Monitoring Experiment Eijsden

27.08.2013  by  Bouke Krom

The monitoring station in Eijsden

The CytoSense and the possibility for manual control inside the station

A scatterplot demonstrating the Smart Grid technology. Every green dot is a picture taken.

In April 2013 a long term, high frequency phytoplankton monitoring project started with a CytoSense running autonomously on the River Meuse measurement platform of the Netherlands Ministry of Infrastructure and the Environment in Eijsden. Every hour the flow cytometer automatically analyses a river sample, the data file is processed automatically - including automatic clustering and classification with EasyClus, and the results are automatically uploaded to a website ( by ThomasRuttenProjects). A poster with photo's and some preliminary results was presented by Machteld Rijkeboer at the ILOW Innovation Symposium held in Utrecht on the 24th of April 2013.
The main goals of the project are:

  • improvement of the value of low frequency data of a network of fixed stations by integration with these new high frequency data;
  • real-time assessment of water quality for early warning purposes;
  • leg up to completely stand-alone operation in buoys, ferry boxes and unmanned field stations.

We at CytoBuoy are testing 2 interesting new features in this setup:


A big problem in random imaging is the large amount of redundant photo's taken of the dominant group of particles and the low number of photo's of interesting but more rare species.  To deal with that statistical problem we previously invented the so-called 'targeted imaging' in which the user can specify a priori a wanted type of particles in data space which should be photographed and the CytoSense evaluates each particle against that criteria before the photo is taken.  This works great but it requires a minimum of a priori knopwledge of the type of water and phytoplankton composition. Therefore we now have developed a similar but opposite type of photo preselection that actually "searches" for new unknown particles. Instead of using a specific area or cluster in data space to target the preselection of the capturing of the photo's we have now implemented a generic approach in which the data space is divided in small bins in a grid and for each bin the total photo count is limited to a user specified number. What happens now is that the bins in the data space of the dominant cluster are filled up with photo's first and quickly until the limit is reached and less rapidly populated bins will start to gain statistical chance to get photographed - if this proces runs a while the added photo's will predominantly represent the more 'rare' and unknown particles.


Usually a calibration with fluorescent beads is performed at certain intervals to enable the comparison of datasets over time, and to assess the system stability in general. To enable such a calibration in an unmanned, autonomous installation we have added a system filled with calibration beads to the CytoSense and the necessary software to execute a bead calibration measurement user specified intervals.  

These systems are started up very recently and already provided us with many valuable results and feedback.