Maritimes Region Coral and Sponge Distribution Models for Benthic Habitat

Random forest and generalized additive models predict the probability of occurrence and biomass for sponges, sea pens, and gorgonian corals across the Maritimes Region. Models demonstrated excellent predictive capacity, with cross-validated AUC values ranging from 0.760 to 0.977. The work by Fisheries and Oceans Canada scientists utilizes trawl catch data and benthic imagery to identify areas of suitable habitat for sensitive benthic taxa.

Use Cases

• Identify potential distribution of the rare sponge Vazella pourtalesi using its modeled probability of occurrence for conservation planning.
• Contrast modeled areas of suitable habitat for large gorgonian corals against known closure areas designated for their protection.
• Compare biomass distribution predictions from random forest models and generalized additive models (GAMs) for taxonomic groups along the continental slope.
• Utilize predicted probability of occurrence layers for sponges and corals in kernel density analyses to refine significant concentration areas.

Strengths

• Models show excellent predictive capacity with cross-validated AUC scores ranging from 0.760 to 0.977.
• Covers the entire spatial extent of Fisheries and Oceans Canada's Maritimes Region, including the Scotian Shelf and Slope.
• Integrates multiple data sources, including DFO multispecies trawl catch data and in situ benthic imagery observations.
• Includes models for specific, ecologically important taxa like the rare sponge Vazella pourtalesi.

Limitations

• Distribution predictions are models based on available surveys and may not capture full ground truth in unsurveyed areas.
• The underlying observation data from targeted surveys are acknowledged to be limited in their spatial extent.
• Model performance varies by taxonomic group, with GAMs providing superior biomass predictions for some groups along the slope.

Provenance

Source

Fisheries and Oceans Canada, Ocean Ecosystems Science Division.

Collection Method

Species distribution modeling using random forest and generalized additive models on trawl catch and benthic imagery data.

Freshness

Data published in July 2023, with a platform record last updated in March 2026.

Geography

Fisheries and Oceans Canada's Maritimes Region, including the Scotian Shelf and Slope.