Stocking density and welfare of cage farmed Atlantic salmon: application of a multivariate analysis
Introduction
There is a trend towards increased concern for the welfare of animals at all levels where they come into contact with humans and recently this concern has expanded to include the welfare of fish. Welfare is a complex concept, this complexity being reflected in the diversity of definitions (Dawkins, 1998, FSBI, 2002).
Whichever definition is used, and for welfare to have any real meaning, the animal concerned must have the capacity for suffering and in the case of fish this is controversial (Rose, 2002). However, recent evidence suggests that external stressors and painful stimuli elicit aversive states in fish, as they do in birds and mammals (Sneddon et al., 2003, Braithwaite and Huntingford, 2004, Chandroo et al., 2004), even though these may differ in degree from those experienced by higher vertebrates. In any event, a wide range of organisations in the United Kingdom and Europe now have fish welfare on their agenda including the Department of the Environment, Food and Rural Affairs, the Scottish Executive, Environment and Rural Affairs Department, the Royal Society for the Prevention of Cruelty to Animals, the Scottish Society for the Prevention of Cruelty to Animals, the Fisheries Society of the British Isles and the Council of Europe. The Farm Animal Welfare Council report (Farm Animal Welfare Council, 1996) made a number of recommendations regarding the welfare of fish under production conditions, but also identified areas where scientific investigation was necessary to provide the information upon which to base welfare guidelines and legislation. Stocking density was identified as one such area, since it is often suggested that farmed fish are held at higher densities than prevail in the wild. However, it would appear that some species of fish elect to shoal at very high densities in the wild (Trevorrow and Claytor, 1998) although estimating the density of wild fish is problematic unless the population has low density and uniform distribution (Parkinson et al., 1994).
Many studies have demonstrated an effect of stocking density on various aspects of the welfare of farmed fish (Wedermeyer, 1997), though the results depend on the species concerned; for example, Arctic charr (Salvelinus alpinus) suffer less physical damage and grow more rapidly at high density (Jorgensen et al., 1993), whereas sea bass (Dicentrarchus labrax, Vazzana et al., 2002) and gilthead seabream (Sparus auratus, Montero et al., 1999) show evidence of reduced welfare at high densities. Studies of the relationship between welfare and stocking density are further complicated by interactions with other variables such as food availability (Robel and Fisher, 1999) or water quality (Ellis et al., 2002). The measurement of stocking density in production cages is in itself difficult for a number of reasons. For example, estimating biomass requires accurate information on growth and survival and converting biomass to density requires knowledge of the volume of space available to the fish, which is constantly changing due to deformation of the net. Furthermore, the fish do not occupy all of the available space (Juell and Fosseidengen, 2004).
To date, there have been no other published studies of the effects of stocking density on welfare in marine production cages. Soderberg et al. (1993) reviewed stocking densities for the “satisfactory growth and health” of juvenile Atlantic salmon (Salmo salar) and found acceptable welfare at a wide range of densities. In the absence of robust empirical data, the Farm Animal Welfare Council (1996) recommended a maximum stocking density of 15 kg m−3 based on the current practice and expert opinion, but called for further research.
Mellor and Stafford (2001) argued that welfare should be improved in an achievable incremental manner rather than aspiring to an unachievable ideal. It is only possible to improve welfare if it can be measured or assessed, but despite recent progress in the evaluation of welfare in terrestrial animals (Spoolder et al., 2003), there is no current consensus on the best way to objectively measure welfare in fish. It has, so far, been difficult to apply systems for assessing welfare in terrestrial animals to fish; this has been partly due to the problems of observing fish especially in production systems. Previous attempts to assess fish welfare have largely concentrated on measurement of individual aspects of welfare such as growth, stress or damage. Such parameters have been used as proximate indicators of welfare in fish (Etscheidt, 1995), but, because they are influenced by factors other than welfare, they are often imprecise or noisy indicators. One possible solution is to use the statistical tool of multivariate analysis to combine simultaneously recorded measures of welfare, on the basis of the observed statistical relationships among them (Manly, 1994). Principal components analysis (PCA) allows trends in multivariate data to be reduced to scores that can then be used as dependent variables in subsequent analysis, without requiring a priori judgements of the relative value of individual variables. Such scores offer the possibility of providing an integrated and objective reflection of diverse measured welfare indicators. This approach to the analysis of complex data sets is only suitable for the research context but may be a very powerful approach for the identification of robust but simple on-farm welfare assessment methods.
This study was initially set up as a balanced design, with replicated cages at three stocking densities. However, although we were successful in generating a range of randomly allocated stocking densities extending beyond those commonly used in salmon aquaculture, the constraints associated with a commercial production site made it impossible to maintain the required densities with sufficient accuracy. We therefore used stocking density as a continuous rather than categorical, independent variable and related this (and other possible influential factors) to the welfare score.
Our specific aims were:
- 1.
To relate a welfare score in Atlantic salmon, derived by multivariate analysis of four commonly used measures of fish welfare, to assessment by farm workers of the health status of their fish.
- 2.
To use this score to examine the relationship between welfare and stocking density in farmed Atlantic salmon.
- 3.
To examine additional effects of time from the start of the study, cage position and time since potentially stressful husbandry event (grading or lice treatment) on welfare.
Section snippets
Subjects and husbandry
The study was conducted in commercial on-growing cages in Loch Duich on the West Coast of Scotland between January and October 2000. The group of cages consisted of 30 cages (estimated 15 m2 and 7 or 8 m deep), of which 24 were stocked. These cages were at the northwesterly end of the loch and comprised two adjacent rows of 15 cages located parallel to the shore and orientated with the long axis of the loch (15 “near-shore” and 15 “off-shore”) (Fig. 1). The fish were of Marine Harvest L70
Results
The first component produced by the PCA was biologically consistent with welfare and accounted for 29.9% of the variability in the data. The additional principal components did not have any biological interpretation relevant to welfare and did not warrant further analysis according to the scree test (Cattell, 1966).
Measuring welfare
The complex nature of welfare is reflected in many physiological and behavioural aspects of an animal, so to evaluate welfare it is necessary to obtain information from a wide range of sources. Although single measures of welfare, such as plasma cortisol, can give useful information about fish status with respect to a single biological system, by integrating a number of relevant measures into a reduced number of values, multivariate analysis potentially provides a useful and important tool for
Acknowledgements
We would like to thank LINK Aquaculture (Project SAL 18), NERC and Scottish Quality Salmon who funded this project, Marine Harvest, who additionally provided the study cages and stock and all the workers at the Loch Duich site for their help and support. Thanks also to the referees for informed, constructive criticism.
References (35)
- et al.
Can fish suffer?: perspectives on sentience, pain, fear and stress
Appl. Anim. Behav. Sci.
(2004) - et al.
Effects of cortisol and stress on the immune system in Atlantic salmon (Salmo salar L.)
Fish Shellfish Immunol.
(1996) - et al.
Effects of stocking density on food intake, growth performance and oxygen consumption in Arctic charr (Salvelinus alpinus)
Aquaculture
(1993) - 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.
Alternative competitive strategies in juvenile Atlantic salmon (Salmo salar): evidence from fin damage
Aquaculture
(2000) - et al.
A comparison of plasma glucose and plasma cortisol as selection markers for high and low stress-responsiveness in female rainbow trout
Aquaculture
(1999) - et al.
Attack site and resultant damage during aggressive encounters in Atlantic salmon (Salmo salar) parr
Aquaculture
(1998) - et al.
Confinement stress in seabass (Dicentrarchus labrax) depresses peritoneal leukocyte cytotoxicity
Aquaculture
(2002) - et al.
Cortisol-induced increase in susceptibility of Atlantic salmon, Salmo salar, to Vibrio salmonicida, together with effects on the blood cell pattern
Aquaculture
(1989) - et al.
Stress hormones and the cellular stress response in salmonids
Fish Physiol. Biochem.
(2000)