Increased topsoil carbon stock across China's forests.

Biomass carbon accumulation in forest ecosystems is a widespread phenomenon at both regional and global scales. However, as coupled carbon-climate models predicted, a positive feedback could be triggered if accelerated soil carbon decomposition offsets enhanced vegetation growth under a warming climate. It is thus crucial to reveal whether and how soil carbon stock in forest ecosystems has changed over recent decades. However, large-scale changes in soil carbon stock across forest ecosystems have not yet been carefully examined at both regional and global scales, which have been widely perceived as a big bottleneck in untangling carbon-climate feedback. Using newly developed database and sophisticated data mining approach, here we evaluated temporal changes in topsoil carbon stock across major forest ecosystem in China and analysed potential drivers in soil carbon dynamics over broad geographical scale. Our results indicated that topsoil carbon stock increased significantly within all of five major forest types during the period of 1980s-2000s, with an overall rate of 20.0 g C m(-2) yr(-1) (95% confidence interval, 14.1-25.5). The magnitude of soil carbon accumulation across coniferous forests and coniferous/broadleaved mixed forests exhibited meaningful increases with both mean annual temperature and precipitation. Moreover, soil carbon dynamics across these forest ecosystems were positively associated with clay content, with a larger amount of SOC accumulation occurring in fine-textured soils. In contrast, changes in soil carbon stock across broadleaved forests were insensitive to either climatic or edaphic variables. Overall, these results suggest that soil carbon accumulation does not counteract vegetation carbon sequestration across China's forest ecosystems. The combination of soil carbon accumulation and vegetation carbon sequestration triggers a negative feedback to climate warming, rather than a positive feedback predicted by coupled carbon-climate models.