Résumé
Question: In conservative biological control, the local landscape is managed in a way that ensures the survival of natural enemies. Our hypothesis is that some components of the landscape, especially secondary host plants of pests, can be identified using trophic chains and food webs published in the literature. This hypothesis was evaluated considering cereal stem borers in a lowland region of Benin characterized by a complex agro-ecological context. A mosaic of crops (maize, sorghum, rice, cotton, gardening...) occupies 49% of the area in the rainy season, the rest being comprised of fallow and natural areas. Methods: The first step was to construct a semantic network using data from 70 scientific papers published between 1957 and2014, among them 11 review articles, concerning lepidopteran cereal borers in Africa. The data introduced in the semantic network are trophic chains, their geographical location and the bibliographic references. The second step was to extract, from this semantic network, the part related to food webs including the cereal borers studied (Busseola fusca, Sesamia calamistis [Noctuidae] and Coniesta ignefusalis [Crambidae]) and their wild host plants (9 species of Cyperaceae and 23 species of Poaceae) observed in the landscape. This subnet included location and bibliographic reference. The last step was to check the consistency of the combination of geographic locations juxtaposed in the subnet, as well as the bibliographical references. Results: The resulting semantic network describes 3004 trophic chains distributed in 40 territories and 13 regions. The extraction of the food web related to the three species of borers helped identify 15 species of plants able to host borers. Amongthem, Rottboellia cochinchinensis is able to host, indirectly via the host borer, two Hymenoptera that are larval parasitoids of the borers, i.e. Goniozus indicus (Bethylidae) and Xanthopimpla stemmator (Ichneumonidae).Conclusion: The pertinence of our analysis is determined by the accuracy of the data provided by the authors. Some studies lacked information on the precise location of the observation or gave an incomplete description of the local ecology. In addition, some reviews did not cite previously published work, raising questions about the value of these reviews or of the uncited studies. (Texte intégral)