First analysis of behavioural responses of humpback whales (Megaptera novaeangliae) to two acoustic alarms in a northern feeding ground off Iceland

Mitigating cetacean entanglement in fishing industries is of global interest. Strategies include the use of acoustic alarms to warn whales of fishing gear. For baleen whales, responses to acoustic alarms are poorly understood. This behavioural response study compared the behaviour of humpback whales (Megaptera novaeangliae) in their feeding grounds off Iceland prior to, during, and after exposure to a low-frequency whale pinger (Future Oceans) and a high-frequency seal scarer (Lofitech ltd.). Linear mixed effects models and binary generalized linear mixed effects models were used to analyze the effect of the alarms on surface feeding, swimming speed, breathing rate, directness and dive time. We observed a significant decrease in surface feeding and a significant increase in swimming speed during exposure to the whale pinger. Changes in dive time between the phases of a trial differed significantly between individuals indicating that responses may depend on individual or behavioural state. We did not find any significant reactions in response to the seal scarer. In addition to the experimental exposures, a trial of whale pingers on a capelin purse seine net was conducted. Results from this trial showed that whales entered the net from the bottom while the pingers were attached at the top, but the encircled whales were able to locate an opening free of pingers and escape without damaging the net. Our results suggest that whale pingers may be a useful entanglement mitigation tool in humpback whale feeding grounds given that a reduction in feeding around nets likely reduces the risk of whales swimming through them. Pingers may also minimize net damage if whales are encircled by aiding the whales in finding their way out. However, given the uncertain long-term consequences of the behavioural changes reported here, whale pingers are most advisable for short-term use in conjunction with other entanglement mitigation measures.

Mitigating cetacean entanglement in fishing industries is of global interest. Strategies include 23 the use of acoustic alarms to warn whales of fishing gear. For baleen whales, responses to 24 acoustic alarms are poorly understood. This behavioural response study compared the 25 behaviour of humpback whales (Megaptera novaeangliae) in their feeding grounds off 26 Iceland prior to, during, and after exposure to a low-frequency whale pinger (Future Oceans) 27 and a high-frequency seal scarer (Lofitech ltd.). Linear mixed effects models and binary 28 generalized linear mixed effects models were used to analyze the effect of the alarms on 29 surface feeding, swimming speed, breathing rate, directness and dive time. We observed a 30 significant decrease in surface feeding and a significant increase in swimming speed during 31 exposure to the whale pinger. Changes in dive time between the phases of a trial differed 32 significantly between individuals indicating that responses may depend on individual or 33 behavioural state. We did not find any significant reactions in response to the seal scarer. In 34 addition to the experimental exposures, a trial of whale pingers on a capelin purse seine net 35 was conducted. Results from this trial showed that whales entered the net from the bottom 36 while the pingers were attached at the top, but the encircled whales were able to locate an 37 opening free of pingers and escape without damaging the net. Our results suggest that whale 38 pingers may be a useful entanglement mitigation tool in humpback whale feeding grounds 39 given that a reduction in feeding around nets likely reduces the risk of whales swimming 40 through them. Pingers may also minimize net damage if whales are encircled by aiding the 41 whales in finding their way out. However, given the uncertain long-term consequences of the Introduction 45 There is global concern over marine mammal bycatch and entanglement in fishing gear (ie. 46 animals becoming incidentally caught in gear and drowning, or escaping, sometimes with 47 gear attached to their body and/or with injuries). Documented impacts of entanglement on 48 cetaceans include injury [1,2,3] , exhaustion of energy budgets [4], emaciation and 49 drowning [5,2]. These impacts at the individual level can lead to increased mortality rates at 50 the population level [6,7] . Entanglement is known to occur involving many different types 51 of fishing gear [8] and is likely to affect most cetacean species [9] . Apart from impacts on 52 cetacean individuals and populations, entanglement also leads to financial losses to the 53 fishing industry due to loss-of-catch, gear damage or loss and downtime for repairs [10,11]. 54 This can be a particularly serious issue in fisheries experiencing large whale, such as 55 humpback whale (Megaptera novaeangliae), entanglements. 56 Technologies have been developed with the intent to mitigate marine mammal entanglement. 57 One such technology is acoustic alarms known as "pingers". These devices can be to attached 58 to fishing gear and emit a tone underwater within the hearing range of target marine 59 mammals [12] . The devices serve to "illuminate" the gear with sound and warn the animals 60 of its presence to encourage them to avoid it. Alternatively, the pingers may simply serve as 61 an annoying, unnatural sound that the animals want to avoid [13]. Since large whales in 62 particular can often escape from or carry away entangling gear, it has been suggested that 63 whales can learn to associate nets and pingers with danger [14] . For cetaceans, specific 64 pinger acoustic alarms have been developed for porpoises, dolphins and beaked whales 65 (odontocetes) as well as baleen whales (mysticetes) with varying degrees of success. The 66 high-frequency porpoise, dolphin and beaked whale pingers have been shown to reduce 67 bycatch of several species [eg. 15, 16, 17, 18], though on the other hand, some studies have found there to be no change or even increased bycatch of some odontocete species with the 69 use of pingers [eg. 19, 20]  In addition to the pinger acoustic alarms, acoustic deterrent devices (ADDs) have also been 85 developed. Primarily used in the aquaculture industry to ward off seals, these devices produce 86 a loud, high-frequency sound designed to scare away the animals [32]. Though not designed 87 originally for use to deter cetaceans, it has been observed that at least some cetacean species 88 react to the loud sound produced by such a device [33,34]. Testing of ADDs on cetaceans 89 has found that odontocetes including harbour porpoises [34]  one-quarter of the coastal Icelandic humpback whale population is estimated to have been 109 entangled in fishing gear at least once [46], and virtually all of the fishing methods in the 110 country have reported issues with humpback whales swimming through, and sometimes 111 becoming entangled in, the gear in the water [47,44]. This has caused gear damage or loss, as 112 well as injury or death to the whales in some cases [46,47,48,49].

113
Currently there are no mitigation methods or regulations in place for minimizing whale 114 entanglement in fishing gear in Iceland, despite growing concern in the local fishing 115 industries. This study conducted the first analysis of behavioural response of free-ranging 116 humpback whales in their northern feeding grounds off the coast of northeast Iceland to the 117 whale pinger acoustic alarm (Future Oceans) and the seal scarer acoustic deterrent device 6 118 (Lofitech AS ltd.). In addition to the experimental exposure of whales to the acoustic alarms, 119 this study conducted the first trial of the whale pingers in the capelin purse seine fishery in 120 Iceland. Results from this study help to decide if acoustic alarms are likely to effectively 121 mitigate humpback whale entanglement in their feeding grounds and shed light on possible 122 adverse effects of the alarms on natural humpback whale behaviour. Oceans whale pinger. This device operates on a single 3.6V lithium battery and activates 145 automatically when in contact with saltwater. When active, the alarm produces a 145 decibel 146 re 1µPa tone at 3kHz for 300 ms at 5 sec intervals (Future Oceans) (Fig 2). The second alarm 147 was the Lofitech AS ltd. seal scarer ADD composed of a control box with a 25m long cable 148 with a transducer unit at the end which produces the sound. This control box is powered by a 149 12V marine battery onboard the boat. When active, the alarm produces a 191 decibel re 1µPa 150 sound between 10-20 kHz for 500ms at random intervals of 5-60 sec (Fig 3). A calibration of 151 both acoustic alarm devices was conducted in a harbour to confirm the manufacturers    individual whale was not exposed to the same device more than once within the same year, to 181 avoid possible habituation to the alarm sound. When photo-identification was complete, the 182 pre-exposure phase (PrE) began with the boat following the focal whale from a distance of 183 approximately 100m for 30 mins to obtain a baseline of behaviour of the individual. The 184 100m distance complies with whale watching criteria set forth in many countries around the 185 world to minimize disturbance to the animal [53] while still being within range to collect all 186 necessary data. Each breath the whale took was recorded as "up" and each terminal dive was 187 recorded as "dive" in Logger 2010. Other information was also noted, including if the whale 188 dove with or without raising the fluke, if the whale appeared to be feeding, and if there were other whales in the area. Furthermore, one researcher used an angle-board and rangefinder to 190 obtain the angle to the whale in relation to the boat and the distance to the whale, and this 191 data was also recorded into Logger 2010. If the distance could not be obtained from the 192 rangefinder, one researcher estimated the distance to the whale when it took a terminal dive.

193
The angle-board, rangefinder, and distance estimation were always done by the same   indicating the whale had a full mouth (Fig 4). A surfacing was also categorized as Y if 210 researchers audibly indicated the whale was feeding in the video even though the surfacing 211 was not visible in the footage.

214
Swimming speed 215 The swimming speed of the focal whale was calculated for each phase of each behavioural 216 trial, when enough data was available. Speed was calculated from each terminal dive to the 217 next terminal dive (and therefore included distance information from when the focal whale 218 was diving and was at the surface).    Since previous findings suggested individual response to sound can depend on behavioural 246 state [54], individual-specific response variation was incorporated into our models by 247 introducing random slopes for the predictor phase for all response variables. We tested if 248 random intercept and slope models fitted the data better than pure random intercept models.   (Table 1). Of these, enough data for analysis was collected on 14 trips resulting in 9 287 WP trials and 7 SS trials.  was not identifiable beyond confirming that it was only used once in the study.

304
There were eleven attempts made to complete a WP trial, resulting in nine usable trials. Out 305 of these eleven attempts, the individual whale was considered lost (disappeared for more than 306 20 minutes) in three cases (WP1, WP7, WP8). Two out of these three cases did not result in 307 enough data to be included in the analysis (WP1, WP8   Table 3). The predictor phase had a significant effect on both speed (p = 0.006; Table 2) and surface 335 feeding (p = 0.019; Table 2). Humpback whale speed during the E phase was 1.7 times higher 336 than during the PoE phase (p = 0.0024; Table 3) and 1.4 times higher than during the PrE 337 phase (p = 0.11; Table 3). No significant differences in humpback whale speed were observed 338 between the PrE and PoE phases (p = 0.62; Table 3). The probability of surface feeding was 339 significantly lower during the E phase than during the PoE phase (p = 0.026; Table 4). The 340 reduction in surface feeding from the PrE to the E phase was marginally significant (p = 341 0.099; Table 4). Rates of surface feeding amounted to 11% and 13% in the PrE and PoE 342 phases and dropped to 4% in the E phase (Fig 7).  Table 2).     358   Table 4). 359 No significant changes in breathing rate (p = 0.42; Table 2) and directness (p = 0.40; Table 2) 360 were detected in response to exposure to whale pinger sound. The model for dive time was 361 the only case in which a random slope model fitted the data significantly better than a random 362 intercept model (p < 0.001; Table 2). Phase of the trial, however, had no significant effect on 363 dive time (p = 0.79; Table 2).  exact same manner as for the WP trial analysis. Random slope models did not fit the data 368 significantly better than random intercept models for any of the response variables (Table 5). 369 Experimental phase did not have a significant effect on any of the response variables (Table  5). Thus, we found no evidence for an individual-specific or shared response of humpback 371 whales to seal scarer alarm.  Purse-seine trial of the whale pingers 386 The captain of the participating capelin purse seine vessel did not report any issues with  be a common response when whales are exposed to anthropogenic noise. This is the first time  underwater feeding activity could provide valuable information for this hypothesis.

480
Disruption of feeding behaviour in these whales is cause for concern for negative impacts on in frequency than the device used in our study. It is possible that the frequency of the seal 548 scarer was just too high for the humpback whales to hear the alarm well enough to exhibit a 549 significant response, confirming that acoustic entanglement mitigation devices need to target 550 the best-estimated hearing range of the whales. However, the surface feeding behavioural 551 response remains unknown for the seal scarer since there was not enough surface feeding 552 observed in the trials to analyze this.

553
The use of the whale pingers on the capelin purse seine net for one season provided a first