DISSOLVED OXYGEN CONCENTRATIONS IN A COMMERCIAL SCALE SALMON FARM UNDER LOW-FLOW CONDTIONS: FIELD MEASUREMENTS AND MODELLING

Introduction

To ensure optimal growth and acceptable fish welfare, dissolved oxygen (DO) concentrations in fish cages need to exceed a temperature-dependent threshold (Remen et al. 2016). As the Norwegian salmon aquaculture industry moves towards larger production units, DO levels are likely to become a more significant concern, and climate change can compound this problem through higher water temperatures. DO concentrations within open aquaculture sea cages depend on the dynamical interplay of factors such as the oxygen consumption of individual fish, spatial distribution of fish biomass in the cage, water temperature, ambient current speed and direction and the flow field within the fish farm. Since DO varies spatially between and within cages, it is insufficient to simply measure DO at a single point in order to assess the availability of oxygen for the fish in the farm (Berntsson et al. 2026). In this study, DO was monitored at multiple locations in a commercial salmon farm, and two mathematical models of high and low complexity, respectively, were evaluated as tools to predict oxygen conditions and to assess the risk of hypoxia inside cages.

Materials and methods

The field study was conducted at a commercial salmon farm in Mid-Norway with 8 cages of 50 m diameter (see Figure 1) stocked with biomass over a 6 week period in May-July 2025. DO and temperature was monitored at 5, 10 and 15 m depth inside one cage stocked with biomss near the center of the farm, at two cage locations without biomass and at a reference location outside of the farm. Current speed and direction was monitored using profiling current sensors at one location near the center of the farm and at a reference location outside the farm. Two mathematical models were used to simulate DO conditions over the study period. One model used simple relations between the oxygen consumption of fish as a function of temperature, the expected vertical distribution of the fish, thecurrent speed and the diameter of the cages in order to estimate DO levels inside cages. The other model was based on Alver et al. (2024), and used a gridded 3D representation of the entire farm to simulate spatial DO distribution as a function of advection, diffusion and the spatially distributed oxygen consumption of the fish. The output of the two models was compared to the measured DO values in order to assess the quality of model predictions. A comparison of the two models was made in order to assess the trade-off of model complexity, computational cost and the ability to accurately predict DO conditions and the risk of hypoxia affecting fish welfare.

Results

The study period included temperatures in the range7-13°C and currents generally below 0.2 m/s at the reference location. Ambient DO levels were in the range 7.5-11.5 mg/l, while measured DO inside the stocked cage ranged down to below 6 mg/l. The 3D model generally predicted DO levels best when using current measurements from within the farm, with RMS deviation on the order of 1 mg/l and overall bias of about 0.1 mg/l at 5 m. The low and high complexity models were compared for similar conditions, meaning a simulation test case where all other cages were considered empty. The overall trends for DO estimated by the low complexity model were often similar to the overall trends displayed by the complex model when considering a time scale of days. There were differences however, especially when investigating hourly variations. The simplified model seems to be more sensitive than the complex model to variations in current velocity, which may be a result of the formulation of the simplified model. A significant limitation of the simple model was the difficulty of representing effects of neighbouring cages. As illustrated in Figure 1, downstream cages tend to see lower DO levels than upstream ones.

Figure 1: Example DO distribution at 6 m from the 3D model at a time when the current flows from right to left through the farm. The hatched white line shows the location of the vertical transect in the lower plot. The yellow x marks indicate locations of DO sensors. The red star indicates the location of the current sensor.

Acknowledgment

This study is part of the PROHAV project, funded by the Norwegian Seafood Research Fund ( #901934).

References

Alver, M.O., Føre, M. and Alfredsen, J.A., 2023. Effect of cage size on oxygen levels in Atlantic salmon sea cages: A model study. Aquaculture, 562, p.738831. doi: 10.1016/j.aquaculture.2022.738831

Alver, M.O., Føre, M., Urke, H.A. and Alfredsen, J.A., 2024. Mathematical modelling of dissolved oxygen levels in a multi-cage salmon farm. Aquaculture, 593, p.741291. doi: 10.1016/j.aquaculture.2024.741291

Berntsson EVC, Alver MO, Liland KH and Stevik TK (2026) Modelling cage-level dissolved oxygen variation within salmon farms. Front. Aquac. 5:1813350. doi: 10.3389/faquc.2026.1813350

Remen, M., Sievers, M., Torgersen, T. and Oppedal, F., 2016. The oxygen threshold for maximal feed intake of Atlantic salmon post-smolts is highly temperature-dependent. Aquaculture, 464, pp.582-592. doi: 10.1016/j.aquaculture.2016.07.037