Savannas are among the most endangered temperate ecosystems. Historically, oak savannas were widespread at the ecotone between eastern deciduous forest and grasslands in the central United States, but their extent has declined dramatically in the past 150 years. In the Cross Timbers region in the southern portion of this ecotone, tree encroachment is a common cause of savanna degradation. Although woodlands can be thinned to restore a savanna overstory, savanna restorations’ understories often lack characteristic herbaceous species. Using a seed addition experiment in a newly thinned southeast Kansas oak savanna restoration, I evaluated 1) the relationship between seed addition density and resulting cover for two native, late-successional grass species and 2) how this relationship differed between habitats at tree canopy edges vs. in openings between oak canopies. At my savanna restoration study site, I sowed Sorghastrum nutans and Panicum virgatum at four densities (0 (control), 50, 250 and 1000 seeds per 0.25 m2 plot) in two habitats (canopy dripline, inter-canopy openings). I quantified added species’ cover and flowering tiller densities for six growing seasons after sowing. I also compared cover and flowering tiller densities in my experimental plots with values for these variables in nearby naturally-occurring oak savannas.
Six growing seasons after seed addition, all sowing densities increased S. nutans cover and flowering tiller density over controls. Adding 1000 seeds produced the greatest S. nutans cover and flowering tiller densities. Sowing P. virgatum seeds did not increase this species’ cover or flowering tiller density. For both species, effects of sowing were similar between canopy edge and open habitats. The lack of an effect of habitat type on seed addition efficacy may reflect similar light availability at the litter surface in both habitats, even though light availability at the top of the herbaceous layer was greater in inter-canopy openings. Control plots had less S. nutans cover than naturally-occurring savannas after six growing seasons, but S. nutans cover in plots sown with 50 seeds was indistinguishable from the savanna. For P. virgatum, control plots and plots sown with 50 seeds had levels of cover that were indistinguishable from the naturally-occurring savanna. For both species, high sowing densities produced levels of cover than exceeded naturally-occurring savannas. For some late-successional grasses, seed addition at low densities may be useful to speed establishment. Seeds sown at tree canopy edges and inter-canopy openings are equally effective in promoting grass establishment early in savanna restoration.