Species, trait and abundance are the essential components of biodiversity at the scales from individual to community. Biotic and abiotic environmental variations affect specie trait, consequently determine the species presence and abundance. The abundance of species with specific traits drives the simultaneous chang es of multiple ecosystem functions. Ecologists proposed different theor ies or hypothes e s to describe the mechanism for the changes of ecosystem functioning. Biomass ratio hypothesis,niche patitioning theory, and green soup hypothesi s highlight different i mportance of dominant traits, trait diversity and community size, respectively. In a word, the quality and quantity of communities drive the changes of ecosystem functioning together. However, althouth the changes of communities drive the simultaneous chan ges of multiple ecosystem functions, most of the present studies investigated individual functions separately, and may underestimate the importance of biodiversity. The synthetic analyses implemented at larger scales confirm ed the driv ing of biodiversity to ecosystem multifunctionality. However, no consensus exists on the relative importance of different components of biodiversity. The recent studies on trait relationships suggest ed the context dependency of ecological functions of plant traits. Therefore,establishing the mechanistic links between traits and environmental variations (or ecosystem functions) is pivotal to understand the functional responses and effects of communities. Ecosystem restoration, regulation and management are usually implemented a t the local scales. Presently, few studies analyse d and compare d the relative importance of the quality and quantity of communities, ecosystem multifunctionality wa s also rarely studied at the local scales.
We implemented our study in the Yangjuangou small watershed of the Loess Pleateau, China to further understand functional responses and effects of communities and the mechanisms for the provision of ecosystem functions or services. W e also assessed the effects of different restoration pratices on the eco system functioning to enhance our ability of regulating and managing ecosystems.Currently, different restoration practices of China s Grain for Green Project in this watershed resulted in a heterogeneaous landscape that comprises forests, shrublands, semi shrub lands,grasslands and croplands. Before the implementation of Grain for Green Project, cropland was the main land use in this watershed, which impl ies that the initial conditions are similar in different ecosystems. Therefore, t his heterogeneaous lan dscape stands for an ideal system to test the different ecological theor ies and assess the different restoration practices. The results from this study are easily extraplotaed considering the similarity of the previsous agricultural practices, the restoration practices, and final landscape pattern.
We investigated and measured the species composition, abundance and trait of different ecosystems, the environmental factors (soil bulk density, carbon and nitrogen concentrations), and the ecosystem functions (s oil carbon and nitrogen storage, community cover and aboveground biomass, litter cover and biomass, annual soil loss and runoff). We determined the dimensionality of 16 traits (including plant height, leaf and root traits), identified the key traits repres enting plant functional varations. We established the mechanistic link among trait, abundance, and ecosystem properties and compared the different intermediate effects of biomass and cover using the fouth corner analysis . We use d the generalized linear mix ed effect models to compare the effect of community size and the different components of biodiversity on soil carbon storage. We used different methods to calculate the multiple trait functional diversity indices, aiming to enhance their ablity to explain the ecosystem functioning. We compared the multifunctionality of different ecosystems and analyzed the relationship between species richness and ecosystem multifuncitonality. We evaluated the effect of different restoration practices on ecosystem functioni ng.
(1)We established a non random trait network of six dimensions and identif ied the six key traits in different dimensions using network analysis. The six traits were force to punch a leaf, plant height, force to tear a root, root mass fraction, specific leaf area, and leaf nitrogen concentration. Some dimensions are not independent, which implies the common ecological and evolutionary pressures exerted on them. Compared with unconstrained ordination, network analysis can objectively and effectively desc ribe the trait relationships.
(2)The fourth corner analysis suggested that abundance was more important than trait to understand the investigated ecosystem properties. Compared with biomass,coverage was more important. Height was negatively associated wi th soil bulk density and soil C N ratio, and positively associated with soil total nitrogen concentration,community cover, aboveground biomass and litter biomass. Root depth and lateral spread were negatively associated with soil bulk density and soil C/N ratio, and positive ly associated with soil organic matter and total nitrogen concentrations,community cover, aboveground biomass, and litter biomass. Leaf area was negatively associated with soil C/N ratio and litter cover.
(3 ) Community size and trait d iversity (or community functional parameters) affected the soil carbon storage together, and the effect of community size was more important than tho se of biodiversity. Trait diversity is more associated with soil carbon storage than species diversity. Sin gle trait functional diversity indices performed similar to or better than multiple trait indices and the ensemble forecast of multiple single trait indices. The performance of multiple trait indices can be improved greatly. The best models to forecast the soil carbon storage consist of community size and the functional diversity indices calculated u s ing specific leaf area
(4)The certain species (such as Robinia pseudoacacia and Populus simonii and species richness siginificantly affected the ecosystem m ultifunctionality. After the forests were excluded from the analysis , ecosystem multifuncitonality increased with species richness.
(5) Besides the soil nitrogen storage, other ecosystem functions of forests, shrublands and grasslands were higher than those of corn Zea mays ) cropland ecosystem a fter the implementation of Grain for Green Project. The multifunctionality of forests (dominated by R pseudoacacia and P simonii , respectively) is higher than those of other ecosystems.
These results confirm that the context dependency of ecological functions of plant traits. The biomass ratio hypothesis, niche patitioning theory, and green soup hypothesis were not mutualy exclusive, and commonly explained the changes of ecosystem functioning . The quantity and qual ity of communities together drive the provision of ecosystem functions. Takeing into account the carry ing capacity of ecosystem s and local social and economic condtions, sustain ing and increasing the provision of multiple ecosystem functions or services sh ould consider the productivity and biodiversity of communities, especially trait diversity.