Aims
Although shrubs are an important component of forests, their role has not yet been considered in forest biodiversity experiments. In the biodiversity–ecosystem functioning (BEF) experiment with subtropical tree species in south-east China (BEF-China), we factorially combined tree with shrub species-diversity treatments. Here, we tested the hypotheses that shrub survival differs between the 10 planted shrub species, with lower survival rates of late- than early-successional species and is affected by environmental conditions, such as topography and top soil characteristics, as well as by biotic factors, represented by tree, shrub and herb layer characteristics.

Methods
We analyzed the survival of 42 000 shrub individuals in 105 plots varying in tree and shrub species richness of the BEF-China project four years after planting. Shrub survival was analyzed with generalized linear mixed effects models at the level of individuals and with variance partitioning at the plot level. Random intercept and random slope models of different explanatory variables were compared with respect to the Bayesian Information Criterion (BIC).

Important Findings
Survival rates differed largely between the 10 shrub species, ranging from 26% to 91% for Ardisia crenata and Distylium buxifolium, respectively. Irrespective of species identity, single abiotic factors explained up to 5% of species survival, with a negative effect of altitude and slope inclination and a positive effect of the topsoil carbon to nitrogen ratio, which pointed to drought as the major cause of shrub mortality. In contrast, neither tree nor shrub richness affected shrub survival at this early stage of the experiment. Among the biotic predictors, only herb layer species richness and cover of the dominant fern species (Dicranopteris pedata) affected shrub survival. Overall, our models that included all variables could explain about 65% in shrub survival, with environmental variables being most influential, followed by shrub species identity, while tree species diversity (species richness and identity) and herb layer characteristics contributed much less. Thus, in this early stage of the experiment the biotic interactions among shrubs and between shrubs and trees have not yet overruled the impact of abiotic environmental factors.