Description
This dataset summarizes species-specific nocturnal vocal activity of four owl species—Mountain Scops-Owl (Otus spilocephalus), Collared Owlet (Taenioptynx brodiei), Collared Scops-Owl (Otus lettia), and Northern Boobook (Ninox japonica)—recorded via passive acoustic monitoring (PAM) from January to December 2023 at Meishan, Yushan National Park, Taiwan. Using an AI-based sound recognition model (SILIC ver. exp32), 705,512 vocalizations were detected and classified from 1,946 hours of nighttime (1600-0800) audio recordings. The dataset provides fine-scale temporal, environmental, lunar and acoustic metadata for each occurrence, supporting future ecological studies, species distribution modeling, and long-term acoustic monitoring of nocturnal forest birds.
Enregistrements de données
Les données de cette ressource données d'échantillonnage ont été publiées sous forme d'une Archive Darwin Core (Darwin Core Archive ou DwC-A), le format standard pour partager des données de biodiversité en tant qu'ensemble d'un ou plusieurs tableurs de données. Le tableur de données du cœur de standard (core) contient 711 353 enregistrements.
2 tableurs de données d'extension existent également. Un enregistrement d'extension fournit des informations supplémentaires sur un enregistrement du cœur de standard (core). Le nombre d'enregistrements dans chaque tableur de données d'extension est illustré ci-dessous.
Cet IPT archive les données et sert donc de dépôt de données. Les données et métadonnées de la ressource sont disponibles pour téléchargement dans la section téléchargements. Le tableau des versions liste les autres versions de chaque ressource rendues disponibles de façon publique et permet de tracer les modifications apportées à la ressource au fil du temps.
Versions
Le tableau ci-dessous n'affiche que les versions publiées de la ressource accessibles publiquement.
Comment citer
Les chercheurs doivent citer cette ressource comme suit:
Wu S, Ko J C, Tsai W (2025). Automatic acoustic detections of four owl species using SILIC in Meishan, Yushan National Park, Taiwan. Version 1.3. Taiwan Biodiversity Research Institute. Samplingevent dataset. https://ipt.taibif.tw/resource?r=owls_ysnp_msc02_silic_exp32&v=1.3
Droits
Les chercheurs doivent respecter la déclaration de droits suivante:
L’éditeur et détenteur des droits de cette ressource est Taiwan Biodiversity Research Institute. Ce travail est sous licence Creative Commons Attribution (CC-BY) 4.0.
Enregistrement GBIF
Cette ressource a été enregistrée sur le portail GBIF, et possède l'UUID GBIF suivante : ac9b5a09-a1b4-416c-9538-ca85078a7210. Taiwan Biodiversity Research Institute publie cette ressource, et est enregistré dans le GBIF comme éditeur de données avec l'approbation du Taiwan Biodiversity Information Facility.
Mots-clé
Samplingevent; Observation
Contacts
- Fournisseur Des Métadonnées ●
- Créateur ●
- Personne De Contact
https://orcid.org/0000-0003-3619-0646
- Créateur
- Fournisseur De Contenu ●
- Créateur
Couverture géographique
Acoustic data were collected using a PAM station located in a sub-montane evergreen broad-leaved forest at Meishan, Yushan National Park, Taiwan (23°16'32"N 120°50'38"E; 1,264 m a.s.l.). This station is part of a long-term soundscape monitoring network established along the Southern Cross-Island Highway, jointly operated by the Taiwan Biodiversity Research Institute and the Yushan National Park Headquarters.
| Enveloppe géographique | Sud Ouest [23,275, 120,843], Nord Est [23,276, 120,844] |
|---|
Couverture taxonomique
We focused on four commonly occurring owl species in Taiwan’s montane forests: Mountain Scops-Owl (MSO, Otus spilocephalus), Collared Owlet (CO, Taenioptynx brodiei), Collared Scops-Owl (CSO, O. lettia), and Northern Boobook (NB, Ninox japonica).
| Species | Otus spilocephalus (Mountain Scops-Owl), Taenioptynx brodiei (Collared Owlet), Otus lettia (Collared Scops-Owl), Ninox japonica (Northern Boobook) |
|---|
Couverture temporelle
| Date de début / Date de fin | 2023-01-01 / 2023-12-31 |
|---|
Données sur le projet
The aim of Sound Identification and Labeling Intelligence for Creatures (SILIC) is to integrate online animal sound databases, PAM databases and an object detection-based model, for extracting information on the sounds of multiple species from complex soundscape recordings. SILIC計畫目的為自動化擷取聲景資料中的動物訊息，以作為生態研究及棲地經營管理之用。
| Titre | SILIC - Sound Identification and Labeling Intelligence for Creatures |
|---|---|
| Identifiant | SILIC |
Les personnes impliquées dans le projet:
Méthodes d'échantillonnage
Acoustic data were collected using an autonomous recording unit (Song Meter Mini; Wildlife Acoustics, Inc.) programmed to record at 44.1 kHz sampling rate and 16-bit resolution. The recorder was installed at a height of approximately 1.5 meters above the forest floor. The device was set to a duty cycle of 1-minute recording followed by a 2-minute pause, yielding 20 minutes of recordings per hour over a 16-hour nightly window. This schedule was maintained continuously for the entire year. Target species included four sympatric owls: Otus spilocephalus, Otus lettia, Taenioptynx brodiei, and Ninox japonica. Automated detection of vocalizations was conducted using the SILIC model (version exp32), which was trained to identify species- and sound type–specific vocalizations from spectrogram features.
| Etendue de l'étude | The study was conducted at the Meishan soundscape monitoring station located in a sub-montane evergreen broadleaved forest at 1,264 m elevation within Yushan National Park, Taiwan (23°16′32″N, 120°50′38″E). Passive acoustic data were collected year-round in 2023, spanning 365 consecutive days from 1 January to 31 December. The recording effort covered the nightly period from 16:00 to 08:00, resulting in a total of 1,946 hours of audio data. This station is part of a long-term ecological monitoring initiative along the Southern Cross-Island Highway jointly operated by the Taiwan Biodiversity Research Institute and the Yushan National Park Headquarters. |
|---|---|
| Contrôle qualité | A confidence threshold of 0.6 was applied to all automated detection outputs to retain only high-confidence vocalizations. To evaluate model precision, 100 randomly selected detections per species were manually validated by an expert, resulting in overall classification accuracy exceeding 94% for all species. Lunar variables—including moon phase, moon altitude, and night-distance (temporal distance from the nearest solar transition)—were calculated using the Python ephem library, based on the geographic coordinates and timestamps of each detection. Meteorological variables (cloud cover, wind speed, air temperature, atmospheric pressure, and precipitation) were obtained from the ERA5 reanalysis dataset and spatially interpolated to the recorder’s location. All detection timestamps were converted to local time (UTC+8), and derived hourly covariates were cross-checked for temporal alignment, missing values, and consistency with modeled nocturnal periods. |
Description des étapes de la méthode:
- Deployment: The autonomous recorder was deployed in Meishan, Yushan National Park, and operated continuously from 1 January to 31 December 2023.
- Recording Schedule: The recorder followed a 1-minute on / 2-minute off cycle from 16:00 to 08:00 daily.
- Data Extraction: Audio recordings were processed using the SILIC model (ver. exp32), a convolutional neural network-based sound recognition tool.
- Species Identification: Species-specific call types for each owl species were predefined in the SILIC model (version exp32).
- Detection Filtering: Detections with a confidence score below 0.6 were discarded.
- Manual Validation: 100 detections per species were manually checked to evaluate model precision.
- Data Aggregation: Detections were aggregated into hourly vocal activity rate (VAR) measures.
- Environmental Covariates: Hourly weather and lunar data were matched to each detection timestamp using astronomical and reanalysis datasets.
- Data Structuring: The final dataset was formatted using the Event Core with Occurrence and MeasurementOrFact extensions, following Darwin Core standards.
Citations bibliographiques
- Wu, S.-H., Chang, H.-W., Lin, R.-S., & Tuanmu, M.-N. (2022). SILIC: A cross database framework for automatically extracting robust biodiversity information from soundscape recordings based on object detection and a tiny training dataset. Ecological Informatics, 68, 101534. https://doi.org/10.1016/j.ecoinf.2021.101534
Métadonnées additionnelles
| Identifiants alternatifs | ac9b5a09-a1b4-416c-9538-ca85078a7210 |
|---|---|
| https://ipt.taibif.tw/resource?r=owls_ysnp_msc02_silic_exp32 |