Non-crop habitats are essential for sustaining biodiversity of beneficial arthropods in agricultural landscapes, which can increase ecosystem services provision and crop yield. However, their effects on specific crop systems are less clear, such as soybean in South America, where the responses of pests and natural enemies to landscape structure have only recently been studied.

Here, we analyzed how native forest fragments at local and landscape scales influenced arthropod communities, herbivory, and yield in soybean fields in central Argentina. To do this, we selected soybean fields located in agricultural landscapes with varying proportions of forest cover. At two distances (10 and 100m) from a focal forest fragment, we sampled natural enemy and herbivore arthropods, and measured soybean herbivory and yield. We focused on herbivore diversity, abundance of key soybean pests in the region (caterpillars and stink bugs), and their generalist and specialist natural enemies.

Higher abundance of predators, lower herbivory rates, and increased yield were found near forests, while overall forest cover in the landscape was positively related with parasitoid and stink bug abundance, soybean yield, and negatively with herbivory. Moreover, yield was positively linked to richness and abundance of generalist and specialist enemies and independent of herbivory according to piecewise Structural Equation Models.

Synthesis and applications. Our results show positive effects of native forests on biodiversity and yield in soybean crops, highlighting the need for conservation of forest fragments in agricultural landscapes. Moreover, the relation between natural enemies and crop yield suggests that Chaco forests support a diverse and abundant community of natural enemies that can provide sustained levels of ecosystem services and result in positive effects for farmers.

Study sites

Twelve landscape circles with varying amounts of soybean fields and forest fragments directly adjacent to soybean fields were selected. The center of each circle was located on a forest-soybean boundary. Forest cover was calculated at three scales (concentric circles of 0.5, 1, and 1.5 km diameter surrounding the focal fields). Sampling was carried out within the soybean fields at two distances, 10 m and 100 m from the forest edge.

Arthropod sampling

Natural enemies and herbivores on soybean fields were collected using two methods. Yellow pan traps were used collect parasitoids and flying herbivores and predators, while the beating-sheet method was used to sample foliage for caterpillars, stink bugs, and other predatory species. Two yellow plastic pan traps were placed on the ground between soybean rows (20 m from each other) at each site and distance from the forest at the end of soybean flowering. Beating-sheet sampling was repeated twice at two different soybean stages: flowering and pod filling. Soybean foliage was beaten against a 1 m long white vertical beating-sheet attached to a plastic trough. Thirty repetitions per distance and site were conducted at each stage. Repetitions were performed at ten sampling points, separated by 5 m. At each point, three soybean rows were sampled (central and adjacent rows). Data from all the repetitions was pooled to get one value per distance.

For natural enemies we combined the information from both sampling methods and calculated total species richness of predators and parasitoids and abundance of predators and parasitoids of stink bugs and caterpillars per distance and field. Almost no soybean pests were collected with yellow pan traps and thus we used only data from beating sheet sampling to calculate herbivore richness and abundance of stink bugs and caterpillars. Pleasee see details on these methods on the article.

Soybean herbivory and yield

For measurements of herbivory, leaves from 10 soybean plants 5 m distant from each other (five leaves randomly selected, i.e. 50 leaves in total) per distance and site were collected. To estimate accumulated leaf consumption, leaves were collected prior to their senescence, at the beginning of the pod-filling stage (Fig. S1; R4-R5 phases; February 10-15). In the laboratory, digital photographs over a white background with a scale were taken. We used the software ImageJ (Schneider, Rasband, & Eliceiri, 2012) to measure leaf area and consumed area, and then calculated the proportion of consumed leaf. For leaves with damage along the edge, the shape of the leaf was digitally reconstructed based on the remaining leaflets or leaves of similar area.

In order to measure soybean seed development and yield, 10 soybean plants (5 m distant from each other) were collected from each site and distance. Plants were collected at the end of the pod/filling stage, prior to harvest (Fig. S1; R8 phase; March 1-10), taken to the laboratory, and left to dry for 10 days at room conditions. We then measured the total number of pods per plant, the number of full pods (i.e. pods containing three or four seeds completely developed), and the total seed weight per plant.