Non-structural carbohydrates (NSC) are critical mediators of plant adaptation to fluctuating light environments. They are tightly coordinated with plant economic strategy, i.e., leaf economics spectrum (LES) and root economics spectrum (RES). However, the role of solar spectral composition in shaping the NSC pool of tree species and their relationships with LES and RES is poorly understood. We examined plant economic traits and NSC of two functional groups, light-demanding vs shade-tolerant tree seedlings, grown under five spectral-attenuation treatments: 1) control, transmitting 95% of solar radiation (CK); 2) attenuating ultraviolet (UV)-B radiation; 3) attenuating all UV radiation; 4) attenuating all UV radiation and blue light; 5) attenuating all UV, blue, and green light. Short-wavelength regions (UV) had strong effects on plant economic traits and the NSC dynamic, irrespective of functional groups. Shade-tolerant species exhibited higher trait plasticity than light-demanding species. Coordination of NSC and economic traits differed between the two groups. Across treatments, leaf and root NSC were negatively correlated with LES for both groups, and with RES for light-demanding species, while they were positively correlated with RES for shade-tolerant species. A more tightly coordinated trait-NSC network was evident in light-demanding species, and UV-A radiation promoted the network tightness. These findings highlight the role of spectral composition in regulating the coordination between above-/belowground functional traits and C dynamics. Different tree species may have employed contrasting strategies to adapt to the solar spectral composition in their habitats.

Linear mixed effect models (LME) were used to explore the effects of spectral treatment and functional group, and their interaction on each functional trait (Table S3), with species and blocks as random factors, using the nlme package (Wang et al., 2020). Pair-wise comparisons were conducted to analyze the effects of specific spectral treatments with the multcomp package (Hothorn et al., 2008) when the effect was significant (P < 0.05). P-values were corrected with the Benjamini-Hochberg (BH) method (Benjamini and Hochberg, 1995). We used redundancy analysis (RDA) to assess the relationship between spectral treatments (using the actual irradiance measured under the filters) and leaf/root traits for the two studied functional groups (light-demanding and shade-tolerant), using the vegan package (Oksanen et al., 2015). Statistical significance was determined by the Monte Carlo permutation method and Bonferroni’s test (permutations = 999, P < 0.05). Prior to analysis, data were log-transformed, and response variables (functional traits and NSC variables) were centered and standardized. The RDA provides a visualization of the primary gradients in the trait-spectra relationships, while the formal statistical tests for the independent effects of each spectral region are presented in subsequent analyses.

Relative Distances Plasticity Index (RDPI) was used to determine the plasticity of functional traits to each spectral region (Valladares et al., 2006; Ma et al., 2024). RDPI was calculated as the Euclidean distance (d) between the trait values exposed to two spectral regions (ij and i′j′, respectively), normalized using the sum of the absolute trait values (x_{i′ j′} + x_ij). Then, LME were used to explore the effect of plant traits, spectral region, and their interaction on RDPI, with species and traits as random factors (Wang et al., 2020). Differences in plant trait plasticity between functional groups (light-demanding vs shade-tolerance) across treatments were tested using t-test.

Principal component analysis (PCA) was conducted to explore leaf and root trait coordination across spectral treatments, related to resource acquisition strategies for each functional group, using the FactoMineR package (Husson et al., 2014). The scores from the first PCA axis (PC1) represented the continuum of the LES (LN, LPhen, SLA, and LCN). The scores of the first two axes of root traits were also selected, with the former dominated by RN and RTD, and the latter dominated by RD and SRL, representing root conservation and foraging and dimensions, respectively. Linear regressions were performed to explore the relationships between PC1 and PC2 scores of LES and RES and concentrations of NSC and its components in leaves and roots, respectively, for each spectral region.

Plant trait networks were constructed to examine the coordination between key economic traits and NSC across spectral treatments and functional groups. This was based on a Pearson adjacency matrix with a significance threshold of P < 0.01 (e.g., |r| ≥ 0.2 at n = 75) (Csardi & Nepusz, 2006). In the network, plant traits and NSC variables were represented as nodes, while trait–trait relationships were represented as edges. Network connectivity and centrality were calculated for each trait and NSC variable using the degree and betweenness functions in IGRAPH (Csardi & Nepusz, 2006). Three parameters [edge density (ED), diameter (D), and average path length (AL)] were used to quantify the tightness, and the average clustering coefficient (AC) was used to quantify the complexity (He et al., 2020). Networks with higher D and AL exhibit greater overall independence among traits, whereas those with higher AC are more extensively partitioned into several distinct components (Armbruster et al., 2014; He et al., 2020). All statistical analyses were performed in R version 4.3.3 (R Core Team, 2024).