PATCH REEF ANALYSIS USING LIDAR-DERIVED METRICS AT BISCAYNE NATIONAL PARK, FLORIDA

This study uses submerged topographic data and coral reef rugosity estimates derived from NASA's Experimental Advanced Airborne Research Lidar (EAARL) for Biscayne National Park, Florida (Brock et al., 2004, Brock et al., 2006). The purpose is to evaluate the capability of NASA EAARL lidar to describe patch reef variability and habitat complexity. Over one thousand patch reefs were analyzed using mean neighborhood statistics and reclassification of lidar and slope data. The area of each reef was divided into the following categories: (1) base or reef footprint, (2) side of the reef, and (3) top of the reef. Different mean metrics were derived for each reef category from the submerged topography and rugosity data. Scatter plots of reef depth versus rugosity, relative relief, shape index, perimeter, area and volume, respectively, suggested that the multivariate data is bi-modal. A mixture of two log-normal distributions suitably approximated the reef depth distribution. The depth at which the two log-normal distributions intersect was used to divide the patch reefs in shallow (less than 7.77m) and deeper reefs (7.77 to 14m), respectively. The results showed that shallow patch reefs had a tendency to be bigger with a smaller relative relief than the deeper patch reefs. Topographic complexity, or rugosity, increased with depth for shallow reefs. In contrast, for deeper reefs, rugosity decreased with depth.

An independent component analysis was carried out on principal components derived from the patch reef metrics to determine if depth was the single most important factor to influence reef physical variability and habitat complexity. Principal components, although uncorrelated, are only partly independent (Hyvarinen & Oja, 2000). Two distinctly different independent components emerged from the analysis of 7 principal components that described over 95% of data variability. We demonstrate that one independent component can be a function of patch reef rugosity while the other independent component is most likely a function of reef geometry and depth. These two independent components divide the patch reefs population in three depth classes, (1) from 2 to 6m, (2) from 6 to 9.5m, and (3) from 9.5 to 14m, respectively. The deepest class correlates with the tail data not modeled by the log-normal mixture distribution.

Independent component analysis is more sensitive than simple multivariate analysis in assessing data variability. Multivariate analysis confirmed two major different populations, shallow and deeper reefs, with divergent rugosity correlations but similar behavior of other reef metrics such as perimeter, area and volume. Independent component analysis suggests that 3 classes may be more appropriate to describe patch reef variability and habitat complexity in Biscayne National Park.

References cited:

Brock, J.C., Wright, C.W., Clayton, T.D., Nayegandhi, A., 2004, LIDAR optical rugosity of coral reefs in Biscayne National Park, Florida, Coral Reefs 23: 48 – 59

Brock, J.C., Wright, C.W., Kuffer, I.B., Hernandez, R., Thompson, Ph., 2006, Airborne lidar sensing of massive stony coral colonies on patch reefs in the northern Florida reef tract, Remote Sensisng of Environment 104: 31 – 42

Hyvarinen, A., Oja, E., 2000, Independent Component Analysis: Algorithms and Applications, Neural Networks 13(4-5):411 – 430