Although there are some examples of change detection on coral reefs,
(Loubersac et al. 1988; Elvidge et al. 2004), decadal-scale observation and analysis is still scarce. The Landsat program, which has been collecting imagery since
1972, is the best suite of sensors to achieve this purpose (Table 3.1). Radiative
transfer simulation to assess the feasibility for change detection by Landsat ETM+
showed that the assessment of the rates of change in three ubiquitous classes
‘sand’, ‘background’ (including rubble, pavement, and heavily grazed dead coral
structure), and ‘foreground’ (including living corals and macroalgae) emerges as
the most reproducible and feasible application (Andréfouët et al. 2001b). For
example, Dustan et al. (2001) explored the use of Landsat TM to detect changes in
Florida. Palandro et al. (2008) extended their study, calculating absolute reflectance at the bottom, and performing classification using four classes (sand, bare
hard-bottom, covered hard-bottom, and coral) to ensure high classification accuracy. Results confirmed consistent degradation of coral reefs in Florida (Fig. 3.4).
In addition to observing changes in coral habitats, time-series imagery can also
be utilized to detect coral reef bleaching. Analysis of aerial photographs taken
during the 1998 bleaching event show that information on bleached corals can be
obtained only by sensors with high (\2 m) spatial resolution (Andréfouët et al.
2002a). A further complication is that analysis of a single image might not correctly identify pixels that contain both healthy and bleached corals, because pixels
that contain other substrate features, such as sand, can have reflectance characteristics similar to those pixels showing partially bleached corals. Comparison of
multi-temporal images thus allows more accurate assessment for detecting
bleached corals in a pixel, because an increase in reflectance will be recorded
where bleaching has taken place. Yamano and Tamura (2004) used 16 normalized
Landsat TM images from 1984–2000 to document the extent of severe bleaching
in 1998. Analysis of the same imagery was also used to validate radiative transfer
simulations that established quantitative limits for detecting coral reef bleaching
by satellite sensors.
3.3.3 Reef Modeling
Modeling is an effective tool to understand key processes and to make predictions
and forecasting possible. The contribution of remote sensing to modeling application can be divided into two categories: (1) validation and (2) input.
Remote sensing results can provide validation of modeling outputs and can
thereby be used to improve modeling accuracy. Examples include: estimation of
population size parameters in population dynamics modeling using an IKONOS
habitat map (Riegl and Purkis 2009); assessing habitat change using IKONOS
imagery for validation of cyclone trajectory modeling (Scopélitis et al. 2007);
validation of sediment transport modeling based on accreted and eroded areas in
atoll islands identified by comparing maps and IKONOS data (Yokoki et al. 2006);
and tuning a local erosion rate coefficient for sediment transport modeling based
66
H. Yamano
(Loubersac et al. 1988; Elvidge et al. 2004), decadal-scale observation and analysis is still scarce. The Landsat program, which has been collecting imagery since
1972, is the best suite of sensors to achieve this purpose (Table 3.1). Radiative
transfer simulation to assess the feasibility for change detection by Landsat ETM+
showed that the assessment of the rates of change in three ubiquitous classes
‘sand’, ‘background’ (including rubble, pavement, and heavily grazed dead coral
structure), and ‘foreground’ (including living corals and macroalgae) emerges as
the most reproducible and feasible application (Andréfouët et al. 2001b). For
example, Dustan et al. (2001) explored the use of Landsat TM to detect changes in
Florida. Palandro et al. (2008) extended their study, calculating absolute reflectance at the bottom, and performing classification using four classes (sand, bare
hard-bottom, covered hard-bottom, and coral) to ensure high classification accuracy. Results confirmed consistent degradation of coral reefs in Florida (Fig. 3.4).
In addition to observing changes in coral habitats, time-series imagery can also
be utilized to detect coral reef bleaching. Analysis of aerial photographs taken
during the 1998 bleaching event show that information on bleached corals can be
obtained only by sensors with high (\2 m) spatial resolution (Andréfouët et al.
2002a). A further complication is that analysis of a single image might not correctly identify pixels that contain both healthy and bleached corals, because pixels
that contain other substrate features, such as sand, can have reflectance characteristics similar to those pixels showing partially bleached corals. Comparison of
multi-temporal images thus allows more accurate assessment for detecting
bleached corals in a pixel, because an increase in reflectance will be recorded
where bleaching has taken place. Yamano and Tamura (2004) used 16 normalized
Landsat TM images from 1984–2000 to document the extent of severe bleaching
in 1998. Analysis of the same imagery was also used to validate radiative transfer
simulations that established quantitative limits for detecting coral reef bleaching
by satellite sensors.
3.3.3 Reef Modeling
Modeling is an effective tool to understand key processes and to make predictions
and forecasting possible. The contribution of remote sensing to modeling application can be divided into two categories: (1) validation and (2) input.
Remote sensing results can provide validation of modeling outputs and can
thereby be used to improve modeling accuracy. Examples include: estimation of
population size parameters in population dynamics modeling using an IKONOS
habitat map (Riegl and Purkis 2009); assessing habitat change using IKONOS
imagery for validation of cyclone trajectory modeling (Scopélitis et al. 2007);
validation of sediment transport modeling based on accreted and eroded areas in
atoll islands identified by comparing maps and IKONOS data (Yokoki et al. 2006);
and tuning a local erosion rate coefficient for sediment transport modeling based
66
H. Yamano