Detecting changes in the distribution of calling bowhead whales exposed to fluctuating anthropogenic sounds

Abstract

This paper describes an analysis approach designed to detect the effects of fluctuating anthropogenic underwater sound on the distribution of calling bowhead whales (Balaena mysticetus) during migration. The anthropogenic sounds in this case were associated with an offshore oil production island (Northstar Island) in the Beaufort Sea northwest of Prudhoe Bay, Alaska, but the method has wider applicability. In autumn, bowhead whales migrate westward at varying distances offshore where some are exposed to Northstar sounds. Anthropogenic effects, if present, were hypothesised to be most pronounced in the southern (proximal) part of the migration corridor. Underwater sound levels were measured continuously ca. 500m from Northstar, and locations of calling whales were determined by a seafloor array of directional acoustical recorders. Weighted quantile regression related the 5th quantile of offshore call distance to anthropogenic sounds and other covariates. Case weights were inversely proportional to both probability of detection and location uncertainty. Due to potential dependencies in call locations, block permutation of uncorrelated whale call clusters was used to assign significance levels to coefficients in the quantile regression model. Statistical model selection was used to determine the anthropogenic sound measures most correlated with the 5th quantile of offshore call distances, after allowing for natural within-season variation quantified by day–night changes, distance of the call east or west of Northstar, and date. Data used to illustrate the method were collected over 29 days in September 2003 and included 25,176 bowhead calls. The estimated offshore distance of the 5th quantile call was 0.67km (95% confidence interval 0.31 to 1.05km) farther offshore when tones associated with Northstar were recorded in the 10–450Hz band during the 15 minutes just prior to each call. The method has been applied successfully to similar data collected near Northstar in other years, and may be useful in other studies that simultaneously collect data on animal locations and fluctuating stimuli.