Estimating Detection Probabilities for Linear Acoustic Monitoring Arrays
Alistair J. Hobday and Doug Pincock
Abstract.—Acoustic monitoring is a widely used low-cost technique for studying the movements of aquatic animals. One common deployment configuration is a linear array of receivers such that passage of a tagged animal across a line of receivers is detected. The typical goal is to estimate the fraction of a population that moves across the line. Receivers in an array can have nonoverlapping or overlapping detection probability envelopes. It has been assumed in the case of nonoverlapping arrays that the detection rate of tagged animals is proportional to the coverage of the line. Unfortunately, the estimation process is not as simple as previously believed. In fact the probability of detection is more Gaussian-shaped rather than uniform with distance from the receiver, and varies over time due to biotic and abiotic noise. Wind-generated noise in particular can have a major influence on the performance of receivers. Range testing, while important, will not solve the problem. In fact, the temporal complexity of this variable detection probability envelope renders a statistical solution improbable. Simulation modeling provides a potential alternative for estimating detection probability for nonoverlapping arrays, however, for most situations, the best solution is to design an overlapping array, with the detection range estimated for the worst environmental conditions.