Environmental impact on visual perception modulates behavioral responses of schooling fish to looming predators.

Aggregation in social fishes has evolved to improve safety from predators. The individual interaction mechanisms that govern collective behavior are determined by the sensory systems that translate environmental information into behavior. In dynamic environments, shifts in conditions impede effective visual sensory perception in fish schools, and may induce changes in the collective response. Here, we consider whether environmental conditions that affect visual contrast modulate the collective response of schools to looming predators. By using a virtual environment to simulate four contrast levels, we tested whether the collective state of minnow fish schools was modified in response to a looming optical stimulus. Our results indicate that fish swam slower and were less polarized in lower contrast conditions. Additionally, schooling metrics known to be regulated by non-visual sensory systems tended to correlate better when contrast decreased. Over the course of the escape response, schools remained tightly formed and retained the capability of transferring social information. We propose that when visual perception is compromised, the interaction rules governing collective behavior are likely to be modified to prioritize ancillary sensory information crucial to maximizing chance of escape. Our results imply that multiple sensory systems can integrate to control collective behavior in environments with unreliable visual information.

Effects on individual level behaviour in mackerel (Scomber scombrus) of sub-lethal capture related stressors: Crowding and hypoxia.

Stress to fish during harvest in wild capture fisheries is known to negatively influence subsequent survival in catches that are released. Therefore, if fisheries are to be conducted sustainably, there is a need to promote good fish welfare during the capture process. Purse seine fishing is a widespread and efficient fishing method. However, capture and release of fish from purse seines (a process called "slipping") can result in extremely high mortality in small pelagic schooling species. The objective of this study was to establish behavioural indicators of sub-lethal stress in Atlantic mackerel (Scomber scombrus) that may be used to set safe threshold limits for use in commercial purse seine fishing, in order to ensure good fish welfare and thereby minimise slipping mortality. Controlled mesocosm scale experiments with schools of mackerel in net pens were undertaken to determine behavioural responses to simulated purse seine capture stressors of "crowding", "hypoxia" and "crowding & hypoxia". Crowding (at 30 kg.m-3) was achieved by reducing the volume of the net pen, while hypoxia (to 40% oxygen saturation) was achieved by surrounding the net pen with a tarpaulin bag to prevent water exchange. Using video analysis, we investigated behavioural responses in nearest neighbour distances, nearest neighbour angular deviations, tail beat amplitude and tail beat frequency (TBF). Of the metrics considered, only TBF showed a response; a significant increase to "crowding" (42% increase) and "crowding & hypoxia" (38% increase) was found. The increase in TBF in response to "hypoxia" alone (29% increase) was not significant. We therefore conclude that increases in tail beat frequency may be used as an indicator of sub-lethal purse seine capture stress in mackerel that may have utility in minimising post slipping mortality.

Automatic interpretation of otoliths using deep learning.

The age structure of a fish population has important implications for recruitment processes and population fluctuations, and is a key input to fisheries-assessment models. The current method of determining age structure relies on manually reading age from otoliths, and the process is labor intensive and dependent on specialist expertise. Recent advances in machine learning have provided methods that have been remarkably successful in a variety of settings, with potential to automate analysis that previously required manual curation. Machine learning models have previously been successfully applied to object recognition and similar image analysis tasks. Here we investigate whether deep learning models can also be used for estimating the age of otoliths from images. We adapt a pre-trained convolutional neural network designed for object recognition, to estimate the age of fish from otolith images. The model is trained and validated on a large collection of images of Greenland halibut otoliths. We show that the model works well, and that its precision is comparable to documented precision obtained by human experts. Automating this analysis may help to improve consistency, lower cost, and increase the extent of age estimation. Given that adequate data are available, this method could also be used to estimate age of other species using images of otoliths or fish scales.

Echo integration using non-vertical sonar beams: The bias caused by non-uniform distribution of fish within the echo beam.

The equivalent beam angle is a key parameter in echo integration, where it is assumed that targets are uniformly distributed within the sampling volume of the transducer beam. For a horizontally oriented sonar, this assumption is violated if the vertical distribution of fish is non-uniform throughout a sample, potentially causing a substantial bias in estimates of fish abundance or biomass. This paper investigates the magnitude of this bias using observations and simulated data, where in each case the vertical distribution of fish within a limited geographical area is estimated.

Effects on schooling function in mackerel of sub-lethal capture related stressors: Crowding and hypoxia.

The selectivity of fishing gears with respect to fish species and size is important, both for fisheries management and fishing operations. Purse seining is an efficient, environmentally friendly fish capture methodology generally targeting single species aggregations, but once a fish school has been selected and surrounded by the seine, there is no selections for individual size, species or catch quantity. A common practice for evaluating the catch is to haul the seine to a point where physical samples or inspections of catch composition can be made. The release process is called slipping and may lead to mortality in the released fish. The objective of this study was to simulate a crowding situation and investigate how the behaviour was affected in response to increased fish density, decreased oxygen levels, or a combination of the two, and to see if there is a behavioural measure that can be used to set safe crowding limits. The experiment was conducted on Mackerel (Scomber scombrus) held in net pens. The volume of the net pen was reduced to increase fish density, and a tarpaulin bag was wrapped around the pen to reduce the oxygen levels. Oxygen, fish density and space occupancy was monitored during the experiment, and the behavioural reactions was assessed using an imaging sonar. The main result was that the schooling function, i.e. the response to a predator model, was significantly reduced during crowding but not in response to hypoxia. There were some indications of a slow recovery of the function post-treatment. We conclude that crowding causes behavioural responses that occur before densities that induce fish mortality. Consequently, there is a behavioural response that could be used as a proxy for setting safe crowding limits.

Collective responses of a large mackerel school depend on the size and speed of a robotic fish but not on tail motion.

So far, actuated fish models have been used to study animal interactions in small-scale controlled experiments. This study, conducted in a semi-controlled setting, investigates robot5 interactions with a large wild-caught marine fish school (∼3000 individuals) in their natural social environment. Two towed fish robots were used to decouple size, tail motion and speed in a series of sea-cage experiments. Using high-resolution imaging sonar and sonar-video blind scoring, we monitored and classified the school's collective reaction towards the fish robots as attraction or avoidance. We found that two key releasers-the size and the speed of the robotic fish-were responsible for triggering either evasive reactions or following responses. At the same time, we found fish reactions to the tail motion to be insignificant. The fish evaded a fast-moving robot even if it was small. However, mackerels following propensity was greater towards a slow small robot. When moving slowly, the larger robot triggered significantly more avoidance responses than a small robot. Our results suggest that the collective responses of a large school exposed to a robotic fish could be manipulated by tuning two principal releasers-size and speed. These results can help to design experimental methods for in situ observations of wild fish schools or to develop underwater robots for guiding and interacting with free-ranging aggregated aquatic organisms.

Investigating the Effect of Tones and Frequency Sweeps on the Collective Behavior of Penned Herring (Clupea harengus).

We experimentally played back tones and sweeps to captive herring (Clupea harengus) in a net pen and measured the collective response of a large and a small group of fish using a camera, echo sounder, and multibeam sonar. The playbacks ranged in frequency from 160 to 500 Hz and 131 to 147 dB re 1 µPa in received sound pressure level. Herring behavior was scored by a team that blindly evaluated the observations. Overall, the responses were modest. Stronger reactions were observed at higher source levels, lower frequencies, and smaller school sizes, but there was no effect on signal rise time.

Experimental evidence of threat-sensitive collective avoidance responses in a large wild-caught herring school.

Aggregation is commonly thought to improve animals' security. Within aquatic ecosystems, group-living prey can learn about immediate threats using cues perceived directly from predators, or from collective behaviours, for example, by reacting to the escape behaviours of companions. Combining cues from different modalities may improve the accuracy of prey antipredatory decisions. In this study, we explored the sensory modalities that mediate collective antipredatory responses of herring (Clupea harengus) when in a large school (approximately 60,000 individuals). By conducting a simulated predator encounter experiment in a semi-controlled environment (a sea cage), we tested the hypothesis that the collective responses of herring are threat-sensitive. We investigated whether cues from potential threats obtained visually or from the perception of water displacement, used independently or in an additive way, affected the strength of the collective avoidance reactions. We modified the sensory nature of the simulated threat by exposing the herring to 4 predator models differing in shape and transparency. The collective vertical avoidance response was observed and quantified using active acoustics. The combination of sensory cues elicited the strongest avoidance reactions, suggesting that collective antipredator responses in herring are mediated by the sensory modalities involved during threat detection in an additive fashion. Thus, this study provides evidence for magnitude-graded threat responses in a large school of wild-caught herring which is consistent with the "threat-sensitive hypothesis".

Simulations of multi-beam sonar echos from schooling individual fish in a quiet environment.

A model is developed and demonstrated for simulating echosounder and sonar observations of fish schools with specified shapes and composed of individuals having specified target strengths and behaviors. The model emulates the performances of actual multi-frequency echosounders and multi-beam echosounders and sonars and generates synthetic echograms of fish schools that can be compared with real echograms. The model enables acoustic observations of large in situ fish schools to be evaluated in terms of individual and aggregated fish behaviors. It also facilitates analyses of the sensitivity of fish biomass estimates to different target strength models and their parameterizations. To demonstrate how this tool may facilitate objective interpretations of acoustically estimated fish biomass and behavior, simulated echograms of fish with different spatial and orientation distributions are compared with real echograms of herring collected with a multi-beam sonar aboard the research vessel "G.O. Sars." Results highlight the important effects of fish-backscatter directivity, particularly when sensing with small acoustic wavelengths relative to the fish length. Results also show that directivity is both a potential obstacle to estimating fish biomass accurately and a potential source of information about fish behavior.

Vortex Formation and Foraging in Polyphenic Spadefoot Toad Tadpoles.

Animal aggregations are widespread in nature and can exhibit complex emergent properties not found at an individual level. We investigate one such example here, collective vortex formation by congeneric spadefoot toad tadpoles: Spea bombifrons and S. multiplicata. Tadpoles of these species develop into either an omnivorous or a carnivorous (cannibalistic) morph depending on diet. Previous studies show S. multiplicata are more likely to develop into omnivores and feed on suspended organic matter in the water body. The omnivorous morph is frequently social, forming aggregates that move and forage together, and form vortices in which they adopt a distinctive slowly-rotating circular formation. This behaviour has been speculated to act as a means to agitate the substratum in ponds and thus could be a collective foraging strategy. Here we perform a quantitative investigation of the behaviour of tadpoles within aggregates. We found that only S. multiplicata groups exhibited vortex formation, suggesting that social interactions differ between species. The probability of collectively forming a vortex, in response to introduced food particles, increased for higher tadpole densities and when tadpoles were hungry. Individuals inside a vortex moved faster and exhibited higher (by approximately 27%) tailbeat frequencies than those outside the vortex, thus incurring a personal energetic cost. The resulting environmental modification, however, suggests vortex behaviour may be an adaptation to actively create, and exploit, a resource patch within the environment.

The dynamics of coordinated group hunting and collective information transfer among schooling prey.

Predator-prey interactions are vital to the stability of many ecosystems. Yet, few studies have considered how they are mediated due to substantial challenges in quantifying behavior over appropriate temporal and spatial scales. Here, we employ high-resolution sonar imaging to track the motion and interactions among predatory fish and their schooling prey in a natural environment. In particular, we address the relationship between predator attack behavior and the capacity for prey to respond both directly and through collective propagation of changes in velocity by group members. To do so, we investigated a large number of attacks and estimated per capita risk during attack and its relation to the size, shape, and internal structure of prey groups. Predators were found to frequently form coordinated hunting groups, with up to five individuals attacking in line formation. Attacks were associated with increased fragmentation and irregularities in the spatial structure of prey groups, features that inhibit collective information transfer among prey. Prey group fragmentation, likely facilitated by predator line formation, increased (estimated) per capita risk of prey, provided prey schools were maintained below a threshold size of approximately 2 m(2). Our results highlight the importance of collective behavior to the strategies employed by both predators and prey under conditions of considerable informational constraints.

Behavior of captive herring exposed to naval sonar transmissions (1.0-1.6 kHz) throughout a yearly cycle.

Atlantic herring, Clupea harengus, is a hearing specialist, and several studies have demonstrated strong responses to man-made noise, for example, from an approaching vessel. To avoid negative impacts from naval sonar operations, a set of studies of reaction patters of herring to low-frequency (1.0-1.5 kHz) naval sonar signals has been undertaken. This paper presents herring reactions to sonar signals and other stimuli when kept in captivity under detailed acoustic and video monitoring. Throughout the experiment, spanning three seasons of a year, the fish did not react significantly to sonar signals from a passing frigate, at received root-mean-square sound-pressure level (SPL) up to 168 dB re 1 µPa. In contrast, the fish did exhibit a significant diving reaction when exposed to other sounds, with a much lower SPL, e.g., from a two-stroke engine. This shows that the experimental setup is sensitive to herring reactions when occurring. The lack of herring reaction to sonar signals is consistent with earlier in situ behavioral studies. The complexity of the behavioral reactions in captivity underline the need for better understanding of the causal relationship between stimuli and reaction patterns of fish.

Observing individual fish behavior in fish aggregations: tracking in dense fish aggregations using a split-beam echosounder.

Acoustic instruments are important tools for observing the behavior of aquatic organisms. This paper presents a simple but efficient method for improving the tracking of closely spaced targets using a split-beam echosounder. The traditional method has been a stepwise approach from the detection of echoes, rejection of apparently multiple targets and then tracking the remainder. This is inefficient because the split-beam angles are not included in the initial detection; rather they are only used in the rejection criteria before the subsequent tracking. A simple track-before-detection method is presented, where the phase angles, echo intensities, ranges, and times are used simultaneously, resulting in better detection and tracking of the individual fish. Two test data sets were analyzed to determine the effectiveness of this method at discriminating individual tracks from within dense fish aggregations. The first data set was collected by lowering a split-beam transducer into a herring layer. The second data set, also collected with a split-beam transducer, was from a caged aggregation of feeding herring larvae. Results indicate the potential of target tracking, using a split-beam echosounder, as a tool for understanding interindividual behavior.

Silent research vessels are not quiet.

Behavior of herring (Clupea harengus) is stimulated by two ocean-going research vessels; respectively designed with and without regard to radiated-noise-standards. Both vessels generate a reaction pattern, but, contrary to expectations, the reaction initiated by the silent vessel is stronger and more prolonged than the one initiated by the conventional vessel. The recommendations from the scientific community on noise-reduced designs were motivated by the expectation of minimizing bias on survey results caused by vessel-induced fish behavior. In conclusion, the candidate stimuli for vessel avoidance remain obscure. Noise reduction might be necessary but is insufficient to obtain stealth vessel assets during surveys.

Tracking individual fish from a moving platform using a split-beam transducer.

Tracking of individual fish targets using a split-beam echosounder is a common method for investigating fish behavior. When mounted on a floating platform like a ship or a buoy, the transducer movement often complicates the process. This paper presents a framework for tracking single targets from such a platform. A filter based on the correlated fish movements between pings is developed to estimate the platform movement, and an extended Kalman filter is used to combine the split-beam measurements and the platform-position estimates. Different methods for gating and data association are implemented and tested with respect to data-association errors, using manually tracked data from a free-floating buoy as a reference. The data association was improved by utilizing the estimated velocity for each track to predict the location of the next observation. The data association was more robust when estimates of platform tilt/roll were used. Other techniques to estimate position and velocity, like linear regression and smoothing splines, were implemented and tested on a simulated data set. The platform-state estimation improved the estimates for methods like the Kalman filter and a smoothing spline with cross validation, but not for robust methods like linear regression and smoothing spline with a fixed degree of smoothing.