Figure 4 Relationships between average plot biomass (a), plant abundance of plot (b) and network complexity are given, respectively. The least square regression is used to fit the relationships. The blue solid circle represent the mean, the blue line is the error bar, and the red quadratic curve is regression linear. Relationship between geometric average of plot biomass and plant abundance and complexity is also given (c). Note: * * * is p<0.001.
N-addition affected primary productivity of plants, but also network structure. The relationship between biomass and complexity was a significant quadratic curve with down bending direction (\(r^{2}=0.313,\ \ P<0.001\), Fig. 4a). Especially, under high nitrogen fertilization (N >= 2 g N·m-2·year-1) at high frequency (once a month) and without mowing treatment, complexity will have a great negative impact on biomass (in Fig. S3). On the contrary, there was a significant quadratic curve, with an opposite direction, between plant abundance and complexity (\(r^{2}=0.171,\ \ P<0.001\), Fig. 4b). There was a complementary relationship between biomass and plant abundance, which the regression line of geometric average of the two remained almost horizontal. In addition, biomass in short networks was significantly more than that in nested networks (\(T_{\text{test}}=2.416,\ \ P<0.05\)). The plants abundance was not able to reflect competitivity of species to resources (\(T_{\text{test}}=-0.784,\ P>0.05\)). However, biomass and plant abundance both increased in the late stage of the experiment (in Fig. S4).

3.3 Network structure affects ecological function