This study was conducted in Mountain Kaka (103° 40′ N, 32° 59′ E, Figure 1 ), located in the middle section of the Minshan Mountains, eastern Tibetan Plateau. The area forms the headstream of Minjiang and Fujiang Rivers, sitting in west Sichuan Province, China (Wang et al., 2021). The vegetation of this region has distinct altitudinal zonation and horizontal distribution with strong floristic transition and abundant plant species. Meanwhile, this region is characterized by contrasting distribution of solar radiation due to the complex undulating terrain (Shi et al., 2022). The mean annual, January (coldest month), and July (warmest month) air temperature is 5.7°C, −7.9°C, and 9.7°C, respectively, and the annual precipitation is 720 mm with peaks during June to August. The mean annual sunshine period is approximately 1,827 h. The soil is classified as Mat-Gryic Cambisol, with the pH value ranging from 6.21 to 7.11 in surface soil (0–10 cm). The grassland type is a typical alpine meadow dominated by Ranunculus tanguticuz, Pedicularis kansuensis, Pyrethrum tatsienense, B. macrophylla, B. vivipara, and C. discoideum (Wang et al., 2021).

During the growing season (mid-June) in 2016, field investigation was conducted at two sites, namely, Mountain A (103° 40′ 16″E, 32° 58′ 57″N, 3828 m a.s.l.) and Mountain B (103° 40′ 12″E, 32° 58′ 48″N, 3797 m a.s.l.). Two surrounding sample plots were set up horizontally at 30 m vertically downward from the summit of two mountains, respectively, dividing into three slope aspects according to the ridge trend, i.e., northeast (30°E), southeast (45°W), and southwest (45°E) slope aspect ( Figure 1 ). Since a large number of dwarf shrubs were distributed at the northwest slope aspect, no sampling plots were set up at this slope aspect to prevent the influence of shrubs on the growth of target species. Three subplots (5 m × 5 m) with 10-m spacing between each other were set up in an area with a relatively even distribution of vegetation in each slope aspect. Four common species, including B. macrophylla, B. vivipara, C. discoideum, and D. littoralis, as target species were collected, since they were distributed mainly in southwest China with a range of altitude from 1,200 to 5,400 m with different life forms, including C3 (B. macrophylla and B. vivipara) and C4 species (C. discoideum and D. littoralis). However, not all species were present at all slope aspects, and we did not find C. discoideum and D. littoralis at the northeast slope aspect at mountain A. For each species, 20–30 fully mature individuals including intact roots were sampled. All the fresh samples were stored in iceboxes and transferred to the laboratory for further measurements.

In total, we sampled 501 plant individuals from two mountains ( Supplementary Table S1 ). For each individual, different organs of root, stem, leaf, and flower were separated, and plant height (cm), individual leaf area (ILA, cm²), individual leaf mass (ILM, g), and specific leaf area (SLA, cm² g⁻¹) were measured following the standard protocols (Perez-Harguindeguy et al., 2016). Sampled fresh leaves were scanned by Canon Scan Gear and calculated by Photoshop CS4 and Matlab 7.9 for their leaf area (Wang et al., 2021). ILA was defined as the ratio of the total leaf area to the number of leaves of each plant individual. Plant height was measured by a ruler. Then, all the samples were dried in an oven at 65°C for a minimum of 48 h. Root mass fraction, stem mass fraction, leaf mass fraction, and flower mass fraction were calculated as the ratio of their respective dry masses to the total dry biomass. ILM was defined as the total leaf dry mass to the number of leaves. SLA was calculated as the ratio of leaf area to leaf dry weight. Total biomass is the sum of belowground biomass (i.e., root mass) and aboveground biomass (i.e., stem mass, leaf mass, and flower mass).

To evaluate the effect of soil properties on functional traits and biomass, we collected soil samples by using a soil auger and an aluminum specimen box. For each subplot, three soil samples were collected from the surface to a depth of 10 cm and mixed into one composite sample. In total, 27 soil samples were collected for each mountain. Soil properties, including soil temperature at the surface (ST, °C) and at 5 cm below the surface (ST5, °C), bulk density (BD, g cm⁻³), soil water content (SWC, %), nitrate nitrogen content (NNC, mg kg⁻¹), ammonium nitrogen content (ANC, mg kg⁻¹), soil organic carbon (SOC, g kg⁻¹), total nitrogen (TN, g kg⁻¹), total phosphorus (TP, g kg⁻¹), and available phosphorus (AP, mg kg⁻¹), were measured ( Table 1 ). Soil temperature was measured by a button thermometer (iButton-TMEX RTE). Soil bulk density was measured by using the cutting ring technique, while the oven-drying method was used for soil water content. The indophenol blue colorimetric and ultraviolet spectrophotometry methods were used to measure ammonium nitrogen content and nitrate nitrogen content, respectively (Wang et al., 2021). Soil organic carbon and total nitrogen were measured by using the dry combustion method on an elementary analyzer. Total phosphorus and available phosphorus were measured by extraction with 0.5 M sodium hydroxide sodium carbonate solution (Wang et al., 2016).

First, the difference in soil properties among slope aspect for each mountain was tested by one-way ANOVA. Second, the coefficient of variation was used to quantify the variation in functional traits for each species, which is defined as the ratio of standard deviation to the mean of each functional trait. Third, one-way ANOVA was used to assess the effect of the slope aspect on the variations in functional traits and biomass allocation for each species. All response variables were log-transformed to best meet model assumptions. Fourth, we used standardized major axis analysis (SMA) to explore the biomass allocation between above- and below ground, and the effect of the slope aspect on variation in biomass allocation strategy. The allometric equation of the form Y = c·X^a was used to test biomass allocation strategy among slope aspects, where Y is the aboveground biomass (i.e., sum biomass of stem, leaf, and flower), X is the belowground biomass, and a and c are allometric coefficients. The equation was logarithmically transformed into a linear equivalent, ln(Y) = ln(c) + a·ln(X) (Niklas & Enquist, 2001).

Finally, we used the structural equation model (SEM) to test the direct and indirect effect of both biotic (i.e., height, ILA, ILM, and SLA) and abiotic factors (i.e., slope aspect, soil properties) on the total biomass of each species. To reduce the dimensionality of soil properties, we ran principal component analysis (PCA) for 10 soil properties. The first three PCA axes explained 84% of the total variation ( Supplementary Table S2 ), which were used in the SEM. The first PC axis (PC1) described lower SWC and higher TP and AP. The second PC axis (PC2) described lower NNC, SOC, and TN, and the third PC axis (PC3) described lower ST. We constructed a hypothetical causal model, which includes several direct paths from biotic and abiotic factors to total biomass and indirect path from abiotic factors to total biomass via biotic factors ( Supplementary Figure S1 ). The slope aspect was treated as a regular numeric variable and coded as 1 (northeast), 2 (southeast), and 3 (southwest) before formulating a structural equation model. To optimize SEM, we removed the non-significant paths (p > 0.05). Then, Fisher’s C statistic (p > 0.05) and Akaike’s information criterion (AICc) were used to estimate the fit of the global model (Shipley, 2009). All data analyses were performed in R v.3.6.0 (R Core Team, 2019). SMA, PCA, and SEM analysis were performed using the R package “smart” (Warton et al., 2012), “vegan” (Oksanen et al., 2013), and ‘‘piecewiseSEM’’ (Lefcheck, 2016), respectively.

Intraspecific trait variation in individual leaf mass had the largest magnitude, followed by individual leaf area, SLA, and height in turn. The trait CV values of four species in individual leaf mass ranged from 43.88% to 88.05% ( Figure 2 ; Supplementary Table S3 ). The CV values of individual leaf area, SLA, and height were ranged from 38.40% to 49.89%,19.86% to 28.12%, and 18.05% to 24.36% ( Figure 2 ; Supplementary Table S3 ), respectively.

Slope aspects had significant effect on the functional traits of four species divergently but species specific ( Figure 3 ). We found the highest SLA of BISVIV and CREDIS, and the lowest SLA of DESLIT existed on the southwest slope aspects in both mountains ( Figures 3C, G ). However, the slope aspect had a minor effect on individual leaf area and mass, and height. The lowest and highest individual leaf area were observed for BISMAC (0.83 ± 0.07 cm², 1.58 ± 0.20 cm²) and DESLIT (1.24 ± 0.09 cm², 1.56 ± 0.12 cm²) at northeast and southwest slope aspects, respectively ( Figures 3A, E ). In addition, the lowest (0.0022 ± 0.0002 g, Figure 3B ) and highest (0.0056 ± 0.0013 g, Figure 3B ) individual leaf mass of BISMAC were presented in northeast and southwest slope aspects, respectively. Similarly, the heights of BISVIV and CREDIS were found to increase in terms of slope aspects from the northeast, southeast, to the southwest ( Figures 3D, H ; Supplementary Table S4 ).

A relatively higher root mass fraction of BISMAC (54.74%–67.07%) and BISVIV (60.90%–79.00%) but lower leaf mass fraction (4.65%–8.95%, 6.85%–12.26%) had been observed. CREDIS and DESLIT reflected lower root mass fraction (16.65%–25.30%, 12.62%–40.88%) and higher leaf mass fraction (38.81%–46.57%, 42.63%–58.42%). A higher flower mass fraction existed in CREDIS (20.82%–27.27%), followed by BISMAC (10.10%–15.16%), BISVIV (1.48%–4.15%), and DESLIT (2.40%–5.51%). The stem mass fraction of BISMAC, BISVIV, CREDIS, and DESLIT ranged from 18.03% to 22.52%, 10.16% to 24.49%, 10.75% to 14.55%, and 13.80% to 25.77%, respectively ( Supplementary Table S5 ).

One-way ANOVA indicated that the mass fraction of each organ of BISVIV and DESLIT was significantly affected by the slope aspect (p < 0.05, Figure 4 ). The slope aspect had a minor effect on both leaf mass fraction of CREDIS and flower mass fraction of BISMAC ( Figure 4 ). The effect of slope aspect on mass fraction was not consistent between the two mountains. For example, the slope aspect exerted a significant effect on root and leaf mass fraction of BISMAC at mountain A but not at mountain B (p < 0.05, Figure 4 ). Similarly, slope aspect had a minor effect on root and stem mass fraction of CREDIS at mountain A but reflected a significant effect on them at mountain B (p < 0.05, Figure 4 ; Supplementary Table S5 ).

Standardized major axis analysis revealed that the above- and belowground biomass allocation of these four species at different slope aspects shared a common slope (p > 0.05, Figure 5 ; Supplementary Table S6 ). We found that most of the species presented an isometric growth relationship between AGB and BGB at different slope aspects ( Figure 5 ). However, the slope of BGB against AGB for CREDIS was significantly <1 at the southeast of mountain A and northeast of mountain B (p < 0.01, Supplementary Table S6 ). Similar results for DESLIT were observed at southwest slope aspect of both mountains A and B (p < 0.05, Supplementary Table S6 ). In addition, we only found that the regression slope of SMA was significantly >1 for BISMAC at the southeast slope aspect at mountain A (p = 0.049, Supplementary Table S6 ).

These four SEMs provided a good fit to the data and accounted for 57%, 44%, 26%, and 36% of the variation in total biomass for BISMAC, BISVIV, CREDIS, and DESLIT, respectively ( Figure 6 ; Supplementary Tables S7–S10 ). The relative importance of biotic and abiotic factors in driving total biomass for four species was not consistent. For BISMAC, soil properties, height, ILA, and ILM had positive direct effects on total biomass. Slope aspect and soil properties affected biomass indirectly via ILA and ILM ( Figure 6A ). For BISVIV, soil properties and height presented a negative and positive effect on biomass, respectively. Soil properties also played a positive indirect effect on biomass via height, ILA, and ILM. Slope aspect exerted an indirect effect on biomass via soil properties ( Figure 6B ). For CREDIS, we found that only ILA exerted a positive direct effect on biomass. Abiotic factors had a minor effect on biomass ( Figure 6C ). For DESLIT, height played a significant positive role on biomass. Slope aspect had a negative direct effect and an indirect effect via height and soil properties on biomass ( Figure 6D ).

Variation in intraspecific plant traits relates to the phenotypic trait plasticity of species, which is determined by environmental conditions and genetics simultaneously (de Bello et al., 2011; Violle et al., 2012; Dong et al., 2020; Li et al., 2021a). Our results showed that BISMAC had higher intraspecific trait variation, especially for ILA and ILM, which suggested that BISMAC may have a relatively higher survival fitness in a stressful alpine ecosystem. However, the intraspecific trait variation in SLA and height is smaller than that of ILA and ILM of each species. The similarity and lower variation in SLA and height may arise as a consequence of the habitat filter in harsh environmental conditions leading to a similar life strategy for different species (Violle et al., 2012; Dong et al., 2020). Our study indicated that the slope aspect had a significant effect on functional traits, especially for SLA ( Figure 3 ). As an important functional trait in leaf economic spectrum, SLA is closely related to resource acquisition and use efficiency (Wright et al., 2004). The significant difference in SLA of BISVIV and CREDIS across slope aspects at both mountains may attribute to two potential reasons. First, limited light at southwest slope aspect allows plants to have a larger SLA to enhance light use efficiency, which is in line with a recent study (Li et al., 2021a). Meanwhile, to increase light use efficiency, plant also tend to possess larger leaf area. A larger leaf area of BISVIV and CREDIS was also found at the southwest slope aspect, despite being non-significant, which may support this scenario to a certain extent for light as a limiting factor at this slope aspect. In addition, we also found higher individuals of BISVIV and CREDIS at the southwest slope aspect at mountains A and B, respectively, which shows that plants allocate more towards photosynthesis production in height to strengthen their competitive ability for light resources (Dolezal et al., 2021). Second, the variation in temperature leads to a change in SLA. We found that a smaller SLA of BISVIV and CREDIS occurred at northeast slope aspect at mountain B. As the lower temperature forces high mountain plants to shift life strategy to become more conservative with lower photosynthetic rate and evapotranspiration rate, to enhance their survival fitness, it can explain why a smaller SLA emerged at the northeast slope aspect with the lowest temperature in this study. Interestingly, but not surprisingly, we found that the variation in SLA of DESLIT among slope aspects presented the opposite pattern to BISVIV and CREDIS, which highlighted the heterogeneity in species response to the same environment (Niinemets, 2006; Cheng and Niklas, 2007; Valladares and Niinemets, 2008). Functional traits can effectively affect growth of individuals and even predict ecosystem functioning like productivity and stability (Cadotte, 2017; He et al., 2019). In majority of the previous studies, community-weighted traits are usually used to predict ecosystem function (Tjoelker et al., 2005; Craven et al., 2018). However, our results provide a new insight that the predictive capability of functional traits for ecosystem functioning may be enhanced when we account for the variation in functional traits under microclimates.

Similarly, slope aspect played a significant effect on mass fraction among organs ( Figure 4 ). To maximize survival and optimize growth in the alpine ecosystem, plants generally allocate more biomass to belowground part (Dolezal et al., 2021). We found that a higher root mass fraction occurred at the low-temperature slope aspect ( Figure 4 ). The root mass fraction of CREDIS and DESLIT ranged from 16.65% to 25.30%, and 12.62% to 40.88%, respectively, which is consistent with a previous study reporting root mass fraction values of plants between 10% and 50% (Poorter et al., 2012; Poorter et al., 2015). However, in our study, the root mass fraction of BISMAC and BISVIV far exceeded this range, especially at the northeast slope aspect. In particular, compared to the south slope aspect, the north slope aspect in northern hemisphere receives less solar radiation, thus resulting in a lower-temperature condition in this slope aspect (Singh, 2018). As such, the slower depletion of carbohydrates and turnover of root stem from colder environments lead to higher accumulation of root mass (Davidson, 1969; Gill and Jackson, 2000; Yang et al., 2010; Pierick et al., 2021). The higher stem mass fraction mainly emerged at the southwest slope aspect, as mentioned above, plants invest more biomass to the stem to increase asymmetric light competition. Consistent with results on the effect of slope aspect on functional traits, the variation in stem mass fraction among slope aspects did not follow a similar pattern between four species, which again highlighted their species-specific response (Niinemets, 2006; Cheng and Niklas, 2007; Valladares and Niinemets, 2008). It also emphasizes the importance of biodiversity for ecosystems with hyper-diverse communities of species with different strategies for responding to environmental fluctuations that will reduce overall community fluctuations through asynchronous responses among populations and thus maintain ecosystem function (also known as species asynchrony, Loreau and de Mazancourt, 2008). Finally, flower mass fraction was also higher at the southwest slope aspect. Investing more resources into flowers in light-limited areas will help plants maintain a higher reproductive capacity in adverse environments (Leck et al., 2008).

Plants adjust their biomass allocation strategies to maintain necessary physiological activities, achieve normal growth, and improve environmental adaptability (Shipley and Meziane, 2002; Mensah et al., 2016). The allocation of plant biomass among organs is not only driven by environmental conditions but also genetics (Poorter et al., 2012; Pallas et al., 2016). We found that most species presented an isometric relationship between above- and belowground biomass at several slope aspects, which is in line with a research focus on community level (Yang et al., 2010). However, both CREDIS and DESLIT presented an allometric relationship, investing more biomass belowground than aboveground at several slope aspects ( Figure 5 ). Meanwhile, our results are also consistent with a study that alpine plants are threatened by low temperatures, even though a shallow snow cover can alleviate the low temperature stress somehow (Dolezal et al., 2016). Additionally, we also found that BISMAC invested more biomass to aboveground at the south east slope aspect, while it invested equal biomass between above- and belowground at the south west slope aspect, which support the optimal partitioning theory that plants are able to achieve optimal acquisition of resources by regulating biomass allocation strategies (Gedroc et al., 1996). It is worth to note that we did not take the density effect into account in this study. For instance, plants tend to invest more biomass to aboveground for a higher competitive capacity for light resources (Poorter et al., 2012; Sun et al., 2021; Tang et al., 2021; Tao et al., 2021). Therefore, the density effect on biomass allocation needs to be considered carefully in future studies.

Our results indicated that both biotic and abiotic factors drove total biomass of plants simultaneously, while their relative importance varied between the four species ( Figure 6 ). The plant height can explain the large variation in biomass, which is in line with recent studies from aquatic plant communities (Gustafsson and Norkko, 2019). The increase in plant height facilitates competition for light resources, which in turn increases productivity (Westoby, 1998; Diaz et al., 2004). For the same reason, the increase in individual leaf area enhanced the assimilation efficiency for biomass accumulation, which supports the leaf economic spectrum hypothesis that exploitative traits tend to have higher productivity (Wright et al., 2004). Unexpectedly, SLA had a neutral effect on biomass. Probably, some unmeasured traits, such as root traits, may increase the traits explanation of biomass in alpine ecosystems with lower temperature stress. In addition, interspecies relationships can also be critical for biotic regulation of biomass allocation, depending on their positive facilitation or negative competition. Thus, these issues should be taken into consideration in future studies to ascertain the relative importance and contribution of biotic and abiotic factors, respectively. In addition, our study also showed that abiotic factors have both direct and indirect effect on biomass. For example, slope aspect affected total biomass directly and indirectly via height for DESLIT. Moreover, consistent with recent research, we found that slope aspect significantly affected soil properties, which thus mediated plant biomass indirectly (Li et al., 2021b; Zhang et al., 2022). Similarly, we also found direct and indirect effects of soil properties on biomass via different path between four species. In summary, our results emphasize the importance of slope aspect for alpine plants not just about plant traits but also biomass allocation regulating their plasticity to merge biotic and abiotic ways.