Chapter 4
Hyperspectral Applications
John D. Hedley
Abstract Hyperspectral approaches are at the technological forefront of optical
remote sensing of coral reef environments. Currently most hyperspectral data
acquisition employs instruments mounted on airplanes, but in the coming years
several planned satellite instruments will increase data availability for hyperspectral analysis of reefs. At the simplest level, hyperspectral data permits classification techniques to derive greater number of classes at higher accuracy than
multispectral data can support. Alternatively, full spectral reflectance profiles at
each pixel allow band-ratio or derivative approaches to look for features of benthic
types that occur at specific wavelengths. But while the feasibility of this approach
is supported by in situ data, there have been relatively few successfully demonstrated image analyses. Beyond this, working with full spectral reflectance profiles
has stimulated exciting new model-based methods that aim to tease apart depth,
benthic type and water quality parameters simultaneously. These methods can also
incorporate uncertainty propagation, so that error bars can be placed on each
derived parameter at every image pixel. Working with hyperspectral data takes
coral reef remote sensing to the edge of what can be achieved by per-pixel optical
analysis. Natural variations in the reflectance of benthic types and water column
properties become limiting and fundamentally confound some objectives. This
prompts future developments to look at analyzing spatial patterns and also to
establish the cost-benefit ratio of the integration of other data, such as sonar and
LiDAR data.
J. D. Hedley (&)
ARGANS Ltd., Tamar Science Park, Derriford PL6 8BT Plymouth, Devon, UK
e-mail: jhedley@argans.co.uk
J. A. Goodman et al. (eds.), Coral Reef Remote Sensing,
DOI: 10.1007/978-90-481-9292-2_4,
Ó Springer Science+Business Media Dordrecht 2013
79
Hyperspectral Applications
John D. Hedley
Abstract Hyperspectral approaches are at the technological forefront of optical
remote sensing of coral reef environments. Currently most hyperspectral data
acquisition employs instruments mounted on airplanes, but in the coming years
several planned satellite instruments will increase data availability for hyperspectral analysis of reefs. At the simplest level, hyperspectral data permits classification techniques to derive greater number of classes at higher accuracy than
multispectral data can support. Alternatively, full spectral reflectance profiles at
each pixel allow band-ratio or derivative approaches to look for features of benthic
types that occur at specific wavelengths. But while the feasibility of this approach
is supported by in situ data, there have been relatively few successfully demonstrated image analyses. Beyond this, working with full spectral reflectance profiles
has stimulated exciting new model-based methods that aim to tease apart depth,
benthic type and water quality parameters simultaneously. These methods can also
incorporate uncertainty propagation, so that error bars can be placed on each
derived parameter at every image pixel. Working with hyperspectral data takes
coral reef remote sensing to the edge of what can be achieved by per-pixel optical
analysis. Natural variations in the reflectance of benthic types and water column
properties become limiting and fundamentally confound some objectives. This
prompts future developments to look at analyzing spatial patterns and also to
establish the cost-benefit ratio of the integration of other data, such as sonar and
LiDAR data.
J. D. Hedley (&)
ARGANS Ltd., Tamar Science Park, Derriford PL6 8BT Plymouth, Devon, UK
e-mail: jhedley@argans.co.uk
J. A. Goodman et al. (eds.), Coral Reef Remote Sensing,
DOI: 10.1007/978-90-481-9292-2_4,
Ó Springer Science+Business Media Dordrecht 2013
79