Atlantic cod individual spatial behaviour and stable isotope associations in a no‐take marine reserve

This is a self-archived version of an original article. This version may differ from the original in pagination and typographic details. Author(s): Title: Year: Version: Copyright: Rights: Rights url: Please cite the original version: CC BY 4.0 https://creativecommons.org/licenses/by/4.0/ Atlantic cod individual spatial behaviour and stable isotope associations in a no‐take marine reserve © 2023 The Authors. Journal of Animal Ecology published by John Wiley & Sons Ltd on behalf of British Ecological Society. Published version Monk, Christopher T.; Power, Michael; Freitas, Carla; Harrison, Philip M.; Heupel, Michelle; Kuparinen, Anna; Moland, Even; Simpfendorfer, Colin; Villegas‐Ríos, David; Olsen, Esben M. Monk, C. T., Power, M., Freitas, C., Harrison, P. M., Heupel, M., Kuparinen, A., Moland, E., Simpfendorfer, C., Villegas‐Ríos, D., & Olsen, E. M. (2023). Atlantic cod individual spatial behaviour and stable isotope associations in a no‐take marine reserve. Journal of Animal Ecology, 92(12), 2333-2347. https://doi.org/10.1111/1365-2656.14014 2023 J Anim Ecol. 2023;00:1–15. | 1wileyonlinelibrary.com/journal/jane 1 | INTRODUCTION Foraging is a behavioural process consisting of multiple stages from searching to recognizing, attacking, capturing, handling and ingesting resources (Holling, 1966; Mangel & Clark, 1986). While foraging, individuals need to trade-off benefits of energy and nutrient consumption with associated costs, such as risks of being preyed upon (Lima & Dill, 1990), excess energy expenditure Received: 24 October 2022 | Accepted: 15 September 2023 DOI: 10.1111/1365-2656.14014 RESEARCH ARTICLE Atlantic cod individual spatial behaviour and stable isotope associations in a no-take marine reserve Christopher T. Monk1,2,3 | Michael Power4 | Carla Freitas1,5 | Philip M. Harrison6 | Michelle Heupel7 | Anna Kuparinen8 | Even Moland1,2 | Colin Simpfendorfer9 | David Villegas-Ríos10 | Esben M. Olsen1,2 1Institute of Marine Research, Flødevigen Marine Research Station, His, Norway; 2Centre for Coastal Research, Department of Natural Sciences, University of Agder, Kristiansand, Norway; 3GEOMAR Helmholtz Centre for Ocean Research Kiel, Kiel, Germany; 4Biology Department, University of Waterloo, Waterloo, Ontario, Canada; 5MARE, Marine and Environmental Sciences Center, Madeira Tecnopolo, Funchal, Madeira, Portugal; 6Department of Biology and Faculty of Forestry and Environmental Management, Canadian Rivers Institute, University of New Brunswick, Fredericton, New Brunswick, Canada; 7Integrated Marine Observing System (IMOS), University of Tasmania, Hobart, Tasmania, Australia; 8Department of Biological and Environmental Sciences, University of Jyväskylä, Jyväskylä, Finland; 9College of Science and Engineering, James Cook University, Townsville, Queensland, Australia and 10Instituto de Investigaciones Marinas, CSIC, Vigo, Pontevedra, Spain Correspondence Christopher T. Monk Email: [email protected] Funding information Academy of Finland, Grant/Award Number: 317495; Canadian Network for Research and Innovation in Machining Technology, Natural Sciences and Engineering Research Council of Canada; Consejo Superior de Investigaciones Científicas; H2020 European Research Council, Grant/Award Number: 770884; H2020 Marie Skłodowska-Curie Actions, Grant/Award Number: 793627; Norges Forskningsråd, Grant/Award Number: 294926; Spanish Ministry of Science and Innovation (Ramón y Cajal Program), Grant/Award Number: RYC2021-032594-I Handling Editor: Jean-Michel Gaillard Abstract 1. Foraging is a behavioural process and, therefore, individual behaviour and diet are theorized to covary. However, few comparisons of individual behaviour type and diet exist in the wild. 2. We tested whether behaviour type and diet covary in a protected population of Atlantic cod, Gadus morhua. 3. Working in a no-take marine reserve, we could collect data on natural behavioural variation and diet choice with minimal anthropogenic disturbance. We inferred behaviour using acoustic telemetry and diet from stable isotope compositions (expressed as δ13C and δ15N values). We further investigated whether behaviour and diet could have survival costs. 4. We found cod with shorter diel vertical migration distances fed at higher trophic levels. Cod δ13C and δ15N values scaled positively with body size. Neither behaviour nor diet predicted survival, indicating phenotypic diversity is maintained without survival costs for cod in a protected ecosystem. 5. The links between diet and diel vertical migration highlight that future work is needed to understand whether the shifts in this behaviour during environmental change (e.g. fishing or climate), could lead to trophic cascades. KEYWORDS acoustic telemetry, Atlantic cod, individual trophic niche, marine reserves, stable isotopes This is an open access article under the terms of the Creative Commons Attribution License, which permits use, distribution and reproduction in any medium, provided the original work is properly cited. © 2023 The Authors. Journal of Animal Ecology published by John Wiley & Sons Ltd on behalf of British Ecological Society. 2 | MONK et al. (O'Brien et al., 1989) and encountering unsuitable thermal habitats (Freitas et al., 2021). The incorporation of behavioural processes as a nonlinear functional response to model density-dependent consumer-resource interaction strengths (Holling, 1959) was a major advancement in predicting food-web dynamics (Berlow et al., 2004; Wootton & Emmerson, 2005). Given the nonlinear basis of food-web interactions, Jensen's inequality implies that population behavioural averages are insufficient measures to predict food web dynamics and individual variation must be accounted for (Bolnick et al., 2011). Indeed, a number of cases have been made for incorporating the role of the individual into food-web modelling (Bolnick et al., 2011; Des Roches et al., 2018; Gibert & DeLong, 2017; Kalinkat, 2014). Building a clear theoretical understanding for how individual behaviour fits into foodweb dynamics will support the prediction of trophic cascades driven by changes in behavioural diversity or, vice versa, changes in behavioural diversity driven by trophic cascades which could conceivably occur when consumption or being predated is consistently coupled with individual behaviour (Laskowski et al., 2022; Sommer & Schmitz, 2020). Consistently structured behavioural differences among individuals that emerge across time and contexts are observed in nearly every animal population (Bell et al., 2009), implying a degree of inflexibility for individual behaviour. Consistent among-individual differences in behavioural traits over time and across contexts are known as “animal personality” or “behaviour types” and have been described for well-studied behavioural axes such as, activity, boldness, sociability, exploration and aggression (Réale et al., 2007). Additionally, personality traits have also been described for other behaviours measured in the wild (Spiegel et al., 2017) such as homerange size and space-use (Villegas-Ríos et al., 2017) or migration timing (Birnie-Gauvin et al., 2021). Relatedly, “individual resource specialists”, that have a substantially narrower individual dietary niche than the population's niche, are frequently observed within overall generalist animal populations (Bolnick et al., 2003). In other words, individual foraging tactics where individuals primarily consume a subset of the prey items consumed by the entire population often coexist (Chavarie et al., 2021), potentially without fitness consequences (Woo et al., 2008). Despite the structured among-individual variation of diet and behaviour within populations, the links between individual resource specialization and personality are so far insufficiently studied. Toscano et al. (2016) have outlined several potential theoretical mechanisms on which testable hypotheses can be developed. For example, in relation to the “locomotor crossover” hypothesis (Huey & Pianka, 1981) active individuals might be expected to forage on less active prey types while inactive individuals might forage on active prey types. Moreover, bolder individuals may forage in riskier habitats (Klefoth et al., 2017), or dominate access to patches in social groups (Kurvers et al., 2010), and mobile and exploratory individuals may have a wider foraging space resulting in access to alternative prey options (Campos-Candela et al., 2018; Spiegel et al., 2017). Lastly, individual behavioural differences may be linked to underlying physiological differences, related to fast-slow life histories (Campos-Candela et al., 2018; Spiegel et al., 2017), which in turn may lead to differing foraging tactics (Nakayama et al., 2017). There is, therefore, a reasonable basis to expect that personality and diet could covary in the wild. Until recently it has been relatively challenging to extract a complete profile of both animal behaviour and diet specialization in the wild over the long term. Diet is often assessed by lethal methods where gut contents are identified and quantified, or by non-lethal methods such as gastric lavage, scat analysis, observations of predator–prey events, or quantifying associations in laboratory environments (Nielsen et al., 2018). However, prey items could easily be missed in the gut based on digestive variability or the timing of sampling, while observations of predation events or predatory behaviour provide only a snapshot of a complete diet, but not a measure of long-term dietary preferences. Fortunately, logging long-term, high-resolution animal behaviour in the wild is rapidly becoming available as biologging technologies advance, and transmitters become smaller and cheaper (Hussey et al., 2015; Kays et al., 2015; Nathan et al., 2022) and long-term diet can be assessed non-lethally in the wild using fatty acids (Budge et al., 2006) or ratios of stable isotopes (Carneiro et al., 2017; Crawford et al., 2008; McKechnie, 2004). The ratio of the stable isotopes of 15N to 14N (i.e. δ15N ‰) in organismal tissue can approximate the trophic position of an individual, while the ratio of the stable isotopes 13C to 12C (i.e. δ13C ‰) can approximate the source (e.g. pelagic, benthic, terrestrial) of primary production for consumed resources (Hobson, 1999; Peterson & Fry, 1987). Both 15N and 14N are taken up by primary producers in proportional amounts to the environmental ratio; however, 14N is preferentially excreted at a relatively predictable rate, enriching δ15N ‰ with consecutive consumption events. Accordingly, δ15N ‰ increases with trophic position. Similarly, 13C and 12C are taken up from the environment by primary producers, but 13C is preferentially used during carbon fixation at a rate dependent on the photosynthetic mechanism (Peterson & Fry, 1987). In coastal ecosystems, pelagic phytoplankton, benthic algae (e.g. brown algae, and diatom mats), eelgrass and terrestrial primary production have distinct δ13C ‰ signatures (Fry & Sherr, 1989). As carbon is transferred up trophic levels, δ13C ‰, remains relatively constant (fractionation rate of 1‰–2‰) (Barnes et al., 2007; Post, 2002) allowing for original sources of primary production to be traced through the food web. As tissues are rebuilt the isotopic ratios also change, reflecting the current diet. As rebuilding rates vary across tissues, by sampling the appropriate tissue one can assess an animal's diet at an approximate timepoint in the past (Peterson & Fry, 1987). The few studies to date comparing personality in the wild and diet using stable isotopes do not yield clear or consistent results. Woo et al. (2008) found that repeatable differences in Brünnich's guillemot (Uria lomvia) flight time, dive depth and dive shape were correlated with differences in prey selection, where foraging tactics did not result in differences in reproductive success. Exploration in captivity predicted resource acquisition (Ersoy 13652656, 0, Downloaded from https://besjournals.onlinelibrary.wiley.com/doi/10.1111/1365-2656.14014 by University Of Jyväskylä Library, Wiley Online Library on [18/10/2023]. See the Terms and Conditions (https://onlinelibrary.wiley.com/terms-and-conditions) on Wiley Online Library for rules of use; OA articles are governed by the applicable Creative Commons License | 3 MONK et al. et al., 2022) by red knots (Calidris canutus) and discovery (Herborn et al., 2010) by blue tits (Cyanistes caeruleus). In freshwater systems, Eurasian perch (Perca fluviatilis) resource consumption from pelagic pathways was higher for individuals that switched more frequently between active and inactive modes (Nakayama et al., 2017), while reliance on benthic pathways for burbot (Lota lota) correlated positively with activity at lower trophic levels and negatively with activity at higher trophic levels (Harrison et al., 2017) and bolder sticklebacks were found to forage in more littoral habitats (Theódórsson & Ólafsdóttir, 2022). In a marine coastal system, Dhellemmes et al. (2021) found that the relationship between individual juvenile lemonshark (Negaprion brevirostris) exploration and isotopic signatures of foraging in high-risk, exposed eel grass habitats fluctuated in direction and magnitude according to predation pressure. Gharnit et al. (2022) also found sex and resource availability specific relationships between exploration (measured in an open field test) and isotopic signatures in wild eastern chipmunks (Tamias striatus). In laboratory and pond environments Kerr and Ingram (2021), Schröder et al. (2016) and Laskowski et al. (2022) observed no impacts of behaviour type on diet, functional response or lower trophic levels, respectively. Hence, some connections between personality and diet have been observed, but no clear pattern has emerged and more effort is required to understand under which circumstances relationships between personality and diet exist, and what mechanisms lie behind the relationships. Both diet and behaviour are known to be commonly associated with body size, and therefore when investigating associations between personality and diet, body size must be accounted for, unless individuals are restricted to a single size-class. Body size is frequently found to relate to diet, particularly in gape limited predators, where larger individuals can consume larger prey, often at higher trophic levels (Gaeta et al., 2018; Segura et al., 2014). Furthermore, body size is correlated with ontogenetic niche shifts (Kimirei et al., 2013; Scharf et al., 2000), which often involve shifts in diet. Predation risk and metabolic demand both scale allometrically (Krause et al., 1998; Werner & Gilliam, 1984). Larger fish are at lower risk from predation and can therefore access alternative food sources located in previously risky habitats. Larger individuals also have greater absolute energy demand to support a higher resting metabolic rate, and therefore may switch to more energetically profitable food sources. Indeed, positive relationships between body size and diet as indicated by stable isotopes are commonly found in fish (McCormack et al., 2019). Similarly, as fish grow their behaviour is expected to shift as well. Larger individuals can escape predation risk, exploring larger areas (Fuiman & Magurran, 1994; Nilsson & Brönmark, 2000). Larger individuals also may also have greater energy stores allowing higher activity rates (Videler & Wardle, 1991). Moreover, larger fish potentially have altered social interactions, as their place in a size-based dominance hierarchy shifts (Ward et al., 2006). Hence, associations between diet and personality can be missed when body size is the dominant cause for variation in diet or behaviour, unless body size is accounted for. We focused our investigation in southern Norway on a coastal population of Atlantic cod, Gadus morhua, a predatory fish species in North Atlantic marine coastal ecosystems. Cod behaviour has been intensively studied given the species' commercial and cultural importance (Meager et al., 2018) and are known to exhibit consistent behavioural differences along a reactive-proactive spectrum connected to spatial behaviour (home range size) in the wild (Villegas-Ríos et al., 2018). Cod exhibit inter-individual variation in feeding tactics, specializing in specific prey types (e.g. sand eel, flatfishes or crustaceans) (Funk et al., 2021; Hüssy et al., 2016) and, therefore, among-individual differences in diet are likely to be observed. Our main objectives were to test whether these among-in- dividual differences in diet are associated with or independent of cod spatial behaviour and if associations were present, whether they may have fitness impacts in terms of survival. Despite being intensely studied, prey selection in cod is poorly understood, in particular with respect to individual variation (Meager et al., 2018). The cod in the coasts of southern Norway have been observed to have very small home range sizes and remain close to steep rocky edges, making rare excursions into the pelagic (Freitas et al., 2021; Villegas-Ríos et al., 2021). The cod also typically move, on a daily basis, from deeper waters in the daytime to shallower waters at night time during which they are assumed to be foraging (Espeland et al., 2010; Freitas et al., 2021). Foraging cod near the rocky edges of fjords are likely consuming small gobies (Gobius niger, Gobiusculus flavescens, Aphia minuta), sculpins (Myoxocephalus Scorpius), amphipods, polychaetes, crustaceans (Carcinus maenas) and wrasses (Ctenolabrus rupestris, Symphodus melops), while cod foraging in the pelagic may consume for example, sprat, herring and other gadids (Pollachius pollachius, Merlangius merlangus, Pollachius virens, Trisopterus minutus) (Hop et al., 1993). In this study we combined high-resolution acoustic telemetry (Nathan et al., 2022) and stable isotope techniques to investigate the relationships between individual behaviour and diet for wild cod in an established no-take marine reserve, with an unexploited food web and cod population. We hypothesized that more active cod with larger activity spaces in shallower habitats and greater diel vertical migration would spend less time foraging on lower trophic organisms such as small gobies and amphipods in benthic habitats and would therefore have a have a more negative δ13C value and a higher δ15N value relative to their counterparts, where an activity space is a spatially and temporally defined area in which an individual engages in normal activities. In addition to investigating fundamental ecological principles relating to foraging and behaviour, we also hypothesized based on the findings of Woo et al. (2008) that diet would not have an impact on individual survival during the tracking period. We focused on activity and activity space size, as these behaviours are directly related to the foraging activity and spatial aspects of foraging mechanisms linking personality and diet specialization, as outlined by Toscano et al. (2016). We also considered depth and diel vertical migration distance, because the vertical dimension is an important component of the cod's environment in our study system (Freitas et al., 2015, 2021). 13652656, 0, Downloaded from https://besjournals.onlinelibrary.wiley.com/doi/10.1111/1365-2656.14014 by University Of Jyväskylä Library, Wiley Online Library on [18/10/2023]. See the Terms and Conditions (https://onlinelibrary.wiley.com/terms-and-conditions) on Wiley Online Library for rules of use; OA articles are governed by the applicable Creative Commons License 4 | MONK et al. 2 | MATERIALS AND METHODS 2.1 | Study region and acoustic telemetry system Our study was conducted in the Tvedestrand Fjord (52.60282°N, 8.948268°E), located on the Skagerrak Sea's coast in southern Norway (Figure 1; See the Appendix for a full description of the fjord). Importantly, a large portion of the fjord (150 ha) has been designated as a no-take marine reserve since June 2012, where all forms of fishing and trapping are prohibited, in addition to two buffer areas where fixed fishing gear is forbidden (Figure 1). A portion of the cod population is resident in the fjord, thought to remain inside the study area for its lifetime (Villegas-Ríos et al., 2021), and thus are probably largely protected from harvest, though harvest remained possible in the buffer areas outside the no-take area. The study site is therefore advantageous for studying long-term natural cod behaviour in the wild. Tvedestrand Fjord has been equipped with a high-resolution Innovasea VPS telemetry system since June 2013. The median system accuracy is 1.8–4.4 m (see Freitas et al., 2016). The original VPS system consisted of 31 VR2W receivers (Innovasea, Halifax, Canada) distributed within the no-take marine reserve. In 2018, the VPS system was expanded to buffer zones located north and south of the no-take reserve, to a total of 55 receivers. Each receiver is fixed at 3 m depth, with moorings and subsurface buoys. The receivers record the time-of-arrival data from unique ultrasonic signals emitted from transmitters surgically implanted into the body cavities of fish. Based on discrepancies in signal arrival times among receivers, the fish positions within the fjord can be calculated by multilateration. Additional transmitters (V16-4X) fastened below each receiver act as “synctags” to correct for clock drift among receivers and four “reference” transmitters (V13-1x & V13T-1x) were placed at fixed locations within the array to measure system performance. The addition of a calibrated pressure sensor inside fish transmitters enabled the addition of a third dimension for each position. 2.2 | Cod sampling Cod were sampled between May 2018 and December 2020 during five sampling events (spring 2018, spring 2019, fall 2019, spring FIGURE 1 Map of the study area showing the location of the 55 acoustic receivers in Tvedestrand Fjord and the location of the full-protected marine reserve and its buffer areas. 13652656, 0, Downloaded from https://besjournals.onlinelibrary.wiley.com/doi/10.1111/1365-2656.14014 by University Of Jyväskylä Library, Wiley Online Library on [18/10/2023]. See the Terms and Conditions (https://onlinelibrary.wiley.com/terms-and-conditions) on Wiley Online Library for rules of use; OA articles are governed by the applicable Creative Commons License | 5 MONK et al. 2020 and fall 2020, see Figure A1). Cod were captured by fykenets distributed along the fjord's coastline between 1 and 10 m depth and soaked for 1–3 days. The captured cods' fork lengths (n = 184) were measured, and a 5 mm dermal punch was used to sample the dorsal muscle for stable isotope analysis. Tissue samples were stored at −25°C until processing. A subset of cod (n = 97) were taken to the nearest shore for sampling and transmitter implantation and anaesthetised with a 9:1 EtOH: clove oil solution added to water at 0.4 mL L−1. Transmitters (V13P; Innovasea Systems, Halifax, Canada; weight in water: 5.5 g; estimated battery life: 1285 days; power output: 147 dB) were surgically implanted into the body cavity of these cod, and incisions were closed with two adsorbable sutures. The acoustic transmitters had a random transmission interval with a mean delay of 180 s (range of 130– 230 s) and a built-in pressure sensor (maximum depth: 68 m; resolution 0.3 m; accuracy: 3.4 m) for transmission of current depth as well as the identity code. Each cod received an external T-bar anchor tag (TBA-2, 30 × 2 mm, Hallprint Pty. Ltd, Holden Hill, South Australia) parallel to the anterior dorsal fin. All cod were released at the capture site immediately after recovery from anaesthesia. All cod sampling and tagging was approved by the Norwegian Food Safety Authority (permit 19556). 2.3 | Estimation of behavioural traits Cod detection data were downloaded from acoustic receivers twice per year. Fine-scale cod positions were subsequently estimated by Innovasea using the VEMCO positioning system and we based our estimation of behavioural traits and mortality on these fine-scale positions. Prior to behavioural analysis, the trajectories of each cod were inspected and subset to filter out unlikely positions (positions with a VPS HPE ≥34 were removed (Freitas et al., 2016)), positions that appeared on land and positions after natural mortality (e.g. that belonged to a preyed cod) to ensure only the live cod behaviour was analysed and used to measure survival. We determined each cod's fate following the methods described by Villegas-Ríos et al. (2020), where cod were classified as surviving, dispersing, or succumbing to natural mortality or harvest, based on the inspection of threedimensional positions. Natural mortality was assigned when the vertical and horizontal movements stabilize at the same time, harvesting was assigned when detections stop before the end of the transmitter battery life and the last detections were not at the edge of the study system (in which case dispersal is assigned), predation was assigned when a clear change in movement pattern suddenly occurs that resembles known behaviours of a cod predator (e.g. a harbour seal, Phoca vitulina), and survival was assigned when the cod displayed the expected vertical and horizontal behaviour throughout the battery life of the transmitter. We estimated four behavioural traits (activity, activity-space size, depth and diel vertical migration distance) on a daily timescale for each cod throughout the entire study period. Days were considered to be between consecutive sunrises (based on the NOAA sunrise calculator (www. esrl. noaa. gov/ gmd/ grad/ solca lc/ ) for all behaviours. Activity was calculated by summing the three-dimen- sional Euclidean distances between consecutive positions along each trajectory. Euclidean distances below 4.4 m were excluded as they could not be distinguished from the telemetry system error (Freitas et al., 2016) and therefore we could not be reasonably certain that the cod was moving. The 2d activity space size of the cod was calculated using the 95% kernel utilization density (kud) area of the cod positions. The activity space sizes were calculated in R using the adehabitatHR package (Calenge, 2011). The daily individual activity space size was only calculated if more than 30 positions were available. Depth was assessed based on the pressure sensor in the acoustic transmitters and summarized as the mean depth. Diel vertical migration distance was calculated as the difference between the mean depth after sunset and the mean depth before sunset. Note that in the case of diel vertical migration distance negative values indicate cod were shallower at night and deeper during daytime. We excluded measures of diel vertical migration distance between March 1 and May 1 each year, as this is the likely spawning period during which diel vertical migration is not a typical behavioural pattern (Espeland et al., 2007). 2.4 | Stable isotopes analysis We used dorsal muscle tissue obtained from muscle biopsies known to have no lasting effects on fish (Bøe et al., 2020). Muscle tissue in cod has been found to change very little 90 days after dietary changes (Ankjærø et al., 2012) and, therefore we can safely interpret the δ13C and δ15N as representing around 3 months of past consumption. As we have only single measures for δ13C and δ15N we considered our isotope measurements to represent individual trophic niche (Dhellemmes et al., 2021; Hussey et al., 2017). The cod dorsal muscle biopsies were processed at the Environmental Isotope Laboratory at University of Waterloo in Canada. The samples were dehydrated at 50°C for 24 h and then crushed into a powder with a mortar and pestle and weighed in tin capsules (Tin Capsules Pressed Standard Weight 5 × 3.5 mm, Elemental Microanalysis Ltd., Okehampton, UK) using an analytical balance (XP205 DeltaRange, Mettler-Toledo GmbH, Greifensee, Switzerland). The stable isotope values from each sample were measured using a Delta Plus Continuous Flow Stable Isotope Ratio Mass Spectrometer (Thermo Finnigan, Bremen, Germany) coupled to a 4010 Elemental Analyser (CNSO 4010, Costech Analytical Technologies Inc., Valencia, USA), with results expressed in delta notation as δ13C and δ15N, relative to the international standards Vienna Pee Dee Belemnite for δ13C (Craig, 1957) and atmospheric nitrogen for δ15N (Mariotti, 1983). Measurement repeatability was assessed by reweighing every tenth sample. Analytical precision (±0.2‰ for δ13C and ±0.3‰ δ15N) was assessed using internal laboratory standards cross-calibrated against the International Atomic Energy Agency standards CH3 and CH6 for carbon and N1 and N2 for nitrogen. The C:N ratios were close to three (3.23 ± 0.12 13652656, 0, Downloaded from https://besjournals.onlinelibrary.wiley.com/doi/10.1111/1365-2656.14014 by University Of Jyväskylä Library, Wiley Online Library on [18/10/2023]. See the Terms and Conditions (https://onlinelibrary.wiley.com/terms-and-conditions) on Wiley Online Library for rules of use; OA articles are governed by the applicable Creative Commons License 6 | MONK et al. mean ± SD, 8 samples >3.5), and therefore we did not conduct any lipid extraction. We also did not observe any correlation between C:N ratios and δ13C values (Figure A2. cor = −0.07, t = −0.96, df = 182, p = 0.34, Pearson's product–moment correlation), and therefore did not correct δ13C values. 2.5 | Stable isotopes context Our assessment that δ15N values are indicative of trophic position relies on several assumptions, namely that the baseline δ15N values in our study system are not heterogeneous across benthic and pelagic modes of primary production, and δ15N values are evenly distributed in space and time. Further, our assessment that δ13C values are indicative of primary production source requires verification that benthic and pelagic primary production can indeed be distinguished at lower trophic levels. We therefore opportunistically collected stable isotopes from lower trophic organisms in the Fjord. We collected seven Ascophyllum nodosum samples in January 2020 and 6 Ascophyllum nodosum, 10 Cerastoderma edule, 6 filamentous green algae, 20 Littorina spp. and 20 Mytilus edulis samples in June 2020 from the study system and measured stable isotopes from these samples using the same methods as we did for cod dorsal muscle samples. We tested for differences in δ15N values between Ascophyllum nodosum and filamentous green algae samples using a t-test and visually inspected all samples for homogeneity of δ15N values across δ13C values in our samples. We also inspected the spatial distribution of δ15N values from these samples as runoff points from farmland or sewage may inflate δ15N value baselines at nearby sample sites (Baker et al., 2017). Further, we inspected for spatial trends along latitude and longitude and temporal trends of δ15N and δ13C values from our cod muscle samples. 2.6 | Statistical analysis We assessed the relationship between fate and individual trophic niche using a cox-proportional hazards model fit to Kaplan– Meier survival curves using the “survival” package (Therneau & Lumley, 2015) in R. In this model we used right censored (survived or dead) survival time. Cod that survived the entire transmitter battery life or dispersed from the study area were considered survivors and right-censored on their last detection day, whereas cod that died from either natural reasons or harvesting were considered dead. The model covariates, δ13C and δ15N median activity, median activity space size, median depth and median diurnal migration distance, were standardized (scaled and centred) for comparability of effect sizes. We confirmed that the proportional hazards assumption was met using the “cox.zph” function within the survival package. To carry over variation in daily behavioural measurements we used multivariate linear mixed effects models to estimate the correlations among cod body size, behaviour and diet (Houslay & Wilson, 2017). Models were fit with Bayesian inference, using the MCMCglmm package in R (Hadfield, 2010). We fit one multivariate model where the sextivariate response included daily measures of individual behaviours (activity, activity space size, depth, diel vertical migration) and single measures of individual δ13C δ15N values. As fixed effects we included cod fork length at the level of all response variables, sampling event fit at the level of both isotope response variables and solar elevation (scaled and centred) and fit at the level of the four behaviour response variables. This was done to control for the effect of body size on behaviour and diet, and sampling period on baseline isotope values and seasonal effects on baseline behaviour respectively. We have assumed that shifts in baseline values of δ15N is homogenous across δ13C values and vice versa. The solar elevation at midnight each day was calculated in degrees above the horizon using the R package oce (Kelley et al., 2022) based on Michalsky (1988). To understand the associations among behaviour and isotope values, we fit an unstructured covariance matrix in the random effects using ID as the grouping variable. For each model we ran 550,000 iterations, with a burn-in of 50,000 and a thinning interval of 175. The model used an uninformative Inverse-Wishart prior (R1: V = diag(6) + 0.002, nu = 6 + 0.002, R2: V = diag(4), nu = 4 + 0.002). To account for the fact that we had repeated measures of behaviours, but did not have repeated measures for body size, δ13C and δ15N, we followed the recommendations of Thomson et al. (2017) and Dingemanse et al. (2021) setting “covu = TRUE” in the residual structure of the prior to allow the sets of residuals to covary with the random effects. Activity and activity space size were log10 transformed to meet normality assumptions, and all response variables were scaled and centred to allow for comparable effect sizes. Model assumption violations were assessed by examining trace plots, Geweke diagnostics and posterior autocorrelation. Correlations between the behaviour traits and isotope measurements were estimated from the median and 95% credible intervals (highest probability density intervals) of the posterior distribution of among-individual covariance between behaviour and the isotope value divided by the product of the square root of each trait's among-individual variances (i.e. the product of each trait's among-individual standard deviation) (Houslay & Wilson, 2017). Significance was assessed by whether the 95% credible interval of the correlation's posterior distribution overlapped zero. As the stable isotope data represents diet prior to sampling and our behavioural data is collected after sampling, an assumption of our models is that behaviour and diet remain stable. We therefore also estimated adjusted repeatability (r), to assess how consistently different the cod behaviours in our study system were. Accordingly, from the multivariate mixed effects model we estimated a posterior distribution for adjusted repeatability by dividing each posterior draw of among-individual behavioural variance by the sum of within-individual variance and among-individual behavioural variance. We then evaluated the 95% credible interval (highest probability density interval) of the posterior of adjusted repeatability. 13652656, 0, Downloaded from https://besjournals.onlinelibrary.wiley.com/doi/10.1111/1365-2656.14014 by University Of Jyväskylä Library, Wiley Online Library on [18/10/2023]. See the Terms and Conditions (https://onlinelibrary.wiley.com/terms-and-conditions) on Wiley Online Library for rules of use; OA articles are governed by the applicable Creative Commons License | 7 MONK et al. 3 | RESULTS The 184 sampled cod (165 unique individuals) ranged between 16 and 72 cm (mean ± SD: 44 ± 13 cm) in fork length. The δ13C values ranged between −19.90‰ and −16.38‰ (mean ± SD: −18.11‰ ± 0.78‰) indicating consumption of varying mixes of prey from both pelagic and benthic pathways in the cod population in Tvedestrand Fjord. The δ15N values ranged over approximately two trophic levels between 12.45‰ and 18.97‰ (mean ± SD: 14.58‰ ± 0.99‰) indicating varying reliance on fish and other higher trophic level prey. Of the 97 cod fit with acoustic transmitters, six left the study system and were assumed to have survived until the dispersal date, 23 died from natural mortality, 25 were consumed by predators, 15 were harvested, 26 remained and survived through the tracking period, one individual died from tagging mortality and one fate could not be determined. One individual, at the northern study system edge, was never detected by a sufficient number of receivers to generate any fine-scale positions. Accordingly, 94 cod generated behavioural data ranging from 5 to 901 days of tracking per individual (median = 179 days). The cod swam per day 362 [0, 14,395] m (median [minimum, maximum]) (mean ± SD), and used an activity space of 0.39 [0.17, 22.0] ha (median [minimum, maximum]) each day (Figure 2). The cod were 11.2 [0, 46.9] m ha (median [minimum, maximum]) deep during the tracking period and the cod migrated -0.3 [-28.8 (descent), 38.8 (ascent)] m ha (median [minimum, maximum]) from night to day (Figure 2). No detectable relationships occurred between individual survival and δ13C, δ15N or behaviour (Table 1). Activity (r = 0.66, 0.58–0.72; mode, 95% CI), activity space size (r = 0.67, 0.59–0.74; mode, 95% CI) and depth (r = 0.53, 0.45–0.62; mode, 95% CI) were repeatable behaviours over the study period, but diel vertical migration (r = 0.17, 0.12–0.24; mode, 95% CI) had lower repeatability. Of the 94 cod generating behavioural data, 78 could be included in our multivariate mixed effects model. This was because behavioural estimates for all four behaviours were required on a daily basis, and for individuals with few detections on a specific day, activity space could not be estimated. Therefore some FIGURE 2 Daily population median values (black points) and standard deviations (polygons) of four cod behaviours (n = 94 fish) based on acoustic telemetry data from the Tvedestrand Fjord. Behaviour from only one individual was collected until fall 2017 (see Figure A1) and therefore no standard deviation is shown. Negative diel vertical migration distance indicates cod were deeper during the day and shallower at night. 13652656, 0, Downloaded from https://besjournals.onlinelibrary.wiley.com/doi/10.1111/1365-2656.14014 by University Of Jyväskylä Library, Wiley Online Library on [18/10/2023]. See the Terms and Conditions (https://onlinelibrary.wiley.com/terms-and-conditions) on Wiley Online Library for rules of use; OA articles are governed by the applicable Creative Commons License 8 | MONK et al. individuals only present in the study system for a shorter time period and with poorer detectability did not have days with all behaviours estimated simultaneously. We did not detect any differences between δ15N values across benthic and pelagic sources of primary production (Figure 3). A t-test comparing δ15N values of filamentous green algae and Ascophyllum nodosum was not significant (t = −1.78, df = 8.95, p = 0.11). Further, we did not observe any locations enriched in δ15N from our samples of baseline and lower trophic organisms (Figure A2) and we did not observe any spatial patterns in δ15N or δ13C values in our cod isotope samples (Figure A3). However, there was a decrease in δ15N values between 2018 and 2020 sampling dates in our cod samples (Figure A4). We detected a positive relationship between cod diel vertical migration distance and δ15N values (Figure 4, Figure A5). As the cod typically exhibited a negative diel vertical migration, moving from deep water during the day to shallow water at night, a positive relationship with distance values indicates that cod exhibiting shorter diel vertical migrations consumed resources from higher trophic levels. We did not detect any other statistically significant correlations between behaviour and δ13C or δ15N as shown by the posterior distributions of the correlations (Figure 4, Figures A6 and A7). Fork length positively predicted δ13C and δ15N values, indicating that cod consume prey from increasingly benthic pathways, and/or higher trophic levels as they become larger (Table A1, Figure 5). We also found some significant effects of sampling period on isotope values (Table A1, fixed effects), where δ15N ‰ was greater in spring 2018 than in other sampling periods. With increasing solar elevation (i.e. in summer) cod increased their activity and depth and reduced their diel vertical migration distance (Table A1, fixed effects). 4 | DISCUSSION As foraging is a behavioural process, among-individual variation in diet must have a basis in behavioural differences. We found some support for our first hypothesis that more active cod with larger activity spaces in shallower habitats and greater diel vertical migration distance would have a have a more negative δ13C value and a higher δ15N value relative to their counterparts, as cod with a shorter diel vertical migration distance consumed prey with a higher contribution from higher trophic levels, suggesting alternative foraging tactics. Our second hypothesis, that diet would not have an impact on individual survival during the tracking period, was also supported as we did not observe any dietary fitness consequences, in terms of fate, suggesting that in our study system prey choices have equivalent costs for survival, or that cod may compensate for any mortality related costs in prey choice. As repeatable cod behaviour was related to diet as measured by stable isotopes, there is a possibility for potential trophic cascades via alterations to the behavioural composition of the population, for example through selective fishing (Audzijonyte et al., 2013; Hočevar & Kuparinen, 2021; Kindsvater & Palkovacs, 2017). The fact that behavioural variation is consistent and structured among individuals in animal populations has been firmly established (Bell et al., 2009; Dall et al., 2012). However, the ecological consequences of such consistent among-individual variation, i.e. animal personality, remain poorly understood (Toscano et al., 2016; Wolf & Weissing, 2012). The connection between repeatable individual differences in cod diel vertical migration distance and δ15N underscores that consistent differences in individual behaviour can have linkages to food web ecology and signifies that changes in frequencies of behaviour types within populations could have community level impacts (Wolf & Weissing, 2012). To fully understand the trophic consequences is beyond the scope of this study, as we do not know the specific prey types consumed and we do not know how changes in cod density would affect foraging patterns. Diel vertical migration is known to be an important cod behavioural pattern (Freitas et al., 2015; Olsen et al., 2012). For coastal cod it may also be a key behaviour underlying diet and body size scaling relationships, as we found diel vertical migration distance was related to both body size and the trophic level of consumed prey. The diel vertical migration patterns we observed TABLE 1 Coefficient estimates, hazard ratios, coefficient standard errors (SE), z values and p values for cox proportional hazard models assessing cod survival after tracking with acoustic telemetry. Model coefficients were standardized for comparability. Coefficient Hazard ratio SE (coef) z value p value δ13C ‰ −0.088 0.916 0.153 −0.578 0.563 δ15N ‰ −0.176 0.838 0.159 −1.108 0.268 Activity 0.281 1.325 0.167 1.680 0.093 Activity space size 0.206 1.228 0.147 1.400 0.161 Depth 0.203 1.225 0.130 1.553 0.120 Diel vertical migration distance 0.085 1.088 0.171 0.495 0.621 FIGURE 3 The δ13C and δ15N values for the 184 cod samples as well as five species indicative of the isotopic baseline in the study system. 13652656, 0, Downloaded from https://besjournals.onlinelibrary.wiley.com/doi/10.1111/1365-2656.14014 by University Of Jyväskylä Library, Wiley Online Library on [18/10/2023]. See the Terms and Conditions (https://onlinelibrary.wiley.com/terms-and-conditions) on Wiley Online Library for rules of use; OA articles are governed by the applicable Creative Commons License | 15 MONK et al. (b, c) show the same measurements along longitude and latitude respectively. Figure A3. An overview of all cod sampling locations in Tvedestrand Fjord (Panel a). Panel b shows the longitude and latitude of δ13C values and δ15N values used to assess spatial biases in isotope values. Figure A4. Cod isotope values (δ13C and δ15N) and fork lengths across sampling periods. Figure A5. The relationship between diel vertical migration distance and δ15N scaled relative to cod body size. Red points represent individual median values. Figure A6. The relationship between activity, activity space size, depth and diel vertical migration distance, and δ13C. Black points represent daily measurements, and red points represent individual median values. Figure A7. The relationship between activity, activity space size, depth and diel vertical migration distance, and δ15N. Black points represent daily measurements, and red points represent individual median values. How to cite this article: Monk, C. T., Power, M., Freitas, C., Harrison, P. M., Heupel, M., Kuparinen, A., Moland, E., Simpfendorfer, C., Villegas-Ríos, D., & Olsen, E. M. (2023). Atlantic cod individual spatial behaviour and stable isotope associations in a no-take marine reserve. Journal of Animal Ecology, 00, 1–15. https://doi.org/10.1111/1365-2656.14014 13652656, 0, Downloaded from https://besjournals.onlinelibrary.wiley.com/doi/10.1111/1365-2656.14014 by University Of Jyväskylä Library, Wiley Online Library on [18/10/2023]. See the Terms and Conditions (https://onlinelibrary.wiley.com/terms-and-conditions) on Wiley Online Library for rules of use; OA articles are governed by the applicable Creative Commons License