A "quick and clean" photographic method for the description of coral reef habitats

The use of scuba-based photo/video methods for characterizing coral reef habitats has gained increased popularity within the last decade, but few work examined the potentiality of surface photography to provide accurate, reliable habitat profiles in contrasted habitats. Photo transects were thus conducted by snorkeling in contrasted reef biotopes (reef flat, reef crest, sandy bottom) from the south-west lagoon of New Caledonia, to develop and test a "quick and clean" approach suitable for addressing monitoring as well as research-oriented programs. Pictures were taken by a snorkeler from the surface over twelve (20 x 1 m) reef crest/reef flat/soft bottoms transects using a standard 8 Mpixel digital-camera with underwater housing. Habitats were characterized from percent covers for 15 categories of local habitat variables related to sediment type and substrate coverage. Exhaustive area analyses using computer-assisted manual digitalizing were used to provide reliable habitat profiles from the digital pictures. Results were subsequently compared with surface estimates derived from random stratified point count techniques, for numbers of points comprised between I and 99 per m(2). Sampling-based randomization techniques allowed us to provide robust, reliable statistical estimates of accuracy and precision over 1000 randomized bootstrap replicates per transect. Results emphasized high accuracy and precision at transect scale whatever the reef biotopes considered, with maximum deviations from reference values of similar to 1 percent cover in almost all cases and associated variances From a practical point of view, using a 9 points/m(2) ratio clearly provided reliable, quantitative descriptions of our reef transects (maximal errors <1.5 percent cover with 95% confidence level). Cost-effectiveness is high, with 15-30 minutes/transect from field data collection (< 10 min) to computation of final percent covers (10-20 min). The method outlined in this paper thus combines high statistical efficiency and logistical ease, and could be used to address more functional perspectives