Combining Traditional Behavioural Methods and Machine Learning to Quantify the Impact of Tourist Boats on The Endangered Hector’s Dolphin (Cephalorhyncus hectori) in New Zealand

Sirine, Yagouta (2026) Combining Traditional Behavioural Methods and Machine Learning to Quantify the Impact of Tourist Boats on The Endangered Hector’s Dolphin (Cephalorhyncus hectori) in New Zealand. MA/MSc, Erdőmérnöki Kar.

PDF MSc Sirine Yagouta.pdf Hozzáférés joga: Csak nyilvántartásba vett egyetemi IP-kről nyitható meg Download (2MB)

Absztrakt (kivonat)

This study using a machine-learning approach to investigate the behavioural responses of Critically Endangered Hector’s dolphins Cephalorhynchus hectori, a coastal species endemic to New Zealand, to the presence of tour boat operations at two South Island sites: the Black Cat at Lyttelton and the Caroline Cat at Timaru during austral summers of 2000-2002 and 2005. Dolphin behaviour was categorised into three main categories, which are stress related, association, and neutral states using both count based and time budget observations, allowing assessment of both the frequency and duration of behavioural responses to vessel presence. This approach highlights the inherent complexity of defining “stress” in wild dolphins, as behaviours traditionally interpreted as stress indicators in many marine mammals may also occur in positive or socially motivated contexts. Distinguishing between negative stress (avoidance, agitation) and positive stress (excitement, social engagement) is therefore essential for accurate interpretation. Consistent with earlier work by Travis (2008), the results indicate that while boat type did not significantly influence the likelihood of dolphin sightings vessel presence did influence behavioural patterns in more subtle, context dependent ways. Our machine learning approach identified sequence of behaviours that are likely associated with the presence of tour boats and showed heterogeneity among individuals in the speed and extent of moving into full stress state. This group of behaviours are likely to serve as a buffer state prior to full escalation into avoidance behaviours such as sustained speed increases (sa>), abrupt heading changes, and sustained swimming away (sa-sa). Revising and reanalysing the data generated by Travis using similar statistical methods, my results similarly show at Timaru, the Caroline Cat was associated with increased stress related behaviours during Season 1, although this effect diminished over time, suggesting potential habituation. At Lyttelton, neutral behaviours decreased in Season 2 while association behaviours increased, indicating that dolphins may actively choose to interact with the Black Cat during tour operations. By integrating traditional ethogram based methods, reconstructing behavioural state transition, and machine learning modelling, this study provides additional evidence at finer grained scale that assesses dolphin responses to tourism in New Zealand waters, and underscore the importance of investigating animal behaviour within a context aware framework, and highlight the need for continued monitoring to understand how tour operations influence both short term behaviour and long term population wellbeing in marine mammals.

Angol cím

Combining Traditional Behavioural Methods and Machine Learning to Quantify the Impact of Tourist Boats on The Endangered Hector’s Dolphin (Cephalorhyncus hectori) in New Zealand

Intézmény

Soproni Egyetem

Kar

Erdőmérnöki Kar

Tanszékcsoport/intézet

EMK - Vadgazdálkodási és Vadbiológiai Intézet

Szak

NEM RÉSZLETEZETT

Témavezető(k)

Helyi kari azonosító

IGGW61

Mű típusa:	Diplomadolgozat (MA/MSc)
Felhasználói azonosító szám (ID):	Sirine Yagouta
Dátum:	09 Júl 2026 07:26
Utolsó módosítás:	09 Júl 2026 07:26
URI:	http://diploma.uni-sopron.hu/id/eprint/16969

Actions (login required)

Tétel nézet