Estimating Whale Abundance from Sparse Hydrophone Arrays.


  • Prof. Danielle Harris – Centre for Research into Ecological and Environmental Modelling, University of St Andrews
  • FCUL (DEIO) – Campo Grande – Bloco C6 Piso 4 – Sala 6.4.30 – 15:00 horas
  • Quarta-feira, 23 de Maio de 2012
  • Referência Projeto: PEst-OE/MAT/UI0006/2011

Estimating cetacean abundance or density from fixed acoustic sensors is a rapidly expanding area of research.  The research challenges encountered vary widely depending on factors such as the species of interest and the type of monitoring equipment used.  In particular, the desire to survey the largest area possible using the fewest instruments may lead to widely distributed or “sparse” hydrophone arrays.  As discussed, the use of sparse arrays can affect which density estimation methods can be used with the collected data.

In this seminar, two case studies are presented where baleen whale density was investigated using two different types of sparse array.

The first case study estimated fin whale (Balaenoptera physalus) density in the northeastern Atlantic, near the Straits of Gibraltar.  An array of ocean bottom seismometers was used, which enabled ranges to calling animals to be estimated.  Point transect sampling, a form of distance sampling (a popular wildlife density estimation method), could therefore be used.  The measured distances to detected animals were used to calculate the average probability of detecting an animal, which allowed missed animals to be accounted for.

The second case study aimed to estimate blue whale (Balaenoptera musculus) density from vocalisations recorded in the northern Indian Ocean, by International Monitoring System hydrophone stations.  In this case, ranges to animals could not be estimated, so the average probability of detection was estimated in a Monte Carlo simulation framework, which incorporated information about the strength of produced calls, the ability of produced calls to propagate through the water column, oceanic noise levels, and the efficiency of the automatic detector that was used to classify and extract the calls from the dataset.