Designing line transect surveys for complex survey regions

Abstract

Line transect surveys are widely used to estimate the density and/or size of cetacean populations. Good survey design is essential for obtaining reliable results using standard (design-based) analysis methods. Even for more complex (model-based) analysis methods, a good survey design is valuable. A ‘good’ design is one (a) that employs randomisation in laying out transects; (b) that is stratified if density is known to vary on a large scale; (c) where each location within a stratum has an equal probability of being surveyed (uniform coverage probability); (d) that produces an even distribution of transects throughout each stratum (e.g. systematic random designs); (e) that produces at least 10-20 transects per stratum; (f) that, given the previous points, gives maximum efficiency per unit effort – for example by minimising time spent travelling between survey lines (off-effort time). We discuss strategies for creating good designs given the constraints inherent in many shipboard surveys of cetaceans: severely limited ship time and complex topography. We advocate the use of computer software, such as the program Distance, to create designs and compare their properties using simulation. We provide a link between the concepts and their implementation through a concrete example of survey design: a multi-species survey of cetaceans in coastal British Columbia. The design uses an equally spaced zig-zag configuration of transects in more open strata combined with sub-stratification to minimise off-effort time. In the highly convex inshore stratum we develop a systematic cluster sampling algorithm, and within the selected clusters use a systematic parallel line layout to ensure equal coverage probability in the long, narrow fjords. To aid those wishing to learn automated design methods, we provide Distance project files online.