Scholarlink

<p>Passive acoustic monitoring offers a non-invasive approach to studying the vocalisation of cryptic species, yet its application to the Asiatic lion (<i>Panthera leo persica</i>) has remained limited. Here, we describe the systematic characterisation of Asiatic lion roars and develop an automated detection model to support acoustic monitoring in the Gir Wildlife Sanctuary, the last stronghold of the free-ranging population of this subspecies. Autonomous recording units (ARUs) were deployed across three survey phases, yielding 1,259&#xa0;h of audio data. Using manually annotated roar clips, we trained two custom detection models via the BirdNET-Analyser framework. Among these, the Leo2 model performed best, achieving a precision of 0.63, a recall of 0.71, and an F1 score of 0.67, and substantially reducing false positives when evaluated on an independent test dataset. Temporal analyses indicated a pronounced peak in roaring activity between 03:00 and 05:00&#xa0;h, along with increased activity after dusk extending into the late-night. An attenuation experiment further demonstrated reliable detection of lion roars up to 2&#xa0;km under natural habitat conditions. To our knowledge, this study establishes the first acoustic reference framework for free-ranging Asiatic lions and demonstrates the feasibility of integrating passive acoustic monitoring with machine learning–based detectors for long-term monitoring, ecological research, and conservation management of this endangered population.</p>

错误：搜索内容不能为空，请输入英文关键词
错误：关键词超出字数限制，请精简
高级检索

Decoding the roar: insights into acoustic detection and attenuation of Asiatic lions’ vocal behaviour

  • 期刊论文

  • 发表日期：

    2026年6月18日

查看全文
  • Mohan Ram,
  • Ram Ratan Nala,
  • Jaipal Singh,
  • Tanisha Dagur,
  • Viral Joshi,
  • Lahar Jhala

摘要

Passive acoustic monitoring offers a non-invasive approach to studying the vocalisation of cryptic species, yet its application to the Asiatic lion (Panthera leo persica) has remained limited. Here, we describe the systematic characterisation of Asiatic lion roars and develop an automated detection model to support acoustic monitoring in the Gir Wildlife Sanctuary, the last stronghold of the free-ranging population of this subspecies. Autonomous recording units (ARUs) were deployed across three survey phases, yielding 1,259 h of audio data. Using manually annotated roar clips, we trained two custom detection models via the BirdNET-Analyser framework. Among these, the Leo2 model performed best, achieving a precision of 0.63, a recall of 0.71, and an F1 score of 0.67, and substantially reducing false positives when evaluated on an independent test dataset. Temporal analyses indicated a pronounced peak in roaring activity between 03:00 and 05:00 h, along with increased activity after dusk extending into the late-night. An attenuation experiment further demonstrated reliable detection of lion roars up to 2 km under natural habitat conditions. To our knowledge, this study establishes the first acoustic reference framework for free-ranging Asiatic lions and demonstrates the feasibility of integrating passive acoustic monitoring with machine learning–based detectors for long-term monitoring, ecological research, and conservation management of this endangered population.