Currently, no single approach is adequate to assess species diversity and community composition in plankton samples. While biomass and abundance of particles within a detected size range can be captured very well with automated image classification, predictions on species identity are often unreliable and/or of low taxonomic resolution. High taxonomic resolution can be obtained with DNA-based approaches, but these perform suboptimal with respect to quantification of abundance or biomass. Appreciating the complementarity of both biodiversity estimation methodologies, we propose to make an integrated pipeline that exploits the strengths and complements the limitations of both approaches. Environmental parameters are also important sample descriptors that relate to the biodiversity at a certain location and will be integrated as well. An extensive collection of validated plankton images generated with a FlowCAM device is already available. Additional context data and DNA samples will be collected monthly in predefined locations in the Belgian part of the North Sea. To develop the pipeline, a deep learning image classification algorithm will be used as starting point, and will be augmented with new input data, side-channel information and an adapted network architecture. As such we will be able to make more accurate plankton image classifications with higher taxonomic resolution than current state- of-the-art procedures.