A Beacon in the Dark: Grey Literature Data Mining and Machine Learning Enlightening Historical Plankton Seasonality Dynamics in the Ligurian Sea

Abstract

The Mediterranean Sea, as one of the world’s most climate-sensitive regions, faces significant environmental changes due to rising temperatures. Zooplankton communities, particularly copepods, play a vital role in marine ecosystems, yet their distribution dynamics remain poorly understood, especially in the Ligurian Sea. Leveraging open-source software and environmental data, this study adapted a methodology to model copepod distributions from 1985 to 1986 in the Portofino Promontory ecosystem using the Random Forest machine learning algorithm to produce the first abundance and distribution maps of the area. Five copepod genera were studied across different trophic guilds, revealing habitat preferences and ecological fluctuations throughout the seasons. The assessment of model accuracy through symmetric mean absolute percentage error (sMAPE) highlighted the variability in copepod dynamics influenced by environmental factors. While certain genera exhibited higher predictive accuracy during specific seasons, others posed challenges due to ecological complexities. This study underscores the importance of species-specific responses and environmental variability in predictive modeling. Moreover, this study represents the first attempt to model copepod distribution in the Ligurian Sea, shedding light on their ecological niches and historical spatial dynamics. The study adhered to FAIR principles, repurposing historical data to generate three-dimensional predictive maps, enhancing our understanding of copepod biodiversity. Future studies will focus on developing abundance distribution models using machine learning and artificial intelligence to predict copepod standing crop in the Ligurian Sea with greater precision. This integrated approach advances knowledge of copepod ecology in the Mediterranean and sets a precedent for integrating historical data with contemporary methodologies to elucidate marine ecosystem dynamics.

Keywords

• machine learning
• copepods
• species distribution models (SDMs)
• Ligurian Sea
• open source
• FAIR data