Fig. 3.

Relative variable importance in the prediction of the initiation of, a exploratory behaviour and, b long distance oriented flight in Pacific black duck after 2500 iterations of the random forest model. Variables are shown in order of their increasing relevance to increasing the accuracy of predicting a change in behaviour. Variables which had a positive influence on the performance of the model are scaled from 1 to 0 in relation to the most important variable for ease of interpretation. Variables showing negative values are those that, when included in an iteration of the model, reduced the performance of the model. Variables of 0 or negative importance can be ignored as they did not help to increase the predictive power of the model