Hyperspectral Applications

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 hyper-spectral analysis of reefs. At the simplest level, hyperspectral data permits clas-sification 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 demon-strated 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.