Further guidance on assessing OWIs

Indicator assessment frequency

This protocol includes over 30 indicators. While having more data can be beneficial, monitoring all indicators may not be practical in commercial farm settings.

While some recommendations for indicator assessment frequency are provided in this document, producers should decide which indicators to monitor at which frequencies, informed by their own site-specific experience, requirements from any certification bodies or regulators, and practical constraints. These indicators should work together to provide a balanced picture of welfare risk.

Routine assessment enables early problem detection. Research on early post-seawater transfer salmon found that welfare scores changed at intervals of 2–3 months, and concluded that this interval was too long to detect important changes, which suggests a higher frequency of monitoring is preferable.¹⁷ Certification schemes also provide guidance: for morphological assessments, ASC requires monthly assessments; the Best Aquaculture Practices Farm Standard recommends weekly checks; Naturland advises at least every seven days during sensitive periods; and the RSPCA requires a minimum of four assessments in the freshwater stage and monthly in seawater.

At a minimum, behavioral indicators, mortality, dissolved oxygen, and temperature should be monitored daily. Depending on the production system, other water quality indicators may also need to be monitored daily (e.g., ammonia in RAS systems). Morphological indicators should be assessed ideally weekly and at a minimum monthly. Monthly assessment allows early detection of trends (e.g., rising deformity prevalence or declining condition factor) before they become serious welfare problems. If monthly sampling is not achievable due to practical constraints, the interval between assessments may be extended, but the sample size should be increased proportionally to compensate for the reduced frequency. For example, if sampling every two months instead of monthly, the sample size should be at least doubled. While a larger sample size will improve the reliability of that individual assessment, it does not compensate for the reduced ability to detect emerging problems in a timely manner.

Morphological OWI assessment should be mandatory during any fish handling event (e.g., grading, transfer, vaccination, harvest preparation). Fish are already being handled during these events, so the additional cost of scoring is minimal, and the sample obtained is often more representative than a dedicated sampling event. Post-mortem assessment of daily mortalities and at harvest is also strongly encouraged.

Producers should increase monitoring frequency of relevant indicators when specific risks are present (e.g., high temperatures, disease outbreaks, or after handling events) or when assessment results suggest emerging welfare problems.

Sampling individual-based indicators

Methodological notes

All indicators in this section are morphological individual-based outcome measures, meaning they require:

1. Sampling of individual fish from the population;
2. Scoring each sampled fish independently;
3. Calculating prevalence (percentage of fish falling into each category of ‘level 0’, ‘level 1’, ‘level 2’, and ‘level 3’); and
4. Using prevalence distributions to classify the population’s welfare status.

Sample size

While classical sampling theory suggests that a sample of 30 is often sufficient regardless of population size, practical aquaculture conditions often violate key statistical assumptions:

• Spatial heterogeneity (where fish are not uniformly distributed)
• Catchability bias (healthy vs. sick fish differ in capture probability)
• Disease clustering (infections often spatially aggregated)
• Behavioral segregation (subordinate/sick fish may occupy different zones)

This means that sampling 30 fish may not adequately represent the welfare status of the entire population, particularly in large-scale operations. To minimize bias in the sample, efforts should be made to sample fish that are representative, e.g., taking the sample right before feeding, using sweep nets (as opposed to netting individual fish at the surface), or conducting stratified sampling.

We recommend scaling the sample size with the number of fish:

The number of fish reasonably able to be sampled will depend on practical factors such as staff time, and achieving the above sample sizes may not always be possible. Further, a post-mortem analysis of morphological indicators is recommended for a substantial sample after harvest, as this does not impose further stress on the fish and can provide accurate prevalence data, informing future production.

We recommend sampling each cage at the site. The number of cages assessed may be reduced over time if consistent results and strong correlations are observed across cages, and cages are assessed on a rotating basis.

Automated assessment

The sampling process itself can be stressful for fish. Where possible, using observations from underwater camera systems (provided they have been validated against traditional sampling assessments) is recommended. Another benefit of using camera-based systems is a much higher sample size, though there is the limitation of often not being able to see the entire fish (e.g., only one lateral side may be visible for any given fish).

Validation of currently existing automated systems against expert human assessment is limited in the published literature. One key publicly available document is a report on the standardization of manual sea lice counts for validation of automatic camera systems by Bui et al. (2024),¹⁸ which should be consulted when validating similar systems for individual-based indicators.

Sampling behavioral indicators

When assessing behavioral indicators, observations should be conducted using methods that do not themselves induce behavioral changes. This means avoiding actions that startle or disturb the fish, such as sudden movements near the cage, loud noises, or casting shadows over the water. Where practical, the use of underwater cameras is recommended as an alternative to direct observation, as cameras allow repeated and non-intrusive assessment (see also “Automated assessment” above).