characteristics of the MABL phenomena with high resolution fully polarimetrc RADARSAT-2 data. The parameters Entropy (H), Anisotropy (A), average Alpha (ƒÑƒn), from the Cloude and Pottier (1997) decomposition have been used as the basis for the development of new classification methods introduced for analysis of polarimetric data (Lee et al.,1999; Ferro-Famil et al.,2001., Touzi et al.,2004). The application of these polarimetric methods have only recently been applied to ocean backscatter. Using RADARSAT-2 SAR images, we present the characteristics of polarimetric parameters of convection in the MABL of the open ocean based on polarimetric theory, and compare the differences of polarimetric characteristics of convection with that of the signature for wind-generated waves in the open ocean. Thus, we obtain polarimetric characteristics to distinguish MABL convection from wind waves in open ocean. The polarimetric parameters make use of the statistical properties of polarimetric data, and take advantage of eigen-characteristics of the covariance matrix. Instead of preselecting the polarimetric channels, all elements of the covariance matrix are analyzed. Thus, the polarimetric parameters that we use directly include the polarimetric backscattering information related to the scattering mechanisms. This work throws new light on the problem of distinguishing among the possible MABL phenomena present at any time or place in a SAR image. Thus, we are able to understand the associated structures and dynamics of these phenomena and to forecast and parameterize their effects.