Marine protected areas are an important tool for management of marine ecosystems. Despite their utility, ecological design criteria are often not considered or feasible to implement when establishing protected areas. In 2001, the Virgin Islands Coral Reef National Monument (VICRNM) in St John, US Virgin Islands was established by Executive Order. The VICRNM prohibits almost all extractive uses. Surveys of habitat and fishes inside and outside of the VICRNM were conducted in 2002-2004. Areas outside the VICRNM had significantly more hard corals, greater habitat complexity, and greater richness, abundance and biomass of reef fishes than areas within the VICRNM. The administrative process used to delineate the boundaries of the VICRNM did not include a robust ecological characterisation of the area. Because of reduced habitat complexity within the VICRNM, the enhancement of the marine ecosystem may not be fully realised or increases in economically important reef fishes may take longer to detect.

Effective management of coral reef ecosystems requires accurate, quantitative and spatially explicit information on patterns of species richness at spatial scales relevant to the management process. We combined empirical modelling techniques, remotely sensed data, field observations and GIS to develop a novel multi-scale approach for predicting fish species richness across a compositionally and topographically complex mosaic of marine habitat types in the U.S. Caribbean. First, the performance of three different modelling techniques (multiple linear regression, neural networks and regression trees) was compared using data from southwestern Puerto Rico and evaluated using multiple measures of predictive accuracy. Second, the best performing model was selected. Third, the generality of the best performing model was assessed through application to two geographically distinct coral reef ecosystems in the neighbouring U.S. Virgin Islands. Overall, regression trees outperformed multiple linear regression and neural networks. The best performing regression tree model of fish species richness (high, medium, low classes) in southwestern Puerto Rico exhibited an overall map accuracy of 75%; 83.4% when only high and low species richness areas were evaluated. In agreement with well recognised ecological relationships, areas of high fish species richness were predicted for the most bathymetrically complex areas with high mean rugosity and high bathymetric variance quantified at two different spatial extents (≤0.01km2). Water depth and the amount of seagrasses and hard-bottom habitat in the seascape were of secondary importance. This model also provided good predictions in two geographically distinct regions indicating a high level of generality in the habitat variables selected. Results indicated that accurate predictions of fish species richness could be achieved in future studies using remotely sensed measures of topographic complexity alone. This integration of empirical modelling techniques with spatial technologies provides an important new tool in support of ecosystem-based management for coral reef ecosystems.

Geographic information system-based analysis was used to derive comprehensive, consistent estimates of the potential area of broadly defined, shallow-water, tropical and subtropical coral ecosystems within the territorial sea and exclusive economic zone of the United States. A coral ecosystem is composed of habitats including unconsolidated sediment, mangrove, hermatypic coral, colonized hardbottom, and submerged vegetation, and major structural zones like reef crest, lagoon, and fore reef. This broad definition reflects the importance of both reef and non-reef habitats and structural zones in the function of these ecosystems. Nautical charts, published by the National Oceanic and Atmospheric Administration’s Office of the Coast Survey, provide a consistent source of 10-fathom (∼18 m) and 100-fathom (∼183 m) depth curve information. The 10-fathom or 100-fathom depth curves are used as surrogates for the potential distribution and extent of shallow-water coral ecosystems in tropical and subtropical U.S. waters. An estimated 36,813 sq·km area has been identified where coral ecosystems can potentially be found in waters less than 10 fathoms (18 m) deep. In addition, an estimated 143,059 sq·km area has been identified where coral ecosystems potentially can be found in U.S. waters at depths down to 100 fathoms (183 m). Results also indicate that previous studies underestimated the extent of potential coral ecosystems for some locations in U.S. tropical and subtropical waters by as much as 100% and that the regional distribution of coral ecosystems has been incorrectly reported.