La Protection Intégrée contre la flore adventice est un bon exemple d’objet d’étude complexe, du fait de la diversité des leviers agronomiques mobilisables, des interactions éventuelles entre ces leviers, de la diversité des espèces adventices, et des temps longs à considérer pour intégrer les effets cumulatifs. Cet article montre comment sont mobilisées la modélisation et l’expérimentation ‘systèmes’ pour appréhender cette complexité. Il illustre également la large gamme de compétences requises, de l’écologie comparative des espèces jusqu’aux sciences de l’environnement et aux sciences de gestion, nécessaires pour étudier l’insertion potentielle des innovations dans l’exploitation agricole. Les résultats indiquent que la maîtrise technique des infestations adventices est possible, avec des bénéfices environnementaux intéressants. La problématique de la diversification des cultures est un enjeu clef de la performance économique des systèmes économes en herbicides. Integrated Weed Management is an example of complex research area, because of (i) the number of potential management measures, (ii) the possible interactions between these measures, (iii) the range of weed species with various traits, and (iv) the long term consideration required for taking into account the cumulative effects of cropping systems. This paper shows how both modeling and a long term system experiment were used for dealing with this complexity. It also illustrates the diversity of skills required, from comparative ecology to characterize the diversity of weed species and predict their behavior as a function of cropping systems, up to environmental sciences and management sciences, which were used for studying the consequences of innovative cropping systems for the farm management. The results indicate that satisfying weed management can be achieved by combining various measures within a fully redesigned cropping system, with a range of environmental benefits. However, some management measures that are efficient for weed control at the field scale, such as delaying the sowing dates of winter cereals for escaping autumn emerging weeds, might be difficult to implement in practice because of induced bottlenecks in the labor organization at the farm scale. Some weed management measures tend to reduce yields, but this might be at least partly compensated for by reduced input costs. The overall economic profitability of Integrated Weed Management-based systems depends mostly on the possibilities for introducing diversifying crops with sufficient potential economic return.