Abstract : This paper describes an ontology evolution activity designed for a data integration system called ONDINE (Ontology based Data INtegration) which proposes a complete workflow to acquire, annotate and query experimental data extracted from scientic documents. The core element of the ONDINE system is an ontology which allows experimental data annotation and querying. In order to adapt to domain changes, new usages and new annotated data, the ontology may change. This paper presents our a priori ontology evolution activity, which takes as input an ontology in a consistent state, denes and applies some changes and manages all the consequences of those changes by producing an ontology in a consistent state. The implementation and evaluation of the evolution activity are presented. Our work is illustrated through an ONDINE system's use case, the annotation of experimental data in the domain of matter transfer.