Our objectives :
♦ Rely on observation networks and data platforms (ROME, ILICO, JERICO, ODATIS, EMODNET) and the hindcasts of coastal models to launch projects on innovative methods of data analysis and modeling (hybrid modeling, machine learning)
♦ Improve our capacity to develop experimental approaches on mesocosms to test hypotheses on ecosystem evolution
♦ Maintain our skills to ensure the evolution, maintenance and transfer of modeling tools developed by DYNECO, particularly in the context of the transition to CROCO modelling platform
Our scientific project places great emphasis on the integration of data and models to represent the systems studied, to understand their functioning, to test hypotheses and hierarchize processes, and to predict possible changes as a function of anthropic and environmental forcing.
Our project is based on the acquisition of data at different spatial and temporal scales, many of which are part of institutional and operational strategies: high frequency observation networks for essential physical and biogeochemical variables; multi-site low frequency observation networks for the benthos and microbial communities. Experimentation is used to test hypotheses and to develop protocols for measuring and analyzing data from sensors.
Statistical or mechanistic modeling relies on two complementary approaches: theory-driven approaches are based on an a priori representation of mechanisms and concepts, and use data to validate or refute hypotheses. Data-driven approaches seek to develop a theory or an understanding of mechanisms based on the assumption that the data are sufficient. Thus, observational data make it possible to describe and measure co-variations between heterogeneous variables and to propose an interpretation in the form of causal relationships. They also allow to prioritize the factors of variability, which can then be taken into account and tested in mechanistic models or experiments.