Modeling variability in replicated surveys at aggregation sites

Surveys of aggregation sites can provide unbiased estimates of annual trends in population size if the proportion of the population counted at these sites does not vary systematically among years. However, counts at these sites tend to be highly variable and resulting trend estimates typically have poor precision. I developed an index based on a simple parametric model for counts of Pacific walruses (Odobenus rosmarus divergens) at haul-out sites in Bristol Bay, Alaska that accounted for the general temporal pattern of variability in the proportion of the population at the sites. Simulations suggested that an index based on mean annual counts was a more sensitive indicator of trend than the model-based index or the currently used index based on maximum annual counts. The model-based index may be more useful for other situations where timing of the aggregation peak is more variable.