京都大学 / 新制・課程博士 / 博士(情報学) / 甲第24936号 / 情博第847号 / 新制||情||142(附属図書館) / 京都大学大学院情報学研究科社会情報学専攻 / (主査)教授 大手 信人, 准教授 小山 里奈, 教授 三田村 啓理, 准教授 市川 光太郎 / 学位規則第4条第1項該当 / Doctor of Informatics / Kyoto University / DFAM

Nightjars are a group of nocturnal and aerial insectivorous birds that have experienced long-term decline likely driven primarily by habitat loss and declines in prey populations. Eastern Whip-poor-will (Antrostomus vociferus) and Chuck-will’s-widow (Antrostomus carolinensis), two nightjar species native to Illinois, declined 69% and 58% since 1966, respectively. Although previous survey efforts have documented presence of Chuck-will’s-widow and Whip-poor-will, their current distribution in the state is not well known. Using Autonomous Recording Units (ARUs) deployed in a uniform, systematic grid, I surveyed 142 locations from May – July 2022 on public and private lands across the southern eleven counties of Illinois to assess Whip-poor-will and Chuck-will’s-widow distribution and estimate species occupancy. I estimated species relationships with proportion of landcover types, forest patch configuration, and proximity to other landcover types. Additionally, I quantified disturbances from the past 15 years to estimate species relationships to the severity and duration of disturbances. I deployed ARUs for 710 survey days collecting 170,400 minutes or 3,000 hours of recordings. Acoustic bird call identification software, BirdNet, was highly accurate at detecting focal species and greatly reduced the time spent manually reviewing acoustic data. BirdNet identified 43,922 calls of Whip-poor-will and 31,447 calls of Chuck-will’s-widow. I detected Whip-poor-will on 78 surveys with 100% accuracy and Chuck-will’s-widow on 75 surveys with 76% accuracy. Whip-poor-will were positively associated with forest patches with large core areas that neighbored pastures. Additionally, Whip-poor-will were likely to occupy landscapes that had experienced low to moderate disturbance within the previous 15 years. Covariates used to model Chuck-will’s-widow occupancy explained little variation in detection or occupancy and there were no significant relationships with any covariate. However, examining non-significant trends suggest similar relationships as Whip-poor-will in the area. Results highlight the efficiency of passive acoustic monitoring for these birds and the need for further investigation into Chuck-will’s-widow species-environmental relationships. In southern Illinois, Chuck-will’s-widow populations appears to be consistent with previous estimates from the 1990s while Whip-poor-will follow the broader trend of decline.

Passive acoustic monitoring (PAM) is a valuable tool in wildlife research, offering non-invasive and cost-effective means to collect acoustic data. However, the processing of PAM recordings can be time-consuming, prompting the use of acoustic indices to expedite analysis. Acoustic indices are numerical values that characterize biological information in sound recordings based on an environment’s acoustic characteristics. While acoustic indices have been correlated with species richness across ecological contexts, their reliability diminishes in areas with heightened vehicular noise. However, it is unclear if index biases caused by vehicular noise are consistent across all traffic levels, and which acoustic indices are most biologically informative in these human-developed contexts. I assessed the direct impact of vehicular noise on nine acoustic indices through controlled manipulation of vehicular noise within 598 computer-generated bird assemblage soundscapes. Using the bird assemblage soundscapes, I also investigated the effects of three high-pass filter treatments (482 Hz, 1 KHz, and 2 KHz) on these acoustic indices under different levels of traffic noise interference. These filtering effects were also assessed within empirical PAM recordings taken from 147 sites in southern Illinois from May into mid-July of 2022 and 220 sites across the state during late-April to mid-July of 2023. Results indicate that proximity to roads and vehicular traffic significantly affect index values, albeit to varying degrees. Four indices – Bioacoustic Index, Acoustic Complexity Index, Acoustic Diversity Index, and Acoustic Evenness Index – exhibited greater resilience to vehicular noise and may be better suited for urban environments. Notably, the Acoustic Diversity Index, Acoustic Evenness Index, and the Number of Frequency Peaks also displayed consistent species richness estimations regardless of vehicular noise level. While filtering had variable interactive effects with vehicular noise, no consistent benefits of filtering were observed across all indices. Nevertheless, the Acoustic Complexity Index, Acoustic Richness Index, and CityBioNet displayed minimal biases when high-pass filters were applied, and CityBioNet demonstrated particularly high correlations with species richness. These findings underscore the importance of understanding index behavior under anthropogenic noise and different filtering methodologies. My findings serve to inform acoustic index implementation within acoustic monitoring efforts, thus expanding access and reliability of these methodologies within human-developed environments.

This research investigated several aspects of passive acoustic monitoring (PAM) which were previously unexplored or understudied. A comparison of autonomous recording units (ARUs) for use with BirdNET for the purpose of bird monitoring was conducted. Four ARUs were compared, including AudioMoth, SM4, SMMicro, and SwiftOne. We found that, of the performance metrics for which ARU choice made a statistically significant difference (P>0.01), which included sensitivity, specificity, F1 harmonic mean, and Matthews Correlation Coefficient, (but not precision: P = 0.94), AudioMoth had the best performance for all statistically significant performance metrics except for specificity, for which SMMicro had the highest. The same audio was then processed using 18 combinations of Overlap and Sensitivity, including default settings. We found that Overlap and Sensitivity values were highly significant (P>0.001) for all performance metrics: precision, sensitivity, specificity, F1 harmonic mean, and Matthews Correlation Coefficient. No individual Overlap-Sensitivity setting combination performed outperformed others in most of the performance metrics; however, in general, as Overlap or Sensitivity increased, the number of true and false positive species reports increased while the number of false negatives decreased. Four confidence-based threshold types were then used to filter BirdNET output to compare threshold performances, comparing two arbitrary thresholds and two species-specific thresholds which were calculated using manual validation data. Of the thresholds tested, one of the arbitrary threshold types and one of the species-specific threshold types achieved a precision ≥ 0.95. We hope this research will help guide PAM decisions regarding ARU choice, BirdNET settings, and threshold type choice.

付記する学位プログラム名: 社会を駆動するプラットフォーム学卓越大学院プログラム / 京都大学 / 新制・課程博士 / 博士(農学) / 甲第25346号 / 農博第2612号 / 新制||農||1107(附属図書館) / 学位論文||R6||N5518 / DGAM / 京都大学大学院農学研究科応用生物科学専攻 / (主査)教授 三田村 啓理, 教授 益田 玲爾, 准教授 市川 光太郎, 准教授 木村 里子 / 学位規則第4条第1項該当 / Doctor of Agricultural Science / Kyoto University / DGAM

Passive acoustic monitoring has been used to investigate many aspects of marine mammal ecology, although methods to estimate absolute abundance and density using acoustic data have only been developed in recent years. The instrument configuration in an acoustic survey determines which abundance estimation methods can be used. Sparsely distributed arrays of instruments are useful because wide geographic areas can be covered. However, instrument spacing in sparse arrays is such that the same vocalisation will not be detected on multiple instruments, excluding the use of some abundance estimation methods. The aim of this thesis was to explore cetacean abundance and density estimation using novel sparse array datasets, applying existing methods where possible, or developing new approaches. The wealth of data collected by sparse arrays was demonstrated by analysing a 10-year dataset collected by the U.S. Navy's Sound Surveillance System in the north-east Atlantic. Spatial and temporal patterns of blue (Balaenoptera musculus) and fin whale (Balaenoptera physalus) vocal activity were investigated using generalised additive models. Distance sampling-based methods were applied to fin whale calls recorded by an array of Ocean Bottom Seismometers in the north-east Atlantic. Estimated call density was 993 calls/1000 km².hr⁻¹ (CV: 0.39). Animal density could not be estimated because the call rate was unknown. Further development of the call localisation method is required so the current density estimate may be biased. Furthermore, analysing a single day of data resulted in a high variance estimate. Finally, a new simulation-based method developed to estimate density from single hydrophones was applied to blue whale calls recorded in the northern Indian Ocean. Estimated call density was 3 calls/1000 km².hr⁻¹ (CV: 0.17). Again, density of whales could not be estimated as the vocalisation rate was unknown. Lack of biological knowledge poses the greatest limitation to abundance and density estimation using acoustic data.

Despite the extensive Brazilian coast areas, little is known on underwater acoustic environments in Brazil. Acoustic environments (or soundscape) are composed by biological, geological and man-made sound sources. Soundscapes are strongly linked to ecosystems dynamics, and follow temporal patters that can vary at daily and seasonal scales. Thus, for soundscape characterization, it is necessary to undertake sound recordings for long periods, which demands innovative analyzing methods. Accordingly, the present research focuses in two principal objectives: (1) to develop methods for analyzing long-term acoustic recordings and, (2) to characterize marine soundscapes of selected points in São Paulo State. Four deployment sites were selected for the underwater acoustic monitoring: a point located at the channel entrance of the Santos Harbor, and three marine Protected Areas (PAs) in Sao Paulo state. As a result, the largest underwater acoustic database from Brazilian seas was acquired. The present work used Power Spectral Density (PSD), Sound Pressure Level (SPL) and Spectrograms to develop an innovative methodology for analyzing long-term acoustic data. In addition, a new visualization tool and a method for automatic detection of dawn and dusk choruses are presented. The achieved results validated the proposed methodology as an effective tool for analyzing long-term acoustic data. The area close to the first site, the vicinity of Santos Harbor, was dominated by ship noise, which values reach levels that can affect some species of fish and marine mammals. The soundscapes of the other three remaining measurement sites were dominated by fish and crustacean choruses, with daily and seasonal patterns (related to sunrise and sunset). For the monitored regions, the present work signifies the first contribution for cataloguing fish choruses, and establishes a baseline for the study of their underwater acoustic environment. Although the proposed methodology has used long-term undersea acoustic datasets as case-study, it can also be extended for monitoring other aquatic or terrestrial ecosystems. Finally, the research indicates to Brazilian environmental agencies and to the related scientific community that passive acoustic monitoring is a noninvasive and cost-effective tool that can be used for the management of PAs and points of economic relevance. / Apesar da ampla área dos mares brasileiros, pouco se conhece sobre paisagens acústicas submarinas no Brasil. Estas paisagens são compostas por sons de origens biológicas, geológicas e as produzidas pelo homem. As paisagens acústicas estão fortemente ligadas à dinâmica dos ecossistemas, mostrando padrões temporais diários e sazonais. Para caracterizar paisagens acústicas é necessário realizar gravações de sons por períodos de tempos prolongados, o que demanda métodos de análise inovadores. Neste sentido, a presente pesquisa visa dois objetivos principais: (1) desenvolver métodos para a análise de gravações acústicas de longa duração, (2) caracterizar a paisagem acústica do litoral do estado de São Paulo. Quatro pontos de coleta foram selecionados para monitoramento acústico passivo: um ponto situado no canal de entrada do Porto de Santos e os outros três em áreas de proteção marinhas (APM) do estado de São Paulo. Como resultado foi obtida a base de dados de sons submarinhos mais extensa dos mares brasileiros. Do ponto de vista da análise destes dados, o presente trabalho baseia-se no cálculo da Densidade Espectral de Potência, Níveis de Pressão Sonora e Espectrogramas, obtendo métodos de análise novedosos a partir técnicas tradicionais. Neste contexto a tese apresenta uma ferramenta para a visualização de dados acústicos e um método para a detecção automática de coros biológicos matutinos e vespertinos. Os resultados obtidos permitiram validar a efetividade dos métodos propostos na descrição e análise de dados acústicos de longa duração. O ambiente acústico nas proximidades do Porto de Santos foi dominado por ruído de embarcações, alcançando valores de níveis sonoros capazes de afetar algumas espécies de peixes e mamíferos marinhos. As paisagens acústicas dos três pontos restantes foram dominadas por coros de peixes e crustáceos, com padrões diários e sazonais (relacionados ao nascer e pôr do sol). O presente trabalho constitui a primeira pesquisa que cataloga coro de peixes e que estabelece uma referência para o estudo do ambiente acústico das regiões monitoradas. Embora os métodos apresentados usaram como estudo de caso dados de sons submarinos, a sua aplicação pode ser estendida para o monitoramento de outros ambientes aquáticos ou terrestres. Por último, a pesquisa mostra aos órgãos ambientais brasileiros que o monitoramento acústico passivo é uma ferramenta eficaz para o manejo e monitoramento de áreas protegidas e pontos de relevância econômica.

Despite the extensive Brazilian coast areas, little is known on underwater acoustic environments in Brazil. Acoustic environments (or soundscape) are composed by biological, geological and man-made sound sources. Soundscapes are strongly linked to ecosystems dynamics, and follow temporal patters that can vary at daily and seasonal scales. Thus, for soundscape characterization, it is necessary to undertake sound recordings for long periods, which demands innovative analyzing methods. Accordingly, the present research focuses in two principal objectives: (1) to develop methods for analyzing long-term acoustic recordings and, (2) to characterize marine soundscapes of selected points in São Paulo State. Four deployment sites were selected for the underwater acoustic monitoring: a point located at the channel entrance of the Santos Harbor, and three marine Protected Areas (PAs) in Sao Paulo state. As a result, the largest underwater acoustic database from Brazilian seas was acquired. The present work used Power Spectral Density (PSD), Sound Pressure Level (SPL) and Spectrograms to develop an innovative methodology for analyzing long-term acoustic data. In addition, a new visualization tool and a method for automatic detection of dawn and dusk choruses are presented. The achieved results validated the proposed methodology as an effective tool for analyzing long-term acoustic data. The area close to the first site, the vicinity of Santos Harbor, was dominated by ship noise, which values reach levels that can affect some species of fish and marine mammals. The soundscapes of the other three remaining measurement sites were dominated by fish and crustacean choruses, with daily and seasonal patterns (related to sunrise and sunset). For the monitored regions, the present work signifies the first contribution for cataloguing fish choruses, and establishes a baseline for the study of their underwater acoustic environment. Although the proposed methodology has used long-term undersea acoustic datasets as case-study, it can also be extended for monitoring other aquatic or terrestrial ecosystems. Finally, the research indicates to Brazilian environmental agencies and to the related scientific community that passive acoustic monitoring is a noninvasive and cost-effective tool that can be used for the management of PAs and points of economic relevance. / Apesar da ampla área dos mares brasileiros, pouco se conhece sobre paisagens acústicas submarinas no Brasil. Estas paisagens são compostas por sons de origens biológicas, geológicas e as produzidas pelo homem. As paisagens acústicas estão fortemente ligadas à dinâmica dos ecossistemas, mostrando padrões temporais diários e sazonais. Para caracterizar paisagens acústicas é necessário realizar gravações de sons por períodos de tempos prolongados, o que demanda métodos de análise inovadores. Neste sentido, a presente pesquisa visa dois objetivos principais: (1) desenvolver métodos para a análise de gravações acústicas de longa duração, (2) caracterizar a paisagem acústica do litoral do estado de São Paulo. Quatro pontos de coleta foram selecionados para monitoramento acústico passivo: um ponto situado no canal de entrada do Porto de Santos e os outros três em áreas de proteção marinhas (APM) do estado de São Paulo. Como resultado foi obtida a base de dados de sons submarinhos mais extensa dos mares brasileiros. Do ponto de vista da análise destes dados, o presente trabalho baseia-se no cálculo da Densidade Espectral de Potência, Níveis de Pressão Sonora e Espectrogramas, obtendo métodos de análise novedosos a partir técnicas tradicionais. Neste contexto a tese apresenta uma ferramenta para a visualização de dados acústicos e um método para a detecção automática de coros biológicos matutinos e vespertinos. Os resultados obtidos permitiram validar a efetividade dos métodos propostos na descrição e análise de dados acústicos de longa duração. O ambiente acústico nas proximidades do Porto de Santos foi dominado por ruído de embarcações, alcançando valores de níveis sonoros capazes de afetar algumas espécies de peixes e mamíferos marinhos. As paisagens acústicas dos três pontos restantes foram dominadas por coros de peixes e crustáceos, com padrões diários e sazonais (relacionados ao nascer e pôr do sol). O presente trabalho constitui a primeira pesquisa que cataloga coro de peixes e que estabelece uma referência para o estudo do ambiente acústico das regiões monitoradas. Embora os métodos apresentados usaram como estudo de caso dados de sons submarinos, a sua aplicação pode ser estendida para o monitoramento de outros ambientes aquáticos ou terrestres. Por último, a pesquisa mostra aos órgãos ambientais brasileiros que o monitoramento acústico passivo é uma ferramenta eficaz para o manejo e monitoramento de áreas protegidas e pontos de relevância econômica.

De nombreux prédateurs marins se nourrissent directement des prises des pêcheurs. Ces interactions, définies comme de la déprédation, engendrent des conséquences socio-économiques considérables pour les pêcheurs ainsi que des implications de conservation pour la faune sauvage. D’un côté, la déprédation endommage le matériel et augmente l’effort de pêche pour atteindre les quotas. D’un autre côté, la déprédation augmente le risque de mortalité des prédateurs marins (prise accidentelle ou rétorsion létale par les pécheurs). La pêcherie à la palangre est la plus impactée par la déprédation, principalement par les odontocètes, ce qui incite à trouver des solutions. La majorité des études se concentrant sur la déprédation s’est principalement basée sur des observations en surface, de ce fait la manière dont les prédateurs retirent les poissons sur les lignes reste confuse. Par ailleurs, l’impact de la déprédation sur le comportement des pêcheurs ainsi que les facteurs expliquant leur détectabilité n’ont reçu que peu d’intérêt. L’objectif de cette thèse est donc d’étudier ces problématiques par un suivi acoustique, une utilisation de balises et une approche en écologie comportementale humaine, en se concentrant sur la pêcherie palangrière française ciblant la légine australe (Dissostichus eleginoides) impactée par la déprédation des orques (Orcinus orca) et des cachalots (Physeter macrocephalus). Les capitaines ont été décrits comme recherchant leur ressource selon la théorie de « l’optimal foraging », mais avec des perceptions de la compétition et du succès de pêche qui divergent. Certains capitaines seraient ainsi plus enclins à remonter les palangres au plus proche et à rester sur une zone, même en présence de compétition, augmentant alors le risque d’interaction. L’acoustique des navires a révélé que certaines manoeuvres (marche arrière par exemple) propagent différemment sous l’eau. La manière dont les capitaines manoeuvrent leur palangrier influencerait ainsi leur détectabilité et donc leur risque d’interaction avec les prédateurs. D’autre part, l’utilisation de capteurs sur les palangres et les animaux a révélé que les orques et les cachalots sont capables de déprédater sur les palangres posées sur le fond marin. Ces observations laissent à penser que les odontocètes sont en mesure de localiser l’activité de pêche bien avant la remontée de la ligne, ce qui pourrait être expliqué par une signature acoustique spécifique du déploiement de la ligne. L’ensemble des résultats de cette thèse suggère que la déprédation sur les palangres démersales est très probablement sous-estimée. Cette thèse apporte également des éléments importants pour la lutte contre la déprédation, en montrant la nécessité de protéger les palangres dans l’intégralité du processus de pêche. / Many marine predator species feed on fish caught by fishers directly from the fishing gear. Known as depredation this interaction issue has substantial socio-economic consequences for fishermen and conservation implications for the wildlife. Costs for fishers include damages to the fishing gear and increased fishing effort to complete quotas. For marine predators, depredation increases risks of mortality (lethal retaliation from fishers or bycatch on the gear). Longline fisheries are the most impacted worldwide, primarily by odontocetes (toothed whales) depredation, urging the need for mitigation solutions to be developed. Most of studies assessing depredation have primarily relied on surface observation data, thus the way odontocetes interact with longlines underwater remains unclear. Besides, the way fishermen respond to depredation during fishing operations, or can influence their detectability to odontocetes, have been poorly investigated. This thesis therefore aimed at investigating these aspects through a passive acoustic monitoring, bio-logging and human ecology approaches, focusing on the French Patagonian toothfish (Dissostichus eleginoides) longline fisheries impacted by killer whales (Orcinus orca) and sperm whales (Physeter macrocephalus). Firstly, this thesis reveals that captains behave as optimal foragers but with different personal perception of competition and fishing fulfilment. Some captains would thus be more likely to stay within a patch or to haul closest longline even in presence of competition, suggesting these captains would show higher interaction rates. Additionally, the propagation of vessels’ acoustics varied depending on the type of manoeuvre (e.g. going backward vs. forward). The way captains use their vessels to navigate may therefore influence their detectability and so their depredation level. Secondly, loggers deployed on both the longlines (accelerometers) and odontocetes (GPS-TDR) revealed that killer whales and sperm whales are able to depredate on longlines while soaking on the seafloor. These observations suggest, therefore, that odontocetes can localise fishing activity before the hauling, which could be partially explained by specific acoustic signatures recorded during the setting process. Altogether, the results of the thesis suggest that depredation rates on demersal longlines are most likely underestimated. The thesis also brings some important insights for mitigation measures, suggesting that countermeasures should start from setting to hauling.

The Littoral Acoustic Demonstration Center has collected passive acoustic monitoring data in the northern Gulf of Mexico since 2001. Recordings were made in 2007 near the Deepwater Horizon oil spill that provide a baseline for an extensive study of regional marine mammal populations in response to the disaster. Animal density estimates can be derived from detections of echolocation signals in the acoustic data. Beaked whales are of particular interest as they remain one of the least understood groups of marine mammals, and relatively few abundance estimates exist. Efficient methods for classifying detected echolocation transients are essential for mining long-term passive acoustic data. In this study, three data clustering routines using k-means, self-organizing maps, and spectral clustering were tested with various features of detected echolocation transients. Several methods effectively isolated the echolocation signals of regional beaked whales at the species level. Feedforward neural network classifiers were also evaluated, and performed with high accuracy under various noise conditions. The waveform fractal dimension was tested as a feature for marine biosonar classification and improved the accuracy of the classifiers. [This research was made possible by a grant from The Gulf of Mexico Research Initiative. Data are publicly available through the Gulf of Mexico Research Initiative Information & Data Cooperative (GRIIDC) at https://data.gulfresearchinitiative.org.] [DOIs: 10.7266/N7W094CG, 10.7266/N7QF8R9K]