Tesi etd-06192025-102858 |
Link copiato negli appunti
Tipo di tesi
Tesi di dottorato di ricerca
Autore
AGABITI, CHIARA
URN
etd-06192025-102858
Titolo
Distribution of sea turtles and their anthropogenic threats: innovative approaches to improve conservation actions
Settore scientifico disciplinare
BIOS-03/A - Zoologia
Corso di studi
BIOLOGIA
Relatori
tutor Casale, Paolo
Parole chiave
- behaviour
- density
- drones
- machine learning
- mortality areas
- sea turtles
Data inizio appello
24/06/2025
Consultabilità
Non consultabile
Data di rilascio
24/06/2028
Riassunto
Knowledge of the distribution and behaviour of sea turtles is essential for effective conservation and management efforts, as these species face numerous anthropogenic threats throughout their life cycle. While indirect methods, such as satellite tracking, nesting counts, capture-mark-recapture, and bycatch data provided valuable insights, they often focus predominantly on adult females. Direct observation techniques, including aerial and boat surveys, are similarly limited by high costs and inherent biases. Recent advancements, including underwater cameras, Unmanned Aerial Vehicles (UAVs, or drones), and machine learning, have demonstrated significant potential to address these limitations, enhancing accuracy, efficiency, and the inclusion of other life stages in research efforts. This PhD project aims to develop innovative tools for conservation stakeholders to address critical knowledge gaps in sea turtle ecology, with a focus on their distribution, density, and underwater behaviour in key Mediterranean foraging grounds. Specifically, the objectives are to: (i) develop innovative and cost-effective approaches for turtle research, (ii) improve fine-scale knowledge of distribution and abundance, (iii) investigate submerged behaviour and ecology, and (iv) assess spatio-temporal patterns of fishing interactions.
First, using advanced biologgers, animal-borne cameras, and machine learning (ML) algorithms, the ecological drivers, proxies and patterns of turtle breathing behaviour were studied. Breath frequency was influenced by dive duration but independent of the specific underwater behaviours preceding emersion. Breath rate increased during short surfacing events but plateaued during longer intervals (> 2.5 min). The integration of cameras was essential for accurately detecting multiple breaths within a single surfacing event.
Second, cutting-edge biologgers with animal-borne cameras and ML algorithms were employed to investigate the behavioural time budgets of turtles in the Pelagian Island Archipelago (PIA), a key foraging area on the Tunisian Shelf. The time allocated to key activities, including resting and foraging, was quantified and reliable proxies of pelagic feeding events were identified. The findings underscored the ecological significance of substrate types in likely shaping benthic behaviours.
Third, UAVs were deployed to estimate the fine-scale density and spatial distribution of loggerhead turtles in the PIA. UAV-based surveys enabled the identification of critical hotspots, revealing one of the highest surface densities globally for loggerhead offshore foraging grounds, demonstrating the potential for fine-scale temporal monitoring, and established a robust methodology for achieving more precise density estimates.
Fourth, the application of UAVs was extended to the Gulf of Manfredonia (GoM), recently recognized as an important neritic foraging ground, combining double simultaneous drone surveys with deep learning algorithms to optimise video analysis. This innovative approach enhanced data accuracy and reduced processing time by approximately 90%, enabling the identification of density hotspots (> 1 turtle km⁻²) which exhibited varying densities, likely influenced by differences in the availability of trophic resources.
Fifth, backtracking oceanographic modelling, stranding data, and decomposition monitoring of carcasses were employed to reconstruct probable mortality zones in the Adriatic Sea, a key neritic foraging ground. By integrating drift simulations and fishing effort data, the study identified spatio-temporal overlaps between turtle mortality and trawling activities in the North-West Adriatic and the GoM and elucidated that only a small proportion (17-25%) of turtles that died offshore ultimately stranded onshore.
This research highlights the importance of integrated, cost-effective methods for addressing threats to loggerhead turtles. These versatile tools, optimised for fine-scale monitoring, offer valuable applications for conservation organisations worldwide, enabling the development of evidence-based strategies. Furthermore, the findings underscore the critical need for interdisciplinary approaches to tackle the complex challenges of sea turtle ecology effectively.
First, using advanced biologgers, animal-borne cameras, and machine learning (ML) algorithms, the ecological drivers, proxies and patterns of turtle breathing behaviour were studied. Breath frequency was influenced by dive duration but independent of the specific underwater behaviours preceding emersion. Breath rate increased during short surfacing events but plateaued during longer intervals (> 2.5 min). The integration of cameras was essential for accurately detecting multiple breaths within a single surfacing event.
Second, cutting-edge biologgers with animal-borne cameras and ML algorithms were employed to investigate the behavioural time budgets of turtles in the Pelagian Island Archipelago (PIA), a key foraging area on the Tunisian Shelf. The time allocated to key activities, including resting and foraging, was quantified and reliable proxies of pelagic feeding events were identified. The findings underscored the ecological significance of substrate types in likely shaping benthic behaviours.
Third, UAVs were deployed to estimate the fine-scale density and spatial distribution of loggerhead turtles in the PIA. UAV-based surveys enabled the identification of critical hotspots, revealing one of the highest surface densities globally for loggerhead offshore foraging grounds, demonstrating the potential for fine-scale temporal monitoring, and established a robust methodology for achieving more precise density estimates.
Fourth, the application of UAVs was extended to the Gulf of Manfredonia (GoM), recently recognized as an important neritic foraging ground, combining double simultaneous drone surveys with deep learning algorithms to optimise video analysis. This innovative approach enhanced data accuracy and reduced processing time by approximately 90%, enabling the identification of density hotspots (> 1 turtle km⁻²) which exhibited varying densities, likely influenced by differences in the availability of trophic resources.
Fifth, backtracking oceanographic modelling, stranding data, and decomposition monitoring of carcasses were employed to reconstruct probable mortality zones in the Adriatic Sea, a key neritic foraging ground. By integrating drift simulations and fishing effort data, the study identified spatio-temporal overlaps between turtle mortality and trawling activities in the North-West Adriatic and the GoM and elucidated that only a small proportion (17-25%) of turtles that died offshore ultimately stranded onshore.
This research highlights the importance of integrated, cost-effective methods for addressing threats to loggerhead turtles. These versatile tools, optimised for fine-scale monitoring, offer valuable applications for conservation organisations worldwide, enabling the development of evidence-based strategies. Furthermore, the findings underscore the critical need for interdisciplinary approaches to tackle the complex challenges of sea turtle ecology effectively.
File
Nome file | Dimensione |
---|---|
La tesi non è consultabile. |