The effect of line-transect placement in a coastal distance sampling survey

Résumé

Distance sampling surveys are commonly used to estimate animal abundance (N). The choice of survey design has only recently received attention in the line-transect research literature, which has tended to focus more on the violation of assumptions. In this study, simulation methods were used to assess the effect of line placement on the accuracy and precision of a line-transect survey for estimating dolphin abundance. In 1998, a vessel-based line-transect survey of Hector’s dolphin (Cephalorhynchus hectori) was carried out around Banks Peninsula (New Zealand). These data were used to generate a spatially realistic dolphin distribution onto which different survey designs were overlaid. Eight types of design were considered, consisting of four types of stratification and two methods for allocating lines: random or systematic. None of the designs showed any evidence of significant bias in the estimate of N(Nˆ). Systematic placement of lines generally provided more precise estimates of N, with an average reduction (over all designs) of 14% in the coefficient of variation of Nˆ, [CV(Nˆ)]. These results correspond well with those expected from classical sampling theory for the case of estimating a population mean. However, these designs also overestimated CV(Nˆ) by 10-28% (mean=22%). Systematic line-placement has several practical advantages over random placement, including more uniform spatial coverage. For coastal cetacean populations with spatial dynamics similar to the one considered here, we recommend the use of systematic line-placement, with the proviso that the estimate of CV(Nˆ) is likely to be overestimated by 10-28%.