1. Introduction

Variation in plant functional traits reflects adaptive strategies to environmental gradients, is shaped by evolutionary history, involves correlated trait evolution, and is influenced by both climate and biotic interactions (Herrera, 1998; Wang et al., 2016, 2022; Chen et al., 2017; Onstein et al., 2018; Koike and Masaki, 2019). Many key traits, such as leaf area, plant height, and seed mass, have been extensively studied within global comparative frameworks (Wright et al., 2004; Moles et al., 2005, 2009; Wang et al., 2016; Pan et al., 2020; Sanchez-Martinez et al., 2025). In contrast, fruit volume, a trait pivotal to plant reproductive ecology due to its influence on dispersal mode, seed predation, germination, and offspring establishment (Eriksson, 2016; Rehling et al., 2021), thereby affecting fitness and population dynamics, has received comparatively little synthesis at large scales. Although Sinnott-Armstrong et al. (2018) revealed broad geographic patterns in fleshy fruit size using fruit length as a proxy, the volume of dry fruits which dominate many temperate floras remains poorly studied. This knowledge gap limits our understanding of how reproductive strategies vary across environments, and constrains predictions of plant responses to ecological and evolutionary pressures. Therefore, investigating the environmental and evolutionary drivers of interspecific fruit volume variation is essential.

The correlated evolution hypothesis and the dispersal syndrome hypothesis provide functional and ecological explanations for interspecific fruit volume variation. The correlated evolution hypothesis predicts that fruit volume covaries with vegetative traits such as plant height and leaf area due to shared resource allocation strategies (Herrera, 2002; Jordano, 2014). Species with greater stature and larger leaves generally exhibit enhanced photosynthetic capacity and carbon assimilation, which may support increased reproductive investment and the development of larger fruits (Wright et al., 2007; Lusk et al., 2019; Brito et al., 2025). Consistent with this, woody plants, which tend to be taller and possess greater functional and photosynthetic capacity, often produce larger, fleshy fruits with nutrient-rich pericarps to attract animal dispersers, whereas herbaceous species typically prioritize rapid reproductive turnover and produce smaller, dry fruits adapted for abiotic dispersal (Primack, 1987; Tiffney and Mazer, 1995; Tabarelli and Peres, 2002; Chen et al., 2004; Bolmgren and Eriksson, 2005; Cortés-Flores et al., 2013, 2019). The dispersal syndrome hypothesis emphasizes the role of frugivores in shaping fleshy fruit traits (Jordano, 1995; Lomáscolo and Schaefer, 2010; Yu et al., 2024); however, our study focused on many temperate floras dominated by abiotic dispersal and does not directly test this mechanism. While these hypotheses collectively address functional and ecological mechanisms of fruit size variation, they generally do not quantify the constraining role of deep phylogenetic history (Wang et al., 2022; Feng and Wang, 2024).

Evolutionary history can strongly structure trait distributions, as evidenced by phylogenetic signal, the statistical tendency for related species to resemble one another (Blomberg et al., 2003; Losos, 2008, 2011b). This pattern is often interpreted within the framework of phylogenetic niche conservatism (PNC), in which evolutionary history limits the rate or extent to which traits adapt to novel environments (Losos, 2008). Many studies have reported strong phylogenetic signals in vegetative and reproductive traits, including specific leaf area, wood density, flower size, fruit type, and seed mass (Chamberlain et al., 2014; Xu et al., 2017; Ma et al., 2019; Guillemot et al., 2022). Such findings suggest that evolutionary history may constrain adaptive flexibility, as species often exhibit conserved responses to climate variation (Wiens, 2007; Wiens et al., 2010; Sanchez-Martinez et al., 2020). Integrating evolutionary history with functional traits may provide important insights into trait variation. However, it remains unclear whether fruit volume exhibits similar levels of phylogenetic constraint, particularly in species-rich and environmentally heterogeneous systems (Yu et al., 2024).

Climatic gradients may further shape fruit volume by mediating resource availability, structuring frugivore assemblages, and filtering community composition (Hampe, 2003). At large scales, high water and energy inputs may support larger fruit development, whereas arid or seasonally stressful habitats tend to favor smaller, drought-tolerant fruits (Chen et al., 2017). Climate also regulates frugivore community composition and activity (e.g., frugivores are more abundant in tropical regions) (Kissling et al., 2009; Machado Mota et al., 2022), thereby shaping selection on fruit volume and phenology (Jordano, 2014). Beyond these direct effects, climate change can also modulate trait conservatism by interacting with phylogenetic constraints (i.e., evolutionary histories reflected in the degree to which phylogenetic relationships explains trait variation) (Du et al., 2015, 2017; Feng and Wang, 2024). In extreme or highly filtered environments, strong environmental selection may modulate phylogenetic inertia, leading to trait convergence or increased trait lability (Cavender-Bares et al., 2009; Jurado-Rivera et al., 2020). However, few studies have tested how climate mediates the combined effects of evolutionary history, plant traits, and broader environmental filtering on fruit volume at large spatial scales.

To fill these gaps, we compiled phylogenetic, functional, and climatic data for 2668 angiosperm species sampled from 22 long-term monitoring stations across China's major ecosystems. Biotic interactions were not explicitly modeled, as most stations are located in temperate regions dominated by dry-fruited species with predominantly abiotic dispersal; nevertheless, we acknowledge that frugivore-fruit interactions represent an important avenue for future work. Based on this framework, we addressed three key questions: (1) How strong is the phylogenetic signal of fruit volume in these temperate floras which are characterized by a high prevalence of abiotic dispersal? (2) What are the relative contributions of phylogenetic constraints, functional traits, and climate to interspecific variation in fruit volume? (3) How does the relative importance of phylogenetic constraints vary along environmental gradients? Using phylogenetic generalized linear models and hierarchical variance decomposition, we revealed the joint phylogenetic and trait drivers of fruit volume. This study not only quantifies the balance between historical constraint and trait correlated evolution in reproductive traits but also offers predictive insights into how plant community composition and associated reproductive strategies may respond to ongoing environmental change.

2. Materials and methods 2.1. Data collection

The data analyzed in this study were sourced from the publicly available Chinese Ecosystem Research Network (CERN) plant species checklist (Zhang et al., 2020), which integrates plant species records from 22 ecological stations across China between 1998 and 2018. The geographic scope of these stations ranges from 87°55′9.4″ E to 133°30′6.9″ E and 21°57′39.4″ N to 47°35′18.5″ N (Fig. S1), encompassing four major ecosystem types: forests, grasslands, wetlands, and deserts (Table S1). All species names were standardized according to the Flora of China (http://www.iplant.cn/foc). After excluding ferns and bryophytes, the final dataset comprised 2668 angiosperm species (from 996 genera and 183 families), represented by 3440 independent records.

Functional trait data for these plant species were compiled from the Flora of China, the Chinese Virtual Herbarium (http://www.cvh.ac.cn/), and the Plant Photo Bank of China (http://ppbc.iplant.cn/). Traits included fruit length (cm), fruit width (cm), fruit height (cm), leaf length (cm), leaf width (cm), plant height (m), fruit type (fleshy/dry), and growth form (woody/herbaceous). For continuous traits, the maximum observed value reported in the databases was used, under the assumption that maximum trait values represent the phenotypic upper limit shaped by natural selection, a rationale commonly adopted in previous studies (Moles et al., 2009; Yu et al., 2024). Categorical traits were defined as follows: fruits were classified as "fleshy" if they possessed a fleshy pericarp, fleshy aril, or other fleshy appendages at maturity; otherwise, they were classified as "dry" (Willson et al., 1989; Wang et al., 2022). Species were considered "woody" if they exhibited a lignified stem or were classified as woody in relevant literature; all others were considered "herbaceous".

Fruit volume was estimated according to fruit type using geometric approximations derived from fruit length (L), width (W) and height (H). The choice of coefficient was informed by fruit morphology to provide biologically meaningful estimates. For fleshy fruits with spherical shapes, as well as fleshy and dry fruits with approximately ellipsoidal morphologies (e.g., achenes, caryopses, capsules, nuts, and the nutlet of samaras), volume was calculated as: V = (π/6) × L × W × H. For elongated dry fruits such as follicles, siliques, and legumes, volume estimation required accounting for their characteristic bilateral tapering. A purely cylindrical model tends to overestimate actual volume, whereas a spindle-shaped model tends to underestimate it. To balance these extremes, we applied a midpoint correction between ellipsoid and spindle approximations, yielding: V = (π/8) × L × W × H. This parameter-free, geometry-based approach avoids arbitrary empirical coefficients and is widely used in macroecological trait syntheses (e.g., Huynh et al., 2020; Kaliniewicz et al., 2022). Although our "midpoint" correction is a pragmatic estimate, a sensitivity analysis using a single coefficient (π/6) for all fruits produced fully congruent results (Fig. S2a), confirming that our conclusions are robust to this methodological choice.

Leaf area was estimated using the ellipse area formula: A = (π/4) × L × W, where L is leaf length and W is leaf width. This geometric approximation is widely applied in ecological morphology and functional trait quantification (Rashidi and Gholami, 2008; Pérez-Ramos et al., 2019). It should be noted that this simplified formula may be less accurate for leaves with non-elliptical morphologies, such as needle-like, deeply lobed, or compound leaves. Nevertheless, in a sensitivity analysis that excluded species with such extreme leaf forms, the relationship between leaf area and fruit volume remained qualitatively unchanged (Fig. S2b), confirming that our core results are robust to potential estimation biases.

The final dataset included a total of 2668 angiosperm species, of which 1085 (40.7%) were fleshy-fruited and 1583 (59.3%) were dry-fruited. Woody species accounted for 1620 (60.7%) and herbaceous species accounted for 1048 (39.3%). Continuous traits exhibited substantial interspecific variation. The mean plant height was 7.20 ± 8.63 m, the mean leaf area was 86.94 ± 172.17 cm², and the mean fruit volume was 11.57 ± 91.69 cm³ (mean ± SD; Fig. 1). As trait values are species-level constants derived from database maxima and do not vary across sites for a given species, the subsequent analyses primarily address interspecific variation and community assembly patterns rather than within-species plasticity or local adaptation.

In addition, eight bioclimatic variables were extracted from the WorldClim 2.1 database at a spatial resolution of 30 arc-seconds, using the geographic coordinates of the 22 stations (Fick and Hijmans, 2017). These variables comprised three precipitation metrics: annual precipitation, precipitation of the wettest month, and precipitation of the driest month; three temperature metrics: annual mean temperature, maximum temperature of the warmest month, and minimum temperature of the coldest month; and two climate variability indices: precipitation seasonality and temperature seasonality (Table S2). These climate variables were used to construct a multidimensional hydrothermal gradient framework.

2.2. Statistical analysis

All statistical analyses were conducted in R v.4.5.1 (R Core Team, 2025). A species-level phylogenetic tree was constructed using the R package "V.PhyloMaker" (v.0.1.0; Jin and Qian, 2019), which mapped all 2668 target species onto a backbone topology of 74, 533 vascular plants. Of these, 2618 species (98.1%) were directly matched to taxonomic information. The remaining 50 unmatched species (belonging to 26 genera) were inserted into genus-level nodes using a conservative polytomy approach with soft polytomies (scenario = "S3"), which helps preserve topological stability while minimizing biases caused by taxonomic uncertainty.

To quantify interspecific similarity in functional traits, Pagel's λ was calculated for continuous traits (fruit volume, plant height, leaf area) using the "phylosig" function from the "phytools" package (v.2.5–2; Revell, 2012). Values of λ range from 0 (no phylogenetic signal) to 1 (strong phylogenetic signal) (Pagel, 1999). For binary traits (fruit type and growth form), Fritz-Purvis D statistic was calculated using the "phylo.d" function from the "caper" package (v.1.0.4; Orme, 2013). D values close to or below 0 indicate strong phylogenetic constraint, while values near 1 suggest a random distribution across the phylogeny (Fritz and Purvis, 2010).

Trait data were missing for fruit volume (411 records, 15.4%), plant height (254 records, 9.5%), and leaf area (87 records, 3.3%). Given the strong phylogenetic signals detected for these traits (Table 1), missing values were therefore imputed using a phylogeny-based multiple imputation approach, implemented through the "phylomice" (https://github.com/pdrhlik/phylomice) and "mice" packages (v.3.18.0; Van Buuren and Groothuis-Oudshoorn, 2011). This method was selected for its ability to integrate phylogenetic covariance directly into a flexible, multivariate imputation framework, which is suitable for datasets like ours exhibiting strong phylogenetic signal and comprising mixed data types. We further validated the robustness of the imputation procedure. All analyses were performed using both the imputed dataset and the original, non-imputed dataset. The results from both datasets showed consistent patterns, confirming the robustness of the imputed values.

Firstly, we investigated the relationships between fruit volume and functional traits (fruit type, growth form, plant height, leaf area). All functional traits are species-level constants derived from database maximum values; they do not vary across sites for the same species. Therefore, the following phylogenetic analyses were performed at the species-level (n = 2668). Phylogenetic generalized linear models (PGLMs) with Gaussian error distribution were fitted using the "phylolm" function in the "phylolm" package (v. 2.6.5; Tung Ho and Ané, 2014), with fruit volume as the response variable and the aforementioned plant traits as predictors, and continuous traits being log₁₀-transformed. Secondly, to assess station-level variation in fruit volume while accounting for repeated species occurrence across sites (n = 3440) and phylogenetic relatedness, we fitted phylogenetic generalized linear mixed models (PGLMMs) using the "pglmm" function in the "phyr" package (v.1.1.0; Li et al., 2020). Log₁₀-transformed fruit volume was used as the response variable. Station was included as a fixed effect, and species identity was included as a random effect to account for repeated occurrences of the same species across multiple stations. Phylogenetic non-independence was incorporated by specifying the species-level random effect with a variance-covariance matrix derived from the phylogeny under a Brownian motion model, constructed using the "vcv.phylo" function in "phytools". To test whether fruit volume differed significantly among stations, we compared the full model (with station as a fixed effect) against a null model containing only an intercept using likelihood ratio tests.

To assess the relative contributions of phylogenetic relationships, traits, and climate to differences in fruit volume, hierarchical partitioning of R² was conducted under the PGLM framework using the "phylolm.hp" package (v.0.0–3; Lai et al., 2025), again using species-level trait values (n = 2668). Multicollinearity was assessed using variance inflation factors (VIFs), all of which were < 5. This method decomposes the total model R² into the independent contributions of phylogenetic relationships and predictor groups. For analyses at the station scale, we fitted separate PGLMs to the species list of each ecological station (five stations were excluded due to absence of fleshy-fruited species or sample sizes fewer than 20; thus, 17 stations were analyzed). Each station was treated as an independent community sample. Because the predictor contributions within a station sum to 100%, forming compositional data, we evaluated differences in contributions among predictors across stations using the Friedman test, a non-parametric repeated-measures method appropriate for such data structure. To further assess the stability of the estimated predictor contributions against potential sampling variability, we conducted a resampling analysis within each station. For stations with sufficient species, we randomly drew 100 species (with replacement) and repeated this process 100 times. For each resampled subset, we recalculated the relative contributions using the same hierarchical partitioning framework. The results (Fig. S3) confirmed a highly consistent pattern across replicates, with phylogeny consistently explaining the largest proportion of variance, followed by plant height and growth form, supporting the robustness of our main conclusions.

Substantial variation was observed in the contribution of predictors to fruit volume across stations (e.g., phylogeny range: 0–73.04%). To identify how climate modulates these contributions, we have therefore re-run the analysis using Spearman's rank (ρ) correlation and calculated 95% bootstrap confidence intervals (1000 replicates) for each correlation coefficient. This robust, non-parametric approach is less sensitive to outliers and provides a direct measure of uncertainty. We calculated correlations for all eight bioclimatic variables. Additionally, we investigated the phylogenetic constraints of fruit volume by separating fleshy and dry fruits, as they have distinct adaptive strategies. This separation helps compensate for the lack of biotic interaction consideration in this study, providing clearer insights into how these groups independently respond to evolutionary and ecological factors.

3. Results

Analysis of phylogenetic signal revealed strong and statistically significant conservatism for all continuous traits: fruit volume (λ = 0.916, P < 0.001), plant height (λ = 0.940, P < 0.001), and leaf area (λ = 0.881, P < 0.001). Similarly, binary traits exhibited strong phylogenetic signals. Growth form (D = −0.241, P_random = 0, P_brownian = 1) and fruit type (D = −0.263, P_random = 0, P_brownian = 1) significantly deviated from a random distribution, indicating strong associations with evolutionary history (Table 1). A parallel analysis performed on the original, non-imputed dataset yielded virtually identical patterns (Table 1), confirming that the strong phylogenetic signals detected here are not an artifact of the imputation procedure.

Phylogenetic generalized linear models (PGLMs) revealed significant differences in fruit volume across trait categories (n = 2668). Fleshy-fruited species had significantly larger fruit volumes than dry-fruited species (β = 0.36, t = 4.82, P < 0.001; Fig. 2a), and woody species exceeded herbaceous species (β = 0.53, t = 7.42, P < 0.001; Fig. 2b). Although a statistically positive relationship was found between plant height and fruit volume (β = 0.47, t = 11.34, P < 0.001; Fig. 2c), as well as between leaf area and fruit volume (β = 0.38, t = 11.76, P < 0.001; Fig. 2d), it is important to note that the latter relationship appears weak and may be influenced by a few outliers. Phylogenetic generalized linear mixed models (PGLMMs) indicated significant variation in fruit volume among ecological stations (χ² = 2043.79, P < 0.001, n = 3440), with the highest values observed at the forest station (Mean ± SD: 20.04 ± 140.47) and the lowest at the desert station (0.16 ± 0.34) (Fig. 2e).

Hierarchical variance partitioning of the multivariate model based on the imputed dataset revealed that, among all predictors, phylogenetic relationships independently explained the largest proportion (64.71%) of the explained variance in interspecific fruit volume, followed by plant height (11.99%), growth form (9.97%), climate (7.23%), leaf area (4.02%), and fruit type (2.08%) (Fig. 3a). A parallel analysis conducted on the original, non-imputed dataset yielded highly consistent estimates (phylogeny: 67.92%; plant height: 11.41%; growth form: 9.04%; climate: 6.07%; leaf area: 3.76%; fruit type: 1.8%; Fig. 3c).

Models constructed for 17 stations using the imputed data showed that phylogeny explained more variance than other factors in most stations, with its average contribution being among the highest (χ² = 25.93, P < 0.001; Fig. 3b). Analysis based on the original dataset supported the same spatial pattern, with phylogeny showing a significantly higher average contribution than all other predictors (Fig. 3d). These results were consistent with the overall trend, but revealed substantial spatial variation in the explanatory power of phylogenetic relationships (range: 0–73.04%) (Fig. 3b). In contrast, the contributions of functional traits such as plant height (0.84–46.50%) and growth form (6.49–49.21%) were highly variable across space, whereas those of leaf area (1.13–23.56%) and fruit type (0.53–27.90%) were comparatively more stable (Fig. 3b).

Of the eight bioclimatic variables examined, only the maximum temperature of the warmest month showed a significant negative correlation with the contribution of phylogenetic relationships to fruit volume variation (Spearman's ρ = −0.57, P = 0.018, n = 17; Fig. 4a and Table S3). No other climate variable, nor the contributions of any functional trait (plant height, growth form, leaf area, or fruit type), was significantly correlated with the climatic gradients tested. It should be noted that, although statistically significant, these correlations are moderate in strength and the data exhibit substantial dispersion (Fig. 4a). Therefore, climate should be interpreted as a modulating factor that influences the relative importance of phylogenetic and functional constraints, rather than as a strong, deterministic driver of fruit volume variation.

After separating fleshy and dry fruits, fruit volume variation remained significantly correlated with functional traits, and phylogenetic constraints explained the largest proportion of interspecific fruit volume variation (70.73% for fleshy fruits, 65.61% for dry fruits), consistent with the overall results (Tables S4 and S5). For dry fruits, the phylogenetic constraint on fruit volume variation was negatively correlated with the maximum temperature of the warmest month (ρ = −0.55, P = 0.022, n = 17; Fig. 4b and Table S3). In contrast, for fleshy fruits, the phylogenetic constraint showed positive correlations with maximum temperature of the warmest month (ρ = 0.50, P = 0.085, n = 13; Fig. 4c and Table S3).

4. Discussion

This study provides a comprehensive assessment of the phylogenetic and ecological drivers underlying interspecific fruit volume variation across 2668 angiosperm species in China. By integrating large-scale trait data, phylogenetic reconstruction, and climate gradients, we demonstrated that fruit volume exhibited a strong phylogenetic signal. Furthermore, our variance partitioning analysis revealed that phylogenetic relationships accounted for 64.71% of the explained variation in fruit volume, followed by plant height (11.99%) and growth form (9.97%), highlighting that evolutionary history is the predominant factor structuring fruit volume distributions across species pools. This result is consistent with the theory of PNC, which posits that closely related species are more ecologically similar than would be expected under a model of trait evolution governed solely by Brownian motion (Wiens and Graham, 2005; Wiens, 2007; Losos, 2008).

Our results demonstrate that fruit volume is strongly associated with key functional traits, particularly growth form, plant height, and leaf area. This suggests that vegetative and reproductive traits may have co-evolved along conserved evolutionary trajectories, rather than evolving independently, consistent with patterns of coordinated trait evolution reported in previous studies (Moles et al., 2009; Moles, 2018; Bogdziewicz et al., 2023; Yang and Swenson, 2023). Such patterns suggest that reproductive traits like fruit volume are embedded within broader functional syndromes shaped by integrated strategies for resource acquisition and dispersal. For instance, fleshy-fruited woody species consistently exhibited larger fruit volumes, likely reflecting co-adapted trait combinations that attract animal dispersers and support greater investment in offspring survival (Stapanian, 1982; Donatti et al., 2007; Guimarães et al., 2008; Delmas et al., 2020).

Importantly, our findings reveal substantial spatial variation in the strength of phylogenetic constraint as a driver of community trait composition. While phylogenetic relationships remained the dominant predictor overall, its explanatory power declined markedly in regions with higher maximum temperatures during the warmest month (ρ = −0.57, P = 0.018, n = 17). This spatial heterogeneity suggests that the role of PNC in structuring community-wide fruit volume distributions is not uniform but is modulated by local climate conditions, likely through climate-driven filtering of the regional species pool. This is consistent with Losos's perspective that PNC is not universal (Losos, 2008). The mechanism behind this climate modulation warrants deeper consideration, as two non-exclusive evolutionary processes could attenuate phylogenetic signal: (1) convergent evolution driven by strong, uniform environmental filtering, which would increase trait similarity among distantly related species (Losos, 2011a); or (2) adaptive radiation or divergent selection facilitated by expanded niche space, which would increase trait disparity within lineages (Losos, 2010). Distinguishing between these mechanisms would require explicit metrics of trait dispersion within communities, which our study does not directly provide; this remains an important avenue for future investigation.

Moreover, warmer regions often harbor greater floristic diversity and may experience accelerated lineage turnover (Qian et al., 2016, 2020), which could reduce trait similarity among close relatives. Our results do not directly test lineage turnover rates; hence, we present this as one plausible, though not exclusive, explanation. Additionally, the greater resource availability and extended growing seasons in warmer and moister regions may facilitate increased reproductive investment, potentially favoring the evolution of larger, more fleshy fruits. These adaptive responses modulate phylogenetic constraints under strong environmental filtering, reinforcing that while the global predominance of phylogeny is maintained, its strength is context-dependent (Chen et al., 2017; Lyu et al., 2021; Messeder et al., 2024).

In addition, biotic interactions may contribute to this spatial pattern. Warmer climates, particularly in tropical and subtropical regions, support more diverse frugivore communities (Fleming et al., 1987). These assemblages vary in body size, gape width, and feeding behavior (Pijl, 1982; Fuzessy et al., 2018), potentially imposing heterogeneous selection pressures on fruit morphology. Such selective variation may lead to greater trait divergence within communities and attenuate phylogenetic constraints (Brodie, 2017; Yu et al., 2024). However, when we analyzed fleshy fruits separately (for which biotic interactions are more relevant), the phylogenetic constraint on interspecific fruit volume variation was higher in warmer regions. This pattern can be understood as a historical legacy of tropical diversification, where many fleshy-fruited lineages originated and diversified under stable climates and persistent frugivore associations, thereby canalizing fruit traits along phylogenetically conserved pathways (Nascimento et al., 2020; Wang et al., 2022). Rather than being diminished by contemporary biotic pressures, phylogenetic conservatism appears to be reinforced through long-term coevolution between plants and dispersers (Valenta and Nevo, 2020; Messeder et al., 2024). This persistence of strong phylogenetic signal in warmer regions likely reflects the deep historical origins of fleshy fruits in the tropics (Biffin et al., 2010; Chen et al., 2017), where evolutionary trajectories were conserved even during northward spread (Wang et al., 2022; Messeder et al., 2024).

An intriguing question arising from our study is whether the contrasting patterns between fleshy and dry fruits in our temperate-dominated flora are universal. In tropical, fleshy-fruit-dominated systems, the interplay between intense biotic selection and deep phylogenetic history might produce even more complex signatures. For dry-fruited lineages, abiotic dispersal may lead to stronger environmental filtering and convergence in colder or drier extremes. Future comparative studies across tropical and temperate floras, with explicit metrics of trait convergence/divergence (e.g., mean pairwise distance, net relatedness index), are needed to test the generality of our findings and disentangle the relative roles of biotic versus abiotic filters in different biogeographic contexts.

Despite the robustness of our findings, several limitations should be acknowledged. First, our functional trait data were compiled from existing national databases and digital resources rather than collected through standardized field measurements across all stations. While this approach enabled us to assemble a large-scale, multispecies dataset, it may introduce some heterogeneity in measurement precision. Second, sampling effort varied across stations, and five sites were excluded from spatial models due to insufficient data, potentially biasing estimates in underrepresented habitats such as deserts. Third, our analyses focused on abiotic and intrinsic factors (climate and functional traits) and did not incorporate direct measures of biotic interactions, such as seed disperser identity or frugivore diversity, which are increasingly recognized as key drivers of fruit evolution (Lomáscolo et al., 2010; Jordano, 2014; Eriksson, 2016).

5. Conclusions

Taken together, our findings highlight the dominant role of evolutionary history in shaping fruit volume, while also revealing the critical influence of local ecological conditions. The weakening of phylogenetic constraint in warmer environments underscores the context dependency of PNC and suggests that environmental filtering and species turnover may modulate deep evolutionary constraints. These results emphasize the importance of integrating phylogenetic and ecological perspectives when studying trait evolution across communities and predicting plant responses to environmental change. By linking macroevolutionary patterns to functional trait strategies, this work contributes to a more nuanced understanding of plant reproductive evolution. Future research incorporating direct measures of biotic interactions, such as dispersal syndromes, frugivore assemblages, and seed traits, as well as explicit tests of trait convergence versus divergence across biomes, will be essential for disentangling the multifactorial drivers of fruit trait variation and clarifying their ecological and evolutionary significance.

Acknowledgements

This study was funded by National Natural Science Foundation of China (32171533, 31971444 and 32560127), Anhui Provincial Natural Science Foundation (2208085J28).

CRediT authorship contribution statement

Yingqun Feng: Conceptualization, Methodology, Software, Data Curation, Visualization, Writing - Original Draft. Jiming Cheng: Conceptualization, Methodology, Software, Data Curation. Chao Zhang: Methodology, Software, Data Curation, Visualization. Yujiao Ma: Methodology, Software, Data Curation, Visualization. Bo Wang: Writing - review & editing, Writing - original draft, Validation, Supervision, Resources, Project administration, Methodology, Investigation, Funding acquisition, Conceptualization.

Data availability

The data that support the findings of this study are openly available in the Figshare repository at https://doi.org/10.6084/m9.figshare.31240807.

Declaration of competing interest

The authors declare that they have no known competing financial interests or personal relationships that could have appeared to influence the work reported in this paper.

Appendix A. Supplementary data

Supplementary data to this article can be found online at https://doi.org/10.1016/j.pld.2026.03.010.