Context <p>Agricultural landscapes must produce food, fodder, fuel, or fiber while simultaneously maintaining non-market ecosystem services. Conservation practices such as planting perennial wildflower strips can improve multifunctionality, but their effectiveness may depend on the surrounding landscape.</p> Objectives <p>Our goals were to (1) model the effect of adding perennial wildflower strips along crop field margins on ecosystem services, and (2) assess the role of natural land cover and landscape complexity (landscape composition, connectivity, fragmentation, and heterogeneity) in the magnitude of ecosystem services.</p> Methods <p>We modeled pollinators, groundwater recharge, sediment runoff, and nutrient runoff in 437 US subwatersheds across a gradient of agricultural to natural land cover. We compared baseline ecosystem service values to a scenario adding wildflower strips along all crop field margins. We assessed the magnitude of the response in relation to natural land cover and landscape complexity.</p> Results <p>Planting perennial strips improved all four ecosystem services, but in distinct ways. Pollinators (wild bees) increased with percent natural area. Landscape-level increase peaked between 20 and 50%, reflecting annual agriculture edge density changes. Annual agriculture cohesion (a connectivity metric) and edge density (a fragmentation metric) were positively but nonlinearly associated with groundwater recharge and sediment runoff, respectively. Nutrient runoff decreased with increasing natural area and edge density.</p> Conclusion <p>The effects of landscape composition and configuration on ecosystem services varied with the service considered, and most relationships were nonlinear or showed threshold responses. Management interventions such as wildflower strips must be placed in their landscape context to ensure that investments in conservation practices are as effective as possible.</p>