Automated underwater microscopy and machine learning to understand and forecast algal blooms
Start Date
23-5-2022 4:15 PM
End Date
23-5-2022 4:30 PM
Abstract
We present an approach for automated in-situ monitoring of phytoplankton and zooplankton communities based on a dual magnification underwater dark-field imaging microscope: the Dual Scripps Plankton Camera (DSPC) system. A DSPC is installed permanently at 3 m depth in Lake Greifensee (Switzerland), delivering images of plankton that are automatically processed and classified using deep-learning models at hourly timescales. The DSPC is able to track the dynamics of ~100 taxa in the size range between ~10 μm to ~ 1 cm, covering virtually all the components of the planktonic food web, including potentially toxic cyanobacteria. Time series collected by the DSPC and associated sensors allow tracking plankton ecological succession patterns and algal bloom dynamics with a temporal frequency and resolution on functional traits that are unprecedented in plankton ecology. The data are robust for water quality monitoring and allow integration with machine learning models to i) study the interaction between abiotic and biotic controls of phytoplankton net growth rates, ii) test ecological hypotheses of ecological processes triggering harmful algal blooms, and iii) develop forecasting models of different types of blooms.
Automated underwater microscopy and machine learning to understand and forecast algal blooms
We present an approach for automated in-situ monitoring of phytoplankton and zooplankton communities based on a dual magnification underwater dark-field imaging microscope: the Dual Scripps Plankton Camera (DSPC) system. A DSPC is installed permanently at 3 m depth in Lake Greifensee (Switzerland), delivering images of plankton that are automatically processed and classified using deep-learning models at hourly timescales. The DSPC is able to track the dynamics of ~100 taxa in the size range between ~10 μm to ~ 1 cm, covering virtually all the components of the planktonic food web, including potentially toxic cyanobacteria. Time series collected by the DSPC and associated sensors allow tracking plankton ecological succession patterns and algal bloom dynamics with a temporal frequency and resolution on functional traits that are unprecedented in plankton ecology. The data are robust for water quality monitoring and allow integration with machine learning models to i) study the interaction between abiotic and biotic controls of phytoplankton net growth rates, ii) test ecological hypotheses of ecological processes triggering harmful algal blooms, and iii) develop forecasting models of different types of blooms.