NL Small Gorgonians Presence Probability tif

Species distribution modelling using a random forest (RF) machine learning approach was used to predict the probability of occurrence and biomass of sponges, sea pens, large and small gorgonians in the Newfoundland and Labrador Region. A suite of 66 environmental predictor variables from different data sources were used. Species occurrence was predicted using all presence and absence data (unbalanced model), and a balanced species prevalence model (i.e. an equal number of presences and absences). The models produced from the unbalanced data were chosen as the better prediction surfaces for all four groups. Three measures of accuracy were used to assess model performance: sensitivity, specificity, and AUC, or Area Under the Receiver Operating Curve. The accuracy measures for the random forest model using all smallgorgonian presenceand absence dataanda threshold equal to species prevalence (0.07)were AUC= 0.859, sensitivity= 0.800 andspecificity=0.800; indicating good model performance.