1. Introduction

Water scarcity is a major challenge for plant life, particularly in drylands. As water scarcity intensifies across the globe, understanding how plants respond to water stress has become more important than ever (Berdugo et al., 2020; Lickley and Solomon 2018; Rowland et al., 2023). How plants deploy their foliage (leaf size, number and distribution) determines their ability to capture sunlight, optimize water use, maintain structural stability, and ultimately survive in water-limited environments (McDowell et al., 2008; Smith et al., 2017; Taneda and Tateno 2004). Plants may allocate biomass preferentially to the stem or foliage within a shoot, producing either many small leaves or a few large ones, so as to optimize critical functions in specific environments (Kleiman and Aarssen 2007; Taneda and Tateno 2004). Across the diversity of plant foliage deployment, the trade-off between leaf size and leafing intensity (i.e., the number of leaves per unit stem size) constitutes a principal pathway of strategic compromise, limiting the simultaneous maximization of both traits (Huang et al., 2016; Kleiman and Aarssen 2007; Yang et al., 2008). Research on this trade-off has greatly advanced our understanding of how alternative foliage deployment strategies are enabled and restrained (Dombroskie and Aarssen 2012; Huang et al., 2016; Milla 2009; Sun et al., 2019; Xiang et al., 2010; Zeng et al., 2024). Yet, the functional significance of leafing intensity and its possible combinations with leaf size in dealing with water stresses has been largely unclear.

Both theoretical and experimental studies (Dudley, 1996; Orians and Solbrig, 1977; Parkhurst and Loucks, 1972) show that leaf size (typically referring to leaf area) generally decreases under water limitation, as smaller leaves offer advantages in maintaining hydrothermal balance and metabolic efficiency. In contrast, little is known about how leafing intensity responds to water stress. Tree forms with small stems bearing numerous leaves are proposed to be advantageous in arid regions, because of architectural plasticity associated with abundant axillary buds and easily substituted stems (Corner, 1949). This hypothesis implies that leafing intensity would be higher under water limitation. However, the opposite may be true, if the argument that stressful habitats with limited resources suppress leafing intensity (Xiang et al., 2010) applies in the case of water shortage. If the latter is the case, it calls into question whether the trade-off between leaf size and leafing intensity operates as a hard rule across a broad range of water availability. These questions call for an examination of whether the axes of trade-off between the two traits within and across habitats converge into a single line, or diverge along multiple pathways.

Leafing intensity, by definition, depends on both leaf number and stem size. On one hand, many studies show that water deficits usually suppress leaf and metamer emergence (Belaygue et al., 1996; Marc and Palmer, 1976; Ogaya and Penuelas, 2006), with little evidence suggesting the opposite. On the other hand, biomass allocation in a shoot tends to enhance structural support (the stem) for photosynthetic machineries (the foliage) under water limitation, due to mechanical and hydraulic constraints (Taneda and Tateno, 2004). Given these constraints, stem size would be larger under higher water stress. Leaf number generally increases with stem size (in terms of dry mass or length) at a rate lower than proportionality (Smith et al., 2017; Yagi, 2004; Yang et al., 2008), indicating that developing more leaves on a shoot becomes disproportionately costly in terms of structural support. As a consequence, large stems usually have lower leafing intensity than small ones. Based on these considerations, we predict that leafing intensity would be lower in drier habitats.

If leaf size is merely a byproduct of leafing intensity, as suggested by the leafing intensity premium hypothesis (Kleiman and Aarssen, 2007), the axes of trade-off between the two traits across habitats should converge into a single line, with larger leaves occurring in drier habitats. This prediction appears in stark contrast with experimental and observational evidence (Dudley 1996; Wright et al., 2017). If a unit change in leaf size itself have a functional significance in dealing with water stress, the axes of trade-off between the two traits across habitats would diverge along multiple pathways. In wetter conditions, where higher photosynthetic gains per unit area and lower risks of dehydration are allowed (Chaves, 1991), a reduction in leaf size can be economically disadvantageous and thus less favored, unless in exchange for an increment in leafing intensity. Hence, we predict that leaf size would decrease more sharply with leafing intensity (i.e., more negative relationships) in drier habitats than in wetter ones. The latter prediction appears more plausible, but requires empirical testing.

This study employed Populus euphratica as an illustrative tree species growing in hyper-arid climates to test the predictions outlined above. First, we assessed how leaf area, mass, number, stem mass and leafing intensity vary across habitats with varying water availabilities. Second, we evaluated how leaf size (in terms of both area and mass) scaled with leafing intensity within individual habitats and tested any predictable shift of the scaling slopes with increasing aridity. Answers to these questions should provide novel insights into the pathways of strategic integration allowed or restrained by the leaf size/number trade-off, and more specifically into the adaptive and/or acclimative advantages of leafing intensity under varying water stresses.

2. Materials and methods 2.1. Study species and area

Populus euphratica Oliv. (Salicaceae) is a deciduous tree species widely distributed in the Eurasian drylands, mainly concentrating along river banks and lake shores of Central Asia and northwest China (Thevs, 2005). This species belongs to a relatively old and oligotypic section of the Populus genus, known as Turanga, which originated in the middle Oligocene (Liu et al., 2022; Wang et al., 2020). It is believed that P. euphratica has ancestral ties to the ancient Mediterranean and North Africa floras during the late Oligocene and early Miocene (Liu et al., 2022). As the ancient Mediterranean Sea retreated and the Qinghai-Tibetan Plateau elevated in the early Quaternary, P. euphratica became a dominant component of riparian forests as deserts expanded across where we called Central Asia in modern times (Niu, 1991). Currently, P. euphratica is among the few arborescent forms that can establish forests spontaneously in the desert landscapes of the Eurasian inland.

Populus euphratica prefers well-lit open lands with moist sandy soils and relies critically on floods and phreatic water for growth and survival (Thevs, 2005). With a monopodial branching habit, the species has a leptocaul (i.e., intricately-branched) rather than pachycaul (i.e., coarsely-branched) form with fairly small leaves (11.8 cm² on average) and tiny samaras (< 0.06 g on average). The leaves of P. euphratica are arranged alternately along the stem, with each metamer bearing a single leaf on a current-year twig. Leaf flush typically occurs in May, and the leaves fall in late October.

We conducted the study on the target species in the lowlands around the Ebinor Lake (43.50–44.99°N, 82.55–83.56°E) and the upper reaches of the Tarim River (37.22–41.66°N, 79.95–82.79°E), Northwestern China. The two areas are respectively situated on the fringes of Gurbantunggut and Taklamakan deserts, straddling the middle part of Tianshan Mountains (Fig. S1). In terms of the Köppen-Geiger climate classification (Peel et al., 2007), a cold desert climate prevails thereupon, with mean annual precipitation < 164 mm and mean annual temperature between 7.8–10.4 ℃. A wealth of natural P. euphratica stands are relatively well-preserved in these areas, due to their role in controlling soil erosion, regulating hydrological cycles and preventing desertification.

2.2. Data collection

In the summers of 2021–2024, we selected 14 sites with varying degrees of climatic and edaphic aridity in the study areas (Fig. S1) to collect twig samples. These sites were almost exclusively dominated by Populus euphratica, with stand densities < 1300 stems per hectare. At each of the 14 sites, we sampled 10–108 trees on ca. 900–2500 m² land area. We selected adult individuals with diameter > 10 cm and with minimal interference from adjacent tree canopies, so as to control for the confounding effects of tree life stage and density on foliage deployment. From each target tree, we used an averruncator to cut 2–5 current-year twigs with fully unfolded leaves but without any furcation. The sampling positions were evenly spaced around the periphery of tree crown and restricted to similar heights between 3 m and 6 m for controlling for the confounding effects of light exposure and height on foliage deployment. In total, we sampled 1834 twigs from 505 trees across the 14 sites.

Once detached from live trees, the twig samples were immediately placed into air-tight bags and then transported back to the laboratory. In the laboratory, we separated leaves from the shoot for each twig. On the sampling day, the area of each fresh leaf was measured using an image processing program (ImageJ v.1.8, https://imagej.nih.gov/ij/), and leaf number per twig was recorded. The dry weight of individual leaf and stem samples was measured after 72 h of drying in an oven at 60 ℃ and 105 ℃, respectively.

We collected 3–10 soil sample replicates from top 30 cm to measure surface soil water content (SWC) at each of the 14 sites. From the records of monitoring wells installed at these sites, we obtained data on groundwater table depth (GWD) during the summer. Annual mean precipitation (PREC), potential evapotranspiration (PET) and vapor pressure deficit (VPD) were extracted from the online database CHELSA (Karger et al., 2017) with respect to the geographic locations (i.e., latitude and longitude) of these sites. The database has a high resolution of 30 arc seconds (about 1 km at the equator), which is fine enough to acquire the climatic heterogeneity over hundreds of kilometers in this study. At these sites, PREC ranged from 7 mm to 48 mm per year, whereas PET ranged between 779 mm and 1278 mm per year (cf. Table S1 for more details). The five environmental variables including SWC, GWD, PREC, PET and VPD were used to characterize water supply regimes at the 14 sites.

2.3. Data analyses 2.3.1. Hierarchical variance decomposition

We first used the coefficient of variation (i.e., standard deviation/mean) to characterize the variability in leaf size (LA and LM), leaf number (LN), stem mass (SM) and leafing intensity (LIM). In addition, we considered specific leaf area (the ratio of leaf area to dry mass, SLA) in the analysis, because of its relevance to foliage deployment. Then we built linear random-effect models to decompose the variance in the six traits attributable to inter-site, inter-tree and among-twig differences. Hierarchical nested random effects (twig within tree and tree within habitat) were incorporated into the model for each trait, and proportional variances from the three sources were partitioned. To improve the normality of the data distribution, all the trait values were log-transformed using a base of 10. Linear random-effect modelling was performed with the R package "lme4" (Bates et al., 2015).

2.3.2. Analyses of inter-habitat variation in foliage deployment traits

We used mixed-effects model of meta-regression to examine how individual traits vary with aridity levels. Meta-regression is conventionally used to examine any effect of a moderator (or a suite of moderators) on the outcomes across many primary study cases addressing a particular well-defined research question (Nakagawa et al., 2023). Here we treated a site as a primary study case and conducted a meta-analysis over 14 "cases". This approach allowed us to account for multiple sources of variance and to apply appropriate weights to various outcomes with different sample sizes and sampling errors.

Assuming that the mean trait value of the k-th site (denoted as θ_k) is sampled from an overarching distribution of trait values, it is modelled as follows:

\widehat{\theta}_k=\theta+\beta x_k+\epsilon_k+\zeta_k

where θ is the true mean of the overarching distribution and β is the fixed effect imposed by the moderator x. ϵ_k is the sampling error at the k-th site through which θ_k deviates from θ, and ζ k is the random effect which describes the effect of unmeasured site-k-specific conditions on θ k . It is important to note that θ is a model parameter, but not a "grand" mean value in descriptive statistics. In a meta-analysis, θ is estimated as the weighted mean over all θ ˆ k from k = 1 to k = K, and the weight for θ ˆ k is set as the inverse of the sum of within-site variance at the k-th site ( s k 2 ) and inter-site variance ( τ 2 ). In the subsequent meta-analyses, we used the DerSimonian-Laird estimator to estimate τ 2 and the restricted maximum likelihood estimator to fit the model. The I² statistic was used to quantify inter-site heterogeneity in trait means. It is defined as the percentage of variability in the outcomes that is not caused by sampling error (Higgins and Thompson, 2002).

For each of the six traits (LA, LM, LN, SM, LIM and SLA) of interest, we separately conducted a meta-regression, with site-specific trait means as the outcome (response). Because we had only 14 "cases" in the meta-regression, the risk of over-fitting would be quite high if we had incorporated too many moderators into the mixed-effects model. Instead, we used two composite measures of aridity as moderators (predictors). A principal component analysis was implemented over the five environmental variables (SWC, GWD, PREC, PET and VPD), and the scores of the first two principal components (accounting for 89.7% of total variance) were extracted as composite variables. The first principal component represented a well-defined gradient of climatic aridity, and the second, a gradient of edaphic aridity (Fig. S2). These meta-regressions were conducted with the R package "meta" (Balduzzi et al., 2019).

2.3.3. Identifying broad-scale patterns of leaf size vs. leafing intensity relationships

We employed standardized major axis (SMA) regression (Warton et al., 2006) to analyze bivariate scaling relationships. In an SMA regression, a power function y = α x β is used to describe a scaling relationship. On double-log axes, the power function has an exact form of linear equation:

log ( y ) = log ( α ) + β log ( x )

where α is the normalization constant and β is the scaling exponent. If |β| = 1, x and y covary in direct proportion (i.e., isometry). If |β| < 1, y increases (or decreases) with x at a rate lower than proportionality (i.e., hypoallometry). If |β| > 1, the relationship is the opposite, referred to as hyperallometry.

Within each site, we inspected bivariate scaling relationships for the following three pairs of trait variables: (ⅰ) LA vs. LIM, (ⅱ) LM vs. LIM, (ⅲ) LN vs. SM. As the variance decomposition showed that among-twig differences accounted for an appreciable portion (> 24%) of total variance in all traits, we decided to perform the analyses at the twig level but not to aggregate trait means at the tree level. In so doing, we preserved the variation that may reflect subtle environmental effects on foliage deployment at a more relevant scale. Moreover, tree-level analyses could be unreliable in some sites with fewer than 25 trees measured. All the analyses of scaling relationships were carried out using the R package "smatr" (Warton et al., 2012).

Based on the scaling exponents (βs) and associated standardized errors in individual sites, we estimated a mean exponent through meta-analysis to capture the central tendency across sites. We tested for the heterogeneity in βs using χ² tests, as commonly practiced in meta-analyses. This kind of test was of similar purpose as the common slope test in SMA analyses (Warton et al., 2006). In addition, we conducted meta-analyses over the site-specific correlation coefficients to show the overall strength of coordinated trait variations of interest. Finally, we tested any predictable shift of the scaling slopes over the two gradients of aridity using meta-regressions. These analyses were also performed with the R packages "meta". Because the moderators (i.e., climatic and edaphic aridity) were the same in all of the meta-regressions, either with trait means or scaling slopes as outcomes, there was an inflated risk of false discovery rate due to multiple non-dependent tests. We used the method described by Benjamini and Hochberg (1995) to adjust p values associated with the moderators. As edaphic aridity was found to be a non-significant moderator in all meta-regression analyses, we reported the results pertaining only to climatic aridity.

3. Results 3.1. Sources of variation in foliage deployment traits

In terms of coefficients of variation, leaf area and mass of Populus euphratica trees showed similar degrees of variation (Table 1). Among the six traits pertaining to foliage deployment, leaf number was least variable, and stem mass was most variable (Table 1). Leafing intensity displayed considerable variation compared with other traits. With respect to the hierarchical sources of variation, inter-habitat variance accounted for the largest proportion (35%–64%) of total variance in most of the six traits, except for leaf number (Table 1). Nonetheless, there were appreciable amounts of variation among trees within habitat and among twigs within tree, accounting for 12%–38% of total variance (Table 1).

3.2. Foliage deployment traits in relation to climatic aridity

As climatic aridity (i.e., low precipitation, high potential evapotranspiration and high vapor pressure deficit, Fig. S2) increased, leaf area and mass showed no significant change (Fig. 1a and b). Leaf number did not vary with climatic aridity, but stem mass increased marginally (Fig. 1c and d). Leafing intensity decreased significantly under drier climates (Fig. 1e). Although specific leaf area showed an apparently decreasing trend as climatic aridity increased, the trend was not significant (Fig. 1f).

3.3. Scaling relationships between leaf size and leafing intensity across and within habitats

Mean leaf size in terms of either area or mass was negatively correlated with mean leafing intensity across habitats (Fig. 2). Leaf mass showed a stronger correlation with leafing intensity than leaf area (Fig. 2a vs. 2b). In individual habitats, both leaf area and mass were also negatively correlated with leafing intensity (Fig. 3a and c, Fig. S3). The habitat-specific coefficients of correlation between leaf area and leafing intensity averaged −0.58 (95%CI = [−0.66, −0.48], Fig. S3), and there were slightly stronger correlations between leaf mass and leafing intensity with a mean coefficient of −0.66 (95%CI = [−0.74, −0.56], Fig. S3).

The scaling exponents for leaf size vs. leafing intensity varied significantly across habitats (p < 0.001, Fig. 3b and d). In aggregate, leaf area scaled with leafing intensity to the power of −0.56 (95%CI = [−0.62, −0.50]), which was significantly higher than −1 (Fig. 3b). That is, a 10-fold increase in leafing intensity coincided with an average of 3.6-fold decrease in leaf area. Similar relationships were found between leaf mass and leafing intensity, with the aggregated power of −0.63 (95%CI = [−0.70, −0.57]) that was also significantly higher than −1 (Fig. 3d). Thereby, a 10-fold increase in leafing intensity coincided with an average of 4.3-fold decrease in leaf mass. Leaf number increased with stem mass at a lower-than-proportional rate, and the scaling exponents averaged 0.40 (95%CI = [0.33, 0.48], Fig. 3e and f).

3.4. Leaf size vs. leafing intensity relationships in response to climatic aridity

The inter-habitat heterogeneity in scaling exponents for leaf size (measured as either area or mass) vs. leafing intensity was partly (50%–65%) explained by water supply regimes, notably by climatic aridity (Fig. 4). Under drier climates, the scaling exponents for leaf size (measured as either area or mass) vs. leafing intensity became more negative (Fig. 4). Nonetheless, the scaling exponents for leaf number vs. stem mass showed no predictable shift with climatic aridity (Fig. 4).

4. Discussion 4.1. Inter-habitat variation in leafing intensity and leaf size

As predicted, we found lower leafing intensities of Populus euphratica under drier climates (Fig. 1e). Because stem mass varied across habitats more pronouncedly than leaf number did (Table 1), the reduction in leafing intensity can be mainly attributed to larger stem mass under higher climatic aridity, even though the latter trend was just marginally significant (Fig. 1d). This suggests that a high investment in structural support for leaf attachment is crucial for trees in coping with water shortage. The underlying mechanism may involve an increased biomass investment required to stiffen xylem cell walls and prevent cavitation induced by low water potential (Mencuccini 2003), which is commonly encountered by water-stressed plants. There was evidence from common garden experiments that hydraulic conductance increased with stem mass per twig (Zhou et al., 2025), underscoring the critical requirement of biomass investment for securing hydraulic machinery.

Earlier studies showed that water deficit typically suppressed leaf and metamer emergence (Belaygue et al., 1996; Marc and Palmer 1976; Ogaya and Penuelas 2006). However, we found no evidence of reduced leaf number per twig in Populus euphratica under higher climatic aridity (Fig. 1c). Notably, the scaling exponents for leaf number vs. stem mass did not change with climatic aridity (Fig. 4c), indicating that the relative rate of leaf emergence to stem growth was not influenced by water availability. In contrast, the average stem mass supporting a leaf (i.e., the inverse of leafing intensity) increased with climatic aridity (Fig. 1e). Therefore, leaf number, as opposed to stem mass, was a less relevant factor determining the climatic response of leafing intensity in P. euphratica across habitats.

In his seminal work on tree branching and leafing patterns, Corner (1949) proposed that leptocaul tree forms with slender twigs bearing numerous leaves (presumably high leafing intensities) would be more advantageous in drought or cold conditions compared to pachycaul forms, due to the abundance of small buds and easily replaced twigs. This theory appears to hold across climatic zones, as pachycaul forms are more common in tropical floras than in non-tropical ones. However, a more localized study in temperate broadleaf forests showed that leafing intensity was lower at higher than at lower altitudes (Xiang et al., 2010), suggesting that increased leafing intensity may not provide a clear advantage in stressful environments, such as cold. Similarly, we found that higher water stresses favored lower leafing intensities and, to a lesser extent, larger stem sizes, for a dryland-dwelling species (Fig. 1). The key point in the disagreement is the difficulty in achieving the putative advantage of high leafing intensity in stressful environments. As our study demonstrated, the investment in structural support for leaf attachment is of great importance. Thus, it appears that high leafing intensities can only be sustained in favorable habitats that promote high meristem activities and/or allow for low investments in structural support for leaf attachment.

While leafing intensity of Populus euphratica decreased with increasing climatic aridity (Fig. 1e), leaf size in terms of either area or mass showed no significant change (Fig. 1a and b). Negative correlations between leaf size and leafing intensity were still evident at the inter-habitat level (Fig. 2). These findings apparently lend some support for the leafing intensity premium hypothesis (Kleiman and Aarssen, 2007), which posits that leaf size may have limited direct adaptive value, but is a byproduct of selection for or against a high leafing intensity. However, it remains unclear whether or not the reduced leafing intensities of P. euphratica under drier climates is an adaptive response.

4.2. Emergent patterns of within-habitat correlations between leaf size and leafing intensity

We found that leaf size and leafing intensity were negatively correlated in individual habitats (Fig. 3a and c). This finding suggests that the trade-off between the two traits can emerge at local scales, even in the absence of strong selective pressures imposed by abiotic environments. Earlier studies showed that there exists a trade-off between meristematic activity associated with leaf production and with individual leaf expansion on a shoot axis (ter Steege et al., 2005; Tisné et al., 2008). This trade-off likely forms the developmental basis of the leaf size/number trade-off. Given the fundamental role of developmental machinery in foliage deployment, it is not surprising that the trade-off between leaf size and leafing intensity can arise within relatively homogenous habitats.

The scaling exponents for leaf size vs. leafing intensity in Populus euphratica were in aggregate higher than −1 (Fig. 3b and d), but did not exhibit the isometric relationship as found in many interspecific studies (Kleiman and Aarssen, 2007; Milla, 2009; Xiang et al., 2010). More specifically, a 10-fold increase in leafing intensity was associated with an average 3.6-fold decrease in leaf area and an average 4.3-fold decrease in leaf mass. If an isometric relationship implies that the benefits of increasing leafing intensity are counterbalanced by the costs of reducing leaf size, the allometric relationship we observed in P. euphratica indicates that the benefits of increasing leafing intensity outweigh the costs of reducing leaf size. In other words, P. euphratica trees can generally afford to increase leafing intensity without severely compromising leaf size as would be expected in an isometric trade-off.

As predicted, we found that leaf size decreased with leaf intensity more sharply under higher climatic aridity, but not at the same rate across habitats (Fig. 4a and b). This indicates that the cost of increasing leafing intensity, in terms of leaf size reduction, is greater under stronger water limitation. A similar pattern was found by Dombroskie and Aarssen (2012) in Acer saccharum under stronger light limitation. In the context of water limitation, our findings have three important implications:

(ⅰ) Even if leaf size did not vary significantly across habitats, the differential rate of leaf size reduction with increasing leafing intensity is functionally significant in relation to water availability. In wetter habitats, a more gradual reduction in leaf size with increasing leafing intensity may help maximize light capture and photosynthetic gains. In contrast, in drier habitats, a steeper decrease in leaf size could help maintain hydrothermal balance and reduce the risk of self-shading and dieback.

(ⅱ) Due to the higher cost gauged from leaf size reduction, the upper limit of leafing intensity is likely lower in drier habitats. This highlights an additional challenge in achieving high leafing intensities under water limitation, beyond the need for greater stem biomass investments.

(ⅲ) While the correlation between leaf size and leafing intensity is consistently negative, there can be a number of permitable combinations of the two traits to achieve functional equilibrium, forming multiple pathways of functional equivalency across habitats. These pathways may diverge depending on water availability, with steeper or gentler reductions in leaf size with increasing leafing intensity.

To what extent the elevated cost of inceasing leafing intensity in terms of leaf size reduction under water limitation can be generalized to other species and geographic locations remains an open question. It is tempting to speculate that this response is widespread, given the vital importance of small size in maintaining leaf hydrothermal and metabolic balance (Dudley, 1996; Orians and Solbrig, 1977; Parkhurst and Loucks, 1972), along with the apperently foundmental meristematic allocation trade-off in foliage deployment (ter Steege et al., 2005; Tisné et al., 2008).

5. Conclusion

Using Populus euphratica as an illustrative tree species growing in hyper-arid climates, we showed that (i) leafing intensity decreased with climatic aridity, primarily due to increased stem mass, and (ii) leaf size decreased more sharply with leafing intensity under drier climates. These results challenge the assumption that higher leafing intensity always confers an advantage ready for environmental stresses through higher developmental flexibility offered by abundant axillary buds. Instead, we propose that lower leafing intensity may be more feasible and sustainable under water limitation, owing to greater structural support for leaf attachment and less compromise in leaf size. This work represents the first investigation of the variation in scaling relationships governing foliage deployment across an explicit gradient of aridity, moving beyond simple bivariate analyses of the spectrum of leaf size and display. We advocate for future research to synthesize existing case studies to elucidate how plant architectural scaling relationships diverge (or converge) over extensive environmental gradients and broad geographic scales. Such efforts hold promise for advancing a refined understanding of the constraints and opportunities shaping strategic integration of plant architecture under diverse environmental regimes.

Acknowledgements

This study was financially supported by the National Natural Science Foundation of China (32460329) and the Bintuan Science & Technology Program (2024AB075) to L.H., the National Natural Science Foundation of China (32360279), an open program from the Key Laboratory of Protection and Utilization of Biological Resources in the Tarim Basin (BRZD2004) and a provincial talent-introduction program of Xinjiang Uygur Autonomous Region to D.H. The authors are grateful to a number of students at Tarim University and Xinjiang University for data collection in the field and laboratory.

CRediT authorship contribution statement

Dong He: Conceptualization, Funding acquisition, Investigation, Writing - original draft, Writing - review & editing. Lu Han: Project administration, Funding acquisition, Investigation, Writing - review & editing.

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.

Data statement

The Dataset is available upon request from the corresponding authors.

Appendix A. Supplementary data

Supplementary data to this article can be found online at https://doi.org/10.1016/j.pld.2025.08.001.