Unveiling links between salmon physiology and online monitored behaviour

SalmonInsight is a four-year research project funded by the Research Council of Norway and conducted by SINTEF Ocean, NINA, NTNU Departments of Engineering Cybernetics and Biology, the University of Gothenburg, and the Swedish University of Agricultural Sciences. The main goal of the project was to provide new knowledge on how physiology and stress in salmon is reflected in observable behavioural expressions, thus forming a foundation for developing future solutions for online monitoring of salmon in sea-cages during production. The total mortality rate during the grow-out phase in Atlantic salmon farming is too high. The first step in improving this situation is to monitor the physiology and stress of salmon during demanding operations (e.g., crowding) and under normal production. Developing technologies for such observations requires knowledge on how the physiology and stress of salmon is reflected in data possible to monitor online. SalmonInsight aimed to answer these questions through a series of experiments where physiological data was collected simultaneously with data possible to obtain in real time in sea cages using state-of-the-art sensors. SalmonInsight included a PhD-candidate who developed a novel biosensing implant that measures acceleration, rotation rates, compass direction, magnetic field strength, temperature, electrocardiogram (ECG) and photoplethysmogram (PPG) in fish. This enables more accurate calculation of an activity proxy while allowing more robust heart rate estimation based on two independent sensing principles. The PPG also opens the prospect of estimating oxygen saturation in arterial blood, a complete novelty in fishes. The first experiment in the project was a controlled laboratory study where sensor tags were used to measure heart rate and activity on individual salmon. The fish were subjected to artificial stress by repeatedly draining their holding tanks. Physiological samples collected before tagging, before and after stressing, and in post-stress recovery were used to correlate sensor values with the physiological response. The results demonstrate that there are changes in heart rate and swimming activity in response to stress events that are consistent with blood proxies for stress levels in Atlantic salmon. This underscores the possibility of on-line stress monitoring in full scale production using tags. In addition, the experiment provided valuable information on fish recovery after surgical tag implantation. The information gathered during the lab trials was essential in the follow-up experiment. In this trial with a similar set-up, tagged fish and their cohabitants were observed before, during and after a crowding operation in meso-scale sea cages using physiological values in addition to the data from the implanted data storage tags. While swimming activity clearly increased in response to the stress situation, a finding that was supported by blood parameter data, the heart rate reactions were more difficult to interpret and did not show the same pattern as seen in the tank trials. In the final experiment salmon were simultaneously implanted with a data storage tag and an intravascular arterial catheter that allowed repeated blood sampling to establish the relationship between sensor values and physiological state. A swim tunnel was used to exercise the fish at defined speeds. Data from measurements of heart rate, external acceleration, and blood physiology were compared to parameters measured using video analyses, including mouth opening and tail beat frequencies, as well as oxygen consumption rates assessed through sensors in the tunnel. Results show that bio-logger data can be used to extrapolate a range of stress-related physiological events when these are accompanied by increases in activity. This highlights the great potential of biosensors for monitoring fish welfare. In summary, the project has brought the field of biosensing implants significantly forward, resulting in a novel biosensing implant, knowledge transfer between institutes and generations, the installation of customised infrastructure, and facilitation of follow up projects. In terms of outreach, the project has resulted in 10 published articles, with 3 more under review, and another 3 to be submitted in the coming weeks. In addition, results have been presented at several national and international conferences and have been published in popular science journals.

The most tangible outcome of the SalmonInsight project is the novel biosensing implant developed by the PhD candidate. The implant has the ability to measure a photoplethysmogram (PPG), allowing the assessment of arterial blood oxygenation and enabling a more robust heart rate measurement. This is especially relevant as the project outcomes showed that current heart rate measurements can be difficult to use as stress indicators. By supplementing PPG data collection in combination with measurements of acceleration, rotation rates, compass direction, magnetic field strength, temperature, and electrocardiogram (ECG) this novel tool may help to bridge the knowledge gap between measured behavioural and physiological parameters and stress. This offers new opportunities for research on the topic of stress, fish welfare, and fish physiology. As such it may support increased sustainability in aquaculture. For the collaborators, the project activities have held many advantages such as the transfer of knowledge between institutes as well as generations, ensuring sustained competence and recruitment into the research community. Moreover, the project created the opportunity to collaborate with institutes/universities in and outside of Norway, establishing good contacts between leading research specialists on the topic. The project has caused interest in the biosensing community through numerous publications and presentations at relevant conferences, leading to visits to relevant research groups outside the project group. Finally, the project facilitated the expansion of existing infrastructure for husbandry and experimentation of fish as well as the installation of new equipment (e.g., fish raceway including its own RAS system) which will accommodate future experiments in the field and increase the competitive advantage of the project partners. The far-reaching impacts of the project include the opportunity for follow-up projects, with some that have already been realised during the project period, building on the acquired competence. This competence further creates the foundation for future collaborations outside the project group. The novel biosensing implant is a tool that will not only be relevant in a basic research context but can facilitate testing of novel aquaculture technology while satisfying the recent regulatory requirements for such testing. As such it may be an important tool for the industry on their quest for future aquaculture technology. The combination of the large number of relevant sensors will eventually make the implant a good candidate for use on sentinel fish, i.e., individuals equipped with a biosensing implant to inform on the conditions experiences by the fish group as a whole. Moreover, the multi-sensory equipment will also make the implant attractive for use in species other Atlantic salmon, opening up for use in a wide range of fresh- and saltwater species.

Although Norwegian salmon aquaculture is a successful and economically efficient industry, cumulative mortality during the grow-out phase has been estimated to about 19%. This indicates a strong need to improve farming processes, and has stimulated a demand for new technologies that can help farmers to better understand and monitor fish health and welfare. In SalmonInsight we will use sensors and acoustic telemetry under laboratory conditions to identify how the physiological and stress dynamics of salmon are linked with behavioural expressions possible to monitor in the field. This knowledge will ultimately be used to design algorithms for acoustic telemetry tags carrying e.g. accelerometers such that they will be sensitive to parameters reflecting variations in fish physiology and stress. The work will feature two laboratory studies aimed at unveiling the relationships between physiological and behavioural expressions in detail. The first laboratory study will feature a simulated crowding operation, where the fish are equipped with sensors of the same type as those used in telemetry tags. Physiological blood parameters will then be manually sampled, while behavioural data will be obtained using visual observation and established computer vision methods. The second laboratory study will be conducted in a swim tunnel using respirometry and cannulation methods in addition to the methods used in the crowding experiments. These studies will uncover links between physiological dynamics, behavioural expressions and sensor data in more detail. We will then develop novel algorithms for acoustic telemetry solutions based on state-of-the-art sensors aimed at detecting the behavioural aspects identified to be most strongly linked with the physiological dynamics of the fish. This will verify the industrial applicability of the method, and produce new knowledge for sustainable and safe farming procedures.