Environmental Models of Cetacean Abundance: Reducing Uncertainty in Population Trends

Trends in population abundance are often used to monitor species affected by human activities. For highly mobile species in dynamic environments, however, such as cetaceans in the marine realm, natu- ral variability can confound attempts to detect and interpret trends in abundance. Environmental variabil- ity can cause dramatic shifts in the distribution of cetaceans, and thus abundance estimates for a fixed re- gion may be based on a different proportion of the population each time. This adds variability, decreasing statistical power to detect trends and introducing uncertainty whether apparent trends represent true changes in population size or merely reflect natural changes in the distribution of cetaceans. To minimize these problems, surveys ideally would be based on species-specific design criteria that optimize sampling within all relevant habitat throughout a species' range. Our knowledge of cetacean habitats is limited, how- ever, and financial and logistic constraints generally force those surveying cetacean abundance to include all species within a limited geographic region. Alternately, it may be possible to account for environmental vari- ability analytically by including models of species-environment patterns in trend analyses, but this will be successful only if such models have interannual predictive power. I developed and evaluated generalized ad- ditive models of cetacean sighting rates in relation to environmental variables. I used data from shipboard surveys of Dall's porpoise ( Phocoenoides dalli ) and short-beaked common dolphins ( Delphinus delphis ) con- ducted in 1991, 1993, and 1996 off California. Sighting rates for these two species are variable and can be partially accounted for by environmental models, but additional surveys are needed to model species-envi- ronment relationships adequately. If patterns are consistent across years, generalized additive models may represent an effective tool for reducing uncertainty caused by environmental variability and for improving our ability to detect and interpret trends in abundance.