Institute of Marine Research



In Norway farmed fish are protected by the same animal welfare legislation as land animals, and these law and regulations demand the fish farmers to have sufficient competence, technology and equipment that secure the animals welfare. However, since there is no established methods to assess or document fish welfare, it is impossible for the fish farmer to know how to comply with these regulations, and impossible for the food authorities to control or enforce them. The authorities and stakeholders are therefore strongly asking for scientific support to develop science-based tools and protocols for fish welfare assessment. To address this need, we will construct an integrating model and tool for overall welfare assessment of the most important farmed fish in Norway, the Atlantic salmon (Salmo salar L.).

Choice of Method

By overall assessment of welfare (OWA) we mean a systematic attempt to assess the welfare status of animals based on observations of the animals, their biological and physical environments, and available scientific knowledge (Bracke et al., 1999abc; Anon. 2001). There are two major approaches of creating a welfare assessment method; risk analysis and semantic modelling (Bracke et al. 2008). The prime objectives of risk analysis are to identify hazards, their consequence and probability of occurrence, and to find critical control points in the production process to avoid welfare and production risks like suffering and disease. In semantic modelling (SM) welfare is defined as the quality of life as perceived by the animals themselves and both positive and negative aspect of welfare are considered, and is therefore a risk benefit analysis (Bracke et al., 1999abc). SM has been designed for the purpose of formalized assessment of animal welfare based on available scientific information, including scientific knowledge and scientific descriptions of housing systems in terms of both environment-based and animal-based measures (Bracke et al., 2008). It was originally developed for assessment of housing systems for dry sows (Bracke et al., 2002ab), but it has also been applied to poultry (DeMol et al. 2006), to tail biting in pigs (Bracke et al., 2004) and to assess enrichment materials for pigs (Bracke, 2008).

Semantic Modelling

The first step of welfare assessment by semantic modelling approach is to collect a list of the species' basic needs, and to collect a list of scientific statements, obtained from a systematic literature review using the criterion that the statements are somehow relevant to assess welfare in aquaculture. The second step is to create a list of measurable or observable attributes (animal-based and environmental welfare indicators) from the scientific statements that can be linked to at least one need, to ensure that all attributes in the model are relevant to welfare from the animal's point of view. The attribute scores are then divided into levels that are mutually exclusive and cover the model's domain. As a result, all characteristics of a farming system, including the animals living there, are described by exactly one level of each attribute. This ensures that a generic calculation rule can be used such that any welfare advantage to for instance a cage system accrues to all systems with the same descriptive property, and only to them.

Each attribute level is linked to at least one scientific statement that provides the scientific basis of the model. The scientific evidence is used to rank the levels within each attribute and to construct weighting scores (WS) using weighting categories (WC) which relate to the different scientific paradigms to measure welfare scientifically. Scientific statements can be used to determine the weight of an attribute level since they identify relationships between the level and welfare performance criteria. When the scientific evidence is not conclusive, one or more levels are disqualified until all attribute levels and their rankings have a scientific basis. WC's classify welfare performance criteria, e.g. pain, illness, survival, fitness, abnormal behaviour, preference and more. The calculation rules for an overall welfare index suggested in Bracke et al. (2002a) calculates the welfare index on a scale from 0 to 10 as a weighted average score from attribute scores and weighting factors. When there is scientific evidence available more complex rules of interaction between attributes can be included into the calculation rules. We hypothesise that by using semantic modelling we will be able to create a science-based model for assessing salmon welfare in aquaculture, and that we will identify deficiencies in available knowledge, i.e. attributes, where there is not enough information to specify/divide it into sufficient levels, attribute levels where there is little, or no, information available, and attribute levels where the information is contradictory.


SWIM is part of the RCN and FHF funded project 199728 SALMOWA.


  1. Anonymous. 2001. Scientists' assessment of the impact of housing and management on animal welfare. Journal of Appl. Animal Welfare Science 4: 3-52.
  2. Bracke M.B.M., B.M. Spruijt, J.H.M. Metz. 1999a. Overall welfare assessment reviewed. Part 1: Is it possible? Netherlands Journal of Agricultural Science 47:279-291.
  3. Bracke M.B.M., J.H.M. Metz, B.M. Spruijt. 1999b. Overall welfare reviewed. Part 2: Assessment tables and schemes. Netherlands Journal of Agricultural Science 47: 293-305.
  4. Bracke M.B.M., B.M. Spruijt, J.H.M. Metz. 1999c. Overall welfare reviewed. Part 3: Welfare assessment based on needs and supported by expert opinion. Netherlands Journal of Agricultural Science 47: 307-322.
  5. Bracke M.B.M., Spruijt, B.M., Metz, J.H.M., Schouten, W.G.P., 2002a. Decision support system for overall welfare assessment in pregnant sows. A: Model structure and weighting procedure. Journal of Animal Science 8: 1819-1834.
  6. Bracke M.B.M., J.H.M. Metz, B.M. Spruijt, W.G.P. Schouten. 2002b. Decision support system for overall welfare assessment in pregnant sows. B: Validation by expert opinion. Journal of Animal Science 8: 1835-1845.
  7. Bracke M.B.M., B. Hulsegge, L. Keeling, H.J. Blokhuis. 2004. Decision support system with semantic model to assess the risk of tail biting in pigs: 1. Modelling. Applied Animal Behaviour Science 87: 31-44.
  8. Bracke M.B.M., 2008. RICHPIG: a semantic model to assess enrichment materials for pigs. Animal Welfare 17: 289-304.
  9. Bracke M.B.M., S.A. Edwards, J.H.M. Metz, J.P.T.M Noordhuizen, B. Algers.2008. Synthesis of semantic modeling and risk analysis methodology applied to animal welfare. Animal 2: 1061-1072.
  10. De Mol R.M., W.G.P. Schouten, E. Evers, H. Drost. 2006. A computer model for welfare assessment of poultry production systems for laying hens. Netherlands Journal of Agricultural Science 54:157-168.