1. Introduction

The mechanisms underlying elevational richness patterns have long fascinated ecologists and biogeographers (Rahbek et al., 2019; Dani et al., 2023). Understanding the drivers of elevational richness patterns not only enhances our knowledge of how biodiversity responds to climate variation but also improves predictions of how future climate change may affect ecological communities (IPCC, 2021; Ma et al., 2022; Pepin et al., 2022). Additionally, the compressed replication of climatic zones over short spatial distances in mountains provides a unique natural laboratory for testing macroecological theories (Testolin et al., 2021; Zhang et al., 2021). Elevational richness patterns emerge from the interplay of local ecological processes (e.g., environmental filtering) and regional processes (e.g., speciation) (Rahbek et al., 2019; Sun et al., 2025; Yang et al., 2025). However, these processes are distorted through the 'observational window' of sampling intensity (Lomolino, 2001; Zhang et al., 2021). While increased sampling improves the accuracy of assessing elevational richness patterns, it requires significant time and effort (Deng et al., 2020). Therefore, monitoring programs require evaluation to establish a cost-effective sampling intensity that yields adequate data for measuring richness.

Sampling intensity (here, sampling intensity is defined as the number of sampling sites per kilometer of elevational range along a transect), directly influences the detection of ecological heterogeneity across elevational gradients. Higher sampling intensity improves the resolution of continuous richness variations and enhances the identification of nonlinear patterns (e.g., mid-elevation peaks; Fig. 1a). Conversely, low sampling intensity may distort peak locations in ecologically complex zones, reducing pattern accuracy (Fig. 1b), and even risk missing critical elevations, potentially generating artificial trends (e.g., spurious declines) that misrepresent true richness patterns (Fig. 1c; Lomolino, 2001). Moreover, sampling intensity significantly affects the interpretation of richness-climate relationships. Insufficient sampling intensity may obscure critical ecological thresholds, where species turnover rates change abruptly along elevational gradients (Jankowski et al., 2013). For instance, when sampling fails to capture transitional zones between biomes (e.g., forest-to-shrubland ecotones), the apparent climatic drivers of richness may be misidentified (Grytnes and McCain, 2007). This sampling artifact could lead to an overestimation of the importance of climatic factors.

However, the influence of sampling intensity may not be uniform and depends critically on the biological characteristics of the study organisms. For instance, growth forms with large body sizes, such as trees, integrate environmental conditions across broad spatial scales, meaning that their distributions reflect coarse-grained habitat variations (Gong et al., 2019; Taylor et al., 2023; Zhou et al., 2025). Therefore, a relatively low sampling intensity can still capture the major trends in richness patterns. By contrast, small organisms, such as herbaceous plants, respond to microclimatic conditions that can vary dramatically over just meters, requiring a much higher sampling intensity to accurately capture their richness patterns (Di Biase et al., 2021; Loidi et al., 2021; Zu and Wang, 2022). However, this difference has not received sufficient attention, and the majority of elevational studies continue to employ standardized sampling schemes, typically using fixed sampling intensity (e.g., 3.3 sampling sites per 1000 m (McCain and Grytnes, 2010)), regardless of the study organisms. This uniform approach risks either oversampling large-bodied growth forms (wasting resources) or critically undersampling small-bodied growth forms (missing important ecological transitions). For example, a sampling design appropriate for detecting tree community turnover might completely miss microhabitat specialists, such as certain herbs that require sampling at 20–50 m intervals to reveal their true distributional limits. Such mismatches between sampling intensity and organismal perception can lead to significant misinterpretations of both richness patterns and their associated climatic drivers. Cheng et al. (2020) demonstrated significant divergences in richness-climate relationships between trees and herbs, whether these differences stem from their distinct sampling requirements remains unresolved.

The influence of sampling intensity also varies significantly among different mountain regions. This is because climatic gradients (such as temperature and moisture changes with elevation) and landscape features (such as slope steepness or valley shapes) differ greatly between mountain ranges. These differences create unique patterns of environmental heterogeneity (Perrigo et al., 2020). In mountains with high environmental heterogeneity (e.g., rugged terrain with many microclimates), species composition often changes rapidly over short elevational distances, requiring higher sampling intensity to accurately capture key richness patterns. If sampling is too sparse, critical details, such as sharp transitions between forest types or isolated high-richness zones, can be missed, leading to misleading trends (Lomolino, 2001; Di Marco and Santini, 2015). In contrast, in regions with more homogeneous conditions, species composition changes gradually, so relatively low sampling intensity is sufficient (Zeidler et al., 2023). Therefore, conducting analyses along the same elevational transect (i.e., within the same mountain area) offers a major advantage by controlling for location-specific effects. All studied groups experience identical landscape complexity and environmental gradients (Sekar et al., 2024).

In this study, we aimed to provide a more nuanced understanding of how sampling intensity affects richness assessment in mountain ecosystems. Through a comparative analysis of richness data collected at varying sampling intensities along the same mountain slope, we examined the elevational richness patterns of different plant growth forms (trees, shrubs, and herbs) and the stability of richness-climate relationships across sampling intensities in the Baima Snow Mountain. We also assessed the suitable sampling intensity (SSI) for each growth form. Furthermore, we synthesized existing studies via meta-analysis to summarize the current state of research on plant diversity elevational patterns and identify key knowledge gaps. The primary objectives of this research were as follows: 1) to quantify whether the sampling intensity influences the detected elevational richness patterns and richness-climate relationships for trees, shrubs, and herbs differently, and 2) to determine the SSI among these three plant growth forms.

2. Material and methods 2.1. Study area

The study area is the Baima Snow Mountain in Yunnan, Southwest China (27°25′–28°35′ N, 98°50′–99°3′ N) (Fig. 2a and b), with the summit reaching an elevation of 5429 m. The western slope at the foot of the mountain has an elevation of 1815 m. This region experiences a subtropical monsoon climate with a mean annual temperature (MAT) of 4.7 ℃ and a mean annual precipitation (MAP) of 640 mm. The interaction between the southwest monsoon and the steep mountain slopes leads to a distinct vertical stratification of vegetation communities, arranged from the bottom to the top as follows: river valley shrub zone (< 2600 m), montane coniferous forest zone (2900–4300 m), alpine shrub and meadow zone (4100–4600 m), alpine scree sparse vegetation zone (4600–4900 m), and alpine ice and snow zone (> 4900 m), with the tree line at an elevation of 4100 m.

2.2. Elevational richness pattern data

In this study, data were collected from two sources: field surveys and published datasets. We conducted a vegetation survey on the western slope of Baima Snow Mountain in July 2023, covering an elevation range from 3740 to 4740 m above sea level. Eleven sites were selected at 100-m elevation intervals, with similar orientation, slope, and soil type at each site. At each site, three replicate tree plots (10 × 10 m) were established to record the tree species. Each tree plot was further subdivided into four shrub plots (5 × 5 m) and three randomly chosen herb plots (1 × 1 m). For trees, species richness was calculated for each 10 × 10 m plot, and the average of the three replicates was used as the site-level tree richness. For shrubs and herbs, species recorded in all subplots within each 10 × 10 m tree plot were pooled (with unique species counted once) to obtain a plot-level richness; the site-level shrub/herb richness was then calculated as the average of the three plot-level values. We used the average rather than the total richness to avoid pseudo-replication, since summing across replicate plots would repeatedly count species occurring in multiple plots and thus artificially inflate richness estimates. This averaging approach provides a robust and comparable estimate of species richness per site by capturing local variability while minimizing the influence of rare species and local heterogeneity (Giladi et al., 2011). Moreover, we integrated the collected data with the published literature from Nie et al. (2022) and Yang et al. (2016) for a comprehensive analysis. Both studies investigated tree, shrub, and herb richness along the western slope (Table S1).

2.3. Climatic data

The climatic data used in this study were derived from long-term meteorological data. Particularly, these data were obtained from three representative meteorological stations (Benzilan Station, 122 Road Station, and Baima Pass Station) reported by Wang et al. (2021), covering the period 2012–2022. The selected climatic factors included MAP, MAT, and actual evapotranspiration (AET). We chose these variables because extensive previous research has consistently demonstrated their strong explanatory power for elevational richness patterns (Peters et al., 2016). These variables encapsulate the critical aspects of climatic influence on elevational richness patterns. MAP and MAT provide a fundamental understanding of climatic conditions, reflecting the direct input of moisture and thermal energy available to ecosystems. AET, on the contrary, represents the combined effect of these inputs and the atmospheric demand for moisture, making it a comprehensive measure of the water balance and energy dynamics. We calculated AET based on Turc's formula (Kluge et al., 2006) as follows:

A E T=M A P /\left[0.9+(M A P / L)^2\right]^{1 / 2}

L=300+25 M A T+0.05 M A T^3

Together, these variables offer a robust framework for interpreting the complex interactions between climate and richness at different elevations.

2.4. Statistical analysis

Our overall approach to the statistical analyses of samples was as follows: a) we created subsample transects by bootstrap resampling to simulate varying sampling intensities; b) we estimated the elevational richness pattern for each subsample transect and assessed the role of climatic factors in shaping the elevational richness pattern; c) we evaluated how the variability of both elevational richness patterns and richness-climate relationships changed with sampling intensity; and d) we identified the inflection points in the relationships between sampling intensity and the variability of both richness patterns and richness-climate relationships.

2.4.1. Bootstrap resampling without replacement

Bootstrap resampling was employed to address the limitation of our original dataset and to quantitatively determine the minimum sampling effort required for reliable inference. Given that our study was initially based on only three transects (each comprising a fixed number of sampling sites), we had only a single estimate of the elevational pattern and its climatic relationships per transect. This made it impossible to assess whether the observed estimates were stable or sensitive to variations in sampling intensity.

In our hierarchical sampling design, plots served as the fundamental and replicated sampling units, nested within sites that represented distinct elevations along the gradient. Transects, on the other hand, functioned as three independent replicas of the entire elevational gradient, each comprising a series of sites. During the bootstrap resampling, this hierarchical structure was strictly preserved. Resampling was performed at the level of sites within each transect. For each iteration, a subset of sites (ranging from five up to the total number of sites in that transect) was randomly selected without replacement, and all plots nested within each selected site were retained. Each subsample thus represented a defined sampling range, which was subsequently used to control for the effect of sampling range in later analyses. This method ensured that (1) each elevational site appeared only once per subsample, accurately reflecting actual field survey conditions; (2) the elevational structure of each original transect was preserved; and (3) artificial inflation of sampling intensity, which can arise from replacement methods, was avoided (LaFontaine, 2021).

The three original transects were resampled separately using termination conditions designed to address different sampling scenarios. Resampling was terminated after either 1000 unique subsample transects or 3 million iterations, whichever occurred first. This dual-threshold approach was implemented for two reasons. First, in cases with large sampling ranges and low sampling intensities, the number of possible unique subsample transects becomes very large; therefore, the 1000-subsample limit ensured computational efficiency while maintaining statistical robustness. Second, for small sampling ranges with high sampling intensities, the number of possible unique subsample transects is naturally limited and may not reach 1000; therefore, the 3 million iteration limit prevented infinite loops while still capturing all meaningful permutations. Together, these criteria ensured thorough yet efficient resampling across all possible sampling configurations in our study.

2.4.2. Estimating elevational richness patterns and their climatic drivers

To assess elevational richness patterns and their climatic drivers, we first interpolated climatic variables at each sampling site based on the three meteorological stations. The observed temperature and precipitation data from these stations were used to fit regression models describing the elevational relationships of climatic variables. For temperature, the MAT and elevation data were used to estimate the elevation-dependent lapse rate following Li and Zhang (2010):

T=T_0+\gamma\left(H-H_0\right)

where T is the MAT (℃) at elevation H (m), T₀ is the MAT at the reference station (Benzilan Station, the lowest station) at elevation H₀ (m), and γ is the temperature lapse rate (℃/m). For precipitation, the MAP and elevation data were used to estimate the regression coefficients a and b following Yu and Yu (1996):

P z=a \cdot l^{-b \cdot(Z-H)^2}

\ln =(P \mathrm{z})=\ln a-b \cdot(Z-H)^2

where Pz is the precipitation (mm) at elevation Z (m), and a and b are the regression coefficients. The resulting lapse rate and precipitation coefficients were then applied to interpolate MAT and MAP at each sampling elevation across the transects. AET was calculated based on the interpolated MAT and MAP.

We assessed elevational richness patterns and climatic drivers using generalized linear models (GLMs) for each growth form (trees, shrubs, and herbs) across all subsample transects. Both linear and quadratic elevation terms were tested to capture potential responses, with model selection performed using the Akaike Information Criterion (AIC; Akaike, 1974). To evaluate the importance of climatic drivers, separate GLMs were fitted for MAT, MAP, and AET, testing linear and quadratic forms.

2.4.3. Assessing how the variability of elevational richness pattern and richness-climate relationship change with sampling intensity

First, it is essential to exclude the influence of the sampling range because the sampling intensity and sampling range are highly correlated in this study (Gallou et al., 2017). For example, when bootstrap resampling is applied to elevational transect data, replicates with broader sampling ranges (e.g., 2000 m) inherently include more unique sampling sites, leading to a higher sampling intensity (number of samples per unit area) simply because of the expanded spatial coverage. Conversely, narrow-range replicates often comprise fewer sites, resulting in a lower sampling intensity. In this study, we employed a stratified analysis to disentangle the specific effect of sampling intensity by controlling for sampling range variation. Particularly, the dataset of subsamples was divided into 18 discrete sampling range groups (500–2200 m), representing the elevational differences among the bootstrap subsamples. Within each group, each subsample included between 5 and the total number of sites whenever possible (Table S1). These groups were then used to analyze how sampling intensity influenced both elevational richness patterns and richness-climate relationship variability.

The number of bootstrap resampling transects varied substantially across different sampling intensities within the sampling range group. For instance, at a sampling intensity of 5, we could extract 1000 bootstrap resampling transects, whereas at an intensity of 20, only 1 transect was possible to extract. Such inconsistency in resampling sizes could significantly bias the results without appropriate correction. To solve this statistical power issue, we used the bias-corrected coefficient of variation of AIC (CVC), which corrects for biases due to unequal and small subsample transect sizes (Cureton, 1968; Vogel and Fennessey, 1993). Higher CVC indicates meaning that the observed richness patterns and climate relationships are unstable and potentially unreliable for inferring true ecological processes. In contrast, lower CVC providing a more reliable reflection of actual ecological relationships (Zhang et al., 2021). The CVC was calculated as follows (Biemann and Kearney, 2010):

S D_N=\left[\sum\left(X_i-\bar{X}\right)^2 / q\right]^{1 / 2}

q=2\left[\mathit{Γ}\left(\frac{N}{2}\right)\right]^2 /\left[\mathit{Γ}\left(\frac{N-1}{2}\right)\right]^2

C V_C=S D_N / \bar{X}

where SDN is the unbiased standard deviation, Xi is the value of the i th individual in the group, X¯ is the group mean, q is the correction factor, Γ is the Gamma function, and N is the group size.

To integrate the results across different sampling range groups, we used a weighted approach based on the number of bootstrap resampling sizes within each sampling range group, thereby adjusting their relative importance. The weighted approach is as follows (Gothi et al., 2021):

\overline{C V_C}=\frac{\sum\limits_{i=1}^k\left(\omega_i \cdot V_{N i}\right)}{\sum\limits_{i=1}^k \omega_i}

where CVCi is the CVC in ith sampling range group, and wi is the corresponding weight, CVC¯ is the CVC of whole Bootstrap resampling dataset.

Finally, to detect the threshold at which CVC changes and the variation trend shifts, we employed segmented regression. The median sampling intensity was set as the initial break point. Through iterative optimization, we identified the transition from high CVC to low. This breakpoint represents the SSI required for the reliable modeling of elevational richness patterns and their relationships with climatic drivers. All statistical analyses were performed using R statistical software (v.4.3.3; R Core Team, 2024), and segmented regression models were fitted using the segmented package (v.2.1–3; Muggeo, 2008).

2.4.4. Meta-analysis of the elevational richness pattern

To investigate the relationship between sampling intensity and elevational species richness patterns, we conducted a systematic meta-analysis of field-based vegetation studies. Our literature search across ISI Web of Science, Google Scholar, and CNKI (1970–2023) used the keywords (elevation* OR altitud) AND (richness OR diversit) AND (gradien* OR patter* OR transec* OR varian*), yielding 152 initial articles. From these, we selected studies that met the following rigorous criteria: (1) field-based vegetation surveys (excluding herbarium or museum records); (2) elevational ranges exceeding 1000 m; (3) clearly reported sampling methods and intensity; (4) focus on woody plants (trees/shrubs) or herbs; and (5) absence of significant sampling bias across elevations; and (6) establishment of at least three plots per sampling site. Finally, we selected 17 studies that satisfied all criteria (Table S3). For each qualifying study, we extracted the total number of sampling sites and the elevation range. Using these data, we calculated the sampling intensity (number of sites per 1000 m elevation band).

3. Results

Our analysis revealed a consistent trend in how the CV_C of the elevational richness pattern and its drivers change with sampling intensity across different plant growth forms (as shown in Fig. 3 for trees, shrubs, and herbs). In all cases, the CV_C initially declined rapidly before stabilizing at a plateau when the sampling intensity varied from 2.3 to 10.0, with a distinct inflection point indicating the transition. However, the rate of change and position of the inflection point varied significantly. The CV_C of tree elevational richness patterns showed the steepest initial decline in CV_C (slope = −35.526) and the lowest inflection point (sampling intensity = 4.0). Shrubs exhibited a more gradual CV_C reduction but the difference was not significant (Figs. S1 and S2), whereas herbs had the slowest decline (slope = −20.748) and were significantly lower than trees and shrubs, with the highest inflection point (sampling intensity = 5.1). Beyond the inflection point, the rate of CV_C decline slowed with increasing sampling intensity, with trees showing the slowest decline, followed by shrubs (though the difference from trees was non-significant), whereas herbs exhibited a significantly faster decline than both trees and shrubs.

In terms of CV_C variation with sampling intensity, the richness-climate relationships exhibited trends similar to those of elevational richness patterns (Fig. 4); however, the rate of decline and position of inflection points varied among growth forms. For all climate factors combined, the inflection point for shrubs occurred at a sampling intensity of 3.7, whereas trees showed an inflection at 4.8, and herbs exhibited a slightly higher inflection point than trees at 5.1. In both the pre- and post-inflection stages, trees and herbs showed the fastest rates of decline, with no significant difference between them, but both were significantly higher than those of shrubs.

Single-factor richness-climate analyses revealed that CV_C variation with sampling intensity differed among growth forms when different climate drivers were examined (Fig. 4), For MAP and AET, the CV_C of herbs showed inflection points at higher sampling intensities (both 5.2), whereas trees exhibited inflection points at lower sampling intensities (5.2 and 4.9, respectively), and shrubs displayed the earliest inflection points (occurring at sampling intensities of 3.7 and 3.1). By contrast, for MAT, the shrub CV_C reached its inflection point at the highest sampling intensity, followed by trees, whereas herbs showed the earliest inflection point. Notably, for MAP, significant differences existed among trees, shrubs, and herbs in both the pre- and post-inflection stages (Figs. S3 and S4), with trees showing the fastest decline rate in CV_C for richness-MAP relationships, followed by shrubs, whereas herbs exhibited the slowest decline as sampling intensity increased. For both MAP and AET, the CV_C of tree richness-climate relationships showed no significant difference from that of shrubs, although both were consistently higher than that of herbs. This pattern was observed across both the pre- and post-inflection stages.

Our meta-analysis revealed that the maximum sampling intensity employed in this study (10 plots/km) exceeded the sampling densities used in the vast majority of published studies on plant elevational diversity patterns (Fig. 5). Specifically, our sampling intensity for trees was greater than that reported in 67% of relevant studies, for herbs in 61%, and for shrubs in 59%. This confirms that our evaluated sampling intensity range (2–10 plots/km) effectively encompasses the densities used across most existing research. We found that only a very small number of studies utilized sampling intensities below the SSI thresholds identified in our assessment for each growth form: specifically, three tree studies fell below the tree SSI (4.0), two herb studies fell below the herb SSI (5.1), and no shrub studies fell below the shrub SSIs (ranging from 3.1 to 4.8 depending on the climate factor). However, our meta-analysis did not identify any existing elevational diversity studies that explicitly designed their sampling intensity differentially based on plant growth form.

4. Discussion

The fundamental role of sampling intensity in shaping the detectability of richness patterns and their drivers has long been recognized (Scheiner et al., 2011; Chase and Knight, 2013; Yu et al., 2013). However, previous studies have primarily focused on differences in sampling intensity across locations (Nottingham et al., 2018), communities (Kessler, 2001), or seasons (Zhang et al., 2021), whereas research on growth form-specific sampling intensity remains scarce. In this study, we observed that the sensitivity of elevational richness patterns to sampling intensity differed significantly among growth forms. With increasing sampling intensity, the accuracy of assessing tree richness patterns along elevation improved most rapidly, requiring the lowest sampling intensity, whereas herb richness patterns showed the slowest improvement in accuracy and required the highest SSI. The SSI marks the point at which the rate of CV_C decline changes substantially, indicating a shift from a phase of rapid improvement to one of diminishing returns, where further increases in sampling intensity provide only marginal gains in stability while demanding considerable time and effort (Ehlers Smith et al., 2018; Broudic et al., 2025). Therefore, our study highlights the need for standardized sampling approaches that account for growth form-specific responses to sampling intensity to accurately characterize richness patterns. These findings have critical implications for designing cost-effective monitoring programs that optimize sampling intensity across different growth forms and reconcile inconsistencies in reported elevational richness patterns across studies with varying methodologies.

The influence of sampling intensity on richness-climate relationships also exhibited growth form-specific variation (Fig. 3), but the accuracy of estimating these relationships was generally higher than that of elevational richness patterns. This difference likely occurs because elevational richness patterns are more directly constrained by local sampling completeness, whereas richness-climate relationships integrate broader-scale climatic gradients that exhibit greater spatial autocorrelation. Elevational patterns primarily reflect localized species turnover along steep environmental gradients, making them highly sensitive to sampling gaps that may miss microhabitat specialists or narrow-range endemics (Deng et al., 2020). Contrastingly, climate-richness relationships emerge from species' aggregated responses to more gradually varying climatic factors (Tripathi et al., 2019), whose spatial structure helps buffer against localized sampling limitations (Whittaker et al., 2001). However, this buffering capacity has limits; when sampling is too sparse, it fails to capture climate-specialized growth forms whose distributions define key portions of the climatic gradient, thereby flattening observed relationships.

Through a meta-analysis of elevational richness studies on plant growth forms over the past two decades, we found no studies that systematically addressed growth form-specific sampling methodologies. This striking gap highlights a critical limitation of current approaches to assessing elevational richness patterns. When uniform sampling protocols are applied across plant growth forms, they inevitably lead to either inefficient resource allocation or compromised data quality issues. For instance, using identical plot sizes and sampling intensities for trees and herbs would result in oversampling for trees (wasting resources without improving accuracy) while simultaneously undersampling herbs (introducing bias through missed microhabitat specialists). Such methodological inconsistencies likely contribute to the unresolved debates surrounding the elevational richness patterns in the literature.

The differences in sampling intensity requirements across growth forms for assessing both elevational richness patterns and richness-climate relationships can be primarily attributed to variations in species turnover rates and ecological tolerance among plant growth forms. This is likely due to their distinct ecological strategies and the spatial scaling of environmental responses. For instance, trees, as long-lived woody plants, generally exhibit the strongest ecological tolerance and slowest species turnover along elevational gradients (Alexander et al., 2018). Their richness patterns are predominantly shaped by broad-scale environmental filters (e.g., temperature and precipitation) that vary predictably with elevation (Easdale et al., 2007). Consequently, the relatively gradual species turnover of trees means that even low-to-moderate sampling intensities can sufficiently capture their richness patterns and climate relationships, as their distributions are less influenced by fine-scale microclimatic heterogeneity. By contrast, herbaceous plants, owing to their small body size, demonstrate higher sensitivity to local microhabitat conditions (Deng et al., 2025) (e.g., soil moisture and light availability) and exhibit faster species turnover across spatial scales (Murphy et al., 2016). This microclimatic specialization leads to greater fine-scale heterogeneity in their distributions, necessitating higher sampling intensities to accurately characterize both the elevational richness patterns and climate relationships. Without sufficient sampling intensity, the distributions of microhabitat-specialized herbs may be overlooked, resulting in underestimated richness and flattened richness-climate relationships.

In addition, differences in SSI among growth forms may stem from variations in community structure. A growing body of evidence suggests that community characteristics (e.g., species evenness and abundance distributions) critically influence sampling intensity requirements (Jeliazkov et al., 2022). Communities with high evenness and/or a predominance of rare species typically require greater sampling intensity to achieve accurate richness estimates (Brose et al., 2003). In our study, the relatively low SSI for trees may reflect their lower species richness in the Baima Snow Mountain, where a few dominant species often account for most individuals. However, this pattern likely shifts in hyperdiverse tropical montane regions, such as the Andes, where tree communities exhibit exceptionally high species richness and a prevalence of rare growth forms (Richter et al., 2009). Under such conditions, substantially higher sampling intensities are necessary to reliably capture both elevational richness patterns and their climatic drivers, as rare species disproportionately contribute to richness estimates in these systems (Durán et al., 2019). Future studies should explicitly compare how sampling intensity requirements vary for the same growth forms (e.g., trees vs. herbs) across mountain systems with contrasting richness regimes (e.g., temperate vs. tropical mountains). Such comparative studies would help disentangle the relative contributions of species traits, community structure, and regional biogeographic history in shaping sampling thresholds, ultimately refining the guidelines for richness assessment across ecosystems.

Notably, our use of inflection points in the CV_C to define SSI was primarily based on cost-effectiveness considerations. However, when studies require higher precision in assessing richness patterns or their environmental responses (e.g., demanding CV_C ≤ 0.05), sampling intensity requirements may substantially exceed SSI thresholds. Interestingly, under such stringent accuracy criteria, our results revealed a reversal in sampling demands: trees unexpectedly required the highest sampling intensity, whereas herbs showed the lowest requirements. This counterintuitive pattern may arise from the fundamental differences in the richness of these functional groups along elevational gradients. For trees, achieving high precision likely necessitates exhaustive sampling to capture their full species pool because their larger spatial distributions mean that individuals are more scattered across landscapes (Gimaret-Carpentier et al., 1998). Conversely, for herbs, their smaller stature and clumped distributions allow relatively complete sampling within smaller areas once basic SSI thresholds are met; their fine-scale environmental specialization is captured through spatial replication rather than plot-area expansion (Sabatini et al., 2022).

However, some researchers have suggested that reported elevational richness patterns may partly result from sampling area (size of each plot) and the sampling strategy (e.g., whether multiple peaks were included within the same mountain), which can also influence the determination of the SSI (Costa et al., 2023). Differences in sampling area (e.g., plot size) can bias richness estimates: small plots tend to underestimate richness by missing rare or patchily distributed species, whereas large plots may overestimate richness by encompassing more habitats and microenvironments (Lomolino, 2001). Similarly, reducing or truncating the sampling range may limit the representation of key ecological zones and the full species pool, thereby distorting the true richness patterns (Nogués-Bravo et al., 2008). Because climatic zones along elevation determine energy and water availability, incomplete coverage of these zones can mislead the interpretation of richness-climate relationships (Kluge et al., 2006). Moreover, complex mountain topography—including variation in slope aspect, gradient, and surface roughness—redistributes light, heat, moisture, soil, and nutrients, generating steep environmental gradients and high habitat heterogeneity (Fang et al., 2004; Lozano-García et al., 2016; Dearborn and Danby, 2017; Qin et al., 2019). These topographic and microenvironmental variations drive species turnover and niche diversification, shaping local richness independently of large-scale elevational trends. Consequently, sampling strategies that integrate multiple peaks or slopes may substantially influence perceived richness patterns by merging sites with distinct climatic and ecological conditions.

In our study, sampling intensity was defined solely as the number of sampling sites per kilometer of elevational range along a transect, with plot size and slope aspect held constant to minimize potential biases. Despite these controls, sampling bias may still influence the observed elevational richness patterns and the determination of the SSI. The meta-analysis provides valuable context by comparing the sampling densities at Baima Snow Mountain with those reported in the literature and highlights the often-overlooked need for growth-form-specific sampling designs. However, due to heterogeneity in biome regions, sampling protocols, and design, the SSI identified here should be interpreted primarily as context-dependent guidance rather than universally applicable targets. Addressing these limitations will require multi-region studies, broader and standardized datasets, and targeted experimental or simulation approaches, representing important directions for future research to improve both the accuracy and generality of elevational richness patterns and SSI estimates.

5. Conclusion

Our study suggests that plant growth forms may differ in the sampling intensity required to accurately assess elevational richness patterns and their underlying drivers. Trees tended to require lower sampling intensity, perhaps reflecting their broader environmental tolerances and capacity to integrate habitat conditions over larger spatial scales (Gong et al., 2019; Taylor et al., 2023). In contrast, herbs appeared to require higher intensity, possibly because their fine-scale microhabitat specialization makes their richness patterns more sensitive to microclimatic variation occurring over short distances (Di Biase et al., 2021; Loidi et al., 2021; Zu and Wang, 2022). These findings indicate that the commonly applied "one-size-fits-all" sampling approaches may not fully capture the variability among growth forms. Recognizing such differences could enhance both the accuracy of individual studies and the comparability of elevational richness pattern research. Future methodological developments would benefit from considering these precision-dependent, taxon-specific scaling relationships to better guide richness inventory designs under varying accuracy requirements and resource constraints. Such considerations are particularly important for conservation prioritization, where even small differences in estimated richness patterns may have disproportionate influence.

Acknowledgements

This work was funded by Chinese Academy of Sciences Hundred Talents Program, Category B, National Postdoctoral Programs, Yunnan Provincial General Project Fund (202401CF070060), the Key R & D Program of Yunnan Province (202403AC100028) and National Natural Science Foundation Regional Project (32360395).

CRediT authorship contribution statement

Hongrui Ling: Writing – review & editing, Writing – original draft, Data curation, Formal analysis, Validation, Visualization. Jianqiang Yang: Writing – review & editing, Data curation, Formal analysis, Validation. Yannan He: Writing – review & editing, Investigation, Validation. Pengwan Zhang: Writing – review & editing, Investigation, Validation. Jiangcinongbu: Writing – review & editing, Investigation, Validation. Sinalaoding: Writing – review & editing, Investigation, Validation. Zhenyu Fan: Writing – review & editing, Investigation. Aoxiang Chang: Writing – review & editing, Investigation. Hang Sun: Writing – review & editing, Investigation, Supervision. Shuang Zhang: Writing – review & editing, Investigation. Zihan Jiang: Writing – review & editing, Writing – original draft, Funding acquisition, Conceptualization, Project administration, Supervision.

Declaration of competing interest

The author Zihan Jiang is Editor for Plant Diversity and was not involved in the editorial review or the decision to publish this article.The other 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.2025.12.006.