Spider web guilds in cacao agroforestry – comparing tree,plot and landscape‐scale management

Aim Owing to their role as insect predators, web‐building spiders can be important biological control agents within agricultural systems. In complex tropical agroecosystems such as agroforests, management determines plant architecture, vegetation composition and associated ant density, but little is known on how these attributes, together with landscape context, determine spider web density. We hypothesized that all three spatial scales and the presence of Philidris ants significantly contribute to the explanation of spider web density with web types being differently affected. Location In 42 differently managed cacao agroforestry systems in Sulawesi, Indonesia. Methods We surveyed the distribution of five spider‐web types on 420 cacao trees to determine how these relate to habitat variables and a numerically dominant ant species at three different spatial scales, comparing tree, plot and landscape features. We fitted linear mixed‐effects model, selected the best model subset using information‐theoretic criteria and calculated the model‐averaged estimates. We used non‐metric multidimensional scaling (NMDS) to determine and visualize guild level responses to the effects of the tree, plot and landscape‐scale variables. Results The five spider guilds preferred different features of cacao tree architecture. Most frequently recorded webs belonged to the line‐ and orb‐web type. At the tree scale, overall web density was positively related to canopy openness. At the plot scale, a higher number of shade trees was related to a higher web density. At the landscape scale, the altitude determined the distribution patterns of web‐building spiders. Presence of Philidris ants was positively associated with density of orb webs, while no pattern was found for other web types. Main conclusions Results suggest spider web density could be increased by pruning of cacao trees while keeping shade trees at high density in cacao plots. The results emphasize the need to consider scale dependency of crop management and web‐guild‐specific responses that may be related to different functional roles of spiders as a high‐density predator group in agroforestry.