At the highest level, on-farm welfare assessments (particularly audits of welfare standards or monitoring routinely carried out by farm staff) currently have access to a limited number of practical, OWIs that directly assess (or attempt to assess) the animals' emotional states. Research into validating new OWIs that attempt to directly assess aquatic animal welfare, or developing similar LABWIs into OWIs, could drastically improve aquaculture's ability to safeguard animal welfare in a comprehensive manner. Behavioural WIs, or even whole-animal measures such as QBA, have demonstrated the most promise in capturing these essential aspects of welfare as they provide a non-invasive and comparatively practical option of assessment. However, logistical hurdles and technical complexities hinder their capabilities for routine use on-site by farm staff and auditors.

Improving actionable insights for WASs - research priorities for behavioural WIs:

Animal behaviour has long been identified as one of the most promising areas for future research, and the incorporation of behavioural WIs into on-farm monitoring routines and management strategies is a priority for many aquaculture species. Against other animal-based indicators, behavioural WIs are already comparatively fast, non-intrusive, and easy to capture within most farm settings, often also providing early warning signs for issues that arise on-site. Certain whole-animal assessments, which represent a snapshot of the entire state of welfare for the animal(s) involved (e.g., QBA), have the obvious benefit of being "complete" measures - provided that the practicality of such measured could be developed enough to be used as OWIs, these could prove to be powerful additions towards WASs.

However, complex swimming behaviours and visual obstructions of individual animals is commonplace when attempting to monitor aquatic animals in their 3-dimensional environment. A number of logistical challenges therefore arise for behavioural assessments:

1) Tracking individuals is often very difficult, sometimes impossible, due to the fact that animals often go in and out of view of the observer during their assessment.

2) Many behavioural WIs are difficult to quantify without appropriate training, and are also heavily dependent on the motivation and skills of the observer using them.

3) Actionable insights are often only obtained after further analysis has been methodically carried out on, for example, collected video data of the animals' behaviours. This makes many behavioural assessments not time-efficient for formal assessments that must be conducted immediately in-situ.

4) Evaluating changes in distinct, physical behaviours (such as fin biting or swimming speeds) also raise risks of misinterpretation in terms of how these changes relate to the animal's current welfare status. The lines between behavioural responses being maladaptive vs. normal coping mechanisms are often blurred; as always, context is crucial. Such WIs not only have to be species-specific, but condition-specific (as previously mentioned, recorded alongside numerous other OWIs that are required to come to a reliable conclusion).

Video monitoring (especially from underwater feed cameras) allows for footage to be reviewed remotely by farm staff (or potentially auditors). As previously mentioned, however, tracking individual behaviours amongst large groups, especially in poor lighting conditions with time-constraints, poor training, and the potential for observer-bias all greatly inhibit the effectiveness of this approach.

The logistical hurdles mentioned above are in part why most widely adopted standards and on-farm monitoring routines include minimal behavioural WIs, despite their unique insights and value to WASs. Fortunately, these indicators are well suited to leverage a number of emerging technologies to overcome these practical limitations.