Multivariate monitoring approaches for animal health in different production systems

PhD Student: Carolina Mendes Galante Merca

Background

This PhD work is part of the DECIDE project, which aims to develop data-driven decision support tools to assist stakeholders in making better-informed decisions. This research focuses on two distinct contexts: the mortality of Atlantic salmon in aquaculture and treatment strategies for Bovine Respiratory Disease (BRD) in fattening farms.

The mortality of maricultured Atlantic salmon has increased in the recent years, being a concern for its sustainable production. In Scotland, approximately 25% of salmon stocked into the sea do not survive. This represents a significant economic burden for the producers and suggests poor animal welfare. Contributing factors to this mortality include infectious pathogens, treatments for sea lice, algal blooms, and the environmental conditions.

In turn, BRD is one of the leading causes of morbidity and mortality within the cattle industry. This complex, multifactorial disease results from the combination of bacterial and viral pathogens along with various stressors. Antimicrobials are the first line for treating this disease, however, their extensive use raises concerns regarding antimicrobial resistance. Efforts should be made towards reducing antimicrobial usage (AMU), ensuring a balance that does not compromise animal health and welfare.

Purpose of the project

The purpose of this PhD was to develop multivariate and hierarchical (multi-level) dynamic monitoring models to address the above mentioned challenges. The idea was to create modelling solutions that could be implemented in the future to help decision-making regarding the increased salmon mortality in Scotland and to enable better targeted treatments for BRD.

Results

In the salmon studies we used the available open-source mortality and environmental data for monitoring monthly mortality of maricultured Atlantic salmon in Scotland, and we provided warnings when mortality significantly exceeded the expected levels. A hierarchical dynamic linear model (DLM), intended to account for the hierarchical inherent structure present on the salmon data, was shown to be the best approach for monitoring salmon mortality in Scotland. Furthermore, our exclusive reliance on open-source data provided additional value to already existing data, and eliminated the need for complex and time consuming data sharing agreements.

In the BRD study we created a proof of concept of a decision support tool which can inform the farmers and the veterinarians about the appropriate timing for conducting collective treatments (metaphylaxis) for BRD, according to different scenarios. The tool was especially beneficial in High-risk scenarios, where the collective treatments triggered by our tool showed a reduction in BRD incidence, disease severity, and AMU.

In conclusion, two potential future decision support tools aiming to address the issues of increased salmon mortality in Scotland and the need for targeted and customized treatment strategies for BRD were created during this PhD work. We anticipate that these tools will assist farmers, veterinarians, and health authorities in making more informed decisions that enhance animal health, promote welfare, and contribute to more sustainable animal production practices.

Future perspective

Further work is ongoing for applying the same modelling approaches but using Scottish weekly mortality data. This weekly dataset, which is not open-source, was collected through a collaborative project with private salmon aquaculture companies operating in Scotland. Additionally, a parallel collaboration is underway to extend these approaches to monthly salmon mortality data from Norway. Regarding the BRD study, the developed decision support tool is being tested to work in real-time based on real-world data from on-farm situations.