Combining artificial intelligence and environmental DNA to predict marine fish range shifts under global change – SHIFTeDNA

Atmospheric and oceanic warming is currently accelerating at an unprecedented rate, with consequences for biodiversity as the modification of species geographic distribution, hereafter range shift. Fish, as marine ectotherms, are particularly sensitive to these temperature changes. These modifications in species assemblages can induce functional reorganizations with disturbances on ecosystem processes, which ultimately affect ecosystem services, and thus human well-being. The polar regions (Arctic and Antarctic) are the scene of the most profound changes, with the arrival of new species from lower latitudes or the local disappearance of species for which cold conditions are no longer guaranteed. The low latitudes (tropics) are also threatened by local extinctions that are difficult to observe in the vast ocean. These difficulties to monitor fish biodiversity in the ocean has created significant gaps in knowledge, both in species occurrences and in models of species distribution. New approaches are urgently needed to better quantify and anticipate species shifts and future assemblage compositions. SHIFTeDNA aims to i) model the distributions of marine fish species, even the rarest, from available OBIS occurrence data in response to climate, habitat and human pressures, and to predict future species range shifts under several climate change scenarios; ii) evaluate predictions of modelled distributional shifts using a global, independent and recent database of environmental DNA (eDNA) occurrences sampled from the tropics to the poles to detect early arrivals of species in cold regions (colonization) or local extinctions in warm regions; iii) adjust predictions of species range shifts till the end of the 21st century, taking into account eDNA occurrences; iv) study the influence of species characteristics (mobile or not, mating type, trophic level and others), conservation status (IUCN status, vulnerability to fishing) and phylogeny on species range shifts to better understand the underlying processes.

To address these objectives, we will exploit publicly available occurrence data of marine fishes via OBIS, and the largest eDNA-derived occurrence database available from a project carried out by a Franco-Swiss consortium involving the project partners (MARBEC-ETH-CEFE) over the past four years across all oceans. To improve the analysis of eDNA data, we will also complete the genetic reference base with >2000 fish samples from aquariums and museums. This new metabarcoding reference database will produce the most complete eDNA occurrence data to date. The resulting eDNA occurrences will be used to update predictions of species shifts at the poles, and extinctions in the tropics, and to feed new models of marine fish species distribution. These results will be made possible through the original application of the most up-to-date advances in artificial intelligence on both metabarcoding bioinformatics pipelines and species distribution models. We will then be able to characterize the factors that influence species range shifts in order to better understand their mechanisms and anticipate their effects. The success of the project relies on a consortium of 4 highly qualified French-Swiss partners with complementary skills (environmental genomics, bioinformatics, modeling, marine ecology, molecular genomics).