Experimental design
The field experiment was set up in April 2011 and has been described elsewhere (Zhang et al., 2015, Zhou et al., 2017, Zhou et al., 2018). Large herbivores were excluded between March and October by fencing the experimental area. A homogeneous area of meadow covering 230 x 100 m was divided into four parts that were given N, P, their combination or neither. Six plots, each 10 x 20 m, were established within each nutrient area. Fertilization treatments consisted of a factorial combination of N and P addition applied annually to fertilized plots in each of three blocks: N, P and NP. Nitrogen was supplied at a rate 15 gN m-2 y-1, phosphorus at a rate of 8 gP m-2 y-1, and nitrogen and phosphorus at a rate of 10 gN m-2y-1 and 8 gP m-2y-1. While we acknowledge that plots within each nutrient area are not independent, previous studies have shown that there were no significant differences among them in term of plant species diversity, community biomass and community composition at the start of the experiment (Zhou et al., 2018). N was applied as ammonium nitrate (NH4NO3) and P as monocalcium phosphate (Ca(H2PO4)2) annually at the end of May. Each plot was subsequently divided into two 10 × 10 m subplots; one randomly assigned subplot was used to measure aboveground plant biomass through time for twenty common species (Table S1), and the other subplot was used to measure aboveground plant biomass and species composition in early August in one randomly selected 0.5 x 0.5 m quadrat.