Spectral reflectance and remote sensing of coral reefs.

摘要

The immediate aim of this dissertation is to investigate spectral reflectance characteristics of, and spectral contrasts between, different classes of coral reef benthos worldwide. Classes include healthy and bleached corals; fleshy brown, green and red algae; non-fleshy calcareous and turf algae; carbonate sand; terrigenous mud; soft/gorgonian corals; and seagrass. I have measured 13,100 reflectance spectra in situ on reefs in the Atlantic, Indian and Pacific Oceans, representing the world's major coral reef biogeographic regions.; Two healthy coral types and the ten other reef classes exhibit characteristic and distinct spectral reflectance features that are conservative across biogeographic regions, and their distinguishing spectral features exist in narrow wavelength ranges (10–20 nm). Classification function analyses for the twelve community-types achieve mean accuracies of 83%, 76% and 71% for full-visible-spectrum data (301-wavelength), 52-wavelength, and 14-wavelength subsets, respectively. Radiative transfer modeling shows that in typically clear coral reef waters, low-albedo substrates such as corals have a depth-of-detection limit on the order of 10–20 m.; Healthy reef-building corals exhibit two basic modes of spectral reflectance at visible wavelengths. (1) The brown mode has characteristic positive reflectance features near 575, 605 and 650 nm which arise solely through absorption of light by zooxanthellae. (2) The blue mode lacks the positive 575 nm feature and often exhibits a negative feature near 580 nm, which is due to absorption by coral-host pigments. Other coral-host pigments produce second-order modifications to the basic reflectance shapes.; Analyses of simulated reflectance spectra for seven remote sensors (AAHIS, AVIRIS, Proto, CRESPO, Ikonos, Landsat-ETM+, SPOT-HRV) indicate that hyperspectral and narrowband multispectral sensors (with wavebands optimized for reef mapping) classify coral, algae and sand at higher accuracies (91–98%) than do broadband multispectral sensors (50–64%). The hyperspectral and narrowband multispectral sensors also have the ability to discriminate between coral and algae across many levels of linearly mixed spectra, while the broadband multispectral sensors do not.; This dissertation demonstrates that basic reef benthic classes are spectrally identifiable and differentiable. These results establish the basis for the ultimate goal of developing remote sensing technology and methodology for scientific study of coral reef ecosystems.