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).