Related Links:

  • Seafloor Mapping Seafloor Mapping
  • Natural Hydrocarbon Seepage Natural Hydrocarbon Seepage
  • Oil Spill Monitoring Oil Spill Monitoring

 

 

  • Oceanographic Studies Oceanographic Studies
  • Data Integration Data Integration
  • Chemosynthetic Studies Chemosynthetic Studies

The ability of SAR to detect features of the ocean’s surface depends on the interactions between the SAR pulse of microwave energy and the sea-surface. The radar return from varying roughness of the surface capillary and short gravity waves produces unique patterns in the radar imagery (Holt and Hilland, 2000).  SAR is therefore useful for detecting surfactant layers produced by floating oil. Ocean slicks are a subset of ocean features detected in SAR data. They are areas of distinctly contrasting brightness against the radar backscatter produced by wind-generated ripples.  Slicks are contiguous areas in which Bragg scattering at wavelength scale of ~0.01 to 0.1 m is suppressed either by layers of oil (Alpers and Espedal, 2004; Hu, et al., 2008), biological surfactants, or floating vegetation (Huehnerfuss, et al., 1983). The application of Bragg scattering theory for ocean imaging has been reviewed (Holt, 2004; Thompson, 2004). As shown in Figure 1, the SAR return from the ocean surface depends on the wavelength of small surface waves (0.7 to 10 cm), the radar electromagnetic wavelength (λr), and the incidence angle of the radar energy.  When sea surface capillary waves are damped by viscoelastic properties of a thin layer of hydrocarbons (~0.1 µm), a larger proportion of radar energy is reflected away from the satellite’s detector. Hydrocarbons on sea surface then can be interpreted as radar-dark regions on SAR images. The persistence of surfactant layers is affected by processes like evaporation, wind, surface currents, photolysis, spreading, flocculation, dissolution (Figure 1).  

Oil slicks can emanate from accidental, transient releases of oil from ships or platforms. Under these circumstances the feature of SAR data will reflect presence of the source and/or dispersion history following the release event. Oil slicks from natural seeps, in contrast, are perennial features, usually relatively small in volume, that are confined to discrete geographic areas (MacDonald, et al., 2002).  Natural seeps are abundant in the Gulf of Mexico (GOM) as well as in other coastal margins world-wide (Kvenvolden, et al., 1992).  The location of many active seeps have been verified by submersible sampling (MacDonald, et al., 1996) or inferred from geophysical data (Roberts, 2006) . Natural seeps therefore produce repeated effects across a wide area of the GOM and provide an excellent source of ocean slicks in multiple SAR images. Because sensor parameter and environmental conditions vary among SAR images, algorithm results can be tested over a range of possible slicks.

Expressions of the thinnest oil slicks can range in width from 60 m to several hundred meters wide and are typically several thousand meters long. The ability of SAR to image natural oil slicks is also strongly influenced by environmental conditions; particularly wind speed and sea state. As described in several studies (De Beukelaer, et al., 2003; Espedal and Wahl, 1999; Fortuny-Guasch, 2003; Pellon de Miranda, et al., 2004), biogenic surfactants, pollution, and oil slicks from natural hydrocarbon seepage are detectable in a wind range from 2 to 10 m·s-1 and in an angle of  incidence range  from 20º to 45º. Ocean SAR image interpretation becomes more problematic when ancillary meteorological and oceanographic data is unavailable. Detection of slicks is challenged when wind speeds are too high for surfactant layers to form. Another challenge is false targets that occur when low wind conditions produce extensive dark areas (low wind features). In addition, the time required to manually delineate and quantify features of interest constrains the number of images that can be studied and standard image processing techniques do not fully solve this problem. Brightness thresholding captures numerous false targets; also, standard methods of edge detection do not perform well where intense noise or SAR processing artifacts are present. Automated routines for semi-supervised pattern recognition and image segmentation have the potential to improve analysis of the SAR archive for basin-wide and long time-series investigations.

Therefore, in this paper we describe a Texture Classifying Neural Network Algorithm (TCNNA), which analyzes the textural content of SAR data in the context of sensor parameters and environmental variables. The algorithm was developed using verified features (ocean slicks) produced by natural oil seeps in the GOM. However the general approach may be applicable to other classes of ocean features.

Download Full Version

 
 
 
| Site Map | Contact Webmaster |
 

Follow us on