reflection (triple bounce) by corner reflectors in the superstructure (Pichel et al.
2004). In general, direct returns from steel vessels with a length greater than or
equivalent to the resolution of a pixel in a SAR image can be detected most of the
time, and vessels with lengths of about half the SAR resolution can be detected
half of the time. Factors affecting the detectability of direct returns from vessels
include: (i) ship characteristics—structural configuration and materials used, orientation of the ship in relation to the radar signal, and ship size; (ii) environmental
conditions—sea state, wind speed, proximity to land and small islands; (iii) radar
characteristics—angle of incidence, polarization, resolution, sensitivity; (iv) SAR
image quality—image processing errors, beam seams, speckle noise, and nadir
ambiguities; and (v) image resolution. In particular, one will see a variation in
vessel detectability in each SAR image from near-range to far-range since the
backscatter from the ocean surface is very large at small angles of incidence
(*20°), and significantly smaller at high angles of incidence (*50°). The vessel
backscatter, however, does not change as dramatically with incidence angle; thus,
there is an increase in contrast between vessel and ocean as incidence angle
increases. Automated vessel detection algorithms therefore have to account for this
change in contrast within the SAR image. Many such algorithms determine
localized background ocean statistics throughout the image and thereby set a
dynamic threshold for detection of a vessel in different parts of the image. This
allows the algorithm to maintain a constant false alarm rate (CFAR) and achieve a
fairly constant detection sensitivity throughout the image (Vachon et al. 2000;
Wackerman et al. 2001).
13.4.2 Example Applications
Wind energy SAR-derived wind imagery is useful for monitoring coastal wind
spatial variations due to: (i) island and coastal landform wind shadowing, island
gaps, and island wake formation (Beal et al. 2005); (ii) katabatic winds (Li et al.
2007); (iii) barrier jets (Winstead et al. 2006); (iv) vortex streets (Li et al. 2008a);
(v) atmospheric fronts (Young et al. 2005); (vi) hurricanes and other storms
(Horstmann et al. 2006; Sikora et al. 2000); (vii) convection cells, thunderstorms,
and roll vortices (Sikora and Ufermann 2004); (viii) coastal and mountain lee waves
(Li et al. 2011); and (ix) other atmospheric marine boundary layer phenomena.
Figure 13.8 shows local wind variations, specifically ocean surface patterns of
atmospheric lower-boundary-layer roll vortices and island wakes, over the Great
Barrier Reef in the vicinity of Rockhampton in eastern Australia. The region
stretches from Curtis Island at the center of the southern edge of the image to Cape
Manifold near the northern edge of the image. The island in the left center of the
image is Great Keppel Island. The wind patterns were calculated from a RADARSAT-1 Standard Mode image covering an area 100 km on a side (original SAR
image resolution of 30 m). Wind directions used in the CMOD5 algorithm were
derived from the Navy Operational Global Atmospheric Prediction System
356
M. L. Heron et al.
2004). In general, direct returns from steel vessels with a length greater than or
equivalent to the resolution of a pixel in a SAR image can be detected most of the
time, and vessels with lengths of about half the SAR resolution can be detected
half of the time. Factors affecting the detectability of direct returns from vessels
include: (i) ship characteristics—structural configuration and materials used, orientation of the ship in relation to the radar signal, and ship size; (ii) environmental
conditions—sea state, wind speed, proximity to land and small islands; (iii) radar
characteristics—angle of incidence, polarization, resolution, sensitivity; (iv) SAR
image quality—image processing errors, beam seams, speckle noise, and nadir
ambiguities; and (v) image resolution. In particular, one will see a variation in
vessel detectability in each SAR image from near-range to far-range since the
backscatter from the ocean surface is very large at small angles of incidence
(*20°), and significantly smaller at high angles of incidence (*50°). The vessel
backscatter, however, does not change as dramatically with incidence angle; thus,
there is an increase in contrast between vessel and ocean as incidence angle
increases. Automated vessel detection algorithms therefore have to account for this
change in contrast within the SAR image. Many such algorithms determine
localized background ocean statistics throughout the image and thereby set a
dynamic threshold for detection of a vessel in different parts of the image. This
allows the algorithm to maintain a constant false alarm rate (CFAR) and achieve a
fairly constant detection sensitivity throughout the image (Vachon et al. 2000;
Wackerman et al. 2001).
13.4.2 Example Applications
Wind energy SAR-derived wind imagery is useful for monitoring coastal wind
spatial variations due to: (i) island and coastal landform wind shadowing, island
gaps, and island wake formation (Beal et al. 2005); (ii) katabatic winds (Li et al.
2007); (iii) barrier jets (Winstead et al. 2006); (iv) vortex streets (Li et al. 2008a);
(v) atmospheric fronts (Young et al. 2005); (vi) hurricanes and other storms
(Horstmann et al. 2006; Sikora et al. 2000); (vii) convection cells, thunderstorms,
and roll vortices (Sikora and Ufermann 2004); (viii) coastal and mountain lee waves
(Li et al. 2011); and (ix) other atmospheric marine boundary layer phenomena.
Figure 13.8 shows local wind variations, specifically ocean surface patterns of
atmospheric lower-boundary-layer roll vortices and island wakes, over the Great
Barrier Reef in the vicinity of Rockhampton in eastern Australia. The region
stretches from Curtis Island at the center of the southern edge of the image to Cape
Manifold near the northern edge of the image. The island in the left center of the
image is Great Keppel Island. The wind patterns were calculated from a RADARSAT-1 Standard Mode image covering an area 100 km on a side (original SAR
image resolution of 30 m). Wind directions used in the CMOD5 algorithm were
derived from the Navy Operational Global Atmospheric Prediction System
356
M. L. Heron et al.