Dairy farming is facing serious environmental challenges, which can have (in)direct impacts on sustainability. Life cycle assessment (LCA) is used to account for the environmental impact. Simulating the effects of farmers’ choices in cattle management has been identified as a new important challenge in farming system modelling. Agent-based models (ABM) offer this possibility. This project aims to develop a decision tool based on a LCA-ABM coupled simulator, which will integrate a multi objective optimizer. This will allow testing the economic and environmental impacts of different herd managements. To design and validate LCA-ABM, phenotypes will be collected on more than 320 dairy farms located in Wallonia and Luxembourg by breeding associations (i.e., animal characteristics, milk composition and yield, feeding and farm specific economic data) or predicted at cow scale from milk mid-infrared spectra (body weight, methane emissions). Additional phenotypes will be predicted in this project from easily recorded traits using machine learning algorithms. Those will be related to farmer behavior (derived from economic data) as well as grazing period. Phenotypes, coupled with standard literature equations from LCA domain, will allow to calculate environmental impacts from individual cows with a higher temporal resolution than normally done in LCA. Finally, the robustness of the assumptions behind the LCA-ABM based decision tool will be assessed using very detailed economic, feeding and production data available on about 10 pilot farms in Luxembourg. The integration of a multiobjective optimiser in such a tool associated to the proposed phenomics approach has never been realised for the dairy sector. The project will mobilise the complementary expertise of two research teams: LIST for the system optimisation under environmental constraints and the environmental LCA and GxABT for the Phenomics approach and the modelling of dairy traits.