STANDARDISED SOUTH ATLANTIC ALBACORE CPUE FOR THE SOUTH AFRICAN BAITBOAT FISHERY, 1985-2002

SUMMARY The historic southern albacore CPUE series for the South African baitboat fleet is updated general linear models and assuming lognormal error structure, i.e. the same procedures as in previous analyses. In addition, three different delta models are fitted. The proportion of positive catches is modelled assuming a binomial error structure in each case, whereas different error structures are tested for each fit to the positive catch records. Examination of standard diagnostic plots suggests that the best fit is obtained when a negative binomial error structure is assumed. The proportion of reported unsuccessful trips (records with zero albacore catch) declined sharply from a mean of 22% for the period 1985 to 1998 to only 2% from 1999 onwards. This sharp drop is interpreted in the models as indicating increased albacore abundance in nearshore waters off southern Africa. However, it is more probably the result of a previously undetected marked change in reporting and/or in fishing practice since 1999. Therefore, CPUE trends for the South African baitboat fleet from 1999 onwards should be treated with caution. In addition, this observation has highlighted a potentially serious flaw in the method followed to prepare the catch and effort records prior to analyses. RESUME