Answers are grounded in this paper and cite their sources.
Investigates how warping factor affects whistle repertoire size in rough-toothed dolphins, suggesting optimal settings for precise measurements.
ARTwarp is a commonly used method for estimating repertoire sizes for species that produce tonal signals. ARTwarp uses adaptive resonance theory neural networks and dynamic time warping (DTW) to organize signals into categories. Categories are then plotted as a discovery curve and the point at which the slope is zero indicates the size of the repertoire. DTW stretches or compresses sections of time-frequency contours in the time domain to facilitate comparisons between contours. The maximum warping factor (WF) is user-selected. Here, we investigate the impact of WF on the estimated size of the whistle repertoire of rough-toothed dolphins (Steno bredanensis). We used the Real-time Odontocete Call Classification (ROCCA) module in PAMGuard to manually trace 1000 whistle contours recorded from 13 single species schools of rough-toothed dolphins in Hawaiian waters. We then ran ARTwarp twice, once with WF = 3 and then with WF = 2. A Vigilance Parameter (similarity threshold for inclusion in a category) of 95% was used for both runs. The size of the whistle repertoire was substantially higher when WF = 2 (140 categories) than when WF = 3 (79 categories). We will discuss causes of these differences and make recommendations for selecting the appropriate WF for repertoire descriptions and species comparisons.
No takes yet. Share an insight, caveat, or question.
Oswald et al. (2025) studied this question.
Synapse has enriched 10 related papers. Consider them for comparative context: