Estimation of bowhead whale (Balaena mysticetus) population density using spatially explicit capture-recapture (SECR) methods

• Main Author: Cheoo, Gisela Vitória
• Other Authors: Marques, Tiago André, Thomas, Len
• Format: Master Thesis
• Language: English
• Published: 2019
• Subjects: Baleia-da-Gronelândia; Estimação de densidade; Captura-recaptura espacialmente explícita; Sensores fixos; Acústica passiva; Teses de mestrado - 2019; Domínio/Área Científica::Ciências Naturais::Matemáticas; Balaena mysticetus; bowhead whale; Chukchi
• Online Access: http://hdl.handle.net/10451/39109

Tese de mestrado em Bioestatística, Universidade de Lisboa, Faculdade de Ciências, 2019 Management and conservation of wildlife populations is a major concern. Population density is a key ecological variable when making adequate decisions about them. A variety of methods can be used for estimating density. Capture-recapture (CR, also known as mark- recapture) methods are a popular choice, but ignoring the spatial component of captures has historically led to problems with resulting inferences on abundance. Spatially explicit capture- recapture (SECR) methods use the spatial information to solve two key problems of classical CR: defining a precise study area where captures occur over and reducing un modeled heterogeneity in capture probabilities. Arrays of Directional Autonomous Sea floor Acoustic Recorders (DASARs) recorded calls from the Bearing-Chukchi-Beaufort (BCB) population of bowhead whales during the autumn migration. The available passive acoustic data set was collected over 5 sites (with 3–13 sensors per site) and 8 years (2007–2014), and then processed via both automated and manual procedures. The automated procedure involved computer-processing by a multi-stage detection, classification and localization algorithm. In the manual procedure, calls were detected and classified by trained staff who manually listened to the recordings and examined spectrograms. The resulting manual data presents some pitfalls for density estimation, including non-independence among sensors caused by human intervention. The non-independence leads to an excess of calls being detected in all DASARs on a site. Data from the automated procedure does not suffer the non-independence issue, but the amount of ’singletons’ is approximately 15 times higher than in the manual data. ’Singletons’ are calls detected exclusively in one sensor and we assume they mostly comprise false positives. False positives are sounds classified as coming from the species of interest, but in reality are something else. Considering only automated data ...