Improved estimation of dissolved oxygen levels in commercial aquaculture using the ensemble Kalman filter

Dissolved oxygen is an essential environmental variable for Atlantic salmon in commercial aquaculture. Hypoxic conditions inhibit growth and increase the risk of mortality, leading to reduced welfare and profitability. Accurate and precise monitoring of dissolved oxygen levels is, therefore, a crucial aspect of optimal operation of aquaculture facilities. This has stimulated an ongoing research campaign seeking to utilise engineering cybernetic methodologies to improve oxygen estimation in commercial aquaculture. Previously, a mathematical model of the oxygen levels in sea-cages was developed based on the advection-diffusion equation. While this model estimated oxygen levels that matched those measured at a commercial site, its accuracy and precision would be greatly improved if it could assimilate real measurements into the estimation framework. Such an expansion would also help adapt the framework to future operational use, where measurements from a farm are assimilated into the framework in real time to provide a more accurate representation of the oxygen conditions therein. This study improves the previously developed oxygen estimation framework by applying the ensemble Kalman filter, a sequential Monte Carlo data assimilation method. The results are unambiguous since the filter implementation is more accurate and precise than the stand-alone oxygen model in both a practical and simulated setting. The estimation framework’s run rates are approximately 10 times faster than real-time on a standard laptop, enabling future deployment on the hardware setups commonly used at aquaculture facilities. This work represents a novel use of the ensemble Kalman filter and expands the frontier of environmental variable estimation in aquaculture.