Multi-environment validation of nitrogen-normalized phenotypic indices for auxiliary clonal screening in the timber crop Toona sinensis

Toona sinensis is a high-value cultivated timber tree for which efficient clonal screening across heterogeneous environments is important. We evaluated whether two nitrogen-normalized phenotypic indices provide useful information beyond their component traits: a structural index (S-NNI = LMA/LN) and a height index (H-NNI = H/(LN+RN)). Eighteen clones were tested at three field sites in China (Lu’an, Heze, and Xingtai). Validation combined algebraic-coupling diagnosis, allometric tests, agreement with non-ratio residuals, external criterion validity, grouped leave-one-clone-out cross-validation, and within-site rank changes. S-NNI agreed strongly with the non-ratio LMA residual (ρ = 0.734, 95% CI = 0.559–0.830) and was positively associated with site-adjusted palisade tissue thickness (ρ = 0.313, 95% CI = 0.031–0.581). For leaf thickness and palisade tissue thickness, the S-NNI model produced the highest leave-one-clone-out R² values among six candidate models (0.176 and 0.163, respectively), although its RMSE advantage over the site-only and quadratic models was uncertain. H-NNI showed moderate agreement with the non-ratio height residual (ρ = 0.478) and site-adjusted DBH (ρ = 0.576), increasing DBH prediction R² from 0.865 in the site-only model to 0.913. However, additive and interaction models retaining H, LN, and RN performed better, and H-NNI provided no independent information after adjustment for its components. Within-site rankings changed most for S-NNI in Xingtai and H-NNI in Heze. S-NNI may therefore serve as a parsimonious preliminary descriptor of leaf structural phenotype, whereas H-NNI should be restricted to exploratory detection of ranking discordance. Neither index represents nitrogen-use efficiency or replaces production traits, direct component analysis, or multi-environment stability assessment.