Fig. 2 Experimental setup.
Each experiment consisted of three phases (see Fig. 2 for experimental
setup): 1) Before exposure (phase 1), the research boat had the engine
turned off >300 m from any porpoise. The distance was
>400 m in the 20 knots experiments to ensure porpoises had
as much time to react to the boat as in the 10 knot experiments. Upon
sighting porpoises, we launched the drone to obtain video footage of
their behaviour for 1.5 min. We only analysed data from the last minute
of this phase for comparison with data collected during the exposure
phase. 2) During exposure (phase 2) the boat started to gradually
increase speed over 30 s while moving towards the drone, passing the
porpoise without changing direction. Due to variations in currents and
waves, the maximum speed (10 or 20 knots) varied by up to 2 knots among
trials. We aimed to pass the target porpoise at a distance of 25 m
(henceforth called the closest point of approach, CPA) and continued
until the boat was >300 m away from the porpoise. If the
porpoise was diving, we estimated its location based on the position of
the drone. When comparing porpoise behaviour “before” and “during”
exposure, we used data from the one minute centred around CPA and from
the last minute of phase 1. 3) After exposure (phase 3), we turned off
the engine and observed the porpoise for another 1.5 minutes with the
drone. To limit the influence of external variables on porpoise
behaviour, we only conducted experiments when there were no moving
vessels <1 km from the porpoise. Each group of porpoises was
only approached once, and we focused on one animal in each trial. After
each trial we moved more than one kilometre away and waited for minimum
of 30 min before approaching another porpoise to minimize the risk of
exposing the same animal twice. Throughout the experiment period, the
echo sounder of the boat was switched off.
On the last day of the fieldwork, we measured underwater noise levels at
varying distances from the boat. A stationary recorder (Sound trap,
OceanInstruments, New Zealand; 576 kHz sampling rate, 16 bits, clipping
level 176 dB re 1 µPa p, as determined by relative calibration) was
suspended 2.7 m below the surface and attached to a buoy. This was done
in the same area where exposure experiments were conducted. Following
this preliminary setup, we drove the boat from a distance of
>300 m to pass the recorder at speeds of either 10 or 20
knots. The boat was stopped when it was >300 m away from
the recorder. This process was repeated twice for each boat speed. The
boat’s geographical coordinates were recorded using a portable GPS
(Garmin GPSMAP 78s).
Data handling
We used the Drone Video Measure tool (Version 1.1.1; Egemose, 2021) to
extract location, swimming state and body length of the target porpoise
(one location per second). Body length was measured from the tip of its
nostrum to the fluke notch. This length was used as an indicator of the
age of the porpoise (following Stepien et al. 2023). Each porpoise was
measured up to three times from different locations (see porpoises’ body
length in Table A1 in Appendix A). The swimming state was categorized as
either shallow-dive (porpoise body shape clearly visible; including
breathing animals) or deep-dive (porpoise under the water, body shape
not clearly visible; Fig. 3). To calculate the porpoise’s speed and
horizontal turning angles between successive moves, we applied the
”adehabitatLT” package (Calenge, 2006) to porpoise locations (one per
second). To quantify whether an animal tried to avoid the boat and its
tendency to move away from the boat track, we calculated the distance
between the boat track and the porpoise at two successive boat locations
(i.e., one second apart; Fig. A1 in Appendix A). To ensure that porpoise
tracks were temporally aligned with the boat tracks we compared the
clocks in the boat GPS and in the drone at the point where the boat
became visible in the drone video footage. When needed, we calibrated
the drone clock based on the time difference between the GPS’s. The
location measurement accuracy was 2.4±1.5 m when the drone was 30 m
above the porpoise (Brennecke et al., 2022).