Conceptual framework of the study estimating environmental conditions experienced by studied individual.

Effect of wind direction on wandering albatross take-off. Relative take-off direction to wind direction (black circles, n = 427) significantly distributed around 0° (head-wind), in contrast to cruising direction relative to the wind (gray x-mark, n = 427). The radial axis represents the wind speed.

Environmental effect on take-off. Effort for the take-off (running duration, running speed, flapping number, and flapping frequency) significantly decreased as wave height and wind speed increased (p < 0.01) except the relationship between flapping frequency and wave height (p = 0.026). Solid line shows the linear regression line determined from the LMMs and the number at the right top corner on each graph shows the sample sizes.

(A) Correlation between wind speed and wave height was weak (r = 0.27, n = 185). Bar charts and solid lines written above and right of the scatter plot are normed histograms of wind speed, wave height and curve fitted lines. Based on the peak value of fitted lines scatter plots were divided into four groups, WL: weak wind low wave (open square), WH: weak wind high wave (filled square), SL: strong wind low wave (open circle), and SH: strong wind high wave (filled circle). (B) Take-off effort comparison among four groups (a: running duration, b: running speed, c: flapping number, and d: flapping frequency). Cross mark indicates the mean value.

(A) running duration, (B) running speed, (C) flapping number, and (D) flapping frequency in response to the wave height change under weak wind (dashed line, 2 m/s) and strong wind (solid line, 8 m/s) conditions estimated from the LMM results. Gray area represents 99% CI.

Time series data of horizontal speed (top), lateral acceleration (middle), and dorsoventral acceleration (bottom) signals of the wandering albatross at the moment of take-off. Horizontal speed starts increasing from the beginning of the take-off. Red square shows the detected running phase based on the variance of the lateral acceleration signal. Red bars show the detected flapping behavior after the running phase based on the dorsoventral acceleration signal. Dorsoventral signal during the running phase fluctuates, probably due to the shaking body derived from the running motion, and thus it is not easy to judge the existence of flapping behavior.