Characterisation of the mesozooplankton of the Bay of Calvi using a developed semi-automatic classification system of digital images
Supervisor: Sylvie Gobert (Univ. Liege)
Zooplankton play diverse crucial roles within the marine ecosystem and can also be used as bio-indicators of climate variability since very sensitive to environmental changes. Therefore it is essential to consider long-term plankton series which require an effective and rapid study method. We have therefore developed a supervised learning approach adapted for the (semi-) automatical classification of digital images of the mesozooplankton of the Bay of Calvi by using Zoo/ PhytoImage sofware. Also, a set of nine environmental time series including mesozooplankton biolovume measures was considered in order to identify controlling factors and determine whether the communities were sensitive to Marine Heat Waves. We created a training set of 22 classes (17 of plankton) and the classifier had an estimated accuracy of 88.0% with 10 fold validation evaluation and selected Random Forest. Also comparison of counting estimates derived from automatic classification and traditional methodology revealed that the two counting methods were statistically not different. It was found that main composition of the mesozooplankton was coherent with other studies and that the community was characterized by both seasonal and inter-annual variability. Some predations on copepods patterns where suggested within the community. Further descriptive analysis confirmed some previously determined dynamics such as nutrient increase through convection of deep sea water and give an insight about the environmental dynamics of the area at that period. Marine Heat Waves did not seem to affect zooplankton community. Most importantly suggested here, is the reliability of the automatic classification method as a plankton treatment system for this area. Also, a second training set was created, which can be used as a template for other studies, since it represents the overall specific diversity of the area.