of the mapping process or the technological interpolation method; and 4) general
mapping error. General (total) map error conveys map quality, or ‘fitness for use’ by
end users (Chrisman, 1991). The conventional method of communicating ‘fitness for
use’ for map users is the confusion or error matrix (Richards, 1996). The error matrix
summarizes results by comparing a primary reference class label to the map land-cover
or land-use class for the sampling unit and presents errors of inclusion (commission
errors) and errors of exclusion (omission errors) in a classification.
6. Applications of Remote Sensing in Monitoring Programs
In the following two sections, information is presented from two perspectives to
illustrate practical applications of the concepts discussed in the preceding sections for
linking remote sensing and multi-temporal analysis techniques to coastal and coral reef
environmental indicators. In the first section, the status of remote sensing for mapping
and monitoring coral reefs is reviewed. A local scale example is then provided,
demonstrating the development and transfer to a government agency of a combined
field and remotely sensed system for mapping toxic algal blooms.
6.1 CORAL REEF MONITORING PROGRAMS USING REMOTELY SENSED
DATA
6.1.1 Potential Coral Reef Monitoring Capabilities Using Remote Sensing
Given the large and often inaccessible areas of reef ecosystems on a global basis,
remote sensing remains the only way to obtain synoptic data about coral reef ecosystem
composition and dynamics. Remotely sensed data provide a mapping capability that
would be impossible to replicate using traditional field survey techniques. Remote
sensing permits construction of baseline maps depicting reef location, extent, structure
and composition, and can also be used to provide information about water quality,
temperature and hydrodynamics, all of which may affect reef processes and health.
Landsat image data have been used for reef mapping applications since the mid
1980s (Jupp, Mayo et al., 1985; Kuchler, Jupp et al., 1986; Bour, 1988). It is commonly
accepted that these data are well suited to geomorphic and reef zonation studies, but
finer description of reef habitats (e.g. coral and algal definition) requires higher spatial
and/or spectral resolution imagery (Mumby and Edwards, 2002). However, to analyse
changes in reef substrate composition over time, the opportunities available with
Landsat data are yet to be fully exploited (but see Palandro et al., 2003a, c). Together,
the Landsat time series and frequency of image acquisition provide an information
source incomparable to other data types. However, while this is a more cost-effective
option than high-resolution data, Landsata data cannot provide the spatial or spectral
information required to map small-scale dynamics relevant to individual coral patches.
More recently, the increased availability of high spatial resolution satellite data
(e.g. Ikonos, Quickbird) has presented the opportunity to map reef habitats in greater
detail (see Robinson et al., Chapter 12, this book). Where analysis of Landsat imagery
may be able to generate a benthic habitat map with up to eight classes, Ikonos can
increase the definition to around thirteen classes with a similar accuracy level (Mumby
and Edwards, 2002; Andréfouët et al., 2003). However, this high spatial resolution still
may not be sufficient to provide information about many reef processes operating on a
finer scale. For example, Andréfouët et al. (2003) suggest a pixel size of as little as 15
cm is needed to detect coral bleaching. In mass bleaching events, however, such as
237
Integrated Information Acquisition and Management
mapping error. General (total) map error conveys map quality, or ‘fitness for use’ by
end users (Chrisman, 1991). The conventional method of communicating ‘fitness for
use’ for map users is the confusion or error matrix (Richards, 1996). The error matrix
summarizes results by comparing a primary reference class label to the map land-cover
or land-use class for the sampling unit and presents errors of inclusion (commission
errors) and errors of exclusion (omission errors) in a classification.
6. Applications of Remote Sensing in Monitoring Programs
In the following two sections, information is presented from two perspectives to
illustrate practical applications of the concepts discussed in the preceding sections for
linking remote sensing and multi-temporal analysis techniques to coastal and coral reef
environmental indicators. In the first section, the status of remote sensing for mapping
and monitoring coral reefs is reviewed. A local scale example is then provided,
demonstrating the development and transfer to a government agency of a combined
field and remotely sensed system for mapping toxic algal blooms.
6.1 CORAL REEF MONITORING PROGRAMS USING REMOTELY SENSED
DATA
6.1.1 Potential Coral Reef Monitoring Capabilities Using Remote Sensing
Given the large and often inaccessible areas of reef ecosystems on a global basis,
remote sensing remains the only way to obtain synoptic data about coral reef ecosystem
composition and dynamics. Remotely sensed data provide a mapping capability that
would be impossible to replicate using traditional field survey techniques. Remote
sensing permits construction of baseline maps depicting reef location, extent, structure
and composition, and can also be used to provide information about water quality,
temperature and hydrodynamics, all of which may affect reef processes and health.
Landsat image data have been used for reef mapping applications since the mid
1980s (Jupp, Mayo et al., 1985; Kuchler, Jupp et al., 1986; Bour, 1988). It is commonly
accepted that these data are well suited to geomorphic and reef zonation studies, but
finer description of reef habitats (e.g. coral and algal definition) requires higher spatial
and/or spectral resolution imagery (Mumby and Edwards, 2002). However, to analyse
changes in reef substrate composition over time, the opportunities available with
Landsat data are yet to be fully exploited (but see Palandro et al., 2003a, c). Together,
the Landsat time series and frequency of image acquisition provide an information
source incomparable to other data types. However, while this is a more cost-effective
option than high-resolution data, Landsata data cannot provide the spatial or spectral
information required to map small-scale dynamics relevant to individual coral patches.
More recently, the increased availability of high spatial resolution satellite data
(e.g. Ikonos, Quickbird) has presented the opportunity to map reef habitats in greater
detail (see Robinson et al., Chapter 12, this book). Where analysis of Landsat imagery
may be able to generate a benthic habitat map with up to eight classes, Ikonos can
increase the definition to around thirteen classes with a similar accuracy level (Mumby
and Edwards, 2002; Andréfouët et al., 2003). However, this high spatial resolution still
may not be sufficient to provide information about many reef processes operating on a
finer scale. For example, Andréfouët et al. (2003) suggest a pixel size of as little as 15
cm is needed to detect coral bleaching. In mass bleaching events, however, such as
237
Integrated Information Acquisition and Management