Statistical analyses

The aboveground biomass (AGB) and species richness of each plot in 2010 prior to treatment were used as a baseline against which all treatments comparisons were made. The response ratios (R)²⁸ of the AGB and species richness were calculated to quantify the impacts of N application by comparing the variables of the after-treatment (2011–2015) to that of the pre-treatment (2010). The response ratio was calculated as

where V_A is the value of the variable in each plot with the N application and V_P is the pre-treatment value of the same variable in the same plot.

To exclude the effects caused by other factors, we revised the response ratios by the correction factor (cf)²⁹. The cf was calculated as

where is the mean pre-treatment value of the variable in the control plot (N0) and is the mean after-treatment value of the same variable in the control plot.

To perform statistical analyses, we calculated the relative effects of N addition (E) by transforming the revised response ratios to the natural log, which was calculated as

where R is the response ratios and cf is the correction factor.

Then, we calculated the relative effects of each N treatment on the total AGB, species richness, the AGB of functional groups and the common species in each N treatment year. The relative effects would be zero, positive or negative if no change, increased or decreased effects occurred with the N addition, respectively.

The relative effects of N addition on these variables were tested in a repeated-measure analysis of variance (ANOVA). The year was treated as a repeated effect. N treatments entered the model as class variables, and they were defined as continuous variables. The Student-Newman-Keuls (SNK) post-hoc test was performed to evaluate the differences in the relative effects at different N treatments and among experimental years for each variable. Least-squares regression line was applied in describing the relationship between the relative effects and nitrogen loads. All analyses were carried out using the statistical software SPSS21.0 (IBM Company, Armonk, NY, USA).