Figure 8
Swordfish vertical behavior general additive model results. Time of day is presented as local time in the study area (UTC + 11). Partial effects are on the link function (log) scale, positive effect values represent increasing depth. Time of day was by far the strongest predictor of depth (a), with the cubic cyclic smooth of time of day illustrating a consistent crepuscular shift in depth. The pattern was largely consistent across individual swordfish (b), although two migrated from depth at night consistently earlier and another was consistently shallower at night and deeper during the day (SC0024). The interaction functions of time of day and day of year, and time of day and moon phase (c) exhibit effects consistent with day length driving vertical migration timing and moon phase affecting depth at night. Smooth functions of main effects included (d) indicate limited evidence of consistent effects on depth independent of time of day, although depth across the diel cycle may decrease during winter and at high latitude. Bottom: Model predictions of depth across time of day overall (e), and at night, during new, full, and 3/4 full moon phases (f) with other covariates held constant at median covariate values (153.7°E, 38.6°S, Aug 13). The shaded regions of panel (e) indicate the periods encompassing twilight and golden hour (i.e., sun elevation −18° to 6°) and the tick marks at the top of the plot indicate sunset and sunrise time at the constant covariate time and latitude. Dotted lines (and shaded areas in panel (f) indicate 95% credible intervals. The bars at the top of panel f indicate the timing of moonrise and moonset during the ¾ full and full moon phases nearest median covariate values (see Supplementary Figs. S3 and S4 for predicted depths during half-moon phases and details on moon visibility across phase). Generated in R 4.2.167.
