Proximate cues to phases of movement in a highly dispersive waterfowl, Anas superciliosa

Table 2 Accuracy rates for random forest models using initiation of phases of exploratory behaviour (EX) or initiation of long distance oriented flight (LD) as a response and weather/environmental variables as predictors

Response	Set of variables available for each tree	Area under ROC curve—training	Area under ROC curve—validation	% Classification Accuracy—training	% Classification Accuracy—validation
Initiation of exploratory flight	Bird ID	0.98	0.97	74	69
	Tmax
	MinPress
	Humidity
	Rain_Week
	SolarEx
	Rain_Month
	NDVI
	Humidity
Initiation of long distance oriented flight	Bird ID	0.92	0.78	71	72
	Tmin
	Tmax
	MinPress
	Humidity
	Rain_3Wks
	Rain_Week

The ROC (Receiver Operating Characteristic) curve is a means of measuring the performance of a binary classifier by plotting true positives against false positives, giving an indication of the sensitivity of the model. For explanation of abbreviations see Table 3 below. See Table A2 in Additional file 1 for examples of models using combinations of variables that did not produce classification accuracies high enough to warrant inclusion in the main results

ISSN: 2051-3933