The marine microbiome can accurately predict its chemical and biological environment
The microbiome responds to physicochemical changes in the environment, making it a sensitive indicator of ecosystem status. Monitoring microbial communities in aquatic systems is therefore essential for understanding ecosystem health and responses to change. Traditionally reliant on microscopy, monitoring programmes are increasingly incorporating DNA-based approaches leveraging on advances in high-throughput sequencing. In this study, we evaluate the potential of using DNA metabarcoding to predict abiotic and biotic parameters across the spatiotemporal gradients of the Baltic Sea. The dataset comprises 397 seawater samples integrating prokaryotic (16S rRNA gene) and eukaryotic (18S rRNA gene) metabarcoding data with environmental measurements and plankton microscopy counts. Random Forest models based on metabarcoding data accurately predicted a range of physicochemical parameters and showed performance comparably to more complex machine learning algorithms. Models based on 16S rRNA gene data tended to perform better than those based on 18S rRNA gene data, with amplicon sequence variant-level data yielding the best results. Metabarcoding outperformed plankton microscopy in predicting abiotic factors and effectively predicted the presence of phytoplankton and zooplankton genera using [≤]1 L of water. Models trained on independent datasets accurately predicted several of the physicochemical parameters, but performed weaker on others, highlighting the potential and challenges for their transferability. Furthermore, our predictions closely matched the observed HELCOM indicator values for assessing good environmental status, suggesting the utility of microbiome-based approaches in regional marine monitoring frameworks. These findings underscore the potential of environmental DNA as a tool for ecosystem monitoring and management in dynamic coastal systems.