Fig. 2: Comparison of the fraction of leaf nitrogen in RuBisCO (fLNR) estimated by \({{{{{{\boldsymbol{V}}}}}}}_{{{{{{{\boldsymbol{c}}}}}}}_{{{{{{\bf{max }}}}}}}}^{{{{{{\boldsymbol{25}}}}}}}\) models to fLNR estimated by random forest (RF).
From: Global variation in the fraction of leaf nitrogen allocated to photosynthesis
Five empirical \({V}_{{c}_{{\max }}}^{25}\) models (EM1 to EM5) and two optimal \({V}_{{c}_{{\max }}}^{25}\) models (EO and LUNA) were examined. a fLNR per plant functional type (PFT) and the responses of fLNR to the first principal components (PC1s) of (b) leaf traits, (c) climate variables, and (d) soil variables. In (a), the acronyms and the numbers of half-degree cells for PFTs are: cropland (CRO; n = 10,525), deciduous broadleaf forest (DBF; n = 7525), evergreen broadleaf forest (EBF; n = 4626), evergreen needleleaf forest (ENF; n = 6259), mixed forest (MF; n = 713), grassland (GRA; n = 5771), shrubland (SH; n = 3606) and wetland (WET; n = 940). For each box plot, the cross indicates the mean, the center line indicates the median, the box indicates the upper and lower quartiles and the whiskers indicate the 10th and 90th percentiles of the data. In (b)–(d), Y axis indicates the partial changes in fLNR (ΔfLNR; unit: %). PC1s account for 83.4%, 57.3%, and 60.0% of the variance in leaf traits, climate variables, and soil variables, respectively (Supplementary Table 1). The partial response of fLNR to PC1s is acquired using a generalized additive model where the solid line indicates the mean partial response, and the shadings indicate one standard error augmented by 10.
