Abstract
This thesis reports the results of the first boat-based line-transect survey to be carried out in New Zealand waters. The survey was carried out in January and February 1998, successfully adapting conventional line-transect methods for use on a privately owned 15m catamaran, R. V. Catalyst. The survey proved that line-transect methods could work well on much smaller cheaper vessels than those generally used on surveys overseas. Analysis of dolphin orientation data suggested, however, that abundance estimates were positively biased as a result of dolphins swimming towards the survey vessel. Further, a comparison of the Akaroa Harbour abundance estimate (124, 95% c.i. 85-181) with previous survey work indicated that estimates may have been biased high. The results from just 3 of 115 small boat surveys of Ak:aroa Harbour fall within the uncorrected 95% confidence limits, and the average summer abundance seen on these surveys was 43 animals (95% c.i. 34-52).
During the following field season (November 1998 to February 1999), combined boat and helicopter surveys were carried out in the waters off Banks Peninsula. A correction factor was derived that accounted both for vessel attraction and for dolphins missed on the trackline. The corrected Hector's dolphin abundance estimate from Motunau- Timaru (0-4 nautical miles offshore) is 1198 (95% c.i. 709-2025). Survey coverage was extended during this season from Timaru to Long Point, west of Te Waewae Bay in Foveaux Strait. The corrected abundance estimate for this area (0 - 4 n.mi. offshore) is 399 (232-686). A total inshore abundance estimate of Hector's dolphins from Motunau to Long Point is 1596 (1002-2546).
A further objective was to determine the effect of survey design on abundance estimates, as a step towards developing guidelines for the design of inshore linetransect surveys. Data collected during the 1998 survey were used to develop a hypothetical dolphin population, which was sampled via computer simulations to determine the relative efficiencies of different survey designs. Results indicated that systematic surveys will generally be more precise than those with random line selection, and that the effect of stratification on abundance estimates and precision is negligible. These results provide some guidance for the design of future inshore surveys. Further, the simulation methods developed here could be modified to investigate the viability of adaptive line-transect sampling.