Calanus redness index from artificial intelligence applications to image analysis (CARDINAL)
Note: lthe full title of this project is "Calanus redness index from artificial intelligence applications to image analysis (CARDINAL): creating the link between satellite and individual-scale imagery for ecosystem-based management of a keystone species in subarctic seas".
This project is part of the Université Laval / University of Tromsø research partnership.
Principal investigators
Co-investigators
Denis Laurendeau, Torbjørn Eltoft
Collaborator outside of ULaval
Jean-Olivier Irisson (Université Sorbonne / CNRS)
Abstract
The Subarctic seas are among the most productive regions of the Earth’s ocean, providing essential ecosystem and economical services. Meanwhile, they are strongly impacted by the consequences of the ongoing climate changes that will lead to profound yet difficult to predict perturbations of their rich marine ecosystems. The base of their high productivity is supported by planktonic organisms that have adapted to past environmental conditions. It appears critical to understand how the individual-scale responses to unprecedented environmental conditions will generate large-scale responses affecting ecosystem functioning.
In Subarctic seas, zooplankton contribute to the biological carbon pump, and more specifically to the lipid pump that actively transfers large amounts of lipid stores between the surface and the deep, between the short productive season and the long winter as well as between the primary producers and the higher trophic levels, including some of the world’s largest fish stocks. In the continuum formed from the Northeast Atlantic via the Norwegian Sea to the Barents Sea and Arctic Ocean four lipid-rich copepod species of the genus Calanus interact to produce nonlinear responses of the pelagic ecosystem.
These species should be regarded as a continuum of biological traits (i.e. essential properties that influence individual fitness and ecosystem functions) rather than distinct entities with loosely overlapping roles in the ecosystems. This is the core of the trait-based approach of the functional diversity of zooplankton communities. The main functional traits usually described for zooplankton are body size, feeding and swimming abilities, spawning and diapause strategies. Colour is one trait that has been overlooked until recently since zooplankton are usually transparent. However, Calanus copepods harbour a bright red colouration so inherent to these species that they are called “red feed” in marine fisheries and aquaculture. Colour is the only individual trait that can be measured automatically from images both at the individual level when observed under a stereomicroscope, and at a much larger scale through ocean colour remote sensing. For the past decade, imaging methods for plankton studies have led to the production of massive amounts of images that are particularly amenable to machine learning approaches. In CARDINAL we aim to fully utilize the potential of these images to define how the redness trait is related to large-scale spatio-temporal aggregations of Calanus spp. and marine ecosystem functioning.
To achieve our ambitious goal, we designed CARDINAL as a bridge between two innovative projects aiming at 1) providing the first large-scale observation of zooplankton through ocean colour remote sensing by taking advantage of Calanus redness (SEA PATCHES, S Basedow) and 2) measuring traits automatically from in situ individual images of zooplankton (ARTIFACTZ, F Maps). This new project brings together an expert team of researchers, end-users and graduate students to explore how to merge individual-based and large-scale approaches. This approach could revolutionize the field of marine ecology in a similar way than satellite imagery alone did for the past few decades, once it provided access to subsurface colour-derived biomass and productivity estimates of phytoplankton.