A machine learning conservation apPROach to evaluaTE extinCTion risk in freshwater biodiversity

Accurate assessments of species’ contemporaneous extinction risk (CER) are vital to quantifying the current biodiversity crisis and prioritising conservation efforts.

However, the most comprehensive global dataset of CER - the IUCN Red List of Threatened Species - is taxonomically biased due to the lengthy assessment process, leaving understudied taxa, such as those in freshwaters, under no formal PROTECTion.

Prediction-based models based on novel machine learning methods enable large-scale automated assessments of CER, reducing data deficits rapidly.

The main goal of this project is to identify predictors of CER in freshwater habitats, focusing on the largest family of freshwater gastropods, the Hydrobiidae.

First, we will use a deep-learning approach to automatically predict the Red List status of hundreds of hydrobiid species from multiple regions and ecosystems that have not been evaluated yet, basing the predictions on ecological and macroevolutionary data.

Second, high-throughput sequencing methods will be conducted for the first time in this taxon to compare microevolutionary diversity with population trends derived from long-term field surveys.

Last, by establishing a multifactorial prediction-based method, the project will identify which features (ecological, macro-, microevolutionary or all) are meaningful to inferring CER in freshwater organisms. The implications of this proposal are threefold and relevant to scientific, technological and societal concerns.

Our findings may provide a basis for comparing predictors of CER across taxa.

They will also open up a more integrative framework for conservation actions, moving beyond species-by-species categorisation.

Focussing on the ""Natural Resources, Agriculture & Environment"" area from HORIZON 2021-2027, this project addresses knowledge gaps in species threats and safeguards freshwater resources, illustrating this with understudied taxa.