Document: DOU-3-91-8

Challenges in survey design and data analysis of a long-term study: Lessons from a 17-year study of Bowhead Whales in the Alaskan Beaufort Sea.

DEUTSCHMAN, D.H.* and J.DIFFENDORFER

San Diego State University, San Diego, CA 92182 USA ¹

Abstract:
One of the greatest limitations in making informed management decisions is the absence of long-term data on population dynamics. The Bowhead Whale Aerial Survey Project (BWASP) is an impressive survey in which nearly 2000 endangered bowhead whales were sighted along 3000 aerial transects over 17 years. The BWASP study was designed as a stratified random sample with strata corresponding to sections of the Alaskan Beaufort Sea. Sampling effort, defined as the number of transects flown, varied dramatically. More than 70% of transect were flown in Onshore strata and this proportion varied from year to year and month to month. The number of whales per transect was strongly skewed with most transects resulting in no sightings but with occasional large aggregations. In fact, 2% of the transects accounted for more than 28% of the whales seen in the entire study. In order to reduce the impact of this skewness, sighting data was transformed using a simple presence/absence criterion. Analysis of this transformed sighting variable suggests that whale populations are highly variable. Sighting success ranged from a low of 6% in 1985 to a high of 31% in 1990, 1995, and 1997. Whales were more likely to be found Onshore (18.1%) rather than Offshore (6.9%). This gradient varied from year to year and with time of year. Sampling effort was correlated with sighting success, which suggests significant bias in the selection of transects to be flown. The location of transects were positively autocorrelated within a season further demonstrating sampling bias. These results suggest that more attention needs to be paid to survey design and implementation for long-term population monitoring. Stratified Random Sampling and Adaptive Sampling are attractive alternatives to Simple Random Sampling. More complex survey designs are more vulnerable to the vagaries of logistics and the tyranny of incremental decision making.

Keywords:

This abstract is being presented at: 4:15 PM in session:
Oral Session #46: Modeling Populations and Statistical Ecology.