Optimal risk based short term decision making for aquaculture.

Project main objective is to develop a powerful tool for operative decision support in aquaculture, based on precise data collection and artificial intelligence combined with stochastic modelling and mathematical optimization expertise. The tool will process and systematize large amounts of data and give recommendations on actions that gives the overall best production result. Salmon farming is characterized by significant dynamics and complexity. Operational planning in modern fish farming requires handling considerable amounts of information describing the fish population with individual count, size growth rate and health condition, environmental conditions, operational conditions, market and logistic conditions. Large amounts of data are generated with a high potential value if connected and processed in the right manner. However, until now there has not been developed any specific tool targeting this purpose. Norwegian salmon aquaculture's production equate to 17 million meals of salmon every day at a value close to 200 MNOK. A farmed salmon reaches its peak value towards the last part of the production cycle. During this stage it is also exposed to various risk factors and decisions are time critical, which increases the importance of a tool supporting data- and knowledge driven decisions. More detailed, the project aimed to enhance short-term decision-making in salmon farming, a field characterized by high uncertainty and complex risk factors, including parasitic sea lice, disease, and oxygen levels. The primary goal was to develop a high-precision decision support tool for six-week harvest planning, focusing on optimal fish material utilization and effectively managing risks, particularly those posed by lice treatments and their effects??. A significant achievement was the development of a spatio-temporal forecasting model. This model accounts for the dynamics between lice infestations and treatments, offering vital operational risk management support. Its spatio-temporal nature is novel, including the geographic location and flow distance between different farming sites and how this affects lice treatments. Additionally, the project formulated a decision model that addresses sequential decision-making in aquaculture operations, considering the value of different paths (harvest sequences/cage selections) and their implications for future operations. These models have improved consistency in decision-making, especially in complex and recurring situations, aligning closely with the project's objectives??. For the research field, this project represents a significant step forward in forecasting and managing operational risks related to lice in aquaculture. It has enhanced the understanding and application of complex stochastic models in real-world scenarios. The implications for trade, industry, and society include potential significant cost savings and operational efficiency improvements in the aquaculture industry. These improvements come through reduced mortality, better fish welfare and harvested quality, and increased growth speed of the fish??. The utilization strategy involves applying the forecasting model in aquaculture operations to optimize harvest planning and mitigate risks associated with lice treatments??. The continuing work focuses on exploring the model's potential in different aquaculture settings and refining its predictive accuracy. The expected outcomes include ongoing advancements in operational decision-making and risk management in aquaculture, leading to more sustainable and profitable practice

The work has led to the publication of several research papers during the project, and two papers are in the process of being submitted now at the end of 2023. Optimeering Aqua plans to continue the work done in this project together with the key research partners and explore the commercial viability of implementing the results.

Salmon farming is by its nature an operation with significant uncertainty along several axes involving a number of risk factors (sea lice, disease, low oxygen, etc) that all can have a major impact on production. Towards the end of the production cycle stock value is at the highest level. At this point fish density will normally also be high, which reduces safety margins related to environmental factors like temperature and oxygen levels. High average fish weight implies increasing physiological challenges to individual fish with corresponding increase of mortality risk, especially in relation to operations like sea lice treatment. Optimal harvest planning must bring all these elements into consideration, along with downstream dispositions like wellboat availability, harvesting capacity, market demand development. Harvest planning involves important decisions with high potential impact on profitability and are based on complex considerations. Paradoxically short term harvest planning is today performed manually supported by internally developed spreadsheets. We aim to develop a decision support tool for analyzing potential risk related to alternative operational dispositions, and for providing specific and precise support for optimal harvest planning.