Estimating the detection probability of long-ranging baleen whale song using a single sensor: Towards density estimation

Passive acoustic data can be used to estimate animal density. A key step is quantifying the range-specific detection probability for vocalizations from the target species. A method developed to estimate cetacean density from single hydrophones was applied to pygmy blue whale (Balaenoptera musculus brevicauda) "Sri Lankan" song recorded near Diego Garcia in the Indian Ocean during May 2002. Detection probability was estimated using a Monte Carlo simulation using information about transmission loss, ambient noise levels, song source levels, and the efficiency of the automatic detection process. The effect of varying source levels was explored. Song density estimates were 0.14 song units/1000 km2 h-1 [coefficient of variation (CV), 0.16; mean source level: 179 dB re 1 μPa @ 1 m] and 0.024 song units/1000 km2 h-1 (CV, 0.12; mean source level, 189 dB re 1 μPa @ 1 m). Estimating whale density additionally requires an estimate of the song production rate, which was not available. Nevertheless, estimating song unit density enables different datasets to be compared in a standardized framework. This simulation method is useful for data collected by sparsely distributed instruments, where wide instrument spacing may exclude the use of standard density estimation methods such as spatial capture-recapture and distance sampling.