Abstract:In this study, explainable machine learning techniques are applied to predict the toxicity of mussels in the Gulf of Trieste (Adriatic Sea) caused by harmful algal blooms. By analysing a newly created 28-year dataset containing records of toxic phytoplankton in mussel farming areas and toxin concentrations in mussels (Mytilus galloprovincialis), we train and evaluate the performance of ML models to accurately predict diarrhetic shellfish poisoning (DSP) events. The random forest model provided the best prediction of positive toxicity results based on the F1 score. Explainability methods such as permutation importance and SHAP identified key species (Dinophysis fortii and D. caudata) and environmental factors (salinity, river discharge and precipitation) as the best predictors of DSP outbreaks. These findings are important for improving early warning systems and supporting sustainable aquaculture practices.

What problem does this paper attempt to address?

The problem that this paper attempts to solve is: **Using machine - learning techniques to predict mussel poisoning incidents in the Adriatic Sea, especially diarrhetic shellfish poisoning (DSP) incidents caused by harmful algal blooms (HABs)**. Specifically, by analyzing 28 - year - long monitoring data, including records of toxic phytoplankton and toxin concentrations in mussels, the researchers trained and evaluated the performance of machine - learning models to accurately predict the occurrence of DSP incidents. ### Research Background and Problem Description 1. **Importance and Challenges of Shellfish Farming** - In the past 30 years, shellfish farming has occupied an important position in global marine and coastal aquaculture production, especially in terms of food safety. - However, certain shellfish productions (such as mussels) in the European Union face challenges due to low profits, limited farming progress, and environmental threats (such as harmful algal blooms, diseases, predators, adverse weather conditions, and pollution). 2. **Hazards of Harmful Algal Blooms (HABs)** - HABs can be divided into non - toxic and toxic categories. Toxic HABs produce toxins that accumulate in shellfish and fish, posing a serious threat to human health and food safety. - Among the globally recorded HAB incidents, approximately 48% are caused by toxin - producing species, and these toxins can cause human food poisoning with neurological or gastrointestinal symptoms. 3. **Specific Situation in the Adriatic Sea** - Among the shellfish in the Adriatic Sea, DSP toxins (such as the acid toxin group) and other lipid - soluble toxins (such as yessotoxin and pectenotoxin) are most frequently detected and often exceed the legal limits. - In the Gulf of Trieste in the northern Adriatic Sea, the cultured Mediterranean mussel (Mytilus galloprovincialis) is a major potential source of human poisoning. 4. **Limitations of Existing Monitoring Systems** - Despite having a sound monitoring system, it is still not guaranteed to completely prevent contaminated seafood from entering the market. - Sales are usually suspended only after toxicity is confirmed, which may lead to economic losses and uncertainty. ### Research Objectives To address the above challenges, this study aims to: 1. **Develop a data pre - processing pipeline suitable for the requirements of specific datasets**. 2. **Train selected machine - learning models to directly predict DSP toxicity incidents and evaluate their potential application in an early warning system (EWS)**. 3. **Interpret the obtained machine - learning models and their predictions to improve the credibility and interpretability of the models**. Through these efforts, the research hopes to provide the shellfish industry with a method to predict and adapt to harmful algal bloom incidents in advance, thereby helping authorities to manage and mitigate risks, such as closing shellfish harvesting areas.

Explainable machine learning for predicting shellfish toxicity in the Adriatic Sea using long-term monitoring data of HABs

Coastal water quality prediction based on machine learning with feature interpretation and spatio-temporal analysis

Machine Learning in Extreme Value Analysis, an Approach to Detecting Harmful Algal Blooms with Long-Term Multisource Satellite Data.

Extracted Spectral Signatures from the Water Column as a Tool for the Prediction of the Structure of a Marine Microbial Community

Explainable machine learning models for predicting the acute toxicity of pesticides to sheepshead minnow ( Cyprinodon variegatus )

Hybrid Machine Learning techniques in the management of harmful algal blooms impact

Machine learning-based prediction of harmful algal blooms in water supply reservoirs

Preventing illegal seafood trade using machine-learning assisted microbiome analysis

Machine Learning in management of precautionary closures caused by lipophilic biotoxins

Seasonal Early Warning of Impacts of Harmful Algal Blooms on Farmed Shellfish in Coastal Waters of Scotland

Predicting and analyzing the algal population dynamics of a grass-type lake with explainable machine learning

Predicting coastal algal blooms with environmental factors by machine learning methods

Exploring the response and prediction of phytoplankton to environmental factors in eutrophic marine areas using interpretable machine learning methods

Leveraging explainable machine learning for enhanced management of lake water quality

Machine learning-based prediction of fish acute mortality: Implementation, interpretation, and regulatory relevance

Machine learning to predict dynamic changes of pathogenic Vibrio spp. abundance on microplastics in marine environment

Prediction of algal blooms via data-driven machine learning models: an evaluation using data from a well-monitored mesotrophic lake

Machine learning based marine water quality prediction for coastal hydro-environment management

Satellite-based feature extraction and multivariate time-series prediction of biotoxin contamination in shellfish

Revealing Physiochemical Factors and Zooplankton Influencing Microcystis Bloom Toxicity in a Large-Shallow Lake Using Bayesian Machine Learning

Forecasting harmful algae blooms: Application to Dinophysis acuminata in northern Norway