Observer coverage required for the prediction of incidental capture of seabirds in New Zealand commercial fisheries

This report obtains estimates of observer coverage required to predict the total numbers of incidental seabird captures (albatross and petrel taxa combined) in a fishery with a specified coefficient of variation. Fisheries analysed are determined by the quality and the availability of data and a known history of seabird captures: the trawl fisheries targeting hoki in the Sub-Antarctic, on the Chatham Rise, and off the west coast of the South Island; the trawl fisheries targeting squid in the Auckland Islands part of SQU 6T and on the Stewart-Snares shelf; and the autoliner fisheries targeting ling on the Chatham Rise and in waters south of 46° S.
A random effects model was developed to enable the contribution from the correlation between the numbers of captures per tow/set by the same vessel to be included in the calculation of the coefficient of variation of the predicted total number of seabird captures. The model also permits the estimation of the contribution from the extra variation that arises from the tendency for seabird capture incidents to involve large numbers of birds on rare occasions. Estimates of the variance components were derived from models that were fitted using data from the 1999–00 to the 2002–03 fishing years for the trawl fisheries and from the 2000–01 to the 2002–03 fishing years for the longline fisheries.
The observer coverage fraction is expressed in terms of tows for the trawl fisheries and hooks hauled for the autoliner fisheries. The estimated observer coverage fraction required to attain a coefficient of variation of 30% in the fishery depends on the total effort and on the capture rate through the predicted total number of captures. The approximate coverage fractions required, based on typical estimated numbers of captures, were between 8 – 54% depending on the fishery and numbers of birds captured for trawl fisheries.For the ling autoliner fisheries, the required fraction of observer coverage does not depend to the same extent on the predicted total number of seabird captures because most uncertainty arises from between vessel variation. Through wide ranges of predicted captures the approximate effort coverage fractions required were between 70 - 91% of days depending on the fishery area.
Estimates of observer coverage required for both the trawl and ling autoliner fisheries can only be regarded as approximate since they are always based on variance estimates, which are notoriously uncertain. The ling autoliner fishery estimates however have an additional source of uncertainty that arises because there are few vessels involved in these fisheries and the distribution of the effort between the vessels can vary markedly between seasons. Thus, for the ling autoliner fisheries, the large values for the coverage required by days in the fishery combined with the large uncertainties in these estimates suggests that 100% coverage is desirable.