Abstract
Seed masting and production of empty seeds have often been considered independently as different strategies to reduce seed predation by animals. Here, we integrate both phenomena within the whole assemblage of seed predators (both pre and post-dispersal) and in two contrasting microsites (open vs. sheltered) to improve our understanding of the factors controlling seed predation in a wind-dispersed tree (Ulmus laevis). In years with larger crop sizes more avian seed predators were attracted with an increase in the proportion of full seeds predated on the ground. However, for abundant crops, the presence of empty seeds decreased the proportion of full seeds predated. Empty seeds remained for a very long period in the tree, making location of full seeds more difficult for pre-dispersal predators and expanding the overall seed drop period at a very low cost (in dry biomass and allocation of C, N and P). Parthenocarpy (non-fertilized seeds) was the main cause of seed emptiness whereas seed abortion was produced in low quantity. These aborted seeds fell prematurely and, thus, could not work as deceptive seeds. A proportion of 50% empty seeds significantly reduced ground seed predation by 26%. However, a high rate of parthenocarpy (beyond 50% empty seeds) did not significantly reduce seed predation in comparison to 50% empty seeds. We also found a high variability and unpredictability in the production of empty seeds, both at tree and population level, making predator deception more effective. Open areas were especially important to facilitate seed survival since rodents (the main post-dispersal predators) consumed seeds mostly under shrub cover. In elm trees parthenocarpy is a common event that might work as an adaptive strategy to reduce seed predation. Masting per se did not apparently reduce the overall proportion of seeds predated in this wind-dispersed tree, but kept great numbers of seeds unconsumed.