Food webs are an integral part of every ecosystem on the planet, yet understanding the mechanisms shaping these complex networks remains a major challenge. Recently, several studies suggested that non-trophic species interactions such as habitat modification and mutualisms can be important determinants of food web structure. However, it remains unclear whether these findings generalize across ecosystems, and whether non-trophic interactions affect food webs randomly, or affect specific trophic levels or functional groups. Here, we combine analyses of 58 food webs from seven terrestrial, freshwater and coastal systems to test (1) the general hypothesis that non-trophic facilitation by habitat-forming foundation species enhances food web complexity, and (2) whether these enhancements have either random or targeted effects on particular trophic levels, functional groups, and linkages throughout the food web. Our empirical results demonstrate that foundation species consistently enhance food web complexity in all seven ecosystems. Further analyses reveal that 15 out of 19 food web properties can be well-approximated by assuming that foundation species randomly facilitate species throughout the trophic network. However, basal species are less strongly, and carnivores are more strongly facilitated in foundation species' food webs than predicted based on random facilitation, resulting in a higher mean trophic level and a longer average chain length. Overall, we conclude that foundation species strongly enhance food web complexity through non-trophic facilitation of species across the entire trophic network. We therefore suggest that the structure and stability of food webs often depends critically on non-trophic facilitation by foundation species.

Large carnivores can reduce ungulate numbers by predation and via induced risk effects alter ungulate behavior, indirectly affecting lower trophic levels. However, predator-induced risk effects probably act at different spatial scales, which have often been ignored in trophic cascade studies. We studied how a fine-scale risk factor (distance from tree logs) affects ungulate browsing intensity and how this is modified over a landscape-scale risk gradient (distance from human settlements to wolf core) in the Białowie_ za forest, Poland. We found that landscape-and fine-scale risk factors strongly interacted in determining the strength and magnitude of carnivore-induced risk effects on lower

1. Coastal ecosystem engineers often depend on self-facilitating feedbacks to ameliorate environmental stress. This makes the restoration of such coastal ecosystem engineers difficult. We question if we can increase transplantation success in highly dynamic coastal areas by engineering measures that promote the development of self-facilitating feedback processes. 2. Intertidal blue mussels Mytilus edulis are a typical example of ecosystem engineers that are difficult to restore. A lack of self-facilitating feedbacks at low densities limits establishment success when young mussels are transplanted on dynamic mudflats. 3. In a large field experiment, we investigated the possibility of increasing transplantation success by stimulating the formation of an aggregated spatial configuration in mussels, thereby reducing hydrologically induced dislodgment and the risks of predation. For this, we applied engineering measures in the form of fences that trapped wave dislodged mussels. 4. Mussel loss rates were significantly lower when mussels were placed between both artificial fences, and in high densities (4.2 kg/m 2) compared with mussels placed in areas without fences and in low densities (2.1 kg/m 2). The fences induced the formation of a banded pattern with high local mussel densities, which locally reduced predation. 5. Synthesis and applications. Our results underline the importance of actively promoting the development of self-facilitating processes, such as aggregation into patterns, in restoration projects of ecosystem engineers. In particular, the current study shows that engineering measures can help to initiate these kinds of selffacilitating interactions, especially in highly dynamic areas.

Restoration of coastal ecosystem engineers that trap sediment and dampen waves has proven to be difficult, especially in the wave-exposed and eroding areas where they are needed the most. Environmental stressors, such as hydrodynamic stress and predation, can only be overcome if transplanted organisms are able to establish self-facilitating feedbacks. We investigate if the artificial lowering of multiple environmental stressors can be used to give transplanted juveniles the opportunity to form a self-sustainable system and thereby increase their long-term survival on wave-exposed and eroding shores. We designed a large field experiment using juvenile mussels (Mytilus edulis) as model species on a wave-exposed tidal flat in the Oosterschelde estuary (the Netherlands). We tested if the environmental stress caused by a high predation pressure and wave-driven dislodgement could be reduced by a combination of artificial structures such as fences (to exclude predatory crabs), attachment substrates (such as coir-net or oyster shells), and breakwaters. Despite a low overall mussel survival (29%), we found that under strong hydrodynamic conditions, experimental fences and attachment substrates increased the retention of transplanted mussel seed. However, modification of local hydrodynamic conditions using breakwaters did not improve mussel coverage preservation. Overall, this study highlights the potential of using techniques that lower multiple environmental stressors to create a window of opportunity for establishment in highly dynamic ecosystems.