Eric J. Hochberg

Spectral reflectance of coral reef bottom-types worldwide and implications for coral reef remote sensing

Abstract Coral reef benthic communities are mosaics of individual bottom-types that are distinguished by their taxonomic composition and functional roles in the ecosystem. Knowledge of community structure is essential to understanding many reef processes. To develop techniques for identification and mapping of reef bottom-types using remote sensing, we measured 13,100 in situ optical reflectance spectra (400–700 nm, 1-nm intervals) of 12 basic reef bottom-types in the Atlantic, Pacific, and Indian Oceans: fleshy (1) brown, (2) green, and (3) red algae; non-fleshy (4) encrusting calcareous and (5) turf algae; (6) bleached, (7) blue, and (8) brown hermatypic coral; (9) soft/gorgonian coral; (10) seagrass; (11) terrigenous mud; and (12) carbonate sand. Each bottom-type exhibits characteristic spectral reflectance features that are conservative across biogeographic regions. Most notable are the brightness of carbonate sand and local extrema near 570 nm in blue (minimum) and brown (maximum) corals. Classification function analyses for the 12 bottom-types achieve mean accuracies of 83%, 76%, and 71% for full-spectrum data (301-wavelength), 52-wavelength, and 14-wavelength subsets, respectively. The distinguishing spectral features for the 12 bottom-types exist in well-defined, narrow (10–20 nm) wavelength ranges and are ubiquitous throughout the world. We reason that spectral reflectance features arise primarily as a result of spectral absorption processes. Radiative transfer modeling shows that in typically clear coral reef waters, dark substrates such as corals have a depth-of-detection limit on the order of 10–20 m. Our results provide the foundation for design of a sensor with the purpose of assessing the global status of coral reefs.

Spectral discrimination of coral reef benthic communities

Abstract Effective identification and mapping of coral reef benthic communities using high-spatial and -spectral resolution digital imaging spectrometry requires that the different communities are distinguishable by their spectral reflectance characteristics. In Kaneohe Bay, Oahu, Hawaii, USA, we collected in situ a total of 247 spectral reflectances of three coral species (Montipora capitata, Porites compressa, Porites lobata), five algal species (Dictyosphaeria cavernosa, Gracilaria salicornia, Halimeda sp., Porolithon sp., Sargassum echinocarpum) and three sand benthic communities (fine-grained carbonate sand, sand mixed with coral rubble, coral rubble). Major reflectance features were identified by peaks in fourth derivative reflectance spectra of coral (at 573, 604, 652, 675 nm), algae (at 556, 601, 649 nm) and sand (at 416, 448, 585, 652, 696 nm). Stepwise wavelength selection and linear discriminant function analysis revealed that spectral separation of the communities is possible with as few as four non-contiguous wavebands. These linear discriminant functions were applied to an airborne hyperspectral image of a patch reef in Kaneohe Bay. The results demonstrate the ability of spectral reflectance characteristics, determined in situ, to discriminate the three basic benthic community types: coral, algae and sand.

Capabilities of remote sensors to classify coral, algae, and sand as pure and mixed spectra

Abstract We investigate the abilities of seven remote sensors to classify coral, algae, and carbonate sand based on 10,632 reflectance spectra measured in situ on reefs around the world. Discriminant and classification analyses demonstrate that full-resolution (1 nm) spectra provide very good spectral separation of the bottom-types. We assess the spectral capabilities of the sensors by applying to the in situ spectra the spectral responses of two airborne hyperspectral sensors (AAHIS and AVIRIS), three satellite broadband multispectral sensors (Ikonos, Landsat-ETM+ and SPOT-HRV), and two hypothetical satellite narrowband multispectral sensors (Proto and CRESPO). Classification analyses of the simulated sensor-specific spectra produce overall classification accuracy rates of 98%, 98%, 93%, 91%, 64%, 58%, and 50% for AAHIS, AVIRIS, Proto, CRESPO, Ikonos, Landsat-ETM+, and SPOT-HRV, respectively. Analyses of linearly mixed sensor-specific spectra reveal that the hyperspectral and narrowband multispectral sensors have the ability to discriminate between coral and algae across many levels of mixing, while the broadband multispectral sensors do not. Applying the results of the general mixing analyses to a specific spatial organization of coral, algae, and sand indicates that the hyperspectral sensors accurately estimate areal cover of the bottom-types regardless of pixel resolution. The narrowband multispectral sensors overestimate coral cover by 11–15%, while the broadband sensors underestimate algae cover by 7–29% and overestimate coral cover by 24–103%. We conclude that currently available satellite sensors are inadequate for assessment of global coral reef status, but that it is both necessary and possible to design a sensor system suited to the task.

Sea surface correction of high spatial resolution Ikonos images to improve bottom mapping in near-shore environments

A significant number of high spatial resolution (4 m) Ikonos images acquired over shallow coastal environments present quasi-stochastic sea surface effects that seriously compromise reconnaissance of bottom features. This problem is common in wide field-of-view images where there is limited control on acquisition conditions other than cloud cover. To eliminate most of these wave and glint patterns, we use the near-infrared band, which exhibits maximum absorption and minimal water leaving radiance over clear waters, to characterize the spatial distribution of relative glint intensity, which is then scaled by absolute glint intensities in each of the visible bands. The result is subtracted from the visible bands, thus filtering out glint effects. Corrected visible bands clearly reveal seabed structural features obscured in the original data. Before- and after-correction classifications of an Ikonos image of Lee Stocking Island (Bahamas) reveal an improvement of users accuracies for critical benthic habitat classes such as coral-dominated habitat (46.8% versus 60.5%) or dense seagrass beds (31.7% versus 52.1%). This technique offers potential to use previously discarded sections of high spatial resolution airborne or satellite images of optically shallow water for mapping substrate features.

Change detection in shallow coral reef environments using Landsat 7 ETM+ data

Abstract This paper aims to clarify the potential of the new Landsat 7/Enhanced Thematic Mapper Plus (ETM+) sensors for change detection in coral reef environments. We processed images of two reef sites in Florida and Hawaii acquired over short time intervals (2 weeks and 3 months). During these periods, reefs were not affected by major disturbances (phase shift, strategy shift, bleaching, and hurricanes). This stability allowed us to assess the bias in change detection analysis. Two methods for change detection analysis were applied. The first one estimates the atmospheric conditions (Rayleigh and aerosol radiances, ozone and diffuse transmittances) using an ETM+/SeaWiFS multisensor approach. The second method is an empirical correction based on pseudoinvariant features that compensates for different atmospheric conditions as well as for any sensor (noise) or environmental (water column, sea surface state) conditions. The atmospheric correction alone did not provide an accurate match in images across time due to significant whitecaps and possible sun glint and its products required an empirical adjustment. Therefore, for the images in this study there was not substantial benefit in performing an atmospheric correction compared to an empirical correction alone. Both methods resulted in a minimum uncertainty of 4, 3, and 3 digital counts, respectively, in ETM+ Bands 1–3. Finally, we completed the study of real images by the analysis of ETM+ reflectance spectra for a large variety of coral reef objects. We concluded 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 for the ETM+ sensor.

Decorrelating remote sensing color bands from bathymetry in optically shallow waters

We have developed a simple technique to decorrelate remote sensing color band data from depth in optically shallow water. The method linearizes color band data with respect to depth by subtracting an optically deepwater value from the entire waveband under consideration and taking the natural logarithm of the result. Next, this linearized waveband is rotated about the model 2 regression line computed against a bathymetry band. The rotated color band is decorrelated from water depth. We demonstrate the technique for a small area of Kailua Bay, Oahu, HI, using Quickbird multispectral and Scanning Hydrographic Operational Airborne Lidar Survey LIDAR data. Results indicate that color band data are effectively decorrelated from depth, while bottom reflector variability is maintained, thus providing the basis for further analysis of the depth-invariant wavebands. The primary benefit of our technique is that wavebands are rotated independently, preserving relative spectral information.

Bio-optical modeling of photosynthetic pigments in corals

The spectral reflectance of coral is inherently related to the amounts of photosynthetic pigments present in the zooxanthellae. There are no studies, however, showing that the suite of major photosynthetic pigments can be predicted from optical reflectance spectra. In this study, we measured cm-scale in vivo and in situ spectral reflectance for several colonies of the massive corals Porites lobata and Porites lutea, two colonies of the branching coral Porites compressa, and one colony of the encrusting coral Montipora flabellata in Kaneohe Bay, Oahu, Hawaii. For each reflectance spectrum, we collected a tissue sample and utilized high-performance liquid chromatography to quantify six major photosynthetic pigments, located in the zooxanthellae. We used multivariate multiple regression analysis with cross-validation to build and test an empirical linear model for predicting pigment concentrations from optical reflectance spectra. The model accurately predicted concentrations of chlorophyll a, chlorophyll c2, peridinin, diadinoxanthin, diatoxanthin and β-carotene, with correlation coefficients of 0.997, 0.941, 0.995, 0.996, 0.980 and 0.984, respectively. The relationship between predicted and actual concentrations was 1:1 for each pigment, except chlorophyll c2. This simple empirical model demonstrates the potential for routine, rapid, non-invasive monitoring of coral-zooxanthellae status, and ultimately for remote sensing of reef biogeochemical processes.

Derivation and Integration of Shallow-Water Bathymetry: Implications for Coastal Terrain Modeling and Subsequent Analyses

Satellite and acoustic remote sensing enable the collection of high-resolution seafloor bathymetry data for integration with terrestrial elevations into coastal terrain models. A model of Tutuila Island, American Samoa, is created using depths derived from IKONOS satellite imagery to provide data in the near-shore gap between sea level and the beginning of sonar data at 10–15 m depth. A derivation method gauging the relative attenuation of blue and green spectral radiation is proven the most effective of several proposed in recent literature. The resulting coastal terrain model is shown to be accurate through statistical analyses and topographic profiles.

Multi-scale remote sensing of microbial mats in an atoll environment

Microbial mats are encountered in many coastal environments. They are generally constituted by stratified layers produced by the development of various micro-organisms. A current research programme aims to assess the biotechnological potential of the microbial communities in South Pacific atolls. As part of this project, the characterisation of mats was examined in Rangiroa atoll (French Polynesia) using high resolution (20-30 m) multi-spectral images (SPOT HRV and Landsat ETM+), hyperspectral imagery (CASI (Compact Airborne Spectrographic Imager), 27 bands, 5.5 2 1 m) and in situ reflectance spectra. At atoll scale, mats are successfully inventoried among different geomorphological environments by fuzzy classification/segmentation of the 20-30 m resolution multi-spectral data. Laboratory biochemical analysis (not described here) highlighted and identified the most promising mats for their biotechnological potential. The spectral signatures of these remarkable mats are described at two scales. Major reflectance features of the microbial community were identified by peaks in fourth-derivatives analysis, discriminating the surface layers dominated by the cyanobacteria Schizothrix sp. (orange mats), Scytonema sp. (grey-black mats) and Phormidium sp. (green mats). At mat scale, CASI-derived spectral signatures from heterogeneous 5.5m 2 pixels including various microbial communities, vegetation, sand and water did not provide fine optical distinctions between mats because of mixing effects. This multi-scale analysis provides optical and geomorphological criteria to locate interesting stratified mats in other atolls.

Multi-Scale Remote Sensing of Coral Reefs

Institut de Recherche pour le Developpement, BP A5, 98848 Noumea, New Caledonia University of Hawaii, School of Ocean and Earth Science and Technology, Hawaii Institute of Marine Biology, P.O. Box 1346, Kaneohe, HI, 96744 USA Institute for Marine Remote Sensing, College of Marine Science, University of South Florida, 140 7th Ave. South, St Petersburg, FL, 33701 USA USGS Center for Coastal and Watershed Studies, 600 4th Street South, St. Petersburg, FL, 33701 USA