A Multi-Dimensional Examination of Foraging Habitat Use by Gray Whales Using Long Time-Series and Acoustics Data

Abstract

Gray whales (Eschrichtius robustus) show high site fidelity to feeding and breeding areas. The whales' annual cycle could be learned or be driven by factors such as prey abundance or ocean conditions. Long-term line transect and photo-identification data were analyzed to consider intra- and inter- annual patterns of habitat use and the underlying drivers for foraging areas in Clayoquot Sound, British Columbia. Time series, autocorrelation and weighted means analysis were used on the 20-years of data (1997-2016). A generalized additive model showed that whale use of the area was most strongly influenced by the maximum number of whales, and the date of its occurrence, recorded the previous year. This maximum, when it occurred in the summer, impacted the whale numbers for up to four subsequent years. The annual average number of whales per transect also influenced the proportion of whales known to return in multiple years to forage. Many of these returning whales first used the site to wean and returned in subsequent years to feed. The transect data was also used to contextualize passive acoustic recordings, comparing call type and rate for periods when the whale number, location and social context was known. Calling patterns appeared to be socially derived, with shorter-range knock calls dominant when whales were closer, and lower-frequency moans when foraging occurred when whales were more distant from each other. This suggests that prey-finding and site use may also be influenced by conspecifics.

Keywords: foraging intensity; passive acoustic monitoring; predator-prey; predictive habitat use; site fidelity; top-down forces.

Figure 1

Study site map, showing transect survey route and Autonomous Multichannel Acoustic Recorder (AMAR) deployment location in Cow Bay (49.25629, −126.15928). The likely detection radii of gray whale class three moan calls are also shown. The AMAR location is the black diamond labeled ‘AMAR’ at the center of the detection circles. The smaller circle represents the range of detection exceeded 90% of the time (500 m) and the larger circle the maximum range 10% of the time (9 km).

Figure 2

Boxplot to indicate the number of foraging whales sighted per transect survey in the study site. This is used as a proxy of foraging intensity through the analysis. The overall global mean of all years is indicated by the dashed line at 7.1 whales. The numbers above each box indicate the number of transects that year.

Figure 3

Above: The average number of foraging whales per transect over the summer season (‘Average’, black line) and maximum number of foraging whales in a single transect (‘Maximum’, grey line). Below: The proportion of the whales sighted in a year that were ‘Return’ whales and had been seen in previous years. No photo-identification data for this analysis for 1997, 2001, or 2014 onwards.

Figure 4

Histogram (A), density plot (B), Q-Q plot (C), lag analysis (D), autocorrelation and (E) and partial autocorrelation of residuals (F) for all transects in all years.

Figure 5

Histogram (A), density plot (B), Q-Q plot (C), lag analysis (D), autocorrelation and (E) and partial autocorrelation of residuals (F) for annual average number of whales per transect.

Figure 6

Relative importance of whale-derived and environmental variables in determining the annual average number of whales per transect (A), the number of returning whales (B) and the number of single visit whales (C) derived from the cross-correlation analysis.

Figure 7

Heat map of foraging whale locations from the transect data by month (Black shading). Weighted average mean locations are shown for each year (red diamonds). Data from all transect and all years in (A) May, (B) June, (C) July, (D) August, (E) September, and an aggregate of all years, all months (F) are also shown.

Figure 8

Calling rate (calls per hour per individual) for core call types compared to the known number of whales in Cow Bay for that hour, taken from 243 h of observation data. Call class one knock calls, class two sweeping tones, class three moans, and class four rumbles are shown aggregated and separately.