Trends in dolphin abundance estimated from fisheries data: A cautionary note

The previously published index of relative abundance of the northeastern offshore stock of spotted dolphins, the species most affected by the purse-seine fishery for tunas in the eastern Pacific Ocean, shows a decreasing trend in the last two decades despite dramatic reductions in incidental mortality since the early 1970s. To better understand the behaviour of this index, the effects of changes in data quality and methods of searching on estimation of relative abundance using current methodologies have been studied here. Changes in data quality since the late 1980s have led to a dramatic reduction in the proportion of sightings that are reported on or near the trackline. The decreasing trend in the index in the late 1970s and through the 1980s is strongly influenced by the fit of the detection function to the high proportion of sightings near the trackline that was present in the data during that time period. If this excess of sightings near the trackline is spurious, then much of the decreasing trend in the index over this time period is likely spurious. In addition, part of the decrease in the index in the late 1980s to mid-1990s is probably due to changes in data-collection biases that result from a dramatic increase in the amount of searching that is currently being carried out using helicopters as compared to high-powered binoculars. The results suggest that trends in bias associated with changes in data quality and fishery operations may have contributed to a trend in the index on the order of 1.0-1.5% per year, or approximately 25-33% of the maximum growth rate of the northeastern stock of offshore spotted dolphin. The pervasive nature of these sources of bias, and their potential magnitude relative to the maximum growth rates of the dolphin species involved, make use of this index in population growth models ill-advised. Fishery-derived indices such as these may be most useful for comparing trends in relative abundance between species, when the sources of biases are unlikely to be species-specific.