| 英文摘要: | To the Editor —
Fernández-Martínez et al.1 show that nutrient availability is the chief determinant of net ecosystem production (NEP) and ecosystem carbon-use efficiency (CUEe, the ratio of NEP to gross primary production, GPP) in global forests. But their conclusions depend on an improper treatment of differences in the GPP range of nutrient-rich and nutrient-poor forests (uneven sampling effect) and outliers. A statistical re-analysis of their data sets while simultaneously excluding the uneven sampling effect and outliers indicates no significant control of nutrient availability on carbon balance.
First, NEP and CUEe both have a nonlinear relationship with GPP (Fig. 1a,b), and this indicates that an uneven sampling effect can result in misleading conclusions. Taking nutrient-poor forests as an example, CUEe within the GPP range of 1,000–2,000 gC m−2 yr−1 (16 ± 3%; mean ± s.e.m.) is significantly higher than that within the whole GPP range (6 ± 4%) (t-test, P < 0.05). A generalized linear model analysis indicates that such differences of GPP ranges significantly affect NEP (P < 0.05). Therefore, the statistical analysis of Fernández-Martínez et al.1 should have been based on samples within a same GPP range to exclude the uneven sampling effect. |