Humans require natural resources but extraction processes are often destructive to ecosystems. A capacity to develop ecosystem structures and functions in damaged environments is constrained by poor understanding of both the establishment and the functional consequences of biodiversity, including community stability. This is particularly the case in high-stress anthropogenic ecosystems and hence their reconstruction has been slow. To address this problem I experimentally tested three hypotheses: (1) Biodiversity can be established within anthropogenic environments if species are introduced which are pre-adapted to stressful but naturally occurring analogs of the degraded ecosystems. (2) Biodiversity itself, above and beyond effects of species identity, can drive community functioning and resistance to climatic disturbances by way of facilitation and complementarity effects. (3) The influence of introduced biodiversity on ecosystem functioning is mediated by effects of environmental heterogeneity on the proportion of introduced species that establish within the plant community. Plant-addition experiments conducted on abandoned limestone quarry floors revealed that communities from natural limestone pavements (alvars) are dispersal- rather than microsite-limited. In addition, when the plants were struck by a severe natural drought, there was a significantly greater loss where few species established initially. This pattern was repeated at the population level, suggesting that facilitation can promote community stability in high-stress environments. Biodiversity experiments controlling for effects of species composition, community density, and environmental covariates revealed that species diversity, independent of species identity, promoted increased vegetation cover and conferred resistance to severe microclimate fluctuations including drought and heat-wave conditions. Realized species richness was strongly correlated with cover, and several metrics of microsite heterogeneity explained the efficiency with which introduced richness manifested as realized richness. This suggests that heterogeneity mediates diversity-cover relationships by promoting complementary usage of the regeneration niche. I conclude that land managers wishing to restore high-stress ecosystems can exploit the mechanisms by which biodiversity from analog ecosystems responds to itself and to spatial heterogeneity. Establishing more diverse communities appears to reduce the risk that future disturbances will result in ecosystem reversion back to the degraded state.