Résumé
The organisation of human-populated landscapes results from many interacting processes tied to the historical development of societies and human activities. Agriculture, in particular, has dramatically altered much of the Earth's surface over many millennia. Landscape mosaics can be used to understand these impacts, which are increasing at an accelerating rate on a global scale. On a local-to-regional scale, mosaic concepts are also of practical interest for designing natural enemy-based pest control strategies as an alternative to intensive pesticide use. Considerable empirical and theoretical work has been conducted into these approaches recent years, although it has largely dealt with issues of landscape biodiversity at the species population level, rather than considering the role of species interactions. When viewed as interaction networks, agricultural landscapes can be used to address a more diverse suite of topics, such as those related to network complexity, community demographics, or ecosystem functioning, and how these properties vary with the complexity of the landscape mosaic over time and space. Ultimately, the functioning of an agricultural landscape determines the production of ecosystem services, such as crop production and pest control, yet ecological theory-based management options are still scarce. Consequently, there is a significant gap to be bridged between the current state-of-the-art in theoretical ecology and practical strategies for the management of agricultural landscapes. One way to bridge this gap involves developing metaecosystem models that can account for the movement and internal dynamics of both biotic and abiotic components of ecosystems. By recognising the specifics of agricultural landscapes, that is, their heterogeneity over time and space, the occurrence of frequent perturbations and the importance of nutrient additions at the landscape scale, metaecosystem models can be proposed to answer some current challenges in agroecology.