The Norwegian salmon farming industry farms over 400 million salmon at sea annually, and the industry struggles with the impact of diseases and mortality. A large amount of data is generated but tends to lack standardisation and traceability. The overall aim of this project is to generate a robust, data driven, system for day-to-day decision support for the industry to reduce the mortality based on mortality causes through standardisation and structured dataflow. We will have monthly visit to salmon farms in Norway and Iceland over 6 to 12 months. Fish health personnel will evaluate the cause specific mortality registrations, in addition environmental and production data will be collected. All these sources will be used to validate cause-specific mortality registration as a tool to identify causes of mortality. Further a mortality classification method from human medicine called “computer-coded verbal autopsy” will be used to generate an automated process that can classify the underlying cause of mortality. Input sources such as production databases, environmental sensors, lab. reports, fish health reports and events at the farming site will be used to build the model. We will also build a model for universal data structuring and standardisation of production data to keep track of populations from egg to processing. We will develop a population identifier to tag data in a dataset for different fish groups and prove that the same data set can be anonymized by changing the identifier but preserving data quality. Machine learning methods will be used to build high accuracy predictors of cause specific mortality categories and compare the machine learning techniques with traditional statistical methods. Finally, the development of a young industry operating in the commons need both regulatory- and company driven improvement. In order to allocate resources effectively the drivers for value creation and sustainability should be known and used systematically.
Knowledge from production data in salmon aquaculture represents a great potential for increased growth, value, and sustainability of the industry. High mortality and losses have been identified as one of the most critical challenges for the industry today. Different companies, regulators, research institutions etc does not use the same identifier on fish groups and this makes it difficult to get correct comparisons and aggregations. As a result, they lack a common foundation to identify significant parameters, make better decisions, and more accurate risk analysis. Lack of information makes it difficult to define objective measurements and automatize analysing processes, ultimately reducing possibilities for industry insights and improvements in production. The project aims to generate a robust, data driven, system for day-to-day decision support for the fish owner to reduce the mortality based on mortality causes through standardisation and structured data-flow in production, applicable at all levels in the farming industry. This will be done by validating the cause-specific mortality registration and using it as a tool to identify causes of reduced survival in salmon farming (WP1), here we will work actively with two major companies in Norway and all salmon farming companies in Iceland. The technique of “computer-coded verbal autopsy” will be adapted from human medicine to aid in the automatization of classification (WP2) based on the results from WP1. Further we will build a model for universal data structuring and standardization of production data from salmon farming (WP3), this also relies on production data provided by the farming companies. The same data will be utilised in WP4 for an automated assembling and analyses of data from multiple sources to identify mortality causes by machine learning. And finally, we will identify drivers for value creation in salmon industry and drivers for sustainability of the production (WP5).