Modelling of Atlantic salmon (Salmo salar L.) behaviour in sea-cages: A Lagrangian approach
Introduction
The importance of aquaculture production of marine finfish as a source of marine protein for human consumption is steadily increasing. Since its introduction to aquaculture in the 1970s, Atlantic salmon has become an important species in sea-cage mariculture with more than 1.3 million tons produced in 2006 (FAO, 2008). Improving production of salmon is a constant focus of the industry, both to reduce costs and reduce environmental impacts. The diverse physical and biological influences experienced by caged salmon ultimately affect growth rates and welfare (Juell and Westerberg, 1993, Fernö et al., 1995, Huntingford, 2004). Models that predict the behaviour of fish under particular environmental conditions may therefore be useful tools in managing production.
Fish in captivity are subject to a range of varying environmental conditions. High densities of fish (up to 25 kg m− 3) within the confined space of sea-cages which are typically between 10,000 and 20,000 m3 in volume at Norwegian sea-farms (Sunde et al., 2003) means that social interactions between fish may have a strong influence on their behaviour. Traditionally two different kinds of modelling paradigms have been used to portray biological phenomena. Lagrangian (individual-based) models replicate each individual as an instance of the same model, though often with different parameters yielding a heterogeneous population (e.g. Reuter and Breckling, 1994, Kunz and Hemelrijk, 2003). In contrast, Eulerian models represent populations by differential equations and scalar fields rather than as distinct individuals (e.g. Flierl et al., 1999).
Typical environmental conditions caged salmon are exposed to include food, light, temperature, salinity and oxygen levels, which may vary over short (minutes, hours) and long (days, seasons) time scales (Juell, 1995). Addition of food greatly affects salmon behaviour, with a switch in swimming behaviour, speed and depth within cages (Andrew et al., 2002, Juell and Fosseidengen, 2004). Light levels influence swimming depths of salmon; avoidance of high light intensity has been observed and is probably an inherent predator-avoidance strategy (Fernö et al., 1995). Salmon seem to prefer a particular range of ambient temperatures (Oppedal et al., 2007), resulting in changes in prevailing swimming depth when subjected to temperatures outside their preferred range in a thermally stratified water column. Excessively high or low temperatures may be physically harmful (Saunders et al., 1975, Wilkie et al., 1997). Oxygen and salinity levels have recently been demonstrated to vary greatly within sea-cages at different depths and different times, however, whether salmon within cages actively respond to varying oxygen and salinity levels remains unknown (Johansson et al., 2006).
The social aspect of salmon behaviour is probably stronger in sea-cages than in the wild due to the unnaturally high densities. Caged salmon tend to maintain a certain distance to neighbouring fish and the cage itself (Juell, 1995). During non-feeding periods, salmon generally exhibit a uni-directional circular swimming pattern, tracing the inner perimeter of the cage wall (Fernö et al., 1995).
The main goal of this study was to simulate vertical migration patterns shown by caged populations of Atlantic salmon since vertical movement is of interest both in terms of fish welfare and cage management strategies (e.g. depth of artificial light sources, feeding regimes). We therefore built a Lagrangian model of fish behaviour based on quantitative data and qualitative knowledge on the behaviour of Atlantic salmon in Norwegian fish farms from the literature and then verified the model against three different scenarios from Johansson et al. (2006). The individual-based approach was chosen since it is better suited for modelling small-scale behaviours of individuals. Furthermore, this approach enables a greater degree of individual variation and provides model outputs on both individual and population levels.
Section snippets
Materials and methods
We developed the model as a stand-alone application using the programming language JAVA, focusing on keeping the model as flexible as possible so that it would be easy to facilitate the future addition of more environmental factors and behavioural patterns. Furthermore, the object-oriented nature of JAVA is very suitable for developing individual-based models. The model description follows the ODD (Overview, Design concepts, Detail) protocol for presenting individual based models proposed by
Individual behaviour
Movement trajectories of fish in the simulations to assess social behaviours in the absence of environmental influences indicated that initially randomly distributed fish grouped together and exhibited a circular swimming pattern within the sea-cage (Fig. 2). The trajectories in the first interval (Fig. 2a) revealed relatively unstructured and chaotic behaviours. Fish zigzagged across the cage and no clear cases of individuals swimming in parallel were seen.
Compared to the first period, the
Individual behaviour
Our results indicate that the set of simple behavioural rules represented by the response toward other individuals, cage avoidance and the stochastic component were sufficient to stimulate an initially randomly distributed population of simulated fish to eventually start following the schooling patterns well-known for caged populations of salmon (Juell and Westerberg, 1993). Furthermore, the fish swam in circular patterns along the inner perimeter of the cage towards the end of the simulation.
Acknowledgements
We thank the late Dr. Jon-Erik Juell (Institute of Marine Research) whose research formed the basis of much of the model and who provided advice on aspects of the model’s biological content. Furthermore, we thank Dag Slagstad for advice on designing the environmental model and Jens Glad Balchen for his pioneering work on modelling fish behaviour in the seventies which gave us a starting point from which to formulate our models. Funding was provided by SINTEF Fisheries and Aquaculture through
References (32)
- et al.
Improved growth and nutrient utilisation in Atlantic salmon (Salmo salar) fed diets containing a bacterial protein meal
Aquaculture
(2006) - et al.
Dynamic modelling of pellet distribution in Atlantic salmon (Salmo salar) cages
Aquac. Eng.
(2004) Modeling, prediction and control of fish behavior
Control Dyn. Syst.
(1979)- et al.
Behaviour and growth of Atlantic salmon (Atlantic salmon) subjected to short-term submergence in sea-cages
Aquaculture
(2008) - et al.
Vertical distribution of Atlantic salmon (Salmo salar) in net pens: trade-off between surface light avoidance and food attraction
Aquaculture
(1995) - et al.
From individuals to aggregations: the interplay between behavior and physics
J. Theor. Biol.
(1999) - et al.
A standard protocol for describing individual-based and agent-based models
Ecol. Model.
(2006) - et al.
Effect of environmental factors on swimming depth preferences of Atlantic salmon (Salmo salar) and temporal and spatial variations in oxygen levels in sea cages at a fjord site
Aquaculture
(2006) - et al.
Use of artificial light to control swimming depth and fish density of Atlantic salmon (Salmo salar) in production cages
Aquaculture
(2004) - et al.
An ultrasonic telemetric system for automatic positioning of individual fish used to track Atlantic salmon (Salmo salar) in a sea cage
Aquac. Eng.
(1993)