Development and Validation of an Automated Sperm Whale Click Detection and Classification Pipeline Using Passive Acoustic Monitoring in the Western Ionian Sea (Mediterranean Sea)
(2026) BION03 20252Degree Projects in Biology
- Abstract
- The Mediterranean subpopulation of sperm whales (Physeter macrocephalus) is classified as Endangered by the IUCN, yet the western Ionian Sea, a key corridor connecting the eastern and western Mediterranean sub-populations, remains a data-poor region for cetacean monitoring. Passive acoustic monitoring (PAM) offers an effective approach for long-term surveillance of this deep-diving species, but large-scale acoustic datasets require validated automated pipelines to distinguish sperm whale clicks from confounding signals such as dolphin echolocation sounds, and anthropogenic noise. This thesis developed and validated a multi-stage pipeline for the detection, classification, and acoustic sizing of sperm whales, using continuous recordings... (More)
- The Mediterranean subpopulation of sperm whales (Physeter macrocephalus) is classified as Endangered by the IUCN, yet the western Ionian Sea, a key corridor connecting the eastern and western Mediterranean sub-populations, remains a data-poor region for cetacean monitoring. Passive acoustic monitoring (PAM) offers an effective approach for long-term surveillance of this deep-diving species, but large-scale acoustic datasets require validated automated pipelines to distinguish sperm whale clicks from confounding signals such as dolphin echolocation sounds, and anthropogenic noise. This thesis developed and validated a multi-stage pipeline for the detection, classification, and acoustic sizing of sperm whales, using continuous recordings from a single bottom-mounted hydrophone deployed at 880 m depth off Siracusa, Sicily (July–October 2024; 90 days). Automated click detections were refined through a tiered post-processing filter integrating signal-to-noise ratio, inter-click interval regularity, spectral slope, and centroid frequency. Validated against manually verified files, the filter achieved a precision of 76.7%. Applied to the full three-month dataset, 478 of 4,486 detection-positive files (10.7%) were accepted. Acoustic presence was documented on 69.1% of monitored days, with a pronounced diel pattern (73.9% of clustered clicks during daytime) suggestive of prey-driven foraging activity. Unsupervised clustering further resolved two distinct acoustic click types (97.4% assignment rate), interpreted as near-range and long-range echolocation categories. Inter-pulse interval (IPI) analysis yielded 39 valid body-length estimates ranging from 8.69 to 13.99 m (mean 10.74 m), consistent with a mixed population of adult females, juveniles, and at least one adult male. The validated pipeline provides a transparent, reproducible framework applicable to the broader dataset and contributes baseline data for conservation management and environmental impact assessment under the EU Marine Strategy Framework Directive. (Less)
- Popular Abstract
- Every Click Counts
Mediterranean sperm whales are among Europe’s most elusive and endangered marine mammals. Fewer than 2,500 adults remain, scattered across one of the busiest seas in the world. They spend most of their lives in the dark, hunting squid in dives that can last nearly an hour, and surface only briefly between bouts of feeding. To protect a species we almost never see, we must listen.
Sperm whales (Physeter macrocephalus) produce some of the loudest sounds in the animal kingdom. Their echolocation clicks, used to find prey in the lightless depth, can travel for up to 60 kilometers through seawater and be picked up by underwater. This makes passive acoustic monitoring an ideal tool for studying them. But there is a... (More) - Every Click Counts
Mediterranean sperm whales are among Europe’s most elusive and endangered marine mammals. Fewer than 2,500 adults remain, scattered across one of the busiest seas in the world. They spend most of their lives in the dark, hunting squid in dives that can last nearly an hour, and surface only briefly between bouts of feeding. To protect a species we almost never see, we must listen.
Sperm whales (Physeter macrocephalus) produce some of the loudest sounds in the animal kingdom. Their echolocation clicks, used to find prey in the lightless depth, can travel for up to 60 kilometers through seawater and be picked up by underwater. This makes passive acoustic monitoring an ideal tool for studying them. But there is a catch: a single recorder left on the seafloor for months produces enormous amounts of audio, and within that audio sperm whale clicks share their sound space with dolphin echolocation, and the rumble of passing ships. Turning raw recordings into reliable ecological information means that every click must be detected, sorted, and ideally traced back to an individual whale.
For this thesis, we developed and tested a pipeline using three months of recordings collected off the eastern coast of Sicily, in the western Ionian Sea which is an important corridor for the Mediterranean population. A recorder anchored at 880 m depth, accumulating thousands of files between July and October 2024. The pipeline flags possible clicks by applying a layered filter that uses signal strength, click rhythm, and acoustic fingerprints to tell genuine sperm whale clicks apart from impostors. A final step takes advantage of a quirk of whale anatomy: every click bounces around inside the whale’s massive nose before escaping into the water, and the tiny delay between these internal echoes scales with body length. Measure the delay, and you can measure the whale without ever seeing it.
Unsupervised sorting of the cleaned-up clicks revealed two distinct click voices, interpreted as long-range scanning of the surroundings and close-range targeting of prey. Body-length estimates ranged from 8.7 to 14 meters, capturing a mixed-sex, mixed-age population of adult females, juveniles, and at least one large adult male. These acoustic results were independently confirmed by visual sightings and bio-logger tags deployed during the same field campaign, including a male photographed from a drone at the upper end of the size range.
These findings provide a baseline against which future changes from increasing shipping traffic to potential offshore industrial development can be measured. The European Union’s Marine Strategy Framework Directive requires member states to monitor underwater noise and the species exposed to it and validated acoustic pipelines like the one developed here give regulators the evidence they need.
Master’s Degree Project in Conservation Biology, 60 credits, 2026
Department of Biology, Lund University
Advisor1: Francesco Caruso
Stazione Zoologica Anton Dohrn (SZN), Naples, Italy
Advisor2: Per Carlsson
Advisor3: Johanna Stedt
Lund University, Lund, Sweden (Less)
Please use this url to cite or link to this publication:
https://lup.lub.lu.se/student-papers/record/9226824
- author
- Jang, Haeun
- supervisor
- organization
- course
- BION03 20252
- year
- 2026
- type
- H2 - Master's Degree (Two Years)
- subject
- language
- English
- id
- 9226824
- date added to LUP
- 2026-05-22 13:36:33
- date last changed
- 2026-06-12 15:28:55
@misc{9226824,
abstract = {{The Mediterranean subpopulation of sperm whales (Physeter macrocephalus) is classified as Endangered by the IUCN, yet the western Ionian Sea, a key corridor connecting the eastern and western Mediterranean sub-populations, remains a data-poor region for cetacean monitoring. Passive acoustic monitoring (PAM) offers an effective approach for long-term surveillance of this deep-diving species, but large-scale acoustic datasets require validated automated pipelines to distinguish sperm whale clicks from confounding signals such as dolphin echolocation sounds, and anthropogenic noise. This thesis developed and validated a multi-stage pipeline for the detection, classification, and acoustic sizing of sperm whales, using continuous recordings from a single bottom-mounted hydrophone deployed at 880 m depth off Siracusa, Sicily (July–October 2024; 90 days). Automated click detections were refined through a tiered post-processing filter integrating signal-to-noise ratio, inter-click interval regularity, spectral slope, and centroid frequency. Validated against manually verified files, the filter achieved a precision of 76.7%. Applied to the full three-month dataset, 478 of 4,486 detection-positive files (10.7%) were accepted. Acoustic presence was documented on 69.1% of monitored days, with a pronounced diel pattern (73.9% of clustered clicks during daytime) suggestive of prey-driven foraging activity. Unsupervised clustering further resolved two distinct acoustic click types (97.4% assignment rate), interpreted as near-range and long-range echolocation categories. Inter-pulse interval (IPI) analysis yielded 39 valid body-length estimates ranging from 8.69 to 13.99 m (mean 10.74 m), consistent with a mixed population of adult females, juveniles, and at least one adult male. The validated pipeline provides a transparent, reproducible framework applicable to the broader dataset and contributes baseline data for conservation management and environmental impact assessment under the EU Marine Strategy Framework Directive.}},
author = {{Jang, Haeun}},
language = {{eng}},
note = {{Student Paper}},
title = {{Development and Validation of an Automated Sperm Whale Click Detection and Classification Pipeline Using Passive Acoustic Monitoring in the Western Ionian Sea (Mediterranean Sea)}},
year = {{2026}},
}