Krigsman, L. M., Yoklavich, M. M., Dick, E. J., Cochrane, G. R.

Models and maps: predicting the distribution of corals and other benthic macro-invertebrates in shelf habitats

2012

Ecosphere

3

1

Article 3

"Presence-absence data of benthic macro-invertebrates and associated habitat (i.e., sediment type and depth) were collected using a towed camera sled in selected areas along the coast off southern California. Using this information, we developed generalized linear models (GLMs) to predict the probability of occurrence of five commonly observed taxa (cup corals, hydroids, short and tall sea pens, and brittle stars in the sediment) within the Santa Barbara Channel (SBC). We employed a utility-based approach to validate the models and to identify optimal cutoff thresholds for the GLM predictions, given a relative cost of false negatives and false positives. We present optimal cutoff values under a range of cost ratios. Out-of-sample predictive accuracy, assuming equal costs of false negatives and false positives, ranged from 75% to 89%. Estimated area under the characteristic curve (AUC) in our models ranged from 0.76 for brittle stars in sediment to 0.91 for cup corals. An AUC value above 0.7 is an acceptable level of performance, between 0.8 and 0.9 is excellent, and above 0.9 is outstanding. We developed predictive maps of probability of occurrence using our validated models and a seafloor character map based on fine-scale (2-5 m) data. Cup corals and hydroids had high predicted probabilities of occurrence in areas of hard substrata, while short and tall sea pens were predicted to occur in parts of the SBC that had unconsolidated and mixed sediment. Our model predicted that brittle stars would occur throughout the entire channel on various bottom types. The combination of high-resolution seafloor character maps and predictive models of invertebrate distributions will aid managers in identifying habitats of particular concern and areas with vulnerable, deep-sea corals. Our models will be useful for marine spatial planning and ecosystem-based management, as well as for assessing the effectiveness of essential fish habitat closures and other marine protected areas."

California Seafloor Mapping Project, deep-sea corals, generalized linear models, predictive species distribution maps

10.1890/ES11-00295.1