Reviews in Aquaculture (2013) 5, 33–57 doi: 10.1111/j.1753-5131.2012.01083.

Salmon Welfare Index Model (SWIM 1.0): a semantic model

for overall welfare assessment of caged Atlantic salmon:
review of the selected welfare indicators and model
presentation
Lars H. Stien1, Marc B. M. Bracke2, Ole Folkedal1, Jonatan Nilsson1, Frode Oppedal1,
Thomas Torgersen1, Silje Kittilsen3, Paul J. Midtlyng3, Marco A. Vindas4, Øyvind Øverli4 and
Tore S. Kristiansen1
1 Institute of Marine Research, Bergen, Norway
2 Wageningen UR Livestock Research, Lelystad, The Netherlands
3 Norwegian School of Veterinary Science, Oslo, Norway
4 Norwegian University of Life Sciences, Ås, Norway

Correspondence Abstract
Lars H. Stien, Institute of Marine Research,
PO Box 1870 Nordnes, 5817 Bergen, Norway. A semantic model for overall welfare assessment of Atlantic salmon reared in
Email: lars.stien@imr.no sea cages is presented. The model, called SWIM 1.0, is designed to enable fish
farmers to make a formal and standardized assessment of fish welfare using a
Received 5 March 2012; accepted 20 June set of selected welfare indicators. In order to cover all welfare relevant aspects
2012.
from the animals’ point of view and to create a science-based tool we first
identified the known welfare needs of Atlantic salmon in sea cages and
searched the literature for feasible welfare indicators. The framework of seman-
tic modelling was used to perform a structured literature review and an evalua-
tion of each indicator. The selected indicators were water temperature, salinity,
oxygen saturation, water current, stocking density, lighting, disturbance, daily
mortality rate, appetite, sea lice infestation ratio, condition factor, emaciation
state, vertebral deformation, maturation stage, smoltification state, fin condi-
tion and skin condition. Selection criteria for the indicators were that they
should be practical and measureable on the farm, that each indicator could be
divided into levels from good to poor welfare backed up by relevant scientific
literature. To estimate each indicator’s relative impact on welfare, all the indi-
cators were weighted based on their respective literature reviews and according
to weighting factors defined as part of the semantic modelling framework. This
was ultimately amalgamated into an overall model that calculates welfare
indexes for salmon in sea cages. More importantly, the model identifies how
each indicator contributes (negatively and positively) to the overall index and
hence which welfare needs are compromised or fulfilled.
Key words: animal welfare score, aquaculture, diagnostic, scientific literature, sea cage.

and national projects related to fish welfare have been or

Introduction
are being performed. These use a range of approaches
The problem of how to assess the welfare status of fish is from studies of fish behaviour to microarrays and a num-
an ongoing debate and no consensus has been reached on ber of welfare related indicators have been suggested, but
definitions or assessment methodology (Ashley 2007; without an integrating model and theory, much confusion
Huntingford & Kadri 2008; Segner et al. 2012). However, remains as to how the indicators can be scored, weighted
food and aquaculture authorities ask for methods that and integrated into an overall welfare assessment (OWA).
can be used to assess fish welfare and thus check the fulf- To gain the best possible assessment, an OWA model
ilment of laws and regulations. A number of EU projects should be based on observations of the animals, their

ª 2013 Wiley Publishing Asia Pty Ltd 33

L. H. Stien et al.

biological and physical environments, and the available animal welfare. This includes scientific descriptions of
scientific knowledge (Bracke et al. 1999b; Anon 2001), housing systems in terms of both environment based and
and the selected welfare indicators (WIs) should be spe- animal based measures, and how these affect animal wel-
cies specific, validated, reliable, feasible and auditable fare. Semantic modelling was first applied to assess hous-
(EFSA 2009). There are two closely related approaches for ing systems for dry sows (Bracke et al. 2002a,b), but it
creating OWA models; risk analysis (EFSA 2006a,b; has also been applied to assess overall welfare in laying
Bracke et al. 2008) and semantic modelling (Bracke et al. hens (De Mol et al. 2004, 2006; Shimmura et al. 2011),
2002a,b). The prime objectives of risk analysis are to for tail biting in pigs (Bracke et al. 2004a,b), for enrich-
identify hazards, their consequences and probabilities of ment materials for pigs (Bracke et al. 2007a,b; Bracke
occurrence, and to find critical control points in the pro- 2008), in dairy cattle (Ursinus et al. 2009) and for wal-
duction process to avoid welfare risks, e.g. stress, injuries, lowing in pigs (Bracke 2011; Bracke & Spoolder 2011).
disease and mortality. Semantic modelling follows a prin- In view of the ongoing debate about fish welfare, it is
cipally different approach, focusing on welfare defined as necessary to clarify definitions and underlying assump-
the quality of life as perceived by the animals themselves tions that the semantic modelling of animal welfare rests
and is searching for indicators of the degree of fulfilment on: Welfare is here defined as ‘the quality of life as per-
of the animal’s welfare needs and the effects on the ani- ceived by the animals themselves’, and the ability to expe-
mals’ wellbeing. Since semantic modelling considers both rience welfare is seen as part of the emotional monitoring
positive and negative aspects of welfare it is a risk–benefit system that guides animals (with advanced central ner-
analysis (Bracke et al. 1999a,b,c, 2008). vous systems) in getting what they need and avoiding
This paper describes a first attempt to apply semantic harm and dangers in an effective way. In order to survive
modelling to review commonly used WIs for farmed an animal must fulfil its basic needs; e.g. nutrition, respi-
Atlantic salmon (Salmo salar L.) and to propose a science ration, thermoregulation etc., and to this end, animals
based model and tool for OWA in the sea cage phase. continuously assess their state of need. The qualitative
Atlantic salmon is chosen as the case species given its welfare experience is created by the reward and punish-
great importance in aquaculture and since there is a rea- ment systems in the emotional brain, and involves experi-
sonable amount of scientific knowledge available. The ence, memories and re-evaluation of needs in anticipation
model is named SWIM 1.0, an acronym for Salmon Wel- of physiological, psychological and behavioural require-
fare Index Model, where no. 1 states that it is the farmer’s ments (Berridge 2004; Panksepp 2005; Korte et al. 2007).
version and .0 states that this is the pilot version which There is growing evidence that teleost fish, and hence sal-
may be revised and upgraded later. A web application mon, can feel pain and that they possess functional equiv-
(http://www.imr.no/swim) was constructed in order to alents of the limbic and dopaminergic nervous systems –
facilitate author collaboration when updating the model’s systems that are linked with emotion, memory, spatial
scientific database (statements from the literature) and relationships, primary consciousness, reward, cost–benefit
the model itself. The web application will also support estimation and decision-making (Sneddon 2003; Brai-
updating the model with results from future research, thwaite & Huntingford 2004; Chandroo et al. 2004a,b;
such that SWIM will be a dynamic and up-to-date model. Håstein et al. 2005; Braithwaite & Boulcott 2007; Broom
The model is primarily intended as a tool for fish farmers 2007; Galhardo & Oliveira 2009; Braithwaite 2010; Tor-
to assess fish welfare in sea cages, but will be expanded gersen et al. 2011). In short, there are strong indications
with WIs that can be measured by farm veterinarians that also fish are able to experience states of welfare.
(SWIM 2) and fish welfare experts (SWIM 3). For use by Based on this we assume that salmon experience a
fish farmers it is important that the WIs are limited in continuum of welfare states, which may vary from very
number, feasible and practical to use. The indicators poor to excellent and that are closely related to the
employed in the current version and their weightings in degree of fulfilment of the salmons’ welfare needs, i.e.
the model may change in future versions as knowledge of needs monitored by the emotional brain. An OWA
different WIs expands. should be in accordance with the needs-assessment per-
formed by the animals themselves. However, since we
cannot tap directly into the animal brain, we must assess
The semantic modelling concept
their state of need and emotionality based on observa-
The semantic modelling concept for the purpose of for- tions of the animals and what we know about the way
malized assessment of animal welfare was first introduced they respond to a variety of environmental conditions.
by Bracke et al. (1999a,b,c), and is based on the meaning This implies using scientific knowledge about animal
(semantics) of available scientific information about the physiology and behaviour to surmise their welfare state
animals’ welfare needs and how these are related to (Bracke et al. 1999c).

Reviews in Aquaculture (2013) 5, 33–57

34 ª 2013 Wiley Publishing Asia Pty Ltd
 Salmon Welfare Index Model (SWIM 1.0)

Welfare relevant needs of farmed Atlantic salmon Welfare indicator literature review, ranking of
levels and weighting
We used a slightly modified version of the semantic
modelling procedure described in Bracke et al. (2002b). The next step of the semantic modelling procedure is to
First, based on the list of needs presented in Bracke collect relevant scientific statements, obtained from a sys-
et al. 1999c we formulated a list of known welfare needs tematic literature review (Bracke et al. 2002b). In this
for Atlantic salmon in sea cages (Table 1). The physical review we used the selection criterion that the statements
welfare needs include respiration, osmotic balance, nutri- are relevant to assess the fulfilment of needs of Atlantic
tion, good health and thermoregulation. Behavioural salmon kept in sea cages. Sources include ISI Web of
welfare needs describe motivations to perform specific KnowledgeSM, Google ScholarTM and various books and
behaviours to get an immediate reward or for which the reports on the topic. As far as possible the statements are
mere performance is rewarding and are behaviours that species specific and for the post-smolt sea water adapted
have evolved to fulfil more ultimate goals related to sur- life stage of Atlantic salmon. Based on the review the WIs
vival, growth or reproduction (Jensen & Toates 1993). were scaled on at least two levels from best to worst.
For Atlantic salmon in sea cages we include behaviour According to semantic modelling these levels must be
control, feeding, safety, protection, social contact, explo- mutually exclusive and cover the model’s domain, i.e. in
ration, kinesis, rest, sexual behaviour and body care. To our case on-growing of Atlantic salmon in sea cages. Each
avoid confusion, we must emphasize that the distinction WI-level must also be linked to at least one scientific
between physical and behavioural needs, and also the statement that provides the scientific basis of the weight-
distinctions between needs, is not absolute and that ing of the model: Firstly, the levels are ranked within each
overlaps exist. WI to create indicator scores (IS):

NLi  RLi;j
Linking of welfare indicators to welfare needs ISi;j ¼ ð1Þ
NLi  1
In order to cover all welfare relevant aspects from the
animals’ point of view we searched the literature for fea- where ISi,j is the score of the j-th level of the i-th WI in the
sible welfare indicators suggestive of the fulfilment of model, NLi is the total number of levels of indicator i and
the welfare needs and 17 WIs were selected for inclusion RLi,j is the rank number of level j. Next, the scientific evi-
in the SWIM 1.0 model. All the WIs were linked to at dence is used to assign weighting scores (WS) using
least one of the needs and all welfare needs are linked weighting categories (WC) (Table 3). This is a somewhat
to at least one WI (Table 2). This was done to make subjective, but systematic, scoring based on an assessment
sure that all the indicators concern the degree of fulfil- of the intensity, duration and incidence of the welfare
ment of the welfare needs and that the assessment cov- impact as implied by each scientific statement that has been
ers all welfare relevant needs from the salmons’ point of linked to the WI. The WC’s classify welfare performance
view. criteria, e.g. pain, illness and reduced survival (Table 2).

Table 1 List of Atlantic salmons basic needs, adapted from Bracke et al. (1999c)

Need Explanation and relevance for salmon

Physical needs Respiration Uptake of oxygen and release of carbon dioxide by pumping water over the gills
Osmotic balance Maintaining homeostasis of body cell fluids
Nutrition Intake of food containing the required energy, amino acids, minerals, vitamins etc.
Health Absence of disease, illness and malfunction
Thermal regulation Optimization of metabolism and temperature, including thermal comfort
Behavioural needs Behaviour control Ability of the fish to freely position themselves (including regulation of buoyancy)
and respond to stimuli
Feeding Regular access to food
Safety Possibility to avoid perceived danger
Protection Possibility to keep the body undamaged from physical injury
Social contact Predictable interaction with conspecifics
Exploration Possibility to search for resources and information
Kinesis Being able to swim (physical activity)
Rest Possibility of reducing activity level or ‘sleep’
Sexual behaviour Homeward migration, breeding behaviour, spawning, etc.
Body care Scratching, parasite cleaning, etc.

Reviews in Aquaculture (2013) 5, 33–57

ª 2013 Wiley Publishing Asia Pty Ltd 35
L. H. Stien et al.

Table 2 The most significant links between the selected welfare indicators and the welfare needs of Atlantic salmon in sea cages

Osmotic Thermal Behaviour Social Sexual Body

Needs Respiration balance Nutrition Health regulation control Feeding Safety Protection contact Exploration Kinesis Rest behaviour care

Welfare indicator
Temperature * *
Salinity * *
Oxygen *
Water current * * *
Stocking density * * * *
Lighting * * *
Disturbances * *
Daily mortality *
Appetite * * * *
Sea lice * * *
Condition factor * * *
Emaciation state * * * * *
Sexual maturity * *
stage
Smoltification * *
state
Vertebral * *
deformation
Fin condition * * * *
Skin condition * * *

The weighting factor (WF) of each welfare indicator i in WS is given in parenthesis behind its respective WC. The
the model was subsequently calculated as proposed by De WIs and WCs have been given capital first letters in these
Mol et al. (2006): paragraphs for easy recognition. This is done in detail for
! ! the first WI, i.e. the temperature-indicator, but only for the
X X best and worst level for the remaining indicators. The WSs
WFi ¼ maxðWSwcl Þ  minðWSwcl Þ are expert opinions based on the reviews, but the reader is
wc ILbest;i wc ILworst;i
free to challenge these decisions.
ð2Þ
where ILbest,i is the best indicator level and ILworst,i is the
Temperature (C)
worst indicator level of the i-th welfare indicator, WSwcl
is the weighting score assigned to the indicator level based Temperature governs the metabolic rate of salmon, and
on the scientific statements; wc identifies the weighting thereby acts as a controlling and limiting factor together
categories linked to the indicator level. A special case is with oxygen for the fishes’ physiological performance
made up of WI-levels that are so detrimental for welfare including their capacity for dealing with stressors. The rele-
that welfare is poor (minimum), no matter which levels vance of water temperature as a welfare indicator is evident
are selected for the other indicators. These levels are from tolerance limits and temperature preferences of
called knockout levels, and if present the overall welfare Atlantic salmon in sea cages. A temperature preference in
index (OWI) is defined as 0. Knockout levels are not temperature stratified conditions in sea cages of about
included when calculating WFs. 17C is suggested by Johansson et al. (2006, 2009), which
As much as possible each indicator was reviewed as correspond well with the finding that the Atlantic salmons’
stand alone, i.e. if an indicator level has an effect on selected temperature in a horizontal temperature gradient
another indicator the resulting change in fish welfare is increased with acclimation (5–20C), showing a final pref-
attributed to the second indicator and not the first. As an erence at about 17C (Javaid & Anderson 1967). In the
example, high stocking densities may lead to poor oxygen available range between 11 and 20C, caged Atlantic sal-
levels if the water in the cage is not sufficiently replenished. mon individuals and groups clearly avoided water warmer
The low oxygen level has a direct effect on the fish and this than 18C as well as water colder than 12C (Johansson
is hence the primary WI in this specific example. Each sec- et al. 2006, 2009; Oppedal et al. 2011a,b). The temperature
tion below reviews a WI, and each review section includes tolerance is highly dependent on fish acclimatization states,
a ranking and weighting paragraph. For each weighting the and in general Atlantic salmon can adapt to a range from 0

Reviews in Aquaculture (2013) 5, 33–57

36 ª 2013 Wiley Publishing Asia Pty Ltd
 Salmon Welfare Index Model (SWIM 1.0)

Table 3 Weighting categories used in the weighting procedure of semantic modelling with brief descriptions and ranges of weighting scores
(WSs). Adapted from Bracke et al. (2002b)

Weighting category Brief description Range of WS

HPI Evidence of activation of the HPI (hypothalamic pituitary interrenal) axis indicative of stress )5 to )1
Illness Evidence of health problems, including increased mortality, but excluding skin lesions, fin damage and )5 to )1
abnormalities in body shape (see ‘pain’)
Pain Evidence of pain including skin and fin damage )5 to )1
Reduced survival Evidence of reduced survival related to physiological requirements (other than through specific health )5 to )1
problems), e.g. longevity, deprivation of food, poor
environment
Abnormal behaviour Evidence of disturbed behaviour and or apathy )3 to )1
Aggression Evidence of aggression such as bite marks and attacks )3 to )1
Avoidance Evidence of avoiding stimuli (which are perceived as dangerous ⁄ noxious) )3 to )1
Frustration Evidence of blocked behaviour or deprivation )3 to )1
Negative performance Evidence of decreased performance (that is likely to indicate negative affect), )3 to )1
including (re)production effects, but excluding specific survival aspects related to physiological
necessities, HPI-activation and illness
SAM Evidence of SAM (sympathetic adrenal medullary) activation )3 to )1
(indicative of negative affect), e.g. increased heart rate and (nor)adrenalin levels
Demand Evidence that the fish are willing to spend effort to obtain food or other recourses 1 to 5
Natural behaviour Evidence of (potential positive reward from) behaviour as seen in (semi) natural conditions 1 to 3
Positive performance Evidence of healthy, fit fish 1 to 3
Preference Evidence of choosing one resource over another (e.g. in a preference test) 1 to 3

to 20–23C provided sufficient oxygen levels and gradual seasonal range Atlantic salmon experience in sea cages.
transitions between temperatures are applied (Priede 2002; Atlantic salmon have Positive performance (3) and show
EFSA 2008). An Icelandic stock of Atlantic salmon survived Preference (2) for level 1: 10–15C. 7–10C is ranked as
1 month with water temperatures <0C before mortalities level 2 since Performance and Preference is less compared
started to occur at )1.4C (Skuladottir et al. 1990). On the with level 1. 16–17C ranks as level 3 since here the sal-
opposite end of the scale Goncalves et al. (2006) observed mon is susceptible to harm, but above 3–6C as level 4
increased mortality already at temperatures slightly above since salmon prefer the third to the fourth level. Very
18C in the case of full-strength seawater, and Hevrøy et al. high (‡18C) and low temperatures (£2C) are associated
(2011) found more than 50% reduction in feed intake, with avoidance ()2), negative performance ()3), illness
growth and feed utilization after 2 weeks at 19C compared ()3) and reduced survival ()3) giving a total WS of )11
with salmon at 14C. This shows that the margins are small for level 5. Very high and low temperatures can be lethal
between temperatures that salmon seem to prefer and what if they persist for a long time. Level 6 is therefore a
may be harmful to them (with exponential effects occur- knockout level. Finally, Equation 2 gives a weighting
ring in the upper range). Comparing Atlantic salmon factor of 16 (Eqn 2: WF = (3 + 2))()2)3)3)3), Table 4)
reared at 6, 10, 14 and 18C for 12 weeks following transfer for the temperature WI.
to seawater, Handeland et al. (2008) found that growth
rate, feed intake, feed conversion efficiency (FCE) and
Salinity
stomach evacuation rate were significantly influenced by
temperature and fish size. The highest growth rate was seen During the smoltification process salmon develop tolerance
in the 14C group (1.53% d)1). No differences in growth for brackish and seawater salinity. Adult, non-migratory
were found between the 10 and 18C groups (1.35% d)1 Atlantic salmon is little affected by salinity (Bakke et al.
vs. 1.29% d)1), and lowest growth rates were observed for 1991; Johansson et al. 2006, 2009), unless damage to the
the 6C group (0.78% d)1). However, in a recent study, skin and disease impair their osmoregulatory ability (Grim-
16C induced a long-term reduced growth rate compared nes & Jakobsen 1996; Boxaspen 2006). Mature salmon have
with 10C following vaccination (Grini et al. 2011). altered osmoregulation in adaptation to a hypo-osmotic
Based on this review we propose that the temperature environment before re-entering freshwater in nature (Pers-
WI can be divided into six levels, which can be ranked son et al. 1998) and may therefore experience osmoregula-
for welfare as follows: (1) 10–15C, (2) 7–10C, (3) 16– tory challenges in high salinities. Small salmon display a
17C, (4) 3–6C, (5) £2, ‡18, short term and (6) £2, ‡18, preference for the halocline (Oppedal et al. 2011a) and
long term. These are temperatures within the normal may benefit from access to brackish water (Handeland

Reviews in Aquaculture (2013) 5, 33–57

ª 2013 Wiley Publishing Asia Pty Ltd 37
L. H. Stien et al.

Table 4 Welfare indicators (WI) with levels from best to worst, the associated indicator level score (IS), the sum of the weighting scores assigned
to the best and worst level and the calculated weighting factor (WF), see Eqn 2. Levels with indicator score K are knockout levels, i.e. levels that
result in severely reduced welfare regardless of other WIs

WI # Levels IS S WF

Environment Sea cage Temperature (C) 1 10–15 1.00 5 16

2 7–10 0.75
3 16–17 0.50
4 3–6 0.25
5 £2, ‡18, short term 0.00 )11
6 £2, ‡18, long term K
Salinity 1 Access to brackish water 1.00 1 3
2 Adult fish with no access to brackish water 0.50
3 Small post smolts, maturing or clearly 0.00 )2
impaired fish with no access to brackish
water
Oxygen (%) 1 >80%, all temperatures 1.00 1 17
2 70–80% for warm water (18C),
60–80% (12C),
50–80% cold water (6C) 0.50
3 60–70% for warm water (18C),
40–60% (12C),
30–50% cold water (6C) 0.00 )8
4 <60% for warm water (18C),
<40% (12C),
<30% cold water (6C) K
Water current (BL s)1) 1 <0.9 1.00 1 3
2 0.9 – Ucrit, 0.00 )2
3 ‡Ucrit K
Stocking density (kg m)3) 1 <22 1.00 1 8
2 22–26 0.66
3 26–32 0.33
4 >32 0.00 )7
Lighting 1 Optimal 1.00 2 4
2 Suboptimal 0.00 )2
Disturbances 1 None 1.00 1 11
2 Light 0.67
3 Moderate 0.33
4 Severe 0.00 )10
Animal Mortality (% day)1) 1 At or below 10 percentile curve 1.00 3 21
2 Below benchmark curve 0.75
3 At the benchmark curve 0.50
4 Above the benchmark curve 0.25
5 At or above the 90 percentile curve 0.00 )18
6 At or above the 90 percentile curve, long term K
Appetite 1 Good appetite 1.00 6 11
2 As expected 0.50
3 Poor appetite 0.00 )5
Individual fish Sea lice 1 No lice 1.00 1 11
2 Light infestation 0.66
3 ‡0.05 pre-adult or adult lice cm)2 fish 0.33
4 ‡0.08 pre-adult or adult lice cm)2 fish 0.00 )10
5 ‡0.12 pre-adult or adult lice cm)2 fish K
Condition factor 1 >1.1 1.00 3 6
2 0.9–1.1 0.50
3 <0.9 0.00 )3
Emaciation state 1 Not emaciated 1.00 1 16
2 Potentially emaciated 0.00 )15
3 Distinctly emaciated K

Reviews in Aquaculture (2013) 5, 33–57

38 ª 2013 Wiley Publishing Asia Pty Ltd
 Salmon Welfare Index Model (SWIM 1.0)

Table 4 (Continued)

WI # Levels IS S WF

Vertebral deformation 1 No external signs of vertebral deformities 1.00 1 10

2 ‘Short-tail’ of normal weight 0.50
3 ‘Short-tail’ of low weight. 0.00 )9
Sexual maturity stage 1 Not mature 1.00 1 9
2 Precocious male 0.66
3 Mature male 0.33
4 Mature female 0.00 )8
Smoltification state 1 Fully smoltified 1.00 1 9
2 Parr, access to brackish water 0.75
3 Parr, incomplete smoltification, 10C 0.50
4 Parr, incomplete smoltification, 14C 0.25
5 Parr, incomplete smoltification, 7C 0.0 )8
6 Parr, incomplete smoltification, 20C K
Fin condition 1 Normal healthy fins, nothing to comment 1.00 3 13
2 Scar tissue or slight necrosis 0.66
3 Moderate current skin damage and ⁄ or necrosis 0.33
including splitting and ⁄ or thickening
4 Severe skin damage and ⁄ or necrosis with bleeding 0.00 )10
and ⁄ or inflammation and ⁄ or exposed fin
rays and severe tissue loss
Skin condition 1 Normal healthy skin, nothing to comment 1.00 1 15
2 Scar tissue, healed 0.80
3 Scale loss (dislocated or missing scales) 0.60
4 Superficial wound or ulcer <1 cm2 0.40
5 Superficial wound or ulcer >1 cm2 0.20
6 Penetrating and ⁄ or multiple wounds or 0.00 )14
ulcers possibly infected
7 Large open wounds, life threatening K

et al. 1998) as osmoregulation is relatively costly for them.

Oxygen saturation (%)
Swimming in brackish water may also help the salmon to
avoid sea lice infestation (Lepeophtheirus salmonis) (Hevrøy For this welfare indicator it is necessary to first explain
et al. 2003; Plantalech Manel-La et al. 2009) as the infec- why we use oxygen saturation (%) and not oxygen
tious larvae of sea lice do not tolerate low salinities (Bric- concentration (mg L)1). These measures are of course
knell et al. 2006). Salinity has been suggested as a factor related, but as oxygen solubility decreases with tempera-
regulating swimming depth in adult salmon, but current ture and salinity, the oxygen concentration correspond-
evidence suggests that salinity is unimportant in determin- ing to any level of oxygen saturation varies. Both
ing vertical distributions in immature fully smoltified sea- concentration and saturation are meaningful metrics of
water-transferred Atlantic salmon (Johansson et al. 2006, available oxygen in the water. Any oxygen that is to be
2007; Oppedal et al. 2011a). utilized by fish tissues must be extracted from the
To conclude, there is little evidence that salinity levels water ventilated by the fish over its gills, and at any
have significant effects on the welfare of adult Atlantic given saturation, cooler and less saline water contains
salmon in sea cages. We do, however, suggest three levels more oxygen. However, the diffusion gradient of oxy-
for the Salinity WI: (1) Access to brackish water, (2) gen over the gills depends on oxygen saturation of the
Adult fish with no access to brackish water (in a sea cage water, and at any given concentration of oxygen, the
containing 10–400 000 individuals it is likely that some higher saturation in warmer, more saline water aids
fish have compromised osmotic balance) and (3) Poorly oxygen uptake over the gills. Also, a considerable
smoltified, maturing or impaired fish with no access to strength of using saturation as the operational welfare
brackish water. These fish (level 3) will show Preference indicator is how intuitive inferences can be drawn from
(1) for brackish water, and otherwise have negative per- such readings without any knowledge about tempera-
formance ()1) and reduced survival ()1). In accordance ture, salinity and solubility: 80% oxygen tells us that
with limited evidence for strong effects on fish welfare the the fish is offered 80% of what is found in pure water
calculated WF is only 3 (Eqn 2, Table 4). at equilibrium with air.

Reviews in Aquaculture (2013) 5, 33–57

ª 2013 Wiley Publishing Asia Pty Ltd 39
L. H. Stien et al.

Stevens et al. (1998) found that the routine oxygen ids such as faeces and excess feed (EFSA 2008; MacIntyre
uptake of juvenile Atlantic salmon in freshwater at 12– et al. 2008). The swimming capacity of Atlantic salmon
13C was not limited by water oxygen saturations above depends on factors such as body size and metabolic scope
38%. This is confirmed in recent studies in sea water (Grøttum & Sigholt 1998). Observations from sea cages
(reviewed in Oppedal et al. 2011a) showing that at 18, 12 show that during daytime salmon cruise at 0.3–0.9 BL s)1
and 6C 400 g salmon post-smolt are not able to maintain (Juell 1995; Dempster et al. 2008, 2009; Korsøen et al.
routine metabolic rates below approximately 60%, 40% 2009), while they typically slow down during darkness to
and 30% saturation, respectively. Below these thresholds 0–0.4 BL s)1 (Korsøen et al. 2009). Salmon reared in race-
mortality will commence in farmed salmon if oxygen levels ways with a fixed current (28 cm s)1) for 8 months prior
are not improved. The difference between the routine and to harvest showed nearly 40% higher weight gain compared
the maximum metabolic rate (the maximum theoretically with control fish farmed in ordinary cages (Totland et al.
possible oxygen uptake under the present conditions) acts 1987). Intensity of exercise has been found positively corre-
as a buffer against factors such as stress, disease and feed- lated with disease resistance (Takle et al. 2010) and
ing, which narrow this metabolic scope (e.g. Helfman et al. improved cardio-vascular capacities (Jørgensen & Jobling
1997; Priede 2002). Salmon will therefore migrate vertically 1994; Davison 1997). Although water current typically is
in sea cages to avoid hypoxic zones (Oppedal et al. 2011a). measured as m s)1, in regard to fish welfare it makes more
A summary from several hypoxia trials (WEALTH 2008) sense to measure it as BL s)1. High currents can drive small
concluded that immune responses are reduced at levels salmon (400–800 g) to exhaustion already at 1.6–2.2
below 55% oxygen saturation, and Sundh et al. (2010) BL s)1 (McKenzie et al. 1998; Deitch et al. 2006), although
found that the intestinal function was clearly disturbed at a some can manage 3.0 BL s)1(Lijalad & Powell 2009). We
level of 50% for salmon kept at both 9C and 16C. Fur- were unable to find data on larger Atlantic salmon, but
thermore, studies with full-feeding Atlantic salmon held in studies in Sockeye salmon (Oncorhynchus nerka) indicate a
seawater at 16C and given fluctuating oxygen levels from critical swimming speed Ucrit of about 1.35 BL s)1 for lar-
90 to 70% showed reduced appetite, fluctuating from 90 to ger salmonids (Steinhausen et al. 2008). It should be noted
60% also initiated acute anaerobic metabolism and that the above studies using swimming tunnels were per-
increased skin lesions, fluctuations from 90 to 50% addi- formed on starved fish and that fully fed, commercial fish
tionally initiated acute stress responses, reduced feed con- probably have lower thresholds due to less available scope
version and growth, and fluctuations from 90 to 40% for activity.
additionally caused impaired osmoregulation and mortali- In conclusion, the water flow through sea cages must be
ties (Remen et al. 2012). Moderate environmental hypoxia sufficient to secure replenishment of oxygen. While satura-
also has an effect. Crampton et al. (2003) and Bergheim tion with oxygen per se is a separate WI, water currents also
et al. (2006) found that salmon displayed reduced growth affect swimming speeds of the fish. We suggest dividing the
at 75% oxygen in 9C water and at 85% in 15C water, water current WI into three levels: At level 1 (<0.9 BL s)1)
respectively, compared with fish kept at 100% oxygen. This currents provide exercise and give positive performance
high sensitivity of growth rate to oxygen availability sug- (1), at level 2 (0.9 – Ucrit) welfare may be reduced, and
gests that even modest reductions in oxygen saturation when the water velocity is so high that it exceeds critical
may start causing welfare problems. swimming speed (Ucrit) then water flow may even be lethal
Based on this review we suggest that oxygen levels for the fish (knock-out, level 3). We were not able to find
above 80% do not cause welfare problems for salmon in any literature about swimming speeds between the comfort
sea cages, but instead are associated with Positive perfor- zone and the Ucrits (level 2), but it is reasonable to assume
mance (3). We divide the dissolved oxygen (DO) WI into that forced swimming leads to loss of control and
four level combinations of oxygen saturation and temper- hence frustration ()2) over time. It is also reasonable to
ature (Table 4), including one knockout level. The worst assume that Ucrit in addition to size depends on the state of
level, excluding the knockout is set to: 60–70% and the fish, for instance how adapted it is to high water cur-
18C, 40–60% and 12C or 30–50% and 6C. This rents. The farmer must, in other words, know the ability of
level is associated with avoidance ()3), negative perfor- the fish or use a Ucrit of 1.3 for safe margins. In accordance
mance ()3), illness ()3) and reduced survival ()5). This with scant evidence for the direct effects of water current
gives a total WF of 17 (Eqn 2, Table 4). on fish welfare we get a WF of only 3 (Eqn 2, Table 4).

Water current (measured as body lengths per second) Stocking density (kg m)3)
The water flow through a sea cage replenishes oxygen used Stocking density, defined as the total biomass of the fish
by the fish and flushes out metabolites and suspended sol- divided by the sea cage volume, is typically used by

Reviews in Aquaculture (2013) 5, 33–57

40 ª 2013 Wiley Publishing Asia Pty Ltd
 Salmon Welfare Index Model (SWIM 1.0)

authorities to set upper limits for what is allowed in sea

Lighting
cages (e.g. 25 kg m)3 in Norway). Despite its frequent use
as a production parameter there are relatively few studies Underwater lights are widely used in the industry to reduce
on how different stocking densities affect salmon in sea the incidence of sexual maturation (e.g. Oppedal et al.
cages. Turnbull et al. (2005) examined densities ranging 2011a). Maturation is covered as a separate WI (see below),
from 10 to 34 kg m)3 at a sea farm and found no but the underwater lights have also more direct implica-
negative effects on the salmon, measured as a combined tions for fish behaviour. Atlantic salmon tend to avoid
score of body condition, fin condition, plasma glucose strong surface daylight (Huse & Holm 1993; Fernö et al.
and cortisol, up to an inflection point at about 1995), but are attracted to night-time surface and under-
22 kg m)3, and no substantial negative effect on these water lights (Oppedal et al. 2001, 2007, 2011a; Juell et al.
parameters below 32 kg m)3. These findings were largely 2003; Juell & Fosseidengen 2004). Lighting the cage at night
confirmed in a tank study by Adams et al. (2007) and a stimulates the salmon to maintain daytime swimming
sea cage study by Oppedal et al. (2011b). Adams et al. speeds and schooling behaviour, but the use of only surface
(2007) found negative effects on welfare for a stocking lights may result in fish swimming at very high densities
density of 35 kg m)3 compared with 25 kg m)3, and near the surface (Juell et al. 2003). Using submersible lights
Oppedal et al. (2011b) found declined feed intake, at depths (e.g. 15 m) that allow the salmon to spread out
growth rate, feed utilization and a greater number of both above and below the lights, therefore, improves the
cataracts when the stocking density exceeded welfare of caged salmon (Juell et al. 2003; Juell & Fosseid-
26.5 kg m)3. Unfortunately, these three studies provide engen 2004; Oppedal et al. 2007, 2011a).
limited information about the oxygen saturation of the Based on this we propose to divide the lighting WI
water or the presence of endemic infections, which both into two levels: (1) optimal and (2) suboptimal. Optimal
may have been important reasons for decreased fish is the use of artificial lights at multiple depths. Subopti-
welfare at the higher densities (Johansson et al. 2006; mal is narrow illumination of the cage volume, such as
Oppedal et al. 2011b). A tank study indicates that low moonlight, artificial lights positioned at only a shallow
stocking densities of only 57 individuals may lead to depth or above the surface. Optimal lighting allows the
aggression and reduced welfare (Adams et al. 2007), but salmon to utilize the entire water column and hence con-
this has not been confirmed for low densities in sea tributes to positive performance (1) and preference (1).
cages holding a higher number of individuals (Turnbull Lack of illumination may force the salmon to school at
et al. 2005; Johansson et al. 2006; Oppedal et al. 2011b). high densities near the surface at night time and experi-
Johansson et al. (2006) showed that salmon in sea cages ence frustration ()1) and avoidance ()1) as the other
at high stocking densities (18–27 kg m)3) have limited depth layers are not used. WF is calculated as 4 (Eqn 2,
abilities to position themselves at preferred temperatures Table 4). The lighting WI is defined as optimal during
compared with fish at lower densities (7–11 kg m)3) and the light season of the year.
as a result grew less. Oppedal et al. (2007, 2011b)
showed that salmon may congregate into very tight
Disturbances
schools, with a local density above 180 kg m)3, in order
to avoid high temperatures. This illustrates that crowd- Removing fish from the water, for instance when estimat-
ing of fish may be a response to an underlying factor, ing the level of sea lice infestation, is one of the most
i.e. competition for limited resources within the cage severe stress events, and induces a high cortisol response
and ⁄ or lack of ability to avoid sub-lethal ⁄ lethal condi- (Schreck et al. 1997). However, this is usually done on
tions, which seem to be far more relevant problems than only a few individuals at a time and likely to have little
stocking density per se. effect on the other fish in the cage. Other procedures may
Although the literature shows that salmon may congre- affect the whole group, e.g. delousing by bath (Vigen
gate at extreme densities, we take as given that high over- 2008; Nilsen et al. 2010), grading (Juell et al. 2008) and
all densities limit the fish’s freedom to move in the cage. transportation (Iversen et al. 1998, 2005; Farrel 2005).
Low stocking densities (below 22 kg m)3) will therefore Studies of wellboat-transportation of smolts (Iversen et al.
give more natural behaviour (1). At higher densities wel- 2005) and live-hauling of harvest fish to processing facili-
fare becomes incrementally worse until above 32 kg m)1, ties (Farrel 2005) show that the salmon recover during
where there is a substantial effect on negative perfor- transportation from the initial handling stress of being
mance ()2), pain ()1), illness ()1) and activation of the loaded. This recovery seems to be crucial for avoiding
HPI-axis ()3). We divided the stocking density WI into cumulative and hence long-term stress during their initial
four levels from <22 kg m)3 to above 32 kg m)3 and period in the sea cages (Iversen et al. 2005). Juell et al.
calculated a WF of 8 (Eqn 2, Table 4). (2008) observed that crowding, pumping and sorting of

Reviews in Aquaculture (2013) 5, 33–57

ª 2013 Wiley Publishing Asia Pty Ltd 41
L. H. Stien et al.

salmon in sea cages led to a rapid drop in oxygen levels were not normally distributed and 73% of the recorded
(not critical) during the procedure. For several days the mortalities occurred in only 20% of the cages. The best
fish were also more dispersed in the cages than before the performing sea cages had a mortality rate, defined as the
treatment and they did not congregate as much in the number of dead fish divided by the total number of fish
warm surface layers as before. Appetite was reduced for in the cage multiplied by 100, of about 0.002% day)1,
approximately 5 days, and did not increase with the while the worst cages had periods of mortality rates with
increasing surface temperatures in May, indicating a peaks of up to 2.4% day)1 with an average of 0.1%
strong negative effect of this commercial sorting proce- day)1. Production data of fish mortalities in sea water
dure. During delousing with bath treatment a bottom (2009–2011) from mid-Norway were grouped according
opened or closed tarpaulin ‘skirt’ is placed around the to smolt-groups (n = 127, 65.6 million individuals),
cage to keep the therapeutic chemicals inside the cage. where 11% of the groups had >30% mortality, 55.9% had
Various aspects of this procedure, including the distur- 30–20% mortality, 33.1% had <10% mortality, and the
bance, crowding, changed environment, skirt and the average mortality was 16.1% (Anon 2011a). Disease dur-
treatment substance, may affect the fish. Vigen (2008) ing the sea water phase accounted for 23.5% of the mor-
found that in a group of salmon held at 25 kg m)3 the talities, smolt quality related problems accounted for 38%
oxygen saturation decreased to around 50% within and handling during the sea water phase accounted for
45 min after a skirt was placed around the cage, when no 38.5% (Anon 2011a). In an extensive study of more than
treatment substance was added. After the treatment sub- 88 production cycles in Scotland within one company,
stance (the pyrethroid cypermethrin, Betamax Vet) had Soares et al. (2011) developed benchmark mortality
been added within the skirt, salmon crowded at very high curves. The 50-percentile benchmark curve starts at above
densities (up to 107 kg m)3) near the surface. Oxygen 0.1% day)1 mortality during the first week after transfer,
saturation decreased faster while the swimming speed and between 0.01% and 0.1% during week 2–40, and then
gill ventilation frequency were higher and more variable. <0.01% day)1 until slaughter. Using the 50-percentile
In a compilation of observations during topical delousing curve as a benchmark gives a total mortality of about
with skirts Nilsen et al. (2010) concluded that the salmon 11% at the end of production. This is considerably better
avoided the therapeutant by swimming below the than the total mortality value of 17% reported by the
enclosed volume when the nets were not lifted. Following Norwegian salmon industry and the 21% reported by the
delousing, many farmers have reported poor performance Scottish Industry (Aunsmo et al. 2008a). For the 10- and
of the fish including poor appetite, reduced growth, dis- 90-percentile curves and more detailed description of the
ease outbreaks and increased mortalities. 50-percentile curve see (Soares et al. 2011). The main
We propose to divide the disturbances WI into four lev- causes of mortalities in Soares et al. (2011) were disease
els: (1) none, (2) light, (3) moderate and (4) severe. Level 4 (31%), production factors (accident loss, caught in net,
includes disturbances such as pumping of the fish which cull, failed smolts, jacks, mature, net tear, parr, precocious
may lead to activation of the HPI ()3) axis, abnormal male, transfer, treatment kill, smolt transfer and suspected
behaviour ()3), frustration ()1), negative performance cannibalism) (29%), environment (8%), predation (7%)
()1), illness ()1) and reduced survival ()1). Level 3 and other causes (26%).
includes disturbances such as crowding and topical delous- High daily mortality compared with the benchmark is
ing. Level 2 includes disturbances such as activity around indicative of illness ()5), reduced survival ()5), pain
the cage that only stresses the fish to a mild extent. Level 1, ()5) and negative performance ()3), while low daily
no disturbances, promotes natural behaviour (1) and the mortality indicates positive performance (3). Based on the
total WF is calculated as 11 (Eqn 2, Table 4). mortality benchmark study we suggest dividing the daily
mortality WI into five levels from best (at or below the
10-percentile curve) to worst (at or above 90-percentile
Daily mortality rate (% per day)
curve (Table 4). Long term values at or above the 90-per-
Mortality in farmed animals, including salmon, is an centile will lead to extreme mortality and is accordingly
indicator of disease outbreaks, poor environmental condi- considered to be a knockout level. The WF is calculated
tions, or injuries, all conditions that are related to to 21 (Eqn 2, Table 4).
reduced welfare. Aunsmo et al. (2008a) studied fish mor-
talities in 20 cages (10 sites) in the three first months
Appetite
after transfer and found that the fish died from various
reasons including incomplete smoltification (5.6%), pre- Appetite is defined here as the fish’s willingness to forage,
cocious males (3.3%), trauma (18.2%), specific diseases and the loss of appetite may be a sign of one or more
(65.6%) or unknown reasons (7.6%). Cage mortality rates underlying welfare relevant conditions (Schreck et al.

Reviews in Aquaculture (2013) 5, 33–57

42 ª 2013 Wiley Publishing Asia Pty Ltd
 Salmon Welfare Index Model (SWIM 1.0)

1997; Huntingford et al. 2006). Several studies have mary stress responses, inflammatory responses, changes in
reported a loss of appetite at seawater transfer (Usher appetite, changes in the skin and gills, compromised
et al. 1991; Toften et al. 2003), infection or disease (Rod- immunity, delayed healing of injuries, osmotic problems
ger & McArdle 1996; Damsgård et al. 2004), handling and tissue self-destruction (Nolan et al. 1999; Bowers
(McCormick et al. 1998), a deteriorating environment et al. 2000; Finstad et al. 2000; Ross et al. 2000; Boxaspen
(Bergheim et al. 2006; WEALTH 2008) and high stocking 2006; Skugor et al. 2008). Sea lice initiate short term
density (Oppedal et al. 2011b). Many fish farmers use physiological effects for the host already at 0.01 lice cm)2
appetite to determine feeding levels. It requires experience fish and long term effects at 0.05 lice cm)2 fish (Nolan
in order to interpret the behaviour of the fish. The farmer et al. 1999). Grimnes and Jakobsen (1996) found that
must assess appetite in relation to water temperature and more than 0.15 lice cm)2 fish was lethal, but indicated
fish size. Generally, appetite increases with water tempera- that the actual mortality limit probably is lower. An
ture and decreases with fish size (Austreng et al. 1987). extensive 10 year sampling of wild Atlantic salmon in the
Feed companies usually supply farmers with expected Norwegian sea revealed no fish carrying more than 10
amounts of feed under different water temperatures and adult lice (Holst et al. 2003). Since a wild smolt leaving
fish sizes (see above). The responsiveness to food varies the coast has a weight of about 15 g (Finstad et al. 2000)
with the time of day and season (Kadri et al. 1991; or surface area (including fins) of 95 cm)2 (Tucker et al.
Jørgensen & Jobling 1992; Smith et al. 1993), and it may 2002: fish surface area (cm2) = 0.6131*fish weight
be manipulated using artificial photoperiods (Taranger (g) + 86.144), this implies an upper limit of
et al. 1995; Nordgarden et al. 2003; Oppedal et al. 2003). 0.12 lice cm)2 fish.
Although the feeding regime in general seems to have Infestations of more than 0.12 lice cm)2 fish are lethal
little effect on growth and the feed conversion ratio for the fish (knockout), at lower levels >0.05 lice cm)2
(FCR) (Sveier & Lied 1998), suppressed growth was seen fish the fish will increasingly suffer from illness ()3), pain
in the daily feeding regime of one meal compared with ()1), activation of the HPI ()1) axis, reduced survival
eight meals in the period just following sea transfer ()3) and negative performance ()2) (Table 4). We sug-
(Flood et al. 2011). Today, many Salmon farmers use a gest five levels for the sea lice WI (Table 4), from no lice
camera positioned beneath the feeding area, looking up, as level 1 (positive performance (1)), via light infestation
to assess appetite levels; when the farmer sees pellets as level 2 (only Copepod and Chalimus stages and ⁄ or
reaching down to the camera the feeding is turned off. <0.05 lice cm)2 fish for the pre-adult and adult stages),
Prolonged (weeks to months) poor appetite is clearly to ‡0.08 adult or pre-adult lice cm)2 fish as level 4
indicative of negative performance ()2) and illness ()3), (Table 4) and calculate a WF of 11 (Eqn 2, Table 4).
and good appetite suggests demand (3) and positive
performance (3). For practical application in the SWIM
Condition factor
1.0-model, we suggest dividing the Appetite WI intro
three levels: (1) good appetite, (2) as expected and (3) Condition factor (K) is a standard measurement of fish
poor appetite and calculate a WF of 11 (Eqn 2, Table 4). nutritional status (Bolger & Conolly 1989; Nash et al.
2006) and is calculated as K = (WL)3)100, where W is
the weight in g and L is the length in cm. In general
Sea lice
terms, a skinny salmon may have a K < 0.9 and a fat fish
Farmed Atlantic salmon are parasitized by two species of a K of 1.5 (Tvenning 1991). During the production cycle
sea lice; Lepeophtheirus salmonis (salmon lice) and, to a K changes from just above 1 as smolt (O’Flynn et al.
lesser extent, Caligus elongates (e.g. Pike & Wadsworth 1997; Mørkøre & Rørvik 2001; Oppedal et al. 1999, 2006;
1999). Salmon respond to a sea lice infestation with pri- Fjelldal et al. 2009a, b) to 1.6 nearer slaughter (Oppedal
mary stress responses including elevated blood cortisol et al. 1997, 1999, 2006; Einen et al. 1998; Rørå et al.
and glucose (Bowers et al. 2000; Finstad et al. 2000). 1998; Mørkøre & Rørvik 2001) but this may partly be
These stress responses occur even though at the infective overruled by season phase and delayed by artificial pho-
copepod stage the lice do not yet feed on the salmon (e.g. toperiods (Oppedal et al. 1997, 1999, 2003, 2006; Fjelldal
Finstad et al. 2011). Grimnes and Jakobsen (1996) and et al. 2009a, b). Generally, K decreases during winter and
Finstad et al. (2000) did not find severe effects on the fish spring, and increases during summer and autumn. Peri-
from extreme infections of sea lice (>1 lice cm)2 fish or ods of good growth typically increase K (Juell et al. 1994;
>100 lice fish)1) at the copepod and early chalimus Endal et al. 2000), while periods of poor growth reduce K
stages, but they did find a sudden increase in mortality (e.g. Juell et al. 1994; Einen et al. 1999). Also, sea transfer
after the appearance of the pre-adult stages. Responses to as either spring or autumn smolts may interfere with the
an infestation of pre-adult and adult sea lice include pri- seasonal pattern (Mørkøre & Rørvik 2001), but not

Reviews in Aquaculture (2013) 5, 33–57

ª 2013 Wiley Publishing Asia Pty Ltd 43
L. H. Stien et al.

inevitably (Fjelldal et al. 2009a). Farmed fish display swimming separated ⁄ apart from the main group, and
higher K compared with hybrid and wild salmon given staying at the surface for prolonged periods of time.
similar farming conditions (Fjelldal et al. 2009a). There is We propose to divide the emaciated state WI into
a strong and significant positive correlation between K three levels: (1) not emaciated, (2) potentially emaciated
and total lipid content in Atlantic salmon (Herbinger & and (3) distinctly emaciated (Table 4). No sign of emaci-
Friars 1991; Einen et al. 1998, 1999; Rørå et al. 1998; ation is evidence of a healthy fish, i.e. positive perfor-
Hamre et al. 2004; Peterson & Harmon 2005). K is nega- mance (1). An emaciated fish is very ill and moribund. A
tively correlated with plasma glucose and cortisol (Turn- positive identification of an emaciated fish is therefore a
bull et al. 2005). Very high K (>1.6) indicates spinal knockout level for that individual fish. A potentially ema-
deformation (Gjerde et al. 2005; Witten et al. 2005; Berg ciated fish is a fish showing signs of abnormal behaviour
et al. 2006; Fjelldal et al. 2009b; Hansen et al. 2010), but ()3), negative performance ()3) and illness ()3), and is
the specific level at which this may occur is difficult to fix likely to have reduced survival ()3) and experience pain
due to the variations discussed above. However, within a ()3). Based on this we calculated a WF 16 (Eqn 2,
population, low K individuals tend to be emaciated Table 4).
fish while ‘normal’ K values indicate good health, and
very high K values often indicate deformed individuals.
Vertebral deformation
We propose to divide the condition factor WI into
three levels: (1) >1.1, (2) 0.9–1.1, and (3) <0.9. Salmon The main vertebral deformity in Norwegian salmon farms
with K above 1.1 have lipid reserves indicating positive is pronounced compression of the vertebral column and
performance (3), while salmon with K below 0.9 and 1.1 reduced fork length, commonly referred to as ‘short-tail’
have negative performance ()2) and activation of the (Gil-Martens 2010). Multiple causes of vertebral deformi-
HPI ()1) axis. Extreme high K (>1.6) may be indicative ties have been identified such as environmental condi-
of malformation, but this is addresses by the vertebral tions during egg incubation (Wargelius et al. 2005), fish
deformities WI (see below) and need therefore not be size and temperature at vaccination (Berg et al. 2006),
considered here. Similarly, for extreme low K which is type of vaccination (Aunsmo et al. 2008b), mineral nutri-
addressed by the emaciation state WI (see below). The tion (Fjelldal et al. 2008, 2009b), use of underyearling
WF was calculated as 6 (Eqn 2, Table 4). smolt (Fjelldal et al. 2006) and temperature at transfer to
sea water (Grini et al. 2011). The prevalence of one or
more vertebral deformities determined by radiology in
Emaciation state
harvest sized salmon have been reported in the range of
Fish may become emaciated due to disease (Stephen & 6.6–73.3% (Witten et al. 2005; Fjelldal et al. 2007, 2009a,
Ribble 1995; Kent & Poppe 2002), poor smoltification b; Korsøen et al. 2009). Hansen et al. (2010) found that a
(Duston 1994), ‘wrong’ feeding strategy (at transfer some low severity of deformed vertebrae (<6 vertebrae com-
fish may start to eat zooplankton instead of pellets) (pers. pressed) has little effect on growth, but individuals with
obs.; wild smolt: Rikardsen et al. 2004), sea lice (e.g. Fins- more than 10 deformed vertebrae were shorter and had a
tad et al. 2011), stress (e.g. Huntingford et al. 2006) and higher condition factor than normal fish, while fish with
social constraints (Jobling & Reinsnes 1986; Adams et al. more than 20 deformed vertebrae in addition showed
2000). Emaciated fish are generally small, very thin fish of lower weight than normal fish. Aunsmo et al. (2008b)
poor health, and they may act as a vector for introducing reported that fish with high intra-abdominal lesion scores
disease to the other (more healthy) fish in the cage. As also more frequently had vertebral deformities and
they are feeding poorly, or not at all, it is difficult to treat weighed 62% of non-deformed fish at slaughter.
them orally (Coyne et al. 2006). Emaciated fish are well Dependent on the severity of deformation, external
known to fish farmers (Stien et al. 2009; Anon 2011b), examination is a less sensitive method of assessment than
but there is little published research on the subject. A radiology, the prevalence has, for example, been assessed
study using Floy anchor tags on farmed chinook salmon as 1.3% vs. 12.4% (Fjelldal et al. 2007) and 13–17% vs.
(Oncorhynchus tshuwytscha) individuals that could be cap- 88–94% (Grini et al. 2011). Since this version of the
tured with a dip net from the surface, showed that these SWIM-model is aimed at fish farmers the vertebral defor-
were mainly emaciated and moribund fish (62% died mation WI must be judged by external examination of
within 24 h) (Stephen & Ribble 1995). Characteristic of the individual fish. We therefore suggest dividing the WI
these fish were obvious pathological and clinical abnor- into three levels: (1) no external signs of vertebral defor-
malities (95% of 366 individuals exhibited gross and ⁄ or mity, (2) ‘short tail’ of normal weight, (3) ‘short tail’ of
histopathological abnormalities), and behavioural abnor- low weight compared with the rest of the population.
malities such as swimming into the nets or in circles, Level 1 is linked with positive performance (1), while

Reviews in Aquaculture (2013) 5, 33–57

44 ª 2013 Wiley Publishing Asia Pty Ltd
 Salmon Welfare Index Model (SWIM 1.0)

level 3 indicates negative performance ()3), pain ()3) temperatures (>14C) compared with intermediate
and illness ()3) this gives a WF of 10 (Eqn 2, Table 4). temperatures (10C), while low temperatures (<7C) may
lead to a prolonged period of osmotic stress and
increased mortality (Sigholt & Finstad 1990; Arnesen et
Sexual maturity stage
al. 1998; Handeland et al. 2000, 2003). For intermediate
Sexual maturation leads to allocation of energy towards water temperatures (which are best for welfare) transfer
gonad build-up and migration. Prior to upstream migra- of salmon to full strength seawater before the smoltifica-
tion wild salmon have an energy loss of about 60% of tion process has completed resulted in high mortality
their body reserves (Jonsson et al. 1997; Fleming 1998). (>40%) and stunted growth rates for a period of
In the wild few survive to breed another year (Fleming 1–2 months (Duston 1994), but when transferred to
1998). Consequently, sexual maturation is detrimental for brackish water (20 ppt) mortality was <10% and only
salmon production, where artificial photo-regimes are temporarily stunted growth rates were observed, and with
used to prevent maturation (e.g. Oppedal et al. 2011a). even less saline water (10 ppt) little to no mortality
Sexually mature parr, precocious males, can be present at occurred and no stunting of growth compared with parr
sea transfer and their presence is linked to increased mor- continuing in freshwater (Bjerknes et al. 1992; Duston
tality (Aunsmo et al. 2008a). The energy expended for 1994). For fully smoltified fish there is little effect of
maturation and spawning increases with fish size and salinity on growth rate and mortality (Duston 1994).
females also expend more energy on gonads compared Fully smoltified fish have few problems with osmoregu-
with males (ca 28% vs. ca 4% of total energy reserves, lation in full strength seawater (positive performance (1)).
Fleming 1998). Whether mature salmon have a behavio- Impaired smolts have negative performance ()3) and
ural need to carry out spawning migration is difficult to reduced survival ()5), especially at low temperatures
answer (cf. Huntingford et al. 2006), but it is plausible (<7C), and knockout for high temperatures (>20C).
that there is an increase in aggression (Fleming & Einum This gives six WI levels from worst (incomplete smoltifi-
2011). With regard to altered osmoregulation in adapta- cation at high temperature) to best (fully smoltified) and
tion to a hypo-osmotic environment before re-entering a WF of 9 (Eqn 2, Table 4). As this is a farmer’s version
freshwater in nature, Persson et al. (1998) found that sal- of SWIM, the smoltification state must be judged based
mon caught in the estuary (before entering the river) had on the colouration and shape of the fish. Fully smoltified
already adapted to a hypoosmotic environment and that Atlantic salmon have lost their distinctive parr markings,
during the upriver migration the gill Na+, K+-ATPase gained a more silvery colour and have a more streamlined
activity decreased even further. It is therefore plausible shape (Hoar 1988).
that mature salmon in sea cages to some extent experi-
ence osmoregulatory challenges. Besides the energy drain-
Fin condition
ing effects of maturation, it has been shown that
compared with immature fish mature salmon have a Fin erosion refers to damage to, and loss of, the tissue of
higher prevalence of the parasite Kudoa thyrsites, that is a the rayed fins (Latremouille 2003) and is often found in
cause of post mortem soft flesh (St-Hilaire et al. 1998). farmed salmonids. Being externally visible, fin damage
Mature females have invested heavily in the develop- represents an intuitive and meaningful welfare indicator
ment of gonads and show negative performance ()3) and easily recognized by farmers and informed consumers
ultimately reduced survival ()3). Mature males and espe- (Ellis et al. 2008). While most studies on nociception in
cially mature juvenile males invest less. Maturity linked fish have focused on the head region or the body, Cherv-
aggression ()2) may also reduce welfare. No maturation ova (1997) demonstrated experimentally that fish fins are
is presumed to give a positive performance (1). We pro- capable of nociception. Being live tissue capable of noci-
pose dividing the sexual maturity stage WI into four lev- ception mechanical injury to fin tissue is probably associ-
els and calculate a WF of 9 (Eqn 2, Table 4). ated with pain. In some cases, mechanical fin damage
may reflect aggressive behaviour within the rearing unit
(salmon parr: Turnbull et al. 1996, 1998; Jones et al.
Smoltification state
2010). Damage to the fins of salmonids is, however, more
During the smoltification process salmon parr develop often caused by chronic infection with biofilm forming
tolerance for high salinity, enabling the young salmon bacteria that progressively necrotize the fin edges (Bernar-
(now called smolt) to enter seawater with only minor dis- det et al. 1998), similar to leprosy in humans not neces-
turbances in osmotic balance (e.g. Stefansson et al. 2008; sarily being painful. Poor fin condition is coupled with a
Thorstad et al. 2011). The physiological disturbances high stocking density, poor water quality, decreased con-
during exposure to seawater (33 ppt) are greater at high dition factor and increased plasma glucose and cortisol

Reviews in Aquaculture (2013) 5, 33–57

ª 2013 Wiley Publishing Asia Pty Ltd 45
L. H. Stien et al.

levels (Turnbull et al. 2005; Adams et al. 2007). The fins rickettsiosis (Mauel & Miller 2002) and salmon anaemia
fulfil important functions in both locomotion and intra- (Totland et al. 1996). Several bacteria in the class Flexibac-
specific communication in salmonids (Abbott & Dill teriae often cause skin lesions and fin erosion in both fresh-
1985; Pelis & McCormick 2003) and severe fin erosion water or seawater reared fish (Bernardet 1998; Lorenzen
thus has the potential to affect behaviour. However, the 1999) and it has been shown that many fish pathogenic
evidence is scarce or contradictory, and any functional bacteria secrete proteolytic enzymes that cause massive tis-
impairment following fin erosion has yet to be demon- sue damage (Leung & Stevenson 1988; Ostland et al. 2000).
strated scientifically. The breakdown of the epithelial bar- It should also be mentioned that the skin provides a first
rier during active fin erosion may disrupt osmotic line of defence against pathogens (Segner et al. 2012),
homeostasis and can thus cause severe stress in the fish where the skin mucus prevents aggregation of pathogens
(Clayton et al. 1998). by being continuously replenished and by containing vari-
Fin damage represents injury to live tissue with the ous immune factors (Shepard 1994). Epidermal damage
potential for inflammation and pain ()5). Damaged epi- such as wounds and non-intact mucus layers therefore
thelial structures may also represent invasion routes for represent invasion routes for virus and bacteria (Svendsen
pathogens and thus lead to illness ()3) and negative per- & Bøgwald 1997).
formance ()2). We propose to divide the fin condition Similar to the fin condition WI damage to the skin
WI into four levels ranging from normal healthy fins may cause pain ()5) and represent invasion routes for
(positive performance (3)) without tissue loss to severely pathogens leading to infection and illness ()3) and possi-
damaged fins with tissue loss, which also may be suffering bly reduced survival ()3) in salmon. Even smaller skin
from necrosis, inflammation, bleeding or exposed fin rays damages may lead to long term negative performance
(Table 4). The WF calculated in SWIM 1.0 is 13 (Eqn 2, ()3) due to increased metabolic cost involved in wound
Table 4). repair and osmoregulatory perturbation. Both the size of
the affected area and the depth (whether it is penetrating
or superficial) of skin damage will probably contribute to
Skin condition
the severity of the condition. Thus, the indicator is
The integrity of the skin-scale complex provides a relatively divided into five levels (Table 4) ranging from normal
impermeable barrier to water and electrolytes. Epidermal healthy skin (positive performance (1)) to penetrating
damage such as scale loss, wounds and ulcers can therefore and ⁄ or multiple wounds or ulcers that also may be
result in a loss of body water and changed ion balance, infected, plus a knockout level for large open wounds.
which produces an osmotic stress that potentially can be The WF calculated in SWIM 1.0 is 15 (Eqn 2, Table 4).
life threatening (Bouck & Smith 1979). There is evidence
that ulceration of as little as 10% of the body surface area
Final model
can result in high acute mortality and that the degree of
mortality is directly related to the amount of skin damage The final step of the semantic modelling procedure
(Bouck & Smith 1979). Sub-lethal skin damage might affect (Bracke et al. 2002b) is to assemble the WIs, the levels
the fish energy budget due to increased metabolic cost and their associated ranks into an OWA-model using the
involved in wound repair, and osmoregulatory perturba- following three formulae for calculating the relative
tions. Such chronic effects can affect growth rates and weighting factors (RWFs), indicator welfare scores (IWSs)
fecundity negatively; it may also lead to an increased sus- and the overall welfare index (OWI):
ceptibility to other diseases (Noga 2000). Many situations !1
Xm
or management procedures in salmon aquaculture are RWFi ¼ WFi  WFj ð3Þ
associated with a high risk for mechanical damage to the j¼1

skin. Examples are transport, sorting, vaccination, pump-

ing, strong currents and high densities of fish, jelly fish IWSi ¼ ISi  RWFi ð4Þ
burns, parasites, attack from other fish and predators
(Noble et al. 2012). Virus- or bacterial infections can often X
m
OWI ¼ IWSj ð5Þ
also constitute the underlying cause of skin necrosis or j¼1
ulcerations in fish. In sea farmed Atlantic salmon several
infections are associated with severe or even pathognomic where m is the total number of indicators in the model,
cutaneous symptoms, i.e. winter ulcer disease (infection WFi and WFj (see Eqn 2) are the weighting factors of the
with Moritella viscosa; Lunder et al. 1995; Benediktsdóttir respective indicator i and j, and ISi (see Eqn 1) is the
et al. 2000), atypical furunculosis (atypical Aeromonas sal- indicator score given by the assessor (the fish farmer) for
monicida infections; Wicklund & Dalsgaard 1998), Pisci- indicator i. In the case of one or more knockout levels

Reviews in Aquaculture (2013) 5, 33–57

46 ª 2013 Wiley Publishing Asia Pty Ltd
 Salmon Welfare Index Model (SWIM 1.0)

the OWI is defined as 0. Knockout levels are not included water column, the water current varied between 3 and
when calculating RWFs and IWSs. 12 cm s)1 (i.e. between 0.05 and 0.22 BL s)1), stocking
Although, we originally intended that the WIs should density at about 14 kg m)3, no artificial lighting, only
be at the sea cage level, the literature reviews made clear light disturbances, mortality at 0.11% and the farmer
that the research on many of the WIs predominantly or reported poorer appetite than expected. Using the sea
exclusively were based on analysis of their effects on indi- cage WIs from Table 4 this gives an OWI for the sea cage
vidual fish. For example, not the prevalence of sea lice of 0.37 (Eqn 5, Table 6), on a scale from 0 to 1, where 0
infested fish in a sea cage and the effect on the overall is worst and 1 is best welfare. The low OWI indicates low
fish welfare in the cage, but the effects on the welfare of fish welfare. This was affirmed 2 days later when the
individual fish from different sea lice infestation ratios. farmer collected more than 3300 dead fish, i.e. 2.36% of
We therefore divided the indicators into sea cage specific the fish in the cage. This is a knockout value and if the
WIs: temperature, salinity, oxygen saturation, water cur- assessment had been performed on the sea cage that day,
rent, stocking density, lighting, disturbances, daily mortal- the OWI would have been set to 0.
ity ratio and appetite and individual fish specific WIs: sea For the individual fish specific indicators, an OWA will
lice infestation ratio, body condition, emaciation state, be based on a representative sample of fish from the cage,
vertebral deformation, maturation stage, smoltification but as an example we only look at one imagined repre-
state, fin condition and skin condition (Table 4). Table 5 sentative fish in the current scenario. Figure 1 shows a
shows the RWFs for the sea cage and individual fish spe- salmon with no lice, a K of 1.21 (1.6 kg and 51 cm), not
cific WIs. These RWFs together with their levels and their emaciated, no external signs of deformity, moderate split-
ISs in Table 4 give a model (or schema) for calculating ting of the fins and a normal healthy skin. This specific
an OWA score for a sea cage and for individual fish. The fish gets an OWI of 0.90 (Eqn 5, Table 7) on a scale from
first gives an overall score for the welfare conditions in 0 worst to 1 best possible welfare score.
the sea cage, while the second give scores for the respec-
tive fishes. We call the model Salmon Welfare Index
Table 6 SWIM 1.0 applied on the sea cage in the example scenario.
Model 1.0, abbreviated SWIM 1.0. 1 states that it is the
The OWI is the sum of the IWS (Eqn 5)
farmer’s version and .0 states that this is the pilot version
which may be revised and upgraded later. Sea cage WIs RWF # Level IS IWS

Temperature (C) 0.17 1 10–15 1.00 0.17

Example scenario Salinity 0.03 2 No access to 0.00 0.00
brackish water
This scenario is based on a real world example from a sea Oxygen (%) 0.18 3 40–60% (12C) 0.00 0.00
farm in Western Norway, autumn 2011. The sea cage was Water current 0.03 1 <0.9 1.00 0.03
157 m in circumference, fitted with a 35 m deep cone- (BL s)1)
Stocking density 0.09 1 <22 1.00 0.09
shaped net containing 140 000 fish with an average
(kg m)3)
weight of 2.3 kg and average length of 55 cm. The water Lighting 0.04 2 Suboptimal 0.00 0.00
temperature was 14C, 33 ppt salinity from top to bot- Disturbances 0.12 2 Light 0.67 0.08
tom, oxygen saturation was 50% in large parts of the Mortality (% day)1) 0.22 5 At or above the 90 0.00 0.00
percentile curve
Appetite 0.12 3 Poor appetite 0.00 0.00
Table 5 Relative weighting factors for the sea cage specific WIs and OWI 0.37
for the individual fish specific WIs in SWIM 1.0

Sea cage WIs WF RWF Individual fish WIs WF RWF

Temperature (C) 16 0.17 Sea lice 11 0.12

Salinity 3 0.03 Condition factor 6 0.07
Oxygen (%) 17 0.18 Emaciation state 16 0.18
Water current (BL s)1) 3 0.03 Vertebral deformation 10 0.11
Stocking density 8 0.09 Sexual maturity stage 9 0.10
(kg m)3)
Lighting 4 0.04 Smoltification state 9 0.10
Disturbances 11 0.12 Fin condition 13 0.15
Mortality (% day)1) 21 0.22 Skin condition 15 0.17
Figure 1 Image of the fish used in the example scenario. This fish
Appetite 11 0.12
had an OWI of 0.90 on a scale from 0 worst to 1 best possible wel-
SUM 94 1.00 89 1.00
fare score (Table 7).

Reviews in Aquaculture (2013) 5, 33–57

ª 2013 Wiley Publishing Asia Pty Ltd 47
L. H. Stien et al.

Table 7 SWIM 1.0 applied on the fish from the example scenario. Combining the score of the sea cage and the score of the
The OWI is the sum of the IWS (Eqn 5) individual ‘representative fish’ results in an OWI (Eqn 5)
Individual fish WIs RWF # Level IS IWS given as OWI = (0.37*94 + 0.90*89) ⁄ (94 + 89) = 0.62. The
conclusion is that the fish welfare at the time of sampling
Sea lice 0.12 1 No lice 1.00 0.12 was mediocre. The example representative fish was still fit,
Condition factor 0.07 1 1.0–1.5 1.00 0.07
but the conditions in the sea cage were very poor.
Emaciation state 0.18 1 Not emaciated 1.00 0.18
Vertebral 0.11 1 No external signs 1.00 0.11
deformation of v. deformities First sampling using SWIM 1.0
Sexual maturity stage 0.10 1 Not mature 1.00 0.10
Smoltification state 0.10 1 Fully smoltified 1.00 0.10 This is the first actual sampling using the SWIM 1.0 model.
Fin condition 0.15 3 Moderate splitting 0.33 0.05 The sampling was done at a sea farm in Western Norway,
Skin condition 0.17 1 Normal healthy skin 0.00 0.17 winter 2012. The sea cage was 157 m in circumference, fit-
OWI 0.90
ted with a 45 m deep cone-shaped net containing 100 000
fish with an average weight of 5.8 kg and average length of
79 cm. The water temperature was 7C, 33 ppt salinity and
Table 8 Results from the first SWIM 1.0 sampling of a commercial 100% oxygen saturation from top to bottom of the cage,
salmon sea cage. The OWI is the sum of the IWS (Eqn 5) the water current at the surface varied between 6 and
Sea cage WIs RWF # Level IS IWS
36 cm s)1 (i.e. between 0.07 and 0.38 BL s)1), the stocking
density was at about 20 kg m)3, artificial lighting posi-
Temperature (C) 0.17 2 7–10 0.75 0.13 tioned at 10 m depth, recent severe disturbances occurred
Salinity 0.03 2 Adult fish with no 0.50 0.02
when 30 000 fish were harvested from the cage, the mortal-
access to brackish water
ity was at about 0.01% and the farmer reported poorer
Oxygen (%) 0.18 3 >80%, all temperatures 1.00 0.18
Water 0.03 1 <0.9 1.00 0.03
appetite than expected. Using the sea cage WIs from
current (BL s)1) Table 4 this gives an OWI for the sea cage of 0.59 (Eqn 5,
Stocking 0.09 1 <22 1.00 0.09 Table 8), on a scale from 0 to 1.
density (kg m)3) Ten fish were sampled for the individual fish specific
Lighting 0.04 1 Optimal 1.00 0.04 indicators. Details for each of the sampled fish (weight,
Disturbances 0.12 2 Severe 0.00 0.00
length, condition factor and number of pre- and adult lice)
Mortality (% day)1) 0.22 3 At the benchmark curve 0.50 0.11
are given in Table 9, together with the assigned WI levels
Appetite 0.12 3 Poor appetite 0.00 0.00
OWI 0.59
and calculated OWIs. The OWIs varied from a minimum
of 0.00 (emaciated fish, Fig. 2) to a maximum of 0.88

Table 9 SWIM 1.0 applied on 10 fish from the first SWIM 1.0 sampling The OWI for each fish is the sum of the IWS for the respective WI levels
(Eqn 5)

Details for fish 1–10

Details 1 2 3 4 5 6 7 8 9 10

Weight (kg) 6.20 5.85 9.00 2.90 8.25 8.55 5.90 7.16 8.50 1.05
Length (cm) 79 77 82 70 85 87 83 84 85 54
Condition factor 1.26 1.28 1.63 0.85 1.34 1.30 1.03 1.21 1.38 0.67
Number of lice (#) 4 0 4 4 5 11 7 7 7 45

WI levels for fish 1–10

Individual fish WIs 1 2 3 4 5 6 7 8 9 10

Sea lice 2 1 2 2 2 2 2 2 2 3
Condition factor 1 1 1 3 1 1 1 1 1 3
Emaciation state 1 1 1 2 1 1 1 1 1 3
Vertebral deformation 1 1 1 1 1 1 1 1 1 1
Sexual maturity stage 1 1 1 3 1 3 1 1 1 1
Smoltification state 1 1 1 1 1 1 1 1 1 1
Fin condition 3 4 2 2 3 2 3 3 2 2
Skin condition 3 5 3 1 3 1 3 5 2 4
OWI 0.79 0.72 0.84 0.59 0.84 0.84 0.79 0.73 0.88 0.00

Reviews in Aquaculture (2013) 5, 33–57

48 ª 2013 Wiley Publishing Asia Pty Ltd
 Salmon Welfare Index Model (SWIM 1.0)

skin and fin indexes (Table 9). Combining the score of

the sea cage and the median score of the individual fish
gives a total OWI = (0.59*94 + 0.79*89) ⁄ (94 + 89) = 0.69
(Eqn 5).

Discussion
Methodology
The objectives of this paper were to review basic welfare
indicators of sea-cage farmed Atlantic salmon and to gen-
erate a semantic model (SWIM 1.0) to enable farmers to
assess overall welfare. Although there are many papers
Figure 2 Image of fish number 10 from the first sampling. This fish published on semantic modelling and on welfare assess-
had an OWI of 0.00 on a scale from 0 worst to 1 best possible wel- ment in various species of farm animals, this is the first
fare score (Table 9). time a systematic review of scientific statements has been
performed and presented on farmed fish. A main advan-
tage of reviewing welfare indicators according to the prin-
ciples of semantic modelling is that it gave focus to the
review, as it was necessary to assess each indicator in
terms of what the indicator itself tells about salmon wel-
fare. This prevented long and overlapping essays about
each indicator; special cases, and interactions that are an
inherent part of a complex problem area such as fish wel-
fare in sea cages.
Figure 3 Image of fish number 9 from the first sampling. This fish In order to create an overall model it is necessary to
had an OWI of 0.88 on a scale from 0 worst to 1 best possible wel- reduce complexity. The advantage of the transparency in
fare score (Table 9).
semantic modelling is that it shows where these reductions
are made and where there is scope for further upgrading
with new scientific knowledge. Semantic modelling also
supports transparency of the model itself, allowing criti-
cism of underlying principles and specific choices made.
The semantic-modelling procedure used in SWIM 1.0
was derived from Bracke et al. (2002b) and De Mol et al.
(2006). It started with an extensive literature review for
statements that are somehow relevant for the welfare of
Atlantic salmon farmed in sea cages. This ensures that the
formulation of WI-levels and the calculation of WFs are
done in relation to unbiased scientific statements, i.e.
statements that have not been produced in order to con-
firm preconceived notions of the importance of different
WIs and how welfare should be assessed.
A major criticism of semantic modelling is that it is
subjective; i.e. one has to decide on how to divide the
indicators into levels, which weighting categories are
Figure 4 Image of fish number 2 from the first sampling. This fish appropriate for each indicator and one must assign indi-
had clearly been injured during the recent harvesting of fish from the cator and weighting scores. These decisions are indeed
cage and had an OWI of 0.72 due to the low skin and fin indexes
based on a partly subjective interpretation of the meaning
(Table 9).
of the collected scientific information. This subjectivity is,
however, decreasing with increasing quality and the
(Fig. 3); median OWI for the ten fish was 0.79. As a further amount of available scientific information; more solid
example, fish number 2 is shown in Figure 4. This fish was data reduces the freedom of the interpretation of the
clearly damaged during the recent harvesting of fish from data. The model and the semantic-modelling procedure
the sea cage and got an OWI of only 0.72 due to the low itself are objective, i.e. the information is scientifically

Reviews in Aquaculture (2013) 5, 33–57

ª 2013 Wiley Publishing Asia Pty Ltd 49
L. H. Stien et al.

valid and the semantic-modelling procedure is formalized

References
and has been described and validated in detail elsewhere
(Bracke et al. 2002b, 2008; Bracke 2008, 2011). It is Abbott JC, Dill LM (1985) Patterns of aggressive attack in
designed to take the modeller’s point of view, as much as juvenile steelhead trout (Salmo-gairdneri). Canadian Journal
possible, out of the equation (Bracke et al. 2002b; Bracke of Fisheries and Aquatic Sciences 42: 1702–1706.
2008; Bracke et al. 2008). Adams C, Huntingford F, Turnbull J, Arnott S, Bell A (2000)
Size heterogeneity can reduce aggression and promote growth
in Atlantic salmon parr. Aquaculture International 8: 543–549.
The model Adams CE, Turnbull JF, Bell A, Bron JE, Huntingford FA
This is the first time semantic modelling has been used to (2007) Multiple determinants of welfare in farmed fish,
stocking density, disturbance, and aggression in Atlantic sal-
create OWA for fish and, although there are several risk
mon (Salmo salar). Canadian Journal of Fisheries and Aqua-
assessment schemas for fish farming, there are to our
tic Sciences 64: 336–344.
knowledge no schemas for assessing fish welfare. As the
Anon (2001) Scientists assessment of the impact of housing
SWIM 1.0 model is based on an extensive review of the lit-
and management on animal welfare. Journal of Applied Ani-
erature, including the mentioned risk assessment schemas,
mal Welfare Science 4: 3–52.
it would be a circular argument to compare the model with Anon (2011a) Prosjekt overlevelse fisk. Regionalt tilsynsprosjekt
the literature and risk assessment schemas. Based on our 2011(in Norwegian). Mattilsynet Rapport. Available from:
expertise, however, we believe that the scoring in SWIM http://www.mattilsynet.no/mattilsynet/multimedia/archive/
1.0 has sufficient validity for fish farmers to start using the 00074/Sluttrapport_Prosjek_74570a.pdf
model. It is, for example, generally agreed that daily mor- Anon (2011b) Supersmolt reduserer ‘pinne-forekomst’ hos ørret.
tality is probably one of most important WIs for fish in sea (in Norwegian), Europharma nyhetsbrev 25. Available from:
cages and that salinity probably is one of the least impor- http://www.europharma.no/downloads/NYHETSBREV_april_
tant. That said, however, it is anticipated that the model 11.pdf
will need further maturation and upgrading. It is also Arnesen MA, Johnsen HK, Mortensen A, Jobling M (1998)
important to note that although the SWIM 1.0 model gives Acclimation of Atlantic salmon (Salmo salar L.) smolts to
OWIs as output, its main purpose is its use as a diagnostic ‘cold’ sea water following direct transfer from fresh water.
tool to identify indicators of reduced welfare and which the Aquaculture 168: 351–367.
fish farmer should address in order to improve fish welfare. Ashley PJ (2007) Fish welfare: current issues in aquaculture.
The next step is to visit several farms and at different times Applied Animal Behaviour Science 104: 199–235.
of the production in order to test and calibrate the model, Aunsmo A, Bruheim T, Sandberg M, Skjerve E, Romstad S,
and to gain an overview of how fish welfare varies in com- Larssen RB (2008a) Methods for investigating patterns of
mercial sea cages. These will be very extensive studies that mortality and quantifying cause-specific mortality in sea-
warrant their own publication. farmed Atlantic salmon Salmo salar. Diseases of Aquatic
Organisms 81: 99–107.
We must again emphasize that this is the first version
Aunsmo A, Guttvik A, Midtlyng PJ, Larssen RB, Evensen Ø,
of the model and by its inherent transparency it is open
Skjerve E (2008b) Association of spinal deformity and
for further upgrading. We hope that readers will have
vaccine-induced abdominal lesions in harvest-sized Atlantic
many suggestions for how the model can be improved,
salmon, Salmo salar L. Journal of Fish Diseases 31: 515–524.
for example new WIs, WIs that can be removed, more
Austreng E, Storebakken T, Åsgård T (1987) Growth rate
precise WI-levels, criticism on specific weightings and estimates for cultured Atlantic salmon and rainbow trout.
choices made, studies that should be included as part of Aquaculture 60: 157–160.
the background knowledge database etc. Finally, we plan Bakke H, Bjerknes V, Øvreeide A (1991) Effects of rapid
to make additional SWIM models adapted for use by changes in salinity on the osmoregulation of postsmolt
farm veterinarians (SWIM 2) and fish welfare experts Atlantic salmon (Salmo salar). Aquaculture 96: 375–382.
(SWIM 3), where we can use more advanced WIs that Benediktsdóttir E, Verdonck L, Spröer C, Helgason S, Swings J
require specific expertise or equipment. The goal is that (2000) Characterization of Vibrio viscosus and Vibrio wodan-
the fish welfare community can build on the SWIMs and is isolated at different geographical locations: a proposal for
in time reach a consensus on how to best assess overall the reclassification of Vibrio viscosus as Moritella viscosa
fish welfare in sea cages. comb. nov. International Journal of Systematic and Evolution-
ary Microbiology 50: 479–488.
Berg A, Rødseth OM, Tangerås A, Hansen T (2006) Time of
Acknowledgement vaccination influences development of adhesions, growth
SWIM is part of the Research Council of Norway (RCN) and spinal deformities in Atlantic salmon Salmo salar. Dis-
funded project 199728 SALMOWA. eases of Aquatic Organisms 69: 239–248.

Reviews in Aquaculture (2013) 5, 33–57

50 ª 2013 Wiley Publishing Asia Pty Ltd
 Salmon Welfare Index Model (SWIM 1.0)

Bergheim A, Gausen M, Naess A, Holland PM, Krogedal P, Bracke MBM, Hulsegge B, Keeling L, Blokhuis HJ (2004b)
Crampton V (2006) A newly developed oxygen injection Decision support system with semantic model to assess the
system for cage farms. Aquacultural Engineering 34: 40–46. risk of tail biting in pigs: 2. ‘Validation’. Applied Animal
Bernardet JF (1998) Cytophaga, Flavobacterium, Flexibacter Behaviour Science 87: 45–54.
and Chryseobacterium infections in cultured marine fish. Bracke MBM, Zonderland JJ, Bleumer EJB (2007a) Expert
Fish Pathology 33: 229–238. judgement on enrichment materials for pigs validates preli-
Berridge KC (2004) Motivation concepts in behavioral neuro- minary RICHPIG model. Applied Animal Behaviour Science
science. Physiology and Behavior 81: 179–209. 104: 1–13.
Bjerknes V, Duston J, Knox D, Harmon P (1992) Importance of Bracke MBM, Zonderland JJ, Bleumer EJB (2007b) Expert
body size for acclimation of underyearling Atlantic salmon consultation on weighting factors of criteria for assessing
parr (Salmo salar L.) to seawater. Aquaculture 104: 357–366. environmental enrichment materials for pigs. Applied Ani-
Bolger T, Conolly PL (1989) The selection of suitable indices mal Behaviour Science 104: 14–23.
for the measurement and analysis of fish condition. Journal Bracke MBM, Edwards SA, Metz JHM, Noordhuizen JPTM, Algers
of Fish Biology 34: 171–182. B (2008) Synthesis of semantic modelling and risk analy-
Bouck GR, Smith SD (1979) Mortality of experimentally de- sis methodology applied to animal welfare. Animal 2: 1061–
scaled smolts of coho salmon (Oncorhynchus-kisutch) in 1072.
fresh and salt-water. Transactions of the American Fisheries Braithwaite VA (2010) Do Fish Feel Pain? Oxford University
Society 108: 67–69. Press, Oxford.
Bowers JM, Mustafa A, Speare DJ, Conboy GA, Brimacombe Braithwaite VA, Boulcott P (2007) Pain perception, aver-
M, Sims DE et al. (2000) The physiological response of sion and fear in fish. Diseases of Aquatic Organisms 75: 131–
Atlantic salmon, Salmo salar L., to a single experimental 138.
challenge with sea lice, Lepeoptheirus salmonis. Journal of Braithwaite VA, Huntingford FA (2004) Fish and welfare: do
Fish Diseases 23: 165–172. fish have the capacity for pain perception and suffering?
Boxaspen K (2006) A review of the biology and genetics of sea Animal Welfare 13: 87–92.
lice. ICES Journal of Marine Science 63: 1304–1316. Bricknell IR, Dalesman SJ, O’Shea B, Pert CC, Luntz AJM
Bracke MBM (2008) RICHPIG: a semantic model to assess (2006) Effect of environmental salinity on sea lice Lepeopht-
enrichment materials for pigs. Animal Welfare 17: 289–304. heirus salmonis settlement success. Diseases of Aquatic
Bracke MBM (2011) Review of wallowing in pigs: description Organisms 71: 201–212.
of the behaviour and its motivational basis. Applied Animal Broom DM (2007) Cognitive ability and sentience: which
Behaviour Science 20: 347–363. aquatic animals should be protected? Diseases of Aquatic
Bracke MBM, Spoolder HAM (2011) Review of wallowing in Organisms 75: 99–108.
pigs: implications for animal welfare. Animal Welfare 20: Chandroo KP, Duncan IJH, Moccia RD (2004a) Can fish suf-
347–363. fer? Perspectives on sentience, pain, fear and stress. Applied
Bracke MBM, Metz JHM, Spruijt BM (1999a) Overall welfare Animal Behaviour Science 86: 225–250.
reviewed. Part 2: assessment tables and schemes. Netherlands Chandroo KP, Yue S, Moccia RD (2004b) An evaluation of
Journal of Agricultural Science 47: 293–305. current perspectives on consciousness and pain in fishes.
Bracke MBM, Spruijt BM, Metz JHM (1999b) Overall welfare Fish and Fisheries 5: 281–295.
assessment reviewed. Part 1: is it possible?. Netherlands Jour- Chervova LS (1997) Pain sensitivity and behaviour of fishes.
nal of Agricultural Science 47: 279–291. Journal of Ichthyology 37: 106–111.
Bracke MBM, Spruijt BM, Metz JHM (1999c) Overall welfare Clayton RD, Stevenson TL, Summerfelt RC (1998) Fin erosion
reviewed. Part 3: welfare assessment based on needs and in intensively cultured walleyes and hybrid walleyes. Progres-
supported by expert opinion. Netherlands Journal of Agricul- sive Fish-Culturist 60: 114–118.
tural Science 47: 307–322. Coyne R, Smith P, Dalsgaard I, Nilsen H, Kongshaug H, Bergh
Bracke MBM, Metz JHM, Spruijt BM, Schouten WGP (2002a) Ø et al. (2006) Winter ulcer disease of post-smolt Atlantic
Decision support system for overall welfare assessment in salmon: an unsuitable case for treatment? Aquaculture 253:
pregnant sows. B: validation by expert opinion. Journal of 171–178.
Animal Science 8: 1835–1845. Crampton V, Hølland PM, Bergheim A, Gausen M, Næss A
Bracke MBM, Spruijt BM, Metz JHM, Schouten WGP (2002b) (2003) Oxygen effects on caged salmon. Fish Farming Inter-
Decision support system for overall welfare assessment in national Jun 2003 (pp. 26–27).
pregnant sows. A: model structure and weighting procedure. Damsgård B, Sørum U, Ugelstad I, Eliassen RA, Mortensen A
Journal of Animal Science 8: 1819–1834. (2004) Effects of feeding regime on susceptibility of Atlantic
Bracke MBM, Hulsegge B, Keeling L, Blokhuis HJ (2004a) salmon (Salmo salar) to coldwater vibriosis. Aquaculture
Decision support system with semantic model to assess the 239: 37–46.
risk of tail biting in pigs: 1. Modelling. Applied Animal Davison W (1997) Training and its effects on teleost fish.
Behaviour Science 87: 31–44. Comparative Biochemistry and Physiology 94: 1–10.

Reviews in Aquaculture (2013) 5, 33–57

ª 2013 Wiley Publishing Asia Pty Ltd 51
L. H. Stien et al.

De Mol RM, Schouten WGP, Evers E, Drost WC, Houwers European Food Safety Authority (EFSA) (2009) General
HWJ, Smits AC (2004) Integrale welzijnsbeoordeling leghen- approach to fish welfare and to the concept of sentience in
nen. (Integrated welfare assessment laying hens). Report fish. EFSA Journal 954: 9–27.
239. Agrotechnology & Food Innovations of Wageningen Farrel AP (2005) Bulk oxygen uptake measured with over
University and Research Centre, Wageningen. 60,000 kg of adult salmon during live-haul transportation at
De Mol MR, Schouten WGP, Evers E, Drost WC, Houwers sea. Aquaculture 254: 646–652.
HWJ, Smits AC (2006) A computer model for welfare Fernö A, Huse I, Juell J-E, Bjordal Å (1995) Vertical distribu-
assessment of husbandry systems for laying hens. Nether- tion of Atlantic salmon (Salmo salar L.) in net pens: trade-
lands Journal of Agricultural Science 54: 57–168. off between surface light avoidance and food attraction.
Deitch EJ, Fletcher GL, Peterson LH, Costa LASF, Shears MA, Aquaculture 132: 285–296.
Driedzic WR et al. (2006) Cardiorespiratory modifications, Finstad B, Bjørn PA, Grimnes A, Hvidsten NA (2000) Labora-
and. limitations, in post-smolt growth hormone transgenic tory and field investigations of salmon lice (Lepeophtheirus
Atlantic salmon (Salmo salar). Journal of Experimental Biol- salmonis Krøyer) infestation on Atlantic salmon (Salmo salar
ogy 209: 1310–1325. L.) postsmolts. Aquaculture Research 31: 795–803.
Dempster T, Juell J-E, Fosseidengen JE, Fredheim A, Lader P Finstad G, Bjørn PA, Todd CD, Whoriskey F, Gargan PG,
(2008) Behaviour and growth of Atlantic salmon (Salmo sa- Forde G et al. (2011) The effect of sea lice on Atlantic sal-
lar L.) subjected to short-term submergence in commercial mon and other salmonid species. In: Aas Ø, Einum S,
scale sea-cages. Aquaculture 276: 103–111. Klemetsen A, Skurdal J (eds)Atlantic Salmon Ecology, pp.
Dempster T, Korsøen Ø, Folkedal O, Juell J-E, Oppdal F 253–276. Blackwell Publishing, Oxford.
(2009) Submergence of Atlantic salmon (Salmo salar L.) in Fjelldal PG, Lock E-J, Grotmol S, Totland GK, Nordgarden U,
commercial scale sea-cages: a potential short-term solution Flik G et al. (2006) Impact of smolt production strategy on
to poor surface conditions. Aquaculture 288: 254–263. vertebral growth and mineralisation during smoltification
Duston J (1994) Effect of salinity on survival and growth of and the early seawater phase in Atlantic salmon (Salmo salar
Atlantic salmon (Salmo salar) parr and smolt. Aquaculture L.). Aquaculture 261: 715–728.
121: 115–124. Fjelldal PG, Hansen TJ, Berg AE (2007) A radiological study
Einen O, Waagan B, Thomassen MS (1998) Starvation prior to on the development of vertebral deformities in cultured
slaughter in Atlantic salmon (Salmo salar) I. Effects on Atlantic salmon (Salmo salar L.). Aquaculture 273: 721–728.
weight loss, body shape, slaughter- and fillet-yield, proximate Fjelldal PG, Hansen T, Breck O, Sandvik R, Waagbø R, Berg A
and fatty acid composition. Aquaculture 166: 85–104. et al. (2008) Supplementation of dietary minerals during the
Einen O, Mørkøre T, Rørå AMB, Thomassen MS (1999) Feed early seawater phase increase vertebral strength and reduce
ration prior to slaughter – a potential tool for managing the prevalence of vertebral deformities in fast growing
product quality of Atlantic salmon (Salmo salar). Aquacul- under-yearling Atlantic salmon (Salmo salar L.) smolt.
ture 178: 149–169. Aquaculture Nutrition 15: 366–378.
Ellis T, Oidtmann B, St-Hilaire S, Turnbull J, North B, Fjelldal PG, Glover KA, Skaala O, Imsland A, Hansen TJ
MacIntyre C et al. (2008) Fin erosion in farmed fish. In: (2009a) Vertebral body mineralization and deformities in
Branson E (ed.) Fish Welfare, pp. 121–149. Blackwell, cultured Atlantic salmon (Salmo salar L.): effects of genetics
Oxford. and off-season smolt production. Aquaculture 296: 36–44.
Endal HP, Taranger GL, Stefansson S, Hansen T (2000) Effects Fjelldal PG, Hansen T, Breck O, Sandvik R, Waagbø R, Berg A
of continuous additional light on growth and sexual matu- et al. (2009b) Supplementation of dietary minerals during
rity in Atlantic salmon, reared in sea cages. Aquaculture 191: the early seawater phase increase vertebral strength and
337–349. reduce the prevalence of vertebral deformities in fast-grow-
European Food Safety Authority (EFSA) (2006a) Scientific ing under-yearling Atlantic salmon (Salmo salar L.) smolt.
opinion on the risks of poor welfare in intensive calf farm- Aquaculture Nutrition 15: 366–378.
ing systems. An update of the Scientific Veterinary Commit- Fleming IA (1998) Pattern and variability in the breeding sys-
tee Report on the Welfare of Calves. EFSA-Q-2005-014. tem of Atlantic salmon (Salmo salar), with comparisons to
EFSA Journal 366: 1–36. other salmonids. Canadian Journal of Fisheries and Aquatic
European Food Safety Authority (EFSA) (2006b) Scientific Sciences 55: 59–76.
report on the risks of poor welfare in intensive calf farming Fleming IA, Einum S (2011) Reproductive ecology: a tale of
systems. An update of the Scientific Veterinary Committee two sexes. In: Aas Ø, Einum S, Klemetsen A, Skurdal J (eds)
Report on the Welfare of Calves. EFSA-Q-2005-014 Atlantic Salmon Ecology, pp. 33–65. Blackwell Publishing
(144 pp.), Annex to EFSA (2006a)). Ltd, Oxford.
European Food Safety Authority (EFSA) (2008) Animal welfare Flood MJ, Purser JG, Carter CG (2011) The effects of changing
aspects of husbandry systems for farmed Atlantic salmon – feeding frequency simultaneously with seawater transfer in
Scientific Opinion of the Panel on Animal Health and Wel- Atlantic salmon Salmo salar L. smolt. Aquaculture Interna-
fare. Annex to the EFSA Journal 736: 122 pp. tional 20: 29–40.

Reviews in Aquaculture (2013) 5, 33–57

52 ª 2013 Wiley Publishing Asia Pty Ltd
 Salmon Welfare Index Model (SWIM 1.0)

Galhardo L, Oliveira RF (2009) Psychological stress and welfare Herbinger CM, Friars GW (1991) Correlation between condi-
in fish. Annual Reviews of Biomedical Sciences 11: 1–20. tion factor and total lipid content in Atlantic salmon, Salmo
Gil-Martens L (2010) Inflammation as a potential risk factor salar L., parr. Aquaculture Research 4: 527–529.
for spinal deformities in farmed Atlantic salmon (Salmo Hevrøy EM, Boxaspen K, Oppedal F, Taranger GL, Holm JC
salar L.). Journal of Applied Ichthyology 26: 350–354. (2003) The effect of artificial light treatment and depth on the
Gjerde B, Pante MJ, Baeverfjord G (2005) Genetic variation infestation of the sea louse Lepeophtheirus salmonis on Atlan-
for a vertebral deformity in Atlantic salmon (Salmo salar). tic salmon (Salmo salar L.) culture. Aquaculture 220: 1–14.
Aquaculture 244: 77–87. Hevrøy EM, Waagbø R, Torstensen BE, Takle H, Stubhaug I,
Goncalves J, Carraca S, Damasceno-Oliveira A, Fernandez- Jørgensen SM et al. (2011) Ghrelin is involved in voluntary
Duran B, Diaz J, Wilson J et al. (2006) Effect of reduction anorexia in Atlantic salmon raised at elevated sea tempera-
in water salinity on osmoregulation and survival of large tures. General and Comparative Endocrinology 175: 118–134.
Atlantic salmon held at high water temperature. North Hoar WS (1988) The physiology of smolting salmonids. In:
American Journal of Aquaculture 68: 324–329. Hoar WS, Randall DJ (eds) Fish Physiology, Vol. 11, pp.
Grimnes A, Jakobsen PJ (1996) The physiological effects of sal- 275–343. Academic Press, London.
mon lice infection on post-smolt of Atlantic salmon. Journal Holst JC, Jakobsen P, Nilsen F, Holm M, Asplin L, Aure J
of Fish Biology 48: 1179–1194. (2003) Mortality of seaward-migrating post-smolts of Atlan-
Grini A, Hansen T, Berg A, Wargelius A, Fjelldal PG (2011) tic salmon due to salmon lice infection in Norwegian sal-
The effect of water temperature on vertebral deformities and mon stocks. In: Mills D (ed.) Salmon at the Edge, pp. 136–
vaccine-induced abdominal lesions in Atlantic salmon, 137. Blackwell Science, Oxford.
Salmo salar L. Journal of Fish Diseases 34: 531–546. Huntingford FA, Kadri S (2008) Welfare and fish. In: Branson
Grøttum JA, Sigholt T (1998) A model for oxygen consump- EJ (ed.) Fish Welfare, pp. 19–31. Blackwell Publishing,
tion of Atlantic salmon (Salmo salar) based on measure- Oxford.
ments of individual fish in a tunnel respirometer. Huntingford FA, Adams C, Braithwaite VA, Kadri S, Pottinger
Aquaculture Engineering 17: 241–251. TG, Sandøe P et al. (2006) Current issues in fish welfare.
Hamre K, Christiansen R, Waagbø R, Maage A, Torstensen Journal of Fish Biology 68: 332–372.
BE, Lygren B et al. (2004) Antioxidant vitamins, minerals Huse I, Holm JC (1993) Vertical distribution of Atlantic sal-
and lipid levels in diets for Atlantic salmon (Salmo salar, mon (Salmo salar) as a function of illumination. Journal of
L.): effects on growth performance and fillet quality. Aqua- Fish Biology 43: 147–156.
culture Nutrition 10: 113–123. Iversen M, Finstad B, Nilssen KJ (1998) Recovery from loading
Handeland SO, Berge Å, Björnsson B, Stefansson SO (1998) and transport stress in Atlantic salmon (Salmo salar L.)
Effects of temperature and salinity on osmoregulation and smolts. Aquaculture 168: 387–394.
growth of Atlantic salmon (Salmo salar L.) smolts in seawa- Iversen M, Finstad B, McKinley RS, Eliassen RA, Carlsen KT,
ter. Aquaculture 168: 289–302. Evjen T (2005) Stress responses in Atlantic salmon (Salmo
Handeland SO, Berge Å, Björnsson BT, Lie Ø, Stefansson SO salar L.) smolts during commercial well boat transports, and
(2000) Seawater adaptation by out-of-season Atlantic sal- effects on survival after transfer to sea. Aquaculture 243:
mon (Salmo salar L.) smolts at different temperatures. 373–382.
Aquaculture 181: 377–396. Javaid MY, Anderson JM (1967) Thermal acclimation and
Handeland SO, Björnsson B, Arnesen AM, Stefansson SO temperature selection in Atlantic salmon, Salmo salar, and
(2003) Seawater adaptation and growth of post-smolt Atlan- rainbow trout, Salmo gairdneri. Journal of the Fisheries
tic salmon (Salmo salar) of wild and farmed strains. Aqua- Research Board of Canada 24: 1507–1513.
culture 220: 367–384. Jensen P, Toates FM (1993) Who needs ‘behavioural needs’?
Handeland SO, Imsland AK, Stefansson SO (2008) The effect Motivational aspects of the needs of animals Applied Animal
of temperature and fish size on growth, feed intake, food Behaviour Science 37: 161–181.
conversion efficiency and stomach evacuation rate of Atlan- Jobling M, Reinsnes T-G (1986) Physiological and social con-
tic salmon post-smolts. Aquaculture 283: 36–42. straints on growth of Arctic charr, Salvelinius alpines L.: an
Hansen T, Fjelldal PG, Yurtseva A, Berg A (2010) A possible investigation of factors leading to stunting. Journal of Fish
relation between growth and number of deformed vertebrae Biology 28: 379–384.
in Atlantic salmon (Salmo salar L.). Journal of Applied Ich- Johansson D, Ruohonen K, Kiessling A, Oppedal F, Stiansen
thyology 26: 355–359. JE, Kelly M et al. (2006) Effect of environmental factors on
Håstein T, Scarfe AD, Lund VL (2005) Science based assess- swimming depth preferences of Atlantic salmon (Salmo salar
ment of welfare: aquatic animals. Revue Scientifique et Tech- L.) and temporal and spatial variations in oxygen levels in
nique (International Office of Epizootics) 24: 529–547. sea cages at a fjord site. Aquaculture 254: 594–605.
Helfman GS, Colette BB, Douglas EF (eds) (1997) Oxygen, Johansson D, Juell J-E, Oppedal F, Stiansen JE, Ruohonen K
metabolism and energetics. In: The Diversity of Fishes, pp. (2007) The influence of the pycnocline and cage resistance
51–68. Blackwell Science, Malden, MA. on current flow, oxygen flux and swimming behaviour of

Reviews in Aquaculture (2013) 5, 33–57

ª 2013 Wiley Publishing Asia Pty Ltd 53
L. H. Stien et al.

Atlantic salmon (Salmo salar L.) in production cages. Aqua- Latremouille DN (2003) Fin erosion in aquaculture and
culture 265: 271–287. natural environments. Reviews in Fisheries Science 11: 315–
Johansson D, Ruohonen K, Juell J-E, Oppedal F (2009) Swim- 335.
ming depth and thermal history of individual Atlantic sal- Leung KY, Stevenson RMW (1988) Characteristics and distri-
mon (Salmo salar L.) in production cages under different bution of extracellular proteases from Aeromonas hydrophila.
ambient temperature conditions. Aquaculture 290: 296–303. Journal of General Microbiology 134: 151–160.
Jones HAC, Hansen LA, Noble C, Damsgård B, Broom DM, Lijalad M, Powell MD (2009) Effects of lower jaw deformity
Pearce GP (2010) Social network analysis of behavioural on swimming performance and recovery from exhaustive
interactions influencing fin damage development in Atlantic exercise in triploid and diploid Atlantic salmon Salmo salar
salmon (Salmo salar) during feed-restriction. Applied Animal L. Aquaculture 290: 145–154.
Behaviour Science 127: 139–151. Lorenzen E (1999) Infeksjoner med Flavobacterium og Flexi-
Jonsson N, Jonsson B, Hansen LP (1997) Changes in proxi- bacter (in Norwegian). In: Poppe T (ed.) Fiskehelse og syk-
mate composition and estimates of energetic costs during dommer, pp. 97–107. Universitetsforlaget, Oslo.
upstream migration and spawning in Atlantic salmon Salmo Lunder T, Evensen Ø, Holstad G, Håstein T (1995) ‘Winter
salar. Journal of Animal Ecology 66: 425–436. ulcer’ in the Atlantic salmon Salmo salar. Pathological and
Jørgensen EH, Jobling M (1992) Feeding behaviour and effect bacteriological investigations and transmission experiments.
of feeding regime on growth of Atlantic salmon, Salmo salar. Diseases of Aquatic Organisms 23: 39–49.
Aquaculture 101: 135–146. MacIntyre CM, Ellis T, North BP, Turnbull JF (2008) The
Jørgensen EH, Jobling M (1994) Feeding and growth of exer- influences of water quality on the welfare of farmed rainbow
cised and unexercised juvenile Atlantic salmon in freshwater, trout: a review. In: Branson EJ (ed.) Fish Welfare, pp. 150–
and performance after transfer to seawater. Aquaculture 184. Blackwell Publishing, Oxford.
International 2: 154–164. Mauel MJ, Miller DL (2002) Piscirickettsiosis and piscirickett-
Juell J-E (1995) The behaviour of Atlantic salmon in relation siosis-like infections in fish: a review. Veterinary Microbiol-
to efficient cage-rearing. Reviews in Fish Biology and Fisheries ogy 87: 279–289.
5: 320–335. McCormick SD, Shrimpton JM, Carey JB, O’Dea MF, Sloan
Juell J-E, Fosseidengen JE (2004) Use of artificial light to con- KE, Moriyama S et al. (1998) Repeated acute stress reduces
trol swimming depth and fish density of Atlantic salmon growth rate of Atlantic salmon parr and alters plasma levels
(Salmo salar) in production cages. Aquaculture 233: 269–282. of growth hormone, insulin-like growth factor I and corti-
Juell J-E, Bjordal Å, Fernö A, Huse I (1994) Effect of feeding sol. Aquaculture 168: 221–235.
intensity on food intake and growth of Atlantic salmon, McKenzie DJ, Higgs DA, Dosanjh BS, Deacon G, Randall DJ
Salmo salar L., in sea cages. Aquaculture and Fisheries Man- (1998) Dietary fatty acid composition influences swimming
agement 25: 453–464. performance in Atlantic salmon (Salmo salar) in seawater.
Juell J-E, Oppedal F, Boxaspen K, Taranger GL (2003) Sub- Fish Physiology and Biochemistry 19: 111–122.
merged light increases swimming depth and reduces fish Mørkøre T, Rørvik K-A (2001) Seasonal variations in growth,
density of Atlantic salmon Salmo salar L. In production feed utilisation and product quality of farmed Atlantic sal-
cages. Aquaculture Research 34: 469–477. mon (Salmo salar) transferred to seawater as 0 + smolts or
Juell J-E, Fosseidengen JE, Vågseth T, Kristiansen T (2008) 1 + smolts. Aquaculture 199: 145–157.
Wealth Report. Effects of Environmental Variation and Nash RDM, Valencia AH, Geffen AJ (2006) The origin of Ful-
Crowding and Sorting Events on Swimming and Feeding ton’s condition factor-setting the record straight. Fisheries
Behaviour of Caged Atlantic Salmon. European Commission, 31: 236–238.
Brussels. Nilsen A, Bjøru B, Vigen J, Oppedal F, Høy E (2010) Evaluer-
Kadri S, Metcalfe NB, Huntingford FA, Thorpe JE (1991) ing av metoder for badebehandling mot lakselus i stormerd
Daily feeding rhythms in Atlantic salmon in sea cages. (in Norwegian). Veterinærinstituttets rapportserie 17-2010.
Aquaculture 92: 219–224. Oslo, Norway.
Kent ML, Poppe TT (2002) Infectious diseases of cold-water Noble C, Jones HAC, Damsgård B, Flood MJ, Midling KØ,
fish in marine and brackish water. In: Lim LHS (ed.) Dis- Roque A et al. (2012) Injuries and deformities in fish: their
eases and Disorders of Finfish in Cage Culture, pp. 61–105. potential impacts upon aquacultural production and welfare.
CABI Publishing, Oxford. Fish Physiology and Biochemistry 38: 61–83.
Korsøen ØJ, Dempster T, Fjelldal PG, Oppedal F, Kristiansen Noga EJ (2000) Skin ulcers in fish: Pfiesteria and other etiolo-
TS (2009) Long-term culture of Atlantic salmon (Salmo gies. Toxicologic Pathology 28: 807–823.
salar L.) in submerged cages during winter affects behaviour, Nolan DT, Reilly P, Bonga SEW (1999) Infection with low
growth and condition. Aquaculture 296: 373–381. numbers of the sea louse, Lepeophtheirus salmonis, induces
Korte SM, Olivier B, Koolhaas JM (2007) A new animal wel- stress-related effects in post-smolt Atlantic salmon (Salmo
fare concept based on allostasis. Physiology and Behaviour salar). Canadian Journal of Fisheries and Aquatic Sciences 56:
92: 422–428. 947–959.

Reviews in Aquaculture (2013) 5, 33–57

54 ª 2013 Wiley Publishing Asia Pty Ltd
 Salmon Welfare Index Model (SWIM 1.0)

Nordgarden U, Oppedal F, Taranger GL, Hemre G-I, Hansen tion of river running Atlantic salmon. Journal of Fish Biology
T (2003) Seasonally changing metabolism in Atlantic salmon 52: 334–349.
(Salmo salar L.) I – Growth and feed conversion ratio. Peterson RH, Harmon PR (2005) Changes in condition factor
Aquaculture Nutrition 9: 287–293. and gonadosomatic index in maturing and non-maturing
O’Flynn FM, McGeachy SA, Friars GW, Benfey TJ, Bailey JK Atlantic salmon (Salmo salar L.) in Bay of Fundy sea cages,
(1997) Comparison of cultured triploid and diploid Atlantic and the effectiveness of photoperiod manipulation in reduc-
salmon (Salmo salar L.). ICES Journal of Marine Science 54: ing early maturation. Aquaculture Research 36: 882–889.
1160–1165. Pike AW, Wadsworth SL (1999) Sealice on Salmonids: their
Oppedal F, Taranger GL, Juell J-E, Fosseidengen JE, Hansen T biology and control. Advances in Parasitology 44: 233–337.
(1997) Light intensity affects growth and sexual maturation Plantalech Manel-La N, Thorstad EB, Davidsen JG, Økland F,
of Atlantic salmon (Salmo salar) postsmolts in sea cages. Sivertsgård R, McKinley RS et al. (2009) Vertical movements
Aquatic Living Resources 10: 351–357. of Atlantic salmon post-smolts relative to measures of salinity
Oppedal F, Taranger GL, Juell J-E, Hansen T (1999) Growth, and water temperature during the first phase of the marine
osmoregulation and sexual maturation of underyearling migration. Fisheries Management and Ecology 16: 147–154.
Atlantic salmon smolt Salmo salar L. exposed to different Priede M (2002) Biology of salmon. In: Stead M, Lindsay L
intensities of continuous light in sea cages. Aquaculture (eds) Handbook of Salmon Farming, pp. 1–35. Praxis Pub-
Research 30: 491–499. lishing, Chichester.
Oppedal F, Juell J-E, Taranger GL, Hansen T (2001) Artificial Remen M, Oppedal F, Torgersen T, Imsland A, Olsen RE
light and season affects vertical distribution and swimming (2012) Effects of cyclic environmental hypoxia on physiol-
behaviour of post-smolt Atlantic salmon in sea cages. Jour- ogy and feed intake of post-smolt Atlantic salmon: initial
nal of Fish Biology 58: 1570–1584. responses and acclimation. The Aquaculture 326–329: 148–
Oppedal F, Taranger GL, Hansen T (2003) Growth perfor- 155.
mance and sexual maturation in diploid and triploid Atlan- Rikardsen AH, Haugland M, Bjørn PA, Finstad B, Knudsen R,
tic salmon (Salmo salar L.) in seawater tanks exposed to Dempson JB et al. (2004) Geographic differences in early
continuous light or simulated natural photoperiod. Aquacul- marine feeding of Atlantic salmon post-smolts in Norwegian
ture 215: 145–162. fjords. Journal Fish Biology 64: 1–25.
Oppedal F, Berg A, Olsen RE, Taranger GL, Hansen T (2006) Rodger HD, McArdle JF (1996) An outbreak of amoebic gill
Photoperiod in seawater influence seasonal growth and disease in Ireland. Veterinary Record 139: 348–349.
chemical composition in autumn sea-transferred Atlantic Rørå AMB, Kvåle A, Mørkøre T, Rørvik K-A, Steien SH,
salmon (Salmo salar L.) given two vaccines. Aquaculture Thomassen MS (1998) Process yield, colour and sensory
254: 396–410. quality of smoked Atlantic salmon (Salmo salar) in relation
Oppedal F, Juell J-E, Johansson D (2007) Thermo- and photo- to raw material characteristics. Food Research International
regulatory behaviour of caged Atlantic salmon: Implication 31: 601–609.
for photoperiod management and fish welfare. Aquaculture Ross NW, Firth KJ, Aniping W, Burka JF, Johnson SC (2000)
265: 70–81. Changes in hydrolytic enzyme activities of naı̈ve Atlantic sal-
Oppedal F, Dempster T, Stien LH (2011a) Environmental driv- mon Salmo salar skin mucus due to infection with the sal-
ers of Atlantic salmon behaviour in sea-cages: a review. mon louse Lepeophtheirus salmonis and cortisol
Aquaculture 311: 1–18. implantation. Diseases of Aquatic Organisms 41: 43–51.
Oppedal F, Vågset T, Dempster T, Juell J-E, Johansson D (2011b) Schreck CB, Olla BL, Davis MW (1997) Behavioral responses
Fluctuating sea-cage environments modify the effects of to stress. In: Iwama GK, Pickering AD, Sumpter JP, Schreck
stocking densities on production and welfare parameters of CB (eds) Fish Stress and Health in Aquaculture, pp. 145–170.
Atlantic salmon (Salmo salar L.). Aquaculture 315: 361–368. Cambridge University Press, Cambridge.
Ostland VE, Byrne PJ, Hoover G, Ferguson HW (2000) Necro- Segner H, Sundh H, Buchmann K, Douxfils J, Sundell KS,
tic myositis of rainbow trout, Oncorhynchus mykiss (Wal- Mathieu C et al. (2012) Health of farmed fish: its relation to
baum): proteolytic characteristics of a crude extracellular fish welfare and its utility as welfare indicator. Fish Physiol-
preparation from Flavobacterium psychrophilum. Journal of ogy and Biochemistry 38: 85–105.
Fish Disease 23: 329–336. Shepard KL (1994) Functions of fish mucus. Reviews in Fish
Panksepp J (2005) Affective consciousness: core emotional feel- Biology and Fisheries 4: 301–429.
ings in animals and humans. Cognition and Consciousness Shimmura T, Bracke MBM, De Mol RM, Hirahara S, Uetake
14: 30–80. K, Tanaka T (2011) Overall welfare assessment of laying
Pelis RM, McCormick SD (2003) Fin development in stream- hens: comparing science based, environment based and ani-
and hatchery-reared Atlantic salmon. Aquaculture 220: 525– mal based assessments. Animal Science Journal 82: 150–160.
536. Sigholt T, Finstad B (1990) Effect of low temperature on sea-
Persson P, Sundell K, Björnsson BT, Lundqvist H (1998) Cal- water tolerance in Atlantic Salmon (Salmo salar) Smolts.
cium metabolism and osmoregulation during sexual matura- Aquaculture 84: 167–172.

Reviews in Aquaculture (2013) 5, 33–57

ª 2013 Wiley Publishing Asia Pty Ltd 55
L. H. Stien et al.

Skugor S, Glover KA, Nilsen F, Krasnow A (2008) Local and (Salmo salar L.) following a bath challenge with Vibrio an-
systemic gene expression responses of Atlantic salmon guillarum and Aeromonas salmonicida. Fish and Shellfish
(Salmo salar L.) to infection with the salmon louse (Lepe- Immunology 7: 317–325.
ophtheirus salmonis). BMC Genomics 9: 498. Takle H, Castro V, Grisdale-Helland B, Helland S, Tørud B,
Skuladottir GV, Schioth HB, Gudmundsdottir E, Richards B, Kristensen T (2010) Aerob utholdenhetstrening for bedret
Gardarsson F, Jonsson L (1990) Fatty-acid composition of hjertefunksjon og helse hos oppdrettslaks: oppfølging og vi-
muscle, heart and liver lipids in Atlantic salmon, Salmo sa- dereutvikling av konseptet trening av fisk (In Norwegian).
lar, at extremely low environmental-temperature. Aquacul- Final Project Report for FHF, Norway. ISBN: 978-82-7251-
ture 85: 71–80. 738-9. 46 pp.
Smith IP, Metcalfe NB, Huntingford FA, Kadri S (1993) Daily Taranger GL, Daae H, Jørgensen KO, Hansen T (1995) Effects
and seasonal patterns in the feeding behaviour of Atlantic sal- of continuous light on growth and sexual maturation in
mon (Salmo salar L.) in a sea cage. Aquaculture 117: 165–178. seawater reared Atlantic salmon. In: Goetz F, Thomas P
Sneddon LU (2003) The evidence for pain in fish: the use of (eds) Proceedings of the 5th International Symposium on the
morphine as an analgesic. Applied Animal Behaviour Science Reproductive Physiology of Fish. University of Texas, Austin,
83: 153–162. Texas, 2–8 July, 1995, p. 200.
Soares S, Green DM, Turnbull JF, Crumlish M, Murray AG Thorstad EB, Whoriskey F, Rikardsen AH, Aarestrup K (2011)
(2011) A baseline method for benchmarking mortality losses Aquatic Nomads: the Life and Migrations of the Atlantic
in Atlantic salmon (Salmo salar) production. Aquaculture Salmon. In: Aas Ø, Einum S, Klemetsen A, Skurdal J (eds)
314: 7–12. Atlantic Salmon Ecology, pp. 1–32. Blackwell Publishing,
Stefansson SO, Björnsson BT, Ebbesson LO, McCormick SD Oxford.
(2008) Smoltification. In: Finn RN, Kapoor BG (eds) Fish Toften H, Arnesen A, Jobling AM (2003) Feed intake, growth
Larval Physiology, pp. 639–681. Science Publishers, Inc. and ionoregulation in Atlantic salmon (Salmo salar L.)
Enfield (NH) & IBH Publishing Co. Pvt Ltd, New Delhi. smolts in relation to dietary addition of a feeding stimulant
Steinhausen MF, Sandblom E, Eliason EJ, Verhille C, Farrell and time of seawater transfer. Aquaculture 217: 647–662.
AP (2008) The effect of acute temperature increases on the Torgersen T, Bracke BM, Kristiansen TS (2011) Reply to Dig-
cardiorespiratory performance of resting and swimming gles et al. (2011): ecology and welfare of aquatic animals in
sockeye salmon (Oncorhynchus nerka). Journal of Experimen- wild capture fisheries. Reviews in Fish Biology and Fisheries
tal Biology 211: 3915–3926. 21: 767–769.
Stephen C, Ribble S (1995) An evaluation of surface moribund Totland GK, Kryvi H, Jødestøl KA, Christiansen EN, Tangerås
salmon as indicators of seapen disease status. Aquaculture A, Slinde E (1987) Growth and composition of the swim-
133: 1–8. ming muscle of adult Atlantic salmon (Salmo salar L.) dur-
Stevens ED, Sutterlin A, Cook T (1998) Respiratory metabo- ing long-term sustained swimming. Aquaculture 66: 299–
lism and swimming performance in growth hormone trans- 313.
genic Atlantic salmon. Canadian Journal of Fisheries and Totland GK, Hjeltnes BK, Flood PR (1996) Transmission of
Aquatic Sciences 55: 2028–2035. infectious salmon anaemia (ISA) through natural secretions
St-Hilaire S, Ribble C, Whitaker DJ, Kent M (1998) Prevalence and excretions from infected smolts of Atlantic salmon
of Kudoa thyrsites in sexually mature and immature pen- Salmo salar during their presymptomatic phase. Diseases of
reared Atlantic Salmon (Salmo salar) in British Columbia, Aquatic Organisms 26: 25–31.
Canada. Aquaculture 162: 1–2. Tucker CS, Sommerville C, Wootten R (2002) Does size really
Stien LH, Kristiansen T, Danielsen TL, Torgersen T, Oppedal matter? Effects of fish surface area on the settlement and ini-
F, Fosseidengen JE (2009) Fra utsett til slakt (In Norwe- tial survival of Lepeophtheirus salmonis, an ectoparasite of
gian). In: Agnalt AL, Bakketeig I, Haug T, Knutsen JA, Ops- Atlantic salmon Salmo salar Diseases of Aquatic Organisms
tad I (eds) Kyst og Havbruk 2009, pp. 160–163. Institute of 49: 145–152.
Marine Research, Bergen. Turnbull JF, Richards RH, Robertson DA (1996) Gross, histo-
Sundh H, Kvamme BO, Fridell F, Olsen RE, Ellis T, Taranger logical and scanning electron microscopic appearance of
GL et al. (2010) Intestinal barrier function of Atlantic sal- dorsal fin rot in farmed Atlantic salmon, Salmo salar L.,
mon (Salmo salar L.) post smolts is reduced by common sea parr. Journal of Fish Diseases 19: 415–427.
cage environments and suggested as a possible physiological Turnbull JF, Adams CE, Richards RH, Robertson DA (1998)
welfare indicator. BMC Physiology 10: 22. Attack site and resultant damage during aggressive encoun-
Sveier H, Lied E (1998) The effect of feeding regime on ters in Atlantic salmon (Salmo salar L.) parr. Aquaculture
growth, feed utilisation and weight dispersion in large Atlan- 159: 345–353.
tic salmon (Salmo salar) reared in seawater. Aquaculture Turnbull J, Bell A, Adams C, Bron J, Huntingford F (2005)
165: 333–345. Stocking density and welfare of cage farmed Atlantic sal-
Svendsen YS, Bøgwald J (1997) Influence of artificial wound mon: application of multivariate analysis. Aquaculture 243:
and non-intact mucus layer on mortality of Atlantic salmon 121–132.

Reviews in Aquaculture (2013) 5, 33–57

56 ª 2013 Wiley Publishing Asia Pty Ltd
 Salmon Welfare Index Model (SWIM 1.0)

Tvenning H (1991) Sammenhengen mellom lengde og vekt. in Atlantic salmon (Salmo salar). Development Genes Evolu-
Fiskeoppdrett, Aschehoug, Oslo, p. 25. tion 215: 350–357.
Ursinus WW, Schepers F, de Mol RM, Bracke MBM, Metz WEALTH (2008) Welfare and health in sustainable aquacul-
JHM, Groot Koerkamp PWG (2009) COWEL: A decision ture. Report from EU project WEALTH, Available from:
support system husbandry systems for dairy cattle on wel- http://www.wealth.imr.no
fare. Animal Welfare 18: 454–552. Wicklund T, Dalsgaard I (1998) Occurrence and significance
Usher ML, Talbot C, Eddy FB (1991) Effects of transfer to sea- of atypical Aeromonas salmonicida in non-salmonid fish spe-
water on growth and feeding in Atlantic salmon smolts cies: a review. Diseases of Aquatic Organisms 32: 49–69.
(Salmo salar L.). Aquaculture 94: 309–326. Witten PE, Gil-Martens L, Hall BK, Huysseune A, Obach A
Vigen J (2008) Oxygen variation within a seacage. Master thesis. (2005) Compressed vertebrae in Atlantic salmon (Salmo
Department of Biology, University of Bergen, Bergen, 73 pp. salar): evidence for metaplastic chondrogenesis as a skeleto-
Wargelius A, Fjelldal PG, Hansen T (2005) Heat shock during genic response late in ontogeny. Diseases of Aquatic Organ-
early somitogenesis induces caudal vertebral column defects isms 64: 237–246.

Reviews in Aquaculture (2013) 5, 33–57

ª 2013 Wiley Publishing Asia Pty Ltd 57