Fig. 3
Model outputs for Random Forest classification method. The oceanographic variables (temperature, dissolved oxygen and salinity) and geographic variables (Longitude W, Latitude S, depth) are hierarchically ordered indicating the most important effects on the distribution of Projasus bahamondei (a, b) and Paromola rathbuni (c, d). Mean Decrease Accuracy is a metric that evaluates the importance of each variable. Mean Decrease Gini is a measure of the impurity of decision tree nodes; higher values indicate that the variable is important for reducing the impurity of the nodes.
