LAUSR

Abstract Plants seeded during ecological restoration sometimes persist but more often fail to establish. Biodiversity has been shown to stabilize a number of ecological processes, suggesting biodiverse seed mixes could be designed to stabilize plant establishment outcomes. In particular, it may be possible to design seed mixes to increase chances at least some seeded species will be adapted to whatever environmental conditions arise during establishment. To explore this possibility, we developed a modelling framework and applied it to data from 30 field experiments (15 sites × 2 seeding years) conducted in a big sagebrush (Artemisia tridentata Nutt.) ecosystem. In each experiment, three native and one nonnative grass were sown (600 seeds m−2) in separate plots, and we estimated the probability each species germinated and survived through two growing seasons post-seeding. Applying an optimization algorithm to these survival probabilities allowed us to assign species identities to 600 seeds m−2 in a manner maximizing the number of experimental conditions yielding ≥5 plants m−2, a common plant density goal in grassland restoration. Allocating 353 (216, 555) [point estimate (95% CI)] seeds to Poa secunda J. Presl and 247 (11, 378) seeds to Pseudoroegneria spicata (Pursh) A. Love) maximized our native plant density goal (goal achieved in 12 (10, 14) of 30 experiments), and the allocation to >1 species supports the hypothesis biodiverse seed mixes could be designed to reduce establishment failures. Averaged over experiments, P. spicata survival was roughly half of P. secunda survival, but P. spicata nevertheless contributed to the density goal by compensating for low P. secunda survival in certain experiments. Strategically combining species with different seed/seedling traits can increase chances of achieving adequate plant establishment during ecological restoration.