Unprecedented species loss in diverse forests indicates the urgent need to test its consequences for ecosystem functioning. However, experimental evaluation based on realistic extinction scenarios is lacking. In this study, we assessed the impacts of random and directed species loss on productivity over seven years in a subtropical forest biodiversity experiment in China. We used a total of 469 plots that harbored 1 to 16 tree species on an area of 0.067 ha (Chinese land area unit of 1 mu). We derived four types of effectively non-random extinction scenarios, in which species with the following attributes went extinct first: species with high specific leaf area, evolutionary distinctiveness, regional rarity, or from small families. To explore the mechanisms driving the impacts of directed species loss on productivity, we developed a partitioning method to decompose the net extinction effect into a node-loss (monoculture difference between remaining and lost species), a link-loss (reduced species interactions associated with lost species or nodes) and a link-compensation (increased frequency of species interactions between remaining species or nodes) effect.
Results/Conclusions
We found that species loss is detrimental for stand volume in all scenarios, and that these effects strengthen with age. However, the magnitude of these effects depends on the type of attribute on which the directed species loss is based, with preferential loss of evolutionarily distinct species and those from small families having stronger effects than those that are regionally rare or have high specific leaf area. These impacts were due to both node loss and link loss or compensation. At high species richness (reductions from 16 to 8 species), strong stand volume reduction only occurred in directed but not random extinction. Our results imply that directed species loss can severely hamper productivity already in diverse young forests. It also alarms us that previous biodiversity—ecosystem functioning studies based on random species loss could bias our understanding or predictions of how realistic diversity loss would affect ecosystem functioning.