Task 7.2 Emerging automated, high-frequency imaging techniques for breakthroughs in aquatic food web research in mesocosms
Emerging automated, high-frequency imaging techniques for breakthroughs in aquatic food web research in mesocosms
Partners: Lead: SYKE, co-lead FVB-IGB, Contributors: NORCE, WCL
Duration: Month 1-48
A pertinent challenge in planktonic community ecology has been the bottleneck of acquiring species-level
information of communities. Because of their high turnover, the dynamics of planktonic communities should be studied using high-resolution experimentation, which is still prevented by the time-consuming analysis of species-level abundance data by traditional methods.
Recent technological advances have led to emergence of automated imaging instruments, with fast-improving resolution and output rates, up to tens of thousands of images per hour. It is becoming possible to produce real-time Big Data of plankton communities. This allows testing of core ecological hypotheses, originally derived from macroscopic realms, in the microbial planktonic food webs: ranging from community ecology to biodiversity research and ecosystem functioning.
AQUACOSM partners have initiated research on several high-frequency imaging techniques, in combination with the required application of advanced machine learning techniques, for phytoplankton and zooplankton species identification. Task 7.2 will harness these break-through developments and apply them for the first time to mesocosm experimentation at a large scale.
The automated imaging instruments (SYKE, FVB-IGB) to be tested in the mesocosm settings, and verified against traditional methods (microscopy, particle counting, flow cytometry), including two of Europe’s first three Imaging FlowCytobot instruments (McLane Labs) for phytoplankton and smaller microzooplankton, Cytosense (CytoBuoy), several FlowCam versions (Fluid Imaging Technologies), Mini Deep Focus Plankton Imager (Bellamare) for in situ imaging of mesozooplankton, LISST Holo2 for holographic detection of zooplankton and aggregates, Amnis® ImageStream®X Mark II Imaging Flow Cytometer (MERC) for combined flow cytometry and microscopy of phytoplankton and bacteria, and LISST 200-X (Sequoia) laser diffraction particle counter. Most importantly, the imaging Big Data provided by these instruments will be analysed by the latest image recognition developments in AI (Artificial Intelligence: computer vision and machine learning), applying advanced Deep Learning techniques (all contributors). The success of species identification between different instrument settings, and between proprietary software and the developed Open Source neural network solutions, will be addressed and reported.