Environmental change shapes understory plant diversity and dominance in boreal forests

Abstract

Ongoing environmental change threatens ecosystems worldwide, yet little is known about its effect on understory plant diversity, which underpins ecosystem functioning and sustainability. Here, we use Canada’s National Forest Inventory database to evaluate decade-long changes in local plant diversity within understory communities. Species richness of shrubs and bryophytes increases by 8 and 11% per decade, while species evenness of herbs and bryophytes declines by 14 and 8%, respectively. Temporal increases in species richness and declines in species evenness are both associated with rising temperature, nitrogen deposition, water availability, and increased temperature seasonality. Additionally, the proportion of bryophyte biomass increases, whereas that of shrub biomass decreases over time, with the effects of temperature seasonality and water availability on these temporal shifts strongly dependent on overstory basal area. Species richness is positively associated with biomass across shrubs, herbs, and bryophytes, suggesting that changes in diversity alter understory biomass distribution under environmental changes. Contrary to the common view that climate warming uniformly reduces biodiversity, our findings show that understory communities undergo complex and dynamic shifts in plant diversity and composition. We suggest that environmental change-driven shifts in resource availability and heterogeneity may shape understory composition and species dominance, ultimately influencing forest ecosystem function.

Using Canada’s National Forest Inventory, this study shows boreal understory plant communities are shifting, species richness rises while evenness falls. These changes track warming, nitrogen deposition and moisture, and are moderated by canopy cover.

Introduction

Plant diversity is essential for sustaining ecosystem functions and services that support human well-being¹. However, the Earth is undergoing rapid climate change and increasing nitrogen deposition (hereafter ‘environmental changes’), which is expected to profoundly alter plant diversity in forest ecosystems^2–4. While considerable attention has been paid to changes in overstory plant diversity and composition^2–4, understory plants, which account for over 80% of forest taxonomic diversity in boreal and temperate forests^5,6, have been largely overlooked in assessments of biodiversity and ecosystem functions. Although understory plants account for a small fraction of total forest biomass and carbon stocks due to their small size compared to canopy trees, they contribute disproportionately to ecosystem processes by facilitating tree regeneration and mediating water, nutrient, and carbon dynamics^7–10. Moreover, because of their smaller body sizes and much higher metabolic rates, understory plants are likely to experience more pronounced and rapid responses to environmental changes compared with plants in the overstory layer^11,12.

Anthropogenic environmental changes could shape plant diversity in local communities (hereafter ‘local plant diversity’) by altering neighborhood species interactions, such as competition, in both direct and indirect ways, through changes in resource availability and niche partitioning that govern species coexistence¹³ (Fig. 1a). On the one hand, rising temperatures may enhance plant diversity directly by increasing temporal niche partitioning through an extended growing season, and indirectly by gradually alleviating nutrient limitation through microbially mediated mineralization of organic matter that releases multiple elements in a moderate and balanced manner^11,14–16. In addition, the rising temperature could also facilitate plant migration, allowing plants that thrive in warmer climates to move into previously cold-limited areas, increasing the number of species in local communities¹⁵. Moreover, increased seasonal temperature and precipitation variabilities due to climate change might enhance plant diversity by promoting temporal niche segregation, whereby species with distinct growth timings or resource-use patterns are differentially favored across variable conditions, facilitating coexistence and reducing competitive exclusion^17,18. On the other hand, warming and altered precipitation (e.g., drought) may decrease local plant abundance, biomass and richness due to limited water availability for species coexistence^19,20. In addition to changes in temperature and precipitation, altered light availability may influence plant diversity and biomass, especially for understory plants, given that light is typically the most limiting resource for plants in the forest understory layers²¹. Although increased nitrogen deposition is expected to result in plant diversity loss by disproportionately promoting the dominance of some species with better access to light and surpassing subordinate, less competitive species²², this pattern appears less evident in relatively species-poor northern forests, where fertilization experiments have reported increased species richness in response to nutrient addition²³. Despite these understandings, the effects of environmental change on understory diversity and biomass remain poorly understood, as does the interaction between overstory and understory plants in their responses to these environmental changes.

Fig. 1. The influence of environmental change on understory plant diversity and functions.

a A conceptual diagram showing the influence of environmental change on understory plant diversity and biomass; b The distributions of ground plots from the Canadian NFI for different understory layers.

Different layers in the forest plant community interact dynamically, with the effects of environmental change on their diversity and productivity strongly influencing one another^5,24,25. On the one hand, light availability for understory vegetation depends strongly on the composition and structure of the overstory trees, which can shift under climate warming and outcompete understory vegetation for this essential resource^5,24 (Fig. 1a). On the other hand, overstory trees moderate temperature and humidity levels in the understory, creating a more stable microclimate that buffers understory communities against extreme climatic variability, thereby mitigating the impacts of environmental change and potentially slowing the rapid responses expected in the absence of overstory trees²⁵ (Fig. 1a). Moreover, within the understory layers, shrubs, herbs, and bryophytes might interact through both competition and facilitation, influencing their collective responses to ongoing environmental change^24,26 (Fig. 1a). For example, shrubs can reduce the diversity and biomass of herbs and bryophytes by creating a dense canopy that shades the ground layer and limits light availability²⁴. Bryophytes may suppress the regeneration of shrubs and herbs through allelopathic effects and by modifying microclimatic conditions²⁶, while simultaneously facilitating their growth through their associated N₂-fixing bacteria²⁷. Ongoing climate warming is expected to favor vascular plants, such as shrubs and herbs, while suppressing bryophytes, as the former are better adapted to rising temperatures and drought, outcompeting bryophytes, which are reliant on cool and moist conditions²⁸.

Observational studies documenting changes in the local diversity of plant communities following recent environmental changes remain limited. Two recent meta-analyses, which include comprehensive datasets comprising fish, bird, plant and freshwater invertebrate assemblages, report a temporal increase in local species richness with increasing temperature^29,30. In addition, a 12% increase in the Shannon diversity index of the overstory layer has been observed in boreal forests from 2000 to 2020³. In contrast, two other studies find that plant species richness did not vary over time in response to ongoing environmental changes across forest, grassland, shrubland, tundra and wetland ecosystems^31,32. While several recent studies have documented environment-driven shifts in boreal and hemiboreal understory plant communities, including increases in more thermophilic and nitrophilous species^33–37, comprehensive assessments of temporal changes in local plant diversity, biomass, and interactions (e.g., competition and facilitation) across multiple understory layers (shrubs, herbs, and bryophytes) remain limited. This gap is particularly important given the critical functional role of understory vegetation^7,8 and the rapid warming projected for boreal ecosystems, where biological activity is often constrained by cold conditions³⁸.

Here, we assess temporal changes in understory plant diversity and biomass stock and identify potential environmental change drivers of these shifts by using the first (2000–2007) and second (2008–2017) census data from the Canadian National Forest Inventory (NFI) program. Data from 589 plot inventories are analyzed to broadly represent Canadian boreal and hemiboreal forests (Fig. 1b), assessing the direction and magnitude of decadal changes in understory diversity metrics and biomass between two censuses. Specifically, we hypothesize that: (1) given that biological activity and species distributions in high-latitude forest understories are often limited by low temperatures, the diversity and biomass of understory vegetation would increase over time due to elevated temperatures; and (2) the effects of environmental change on understory diversity and biomass would be less pronounced under dense overstory canopies²⁵. To address these hypotheses, we individually analyze the three primary understory strata (i.e., shrubs, herbs, and bryophytes) and investigate how interactions among these layers mediate environmental change effects to offer novel insights into the dynamics of multi-strata plant communities in a changing environment. In our study, relocation error is minimized because the Canadian National Forest Inventory plots are permanently marked and precisely georeferenced according to standard national protocols, ensuring accurate remeasurement at the same location. Our observational approach with repeated measurements provides robust findings, minimizing the potential biases such as species detectability and differences in sampling protocols, which are common problems in meta-analytic and resurvey studies³⁹.

Results

Environmental changes shape understory diversity

On average, across all plots, shrub and bryophyte species richness increased significantly by 0.7 (95% confidence interval, 95% CI, 0.4–1.0) and 0.9 (CI, 0.6–1.3) species per decade, corresponding to per-decade increases of 8 and 11%, respectively (Fig. 2a). In contrast, herb and bryophyte species evenness decreased significantly by 0.08 (95% CI, −0.11 to −0.06) and 0.04 (95% CI, −0.07 to −0.02) per decade, representing per-decade decreases of 14 and 8%, respectively (Fig. 2a). Bryophyte biomass increased significantly by 0.7 Mg ha⁻¹ per decade (95% CI, 0.3–1.1; a 5% increase), while the biomass of shrubs (95% CI, −0.3 to 0.2) and herbs (95% CI, −0.2 to 0.03) did not change between censuses (Fig. 2b). In addition, the proportion of shrub biomass in the total community biomass within the understory declined (95% CI, −0.10 to −0.004), whereas the proportion of bryophyte biomass increased significantly (95% CI, 0.005–0.01) (Fig. 2b). Overstory tree species richness (95% CI, −0.04 to 0.2), evenness (95% CI, −0.004 to 0.03), and basal area (95% CI, −1.3 to 0.7 m² ha⁻¹ dec⁻¹) remained stable between the censuses (Supplementary Fig. 1). Meanwhile, annual temperature, climate moisture index, the seasonality of temperature and precipitation, and cumulative nitrogen deposition showed consistent increases, while solar radiation, an indicator of light availability, declined (Fig. 2c).

Fig. 2. Temporal trends in understory plant diversity, biomass, and potential environmental change drivers.

a Changes (dec⁻¹) in species richness and evenness of shrubs, herbs and bryophytes (n = 295, 554 and 543, respectively). b Changes (dec⁻¹) in biomass and biomass proportion of shrubs, herbs and bryophytes (n = 371, 438 and 301, respectively). c Changes in annual temperature, climate moisture index, temperature and precipitation seasonality, solar radiation, and cumulative nitrogen deposition (n = 589). Values and error bars are bootstrapped means and 95% confidence intervals. Temporal trends were considered statistically significant at α = 0.05 when the 95% confidence intervals (CIs) did not include zero. ΔTem: temporal changes in annual temperature; ΔCMI: temporal changes in annual climate moisture index; ΔTemS: temporal changes in annual temperature seasonality; ΔPrecS: temporal changes in annual precipitation seasonality; ΔCND: temporal changes in cumulative nitrogen deposition; ΔSA: temporal changes in annual solar radiation.

Decadal changes in species richness of shrubs, herbs, and bryophytes were primarily associated with increases in temperature seasonality (i.e., seasonal variations in temperature) (P < 0.001 for shrubs and bryophytes, P = 0.001 for herbs; Fig. 3a, Supplementary Table 1). Change in shrub richness was also significantly positively associated with rising annual temperature and climate moisture index (P = 0.001 and 0.006, respectively), while bryophyte richness was positively linked to increases in precipitation seasonality (i.e., seasonal variations in precipitation) (P = 0.015) (Supplementary Fig. 2a–c, Supplementary Table 1). In contrast, changes in shrub, herb and bryophyte evenness were all significantly negatively correlated with increasing cumulative nitrogen deposition (P < 0.001, = 0.003 and = 0.001 for shrubs, herbs and bryophytes, respectively; Fig. 3b, Supplementary Fig. 2d, Supplementary Table 1). Changes in herb evenness were also negatively correlated with rising temperature, temperature seasonality, and climate moisture index (P < 0.001, = 0.006, = 0.013, respectively) (Fig. 3b, Supplementary Fig. 2e, f, Supplementary Table 1). The relationships between environmental change drivers and changes in understory richness and evenness were largely unaffected by overstory basal area, except for shrub richness (Supplementary Table 1). The positive relationship between changes in shrub richness and annual temperature was more pronounced in forests with low overstory basal area, highlighting the interactive effects of canopy structure and climate warming on shrub richness (Interaction effect: P = 0.012) (Supplementary Fig. 2a, Supplementary Table 1).

Fig. 3. Effects of dominant environmental change drivers on species richness, evenness, biomass, and biomass proportion of understory plants.

a Changes in species richness; b changes in species evenness; c changes in biomass; d changes in biomass proportion. Red line shows the model-predicted mean; gray shaded band shows bootstrap 95% confidence intervals. Solid and dashed lines represent the significant (P ≤ 0.05) and nonsignificant (P > 0.05) relationship. Slope estimates are partial dependencies derived from the most parsimonious model introduced (details in Supplementary Tables 1 and 2). The dominant environmental change driver was selected based on the highest r² value. The significance (P) is presented for each term tested. BP: biomass proportion. ΔTemS: temporal changes in annual temperature seasonality; ΔTem: temporal changes in annual temperature; ΔCMI: temporal changes in annual climate moisture index; ΔCND: temporal changes in cumulative nitrogen deposition; ΔSA: decadal changes in annual solar radiation. We use two-sided t-test to calculate P-values.

Environmental changes shape understory biomass

Changes in shrub biomass significantly decreased with increasing solar radiation (P = 0.034), while changes in herb biomass decreased with both increasing temperature and solar radiation (P = 0.010 and 0.032, respectively) (Fig. 3c, Supplementary Fig. 3a, Supplementary Table 2), despite neither shrub nor herb biomass showing significant differences between the two censuses (Fig. 2b). In terms of biomass proportion within the total understory community, increases in shrub biomass proportion decreased significantly with increasing cumulative nitrogen deposition (P = 0.002), but showed a marginally significant increase in response to rising temperatures (P = 0.069) (Fig. 3d, Supplementary Fig. 3b, Supplementary Table 2). Changes in herb biomass proportion significantly decreased with increasing temperature, climate moisture index and temperature seasonality (P = 0.009, 0.021, and 0.031, respectively) (Fig. 3d, Supplementary Fig. 3c, d, Supplementary Table 2).

The relationships between changes in bryophyte biomass and environmental change drivers varied significantly with overstory basal area (Supplementary Table 2). When overstory basal area was high (≥30 m² ha⁻¹), bryophyte biomass changes were positively associated with climate moisture index, temperature seasonality, and mean temperature, whereas these relationships were negative under overstory basal area was low (≤5 m² ha⁻¹) (Interaction effect: P < 0.001, = 0.006, and = 0.037, respectively) (Fig. 4a–c, Supplementary Table 2). Additionally, changes in shrub biomass proportion were positively correlated with temperature seasonality but negatively with precipitation seasonality under low overstory basal area (Fig. 4d, e, Supplementary Table 2). In contrast, under high basal area, shrub biomass proportion changes were negatively associated with temperature seasonality but positively with precipitation seasonality (interaction effects: P < 0.001 and P = 0.039, respectively) (Fig. 4d, e, Supplementary Table 2). Herb biomass proportion changes were negatively associated with solar radiation under low overstory basal area, while positively correlated under high overstory basal area (Interaction effect: P = 0.008) (Fig. 4f, Supplementary Table 2). Changes in bryophyte biomass proportion were positively associated with temperature seasonality and climate moisture index under high overstory basal area but negatively associated under low overstory basal area (Interaction effect: P = 0.001 and 0.009, respectively) (Fig. 4g, h, Supplementary Table 2).

Fig. 4. Overstory basal area-dependent responses of understory biomass to environmental change drivers.

a–c Changes in bryophyte biomass in response to climate moisture index, temperature seasonality, and mean temperature; d, e changes in shrub biomass proportion in response to temperature seasonality and precipitation seasonality; f changes in herb biomass proportion in response to solar radiation; g, h changes in bryophyte biomass proportion in response to temperature seasonality and climate moisture index. Colored lines represent the overstory basal area-specific responses, with their bootstrapped 95% confidence intervals shaded in the corresponding color. The overstory basal area was used as the index of canopy density here and analyzed as a continuous variable, but it was illustrated based on the three meaningful levels of breakpoints (mean, mean plus and minus standard deviation (SD)). The significance (P) of the interaction effect between overstory basal area and environmental change drivers is presented. BP: biomass proportion. ΔTemS: temporal changes in annual temperature seasonality; ΔTem: temporal changes in annual temperature; ΔCMI: temporal changes in annual climate moisture index; ΔSA: temporal changes in annual solar radiation. We use two-sided t-test to calculate P-values.

Interactions among shrubs, herbs and bryophytes

After accounting for environmental change drivers, stand age, overstory basal area, and historic climate conditions, changes in species richness among the three understory layers were significantly positively correlated (Fig. 5a). Similarly, changes in herb species evenness were significantly positively associated with changes in the evenness of shrubs and bryophytes, while changes in bryophyte biomass were significantly positively correlated with changes in the biomass of shrubs and herbs (Fig. 5b, c). In contrast, changes in biomass proportions among the three understory layers were significantly negatively correlated between each of their pairs (Fig. 5d). However, as the biomass proportions of the three understory layers are interdependent, these correlations should be interpreted as descriptive patterns rather than causal relationships.

Fig. 5. Interrelationships among shrubs, herbs, and bryophytes in species richness, evenness, biomass, and biomass proportions, accounting for environmental change drivers, historical climate conditions, overstory basal area, and stand age.

a Changes in species richness; b changes in species evenness; c changes in biomass; d changes in biomass proportion. Red line shows the model-predicted mean; gray shaded band shows bootstrap 95% confidence intervals. Slope estimates reflect partial dependencies, and significance (P-values) are provided for each tested term. BP: biomass proportion. We use two-sided t-test to calculate P-values.

Changes in shrub species richness and evenness were not regulated by historic climate conditions, while changes in shrub biomass shifted from positive to negative with increasing historic mean annual temperature and climate moisture index (Supplementary Fig. 4a, b). Changes in herb species richness were significantly negatively associated with mean annual climate moisture index (Supplementary Fig. 4c). Changes in both herb and bryophyte species evenness were significantly positively correlated with mean annual temperature (Supplementary Fig. 4d, f), while changes in biomass for both herbs and bryophytes were negatively linked with mean annual climate moisture index (Supplementary Fig. 4e, g). Moreover, changes in diversity indices and biomass of understory vegetation were largely consistent across dominant overstory tree types (coniferous, broadleaved, and mixedwood forests), except for shrub evenness and bryophyte biomass, which increased with an increasing proportion of coniferous trees in the overstory (Supplementary Fig. 5).

Discussion

Our study provides large-scale observational evidence that environmental changes are associated with shifts in understory vegetation diversity and biomass in boreal and hemiboreal forests, while also revealing critical interactions among the three primary understory layers on a continental scale. Specifically, our analyses unveiled temporal increases in species richness of shrubs and bryophytes, with concurrent declines in species evenness for herbs and bryophytes over the studied period. Our findings challenge the common view of uniform biodiversity loss under climate warming^40,41, suggesting that trends in local understory plant diversity may not necessarily mirror global patterns of species extinction. This highlights the importance of considering scale-dependent and context-specific responses of biodiversity to environmental changes⁴². Furthermore, the divergent responses of diversity metrics, such as increasing species richness and decreasing evenness, highlight that metrics based on presence/absence and those reflecting composition and dominance respond to environmental change in divergent ways, contrasting with the parallel responses observed in grasslands¹³. Moreover, our results reveal strong positive relationships between species richness and biomass within each understory layer (shrubs, herbs, and bryophytes) (Supplementary Fig. 6a), indicating that synergistic interactions, such as resource sharing and microhabitat creation, may play an important role in shaping plant diversity and biomass under changing environmental conditions.

Increases in species richness for shrubs and bryophytes over time support our first hypothesis, while the increase observed for herbs was not statistically significant (P = 0.059). These increases in understory plant species richness are likely attributable to elevated temperature and water availability, particularly temperature and precipitation seasonality, as suggested by the observed relationships between changes in richness and environmental change drivers. Seasonal variations in temperature and precipitation increase temporal resource heterogeneity, expanding niches and reducing competition, thereby facilitating species coexistence^5,18,43. Temperature often serves as a surrogate for thermal energy and influences the number of species by facilitating a wider range of energetic lifestyles in warmer conditions¹¹. Rising temperatures could prompt species to progressively migrate northward, leading to the establishment of more thermophilic understory plant species in cold, high-latitude forests; however, the pace of such migration is likely to be slow and was likely not a major factor in the changes observed in our study^15,25. Moreover, the warming effect on changes in shrub richness was stronger in sites with less dense canopy cover, consistent with our second hypothesis. Although dense canopy cover might cause light limitation, our results highlight the buffering role of overstory trees in mediating the impacts of environmental change on understory composition²⁵. Like temperature, increased water availability creates additional ecological niches for species with diverse life history strategies to fulfill their water requirements, e.g., drought-intolerant species²⁰. Our findings suggest that climate warming and increasing climate variability may enhance understory species richness at local scales by promoting resource availability and environmental heterogeneity in cold, high-latitude forests. This contrasts with plant diversity declines in warmer, drier regions, where warming intensifies water stress and environmental filtering^16,20, suggesting that the same environmental change driver can have opposite effects, depending on whether a region’s temperature or moisture is the limiting factor^19,44. In colder regions, such as in boreal and hemiboreal forests³, warming alleviates the temperature limitation. However, increased temperature exacerbates moisture limitations in drier regions¹⁶.

In contrast with our hypothesis, we found that the evenness of the herb and bryophyte species decreased over time due to elevated temperature, water availability, temperature seasonality, and especially cumulative nitrogen deposition. Increased nitrogen availability, temperature and water availability could create more favorable conditions for some species (e.g., thermophilic and nitrophilous species) than others, potentially decreasing evenness if certain species become dominant due to their better adaptation to new conditions^36,37,45,46. Elevated temperature and seasonality would also affect plant phenology, such as the timing of flowering and leaf-out, allowing certain plants a competitive advantage through better adaptation to the new condition, consequently influencing plant species evenness⁴⁷. Despite the increase in local species richness during the study, enhanced species dominance may gradually exclude drought-tolerant species, alter habitats, and monopolize resources, ultimately reducing species richness and impairing ecosystem functioning in the future^48–50.

We observed no significant temporal changes in aboveground biomass for shrubs and herbs, but bryophyte biomass increased over time. However, analysis of the distribution of biomass among the three understory layers revealed a decline in shrub biomass proportion and an increase in bryophyte biomass proportion, suggesting that environmental change enhanced bryophyte dominance, contrary to previous experimental expectations^27,28. Importantly, overstory basal area strongly influenced the effects of environmental change on shrub and bryophyte biomass or their biomass proportions. In contrast with our second hypothesis, warming, increased moisture availability, and heightened temperature seasonality increase bryophyte biomass and its proportion in sites with denser overstory, where reduced evapotranspiration ensures stable moisture availability, and lower incoming solar radiation creates conditions that favor bryophytes due to their small size, tolerance to desiccation, and ability to thrive in shaded environments, offering them a competitive advantage over shrubs⁵¹. Increases in bryophyte biomass and the thickness of the bryophyte layer could further physically inhibit the establishment of shrubs and compete with shrubs for resources^26,52. However, shrub proportion increased more in response to heightened temperature seasonality and reduced precipitation seasonality in sites with sparse overstory (such as ecotones toward the tundra), where reduced shading provides higher light availability, which is critical for shrubs and allows them to benefit more from temperature fluctuations compared to bryophytes^24,43,46,52. In contrast, increased precipitation seasonality might increase drought risk by creating longer or less predictable dry periods, which can reduce shrub biomass proportion, particularly in stands with low overstory cover where the absence of canopy buffering exposes the understory to greater moisture variability^25,53. Since the sensitivity of understory vegetation to environmental change varies with the density of overstory trees, forest management strategies could be tailored to mitigate environmental change-driven shifts in understory composition and dominance.

Furthermore, positive correlations in temporal changes of plant diversity and biomass among the three understory layers indicate that shrubs, herbs, and bryophytes are closely associated and interdependent^24,51,52. Interactions among shrubs, herbs, and bryophytes are not limited to competition but are significantly shaped by facilitation, where diverse shrub plants can create a variety of ecological niches for herbs and bryophytes^24,51,52,54, while increased bryophyte production can enhance soil nutrient availability, benefiting shrubs and herbs^7,27. The close associations and interdependencies among shrubs, herbs, and bryophytes mean that the effect of changes in one vegetation layer (e.g., reduced bryophyte growth due to drought) can cascade through the system, potentially reducing biomass stability and ecosystem functions across the different layers of the understory community. The positive interactions among these functional groups may be partly driven by severe light limitation under dense canopies, which promotes mutual adaptations that facilitate coexistence²⁴. In addition, simple bivariate plots showed that the biomass of shrubs, herbs and bryophytes was positively correlated with species richness (Supplementary Fig. 6a) but did not change with species evenness (Supplementary Fig. 6b), which is in line with previous studies⁵⁵. These findings collectively emphasize the importance of considering these interactions to sustain biodiversity and associated ecosystem functions across all understory components and enhance resistance to changing climatic conditions.

The temporal changes in species richness and evenness of shrubs and bryophytes were generally held regardless of the local historical climate. However, temporal increases in herb richness were more pronounced in drier sites, while decreases in herb evenness were more prominent in colder sites experiencing a greater extent of climate warming. This pattern suggests that in particularly harsh environments, such as those characterized by cold or drought, changes in temperature and precipitation may enable certain species, typically limited by abiotic factors, to become dominant across a broader range of locations^46,56. Moreover, temporal increases in biomass across all three understory layers were more pronounced in drier sites, suggesting that increased water availability has a greater effect in alleviating water stress and enhancing growth and diversity where water is more limiting¹⁹. Temporal increases in bryophyte biomass proportion were more pronounced in drought-affected sites, suggesting that changes in water availability, rather than temperature, may play the dominant role in shaping competition between shrubs and bryophytes^56,57. The temporal changes in species richness and evenness were consistent across different overstory types, except for shrub evenness. While the mean shrub evenness did not change over time, we observed a temporal increase in shrub evenness, especially in coniferous forests. This increase may be attributed to a shift toward deciduous broadleaved trees within these forests, driven by the faster germination rates and stronger competitive abilities for sunlight of fast-growing, early-successional broadleaved species, thus facilitating their invasion into boreal regions⁴. The increase of broadleaved species seedlings in coniferous forests would decrease the dominance of coniferous seedlings, consequently increasing species evenness.

Our models explained only a modest proportion of the variation in understory diversity and biomass, suggesting that other ecological drivers not captured by our dataset likely contributed to the observed patterns. This points to the likely influence of additional factors such as microclimatic conditions, species interactions, or other environmental drivers not included in our models. Therefore, although the reported associations are statistically significant, they should not be interpreted as evidence of direct causality. In addition, the absence of species-level and trait-based analyses limits our ability to determine which taxa and functional strategies contribute to the observed changes in richness and evenness. We initially attempted to include species-level data, but a substantial proportion of unresolved species codes introduced uncertainty. Moreover, available trait data (e.g., leaf area, height, drought tolerance) covered only a small subset of understory species. Future studies incorporating validated species identity with species functional traits will be essential for advancing mechanistic understanding of how understory plant communities respond to environmental change.

In conclusion, our findings from repeated measurements in boreal and hemiboreal forests show that warming, increased moisture availability, greater cumulative nitrogen deposition and higher climatic seasonality might simultaneously increase understory species richness and species dominance (reduced species evenness), suggesting that in these systems, increasing the abundance of dominant species does not lead to high competitive exclusion of other species. Since species evenness significantly influences ecosystem functions, such as soil carbon accumulation^48,49, its decline in response to global warming and increased precipitation could become a primary driver of diminished forest ecosystem functioning and services in the future. Our results also highlight how environmental change-driven shifts in resource availability and heterogeneity may promote greater understory richness while interactions among overstory trees, shrubs, herbs and bryophytes, including both competition and facilitation, shape local plant communities. The rapid changes in species richness, evenness and biomass across understory layers indicate the profound and immediate impacts of environmental change on forest ecosystems, with significant implications for ecosystem functioning, human well-being, and future climate dynamics.

Methods

Study area and available data

We used plot-level data from the Canadian National Forest Inventory (NFI) database (https://nfi.nfis.org) to determine how understory diversity and biomass change over time⁵⁸. The NFI database encompasses a network of permanent ground plots covering much of Canada’s forests across boreal and hemiboreal biomes, which were established by Canadian provincial authorities between 2000 and 2008 (first measurement) and subsequently remeasured between 2008 and 2017 (second measurement) following the same standard ground sampling guidelines⁵⁹. For inclusion in our analysis, we selected only ground plots located in forest stands that had not experienced harvesting, thinning, fertilization, or other forms of human management at or between the two sampling times. All selected plots had two measurements and complete data on forest canopy composition, understory composition, and biomass. In total, 589 plots (43°30’–68°00’ N, 53°24’–134°18’ W) met these criteria (Fig. 1b). Each of these represents a unique, spatially independent plot from the Canadian National Forest Inventory. Most of these plots were established in natural disturbance-driven ecosystems (427 plots), while others were harvested (62 plots) or stands of unknown origin (100 plots).

The ground plots comprised several sub-plots associated with the vegetation layers. Within each ground plot, a Large Tree Plot was established with a radius of 11.28 m and an area of 400 m² (0.04 ha). All canopy trees (tree stems ≥ 9.0 cm in diameter at breast height) in the Large Tree Plot were systematically numbered, tagged, identified for species, and measured for height and DBH. Within the Large Tree Plot, an Ecological Plot was set up with a radius of 5.64 m to measure plant species presence, composition, and cover for shrubs, herbs and bryophytes. Four 1 m² Microplots were established outside of the Large Tree Plot (but within a 15 m radius) to determine the biomass of shrubs, herbs and bryophytes. To avoid repeated destructive sampling, microplot locations were shifted within approximately 2 meters during remeasurements (Supplementary Fig. 7). While this introduces a small degree of spatial variability, it represents the best practice for assessing temporal changes in understory biomass across large-scale forest monitoring networks. Biomass samples from the shrub, herb, and bryophyte layers were collected separately and then subjected to oven-drying and subsequent weighing. A plant is considered in or out of the Microplots depending on the germination point where the plant enters the soil. In cases where a plot splits a large grass clump, making germination points unclear, the portion within the plot boundaries was clipped. Shrubs are woody perennial plants, including trees and shrubs, with a total plant height of less than 2.0 m. The herb layer included all herbaceous species, such as forbs (including ferns and fern allies), grasses, sedges, and rushes, while the bryophyte layer included mosses, liverworts, hornworts, and noncrustose lichens. Additionally, slime molds and mushrooms were incorporated into the bryophyte samples. Plant samples were cut into smaller pieces and bagged. All bagged samples were oven-dried in a forced-air drying oven at 70 °C for 72 h and weighed to the nearest 0.1 g.

After excluding missing values for each vegetation layer (i.e., where a vegetation layer was not surveyed), 471, 583 and 579 plots were included for the shrub, herb and bryophyte layers, respectively, in the statistical analysis⁶⁰. Plots where a layer was assessed but no species were recorded were retained and treated as true zeros. The temporal changes in canopy tree basal area (m² ha⁻¹ decade⁻¹), understory diversity metrics such as species richness and species evenness (decade⁻¹), and understory biomass (Mg ha⁻¹ decade⁻¹) were calculated as the difference in diversity metrics and biomass between two consecutive censuses divided by the census length in decades. The interval between the two measurements varied across plots, ranging from 5.0 to 17.9 years (mean = 10.6 years; SD = 2.9, 349 plots more than ten years), with 349 plots remeasured after more than 10 years, and was used to calculate plot-specific, per-decade changes in environmental variables (Supplementary Table 3, Supplementary Fig. 8).

Local environmental change drivers, nitrogen deposition rate and stand age

The temporal changes in annual average temperature (ΔTem, °C decade⁻¹), climate moisture index (ΔCMI, cm decade⁻¹), temperature seasonality (ΔTemS, decade⁻¹), precipitation seasonality (ΔPrecS, decade⁻¹), cumulative nitrogen deposition rate (ΔCND, g m⁻² decade⁻¹), and solar radiation (ΔSA, MJ m⁻² decade⁻¹) were calculated as the difference between values in the actual census years (i.e., the years each plot was surveyed in the first and second census), divided by the plot-specific interval in years and scaled to per-decade values. The climate moisture index was calculated as mean annual precipitation minus potential evapotranspiration⁶¹. Temperature and precipitation seasonality were estimated as the standard deviation of monthly mean temperatures and precipitation, respectively, expressed as a percentage of the mean annual temperatures and precipitation. Utilizing degrees Kelvin for temperature in this calculation prevented the potential issue of dividing by zero. All temperature, solar radiation, mean precipitation and potential evapotranspiration data were acquired by BioSIM (https://cfs.nrcan.gc.ca/projects/133), which generates scale-free climate data from geographic coordinates (latitude, longitude and elevation)⁶². Annual N deposition rate data were extracted at 0.5° gridded from the Inter-Sectoral Impact Model Intercomparison Project (ISIMIP) database (https://www.isimip.org)^63,64. Cumulative nitrogen deposition was calculated as the sum of annual values from 2000 to the year of field sampling⁶⁵. Meanwhile, annual temperature, climate moisture index, cumulative nitrogen deposition, and temperature and precipitation seasonality consistently increased (Fig. 2c). In contrast, solar radiation declined (Fig. 2c). Because field-measured climate data are unavailable at the national scale, we used model-derived datasets from BioSIM and ISIMIP, which are based on interpolated meteorological station data. While these data are not free from uncertainty and may not capture fine-scale microclimatic variability, they provide the most consistent and spatially comprehensive estimates currently available for assessing long-term environmental trends across large geographic extents. Given the national-scale scope of our study and the standardized NFI plot network, these datasets represent the most suitable climate inputs for evaluating temporal changes in vegetation.

To illustrate the robustness of observed environmental change trends and assess whether extreme values influenced them, we additionally calculated standardized anomalies (deviations from the 2000–2018 mean) for annual temperature, climate moisture index, temperature and precipitation seasonality, nitrogen deposition, and solar radiation. While these anomalies are derived from the same underlying data sources as our main environmental change metrics, they offer a complementary way to visualize long-term patterns and confirm that the reported trends are not driven by a few extreme years. Consistent results between direct change estimates and standardized anomalies support the reliability of the underlying environmental trends(Supplementary Fig. 9).

Local historical climate, diversity metrics and stand age

To represent spatial variation in climate, we calculated the mean annual climate moisture index (MACMI, cm) and mean annual temperature (MAT, °C) using the BioSIM software (2000–2018)⁶² (Supplementary Fig. 10). The stand age for each plot was determined according to the last stand-replacing fire date or by coring three dominant/co-dominant trees of each tree species. We used species richness and evenness to measure understory species diversity. Species richness (R) was calculated as the number of all live understory species within each plot. Understory species evenness was calculated using Pielou’s evenness index (J′): $J^{\prime }=H/\ln \left(R\right)$ and weighted by the percent cover of constituent species within each plot⁶⁶. Here, H is the Shannon index, calculated as $H=-{\sum }_{i=1}^R{p}_i\times \ln \left(p_i\right)$, where p represents the i-th species’ percent cover in a plot and i is the number of species, R is the total number of species observed, and ln(·) is the natural logarithm function⁶⁷. Stand age was represented by the middle stand age between the first and second NFI measurements.

Statistical analyses

The species richness, species evenness, biomass, and biomass proportion of shrubs, herbs, and bryophytes were analyzed separately as response variables. For each plot, we calculated ΔRichness, ΔEvenness, ΔBiomass, and ΔBiomass Proportion as the changes in species richness, species evenness, biomass, and biomass proportion between two consecutive censuses, divided by the census interval in decades. To examine the effects of environmental change drivers on understory species richness, evenness, biomass, and biomass proportion and to evaluate whether overstory density (quantified as overstory basal area, OBA) mediates these effects, we applied the following linear model:

$$\mathrm{\Delta }\mathrm{Richness},\mathrm{\Delta }\mathrm{Evenness},\mathrm{\Delta }\mathrm{Biomass}\mathrm{or}\mathrm{\Delta }\mathrm{Biomass}\mathrm{proportion}={\beta }_0+{\beta }_1\cdot \mathrm{\Delta }\mathrm{Tem}+{\beta }_2\cdot \mathrm{\Delta }\mathrm{CMI}+{\beta }_3\cdot \mathrm{\Delta }\mathrm{TemS}+{\beta }_4\cdot \mathrm{\Delta }\mathrm{PrecS}+{\beta }_5\cdot \mathrm{\Delta }\mathrm{CND}+{\beta }_6\cdot \mathrm{\Delta }\mathrm{SA}+{\beta }_7\cdot \mathrm{OBA}+{\beta }_8\cdot \mathrm{\Delta }\mathrm{Tem}\times \mathrm{OBA}+{\beta }_9\cdot \mathrm{\Delta }\mathrm{CMI}\times \mathrm{OBA}+{\beta }_{10}\cdot \mathrm{\Delta }\mathrm{TemS}\times \mathrm{OBA}+{\beta }_{11}\cdot \mathrm{\Delta }\mathrm{PrecS}\times \mathrm{OBA}+{\beta }_{12}\cdot \mathrm{\Delta }\mathrm{CND}\times \mathrm{OBA}+{\beta }_{13}\cdot \mathrm{\Delta }\mathrm{SA}\times \mathrm{OBA}+{\beta }_{14}\cdot \mathrm{Stand}\mathrm{age}+{\beta }_{15}\cdot \mathrm{MAT}+{\beta }_{16}\cdot \mathrm{MACMI}+\mathrm{\pi }50\mathrm{km}+\mathrm{\epsilon }$$ 1

where β_i and ɛ are coefficients and sampling error, respectively. We used basal area as a proxy for overstory density and understory light availability because it is consistently available across the network and has been shown to predict understory conditions in large-scale studies^5,68,69. Although shading differs among tree species, implementing species-weighted basal area or basal area × species interactions is not feasible at the national scale, given more than 150 tree species. We acknowledge that basal area does not capture all species-level variation in canopy architecture and light penetration, which remains a limitation and an avenue for future refinement.

We evaluated the validity of the assumptions of the normality of residuals, homoscedasticity, and linearity using diagnostic plots. This analysis assumes normally distributed data, while our data are left-skewed. We thus bootstrapped the fitted coefficients by 1000 iterations⁷⁰ and used bootstrapped 95% confidence intervals for all figures. We compared the bootstrapped estimates with those of the linear models and found that both methods yielded qualitatively similar trends. Collinearity among explanatory variables was tested by variance inflation factors (VIFs). We found that all predictors had VIF < 5, which suggests that multicollinearity was unlikely to be a concern in the most parsimonious models⁷¹ (Supplementary Tables 1, 2). Following previous studies^72,73, we tested for spatial autocorrelation of residuals using Moran’s I test. Significant spatial autocorrelation was detected in most cases for changes in diversity or biomass across the three understory layers. To address this, we calculated the spatial distance at which the spatial effect became nonsignificant, determining that a distance of 50 km was appropriate^72,73. This distance was used for group plots, with each group (π_50km) included as a random factor in the models. All continuous predictors were standardized by centering on their mean values and scaling by their standard deviations to ensure comparability across variables and improve model interpretability. When scaled in this manner, β₀ is the overall mean ∆Diversity or ∆Biomass at the mean ΔTem, ΔCMI, ΔTemS, ΔPrecS, ΔCND, ΔSA, OBA, Stand age, MAT and MACMI. Historical climate conditions (MAT and MACMI) and stand age were used as covariates as they significantly affect understory diversity and biomass^5,24,74. Plot-level soil nutrient data were unavailable; by analyzing within-plot temporal differences, we reduced the influence of fixed site properties (including inherent soil fertility), but we note that unmeasured plot-scale soil conditions may still influence the estimate.

To prevent overfitting⁷⁵, we selected the most parsimonious model based on the lowest AIC among all alternatives using the ‘dredge’ function in the muMIn package⁷⁶ (Supplementary Table 4). To test the sensitivity of our findings to extreme values, we excluded observations falling beyond ±3 standard deviations from the mean for each environmental change driver. Refitting the model with this truncated dataset yielded consistent main and interaction effects in both direction and significance (Supplementary Tables 5, 6), confirming the robustness of our conclusions.

To examine interactions among the three understory layers under ongoing environmental changes, we included changes in richness, evenness, biomass, and biomass proportion of lower layers (e.g., bryophytes and herbs) as predictors in the most parsimonious models for upper layers (e.g., herbs and shrubs). In addition, to assess whether the temporal changes in understory vegetation diversity and biomass are context-dependent, we analyzed the temporal differences in these parameters relative to MAT and CMI, employing the most parsimonious models for visualization. In addition to historical climate factors, we also tested whether the temporal changes in understory vegetation diversity and biomass differed among forest types (broadleaved, mixedwood and coniferous). Similar to Chen and Luo⁷⁷, forest types were classified based on stand composition: stands dominated by a single leaf type (broadleaved or coniferous) with ≥80% basal area of one leaf-type species and mixedwood stands without any leaf-type species exceeding 80% basal area. To understand how local understory diversity contributes to vegetation biomass and ecosystem functioning, we examined the bivariate relationships between understory vegetation biomass and different diversity metrics.

We used partial regressions to graphically illustrate the effects of overstory basal area on relationships between environmental change drivers and changes in understory diversity and biomass. We calculated overstory basal area-dependent environmental change drivers effects as β₀ + β₆·OBA + β₇·ΔTem (or ΔCMI, ΔTemS, ΔPrecS, ΔCND, ΔSA) × OBA for the mean, and mean plus and minus one standard deviation of overstory basal area, respectively. All statistical analyses were performed in R 4.3.3⁷⁸.

Inclusion and ethics

For this research, local researchers were included throughout the research process, including study design, implementation, data ownership, and authorship. Contributors who do not meet all criteria for authorship have been listed in the “Acknowledgements” section. All roles and responsibilities were agreed upon amongst collaborators ahead of the research. In the citations, we have considered local and regional research relevant to our study. This study does not involve human research participants or animals and does not require approval by a local ethics review committee.

Author contributions

Xinli Chen, P.B.R., and S.X.C. were responsible for the conception and design of the project. Xinli Chen and A.R.T. compiled data. Xinli Chen analyzed the data and wrote the first draft of the manuscript. Xinli Chen, P.B.R., Xin Chen, M.H., A.R.T., D.L., and S.X.C. contributed to reviewing and editing. S.X.C. supervised the work and acquired funding. All authors approved the final manuscript.

Peer review

Peer review information

Nature Communications thanks Per-Ola Hedwall and the other anonymous reviewer(s) for their contribution to the peer review of this work. A peer review file is available.

Data availability

All data and the full analysis code required to reproduce every figure and table in the main text and Supplementary Information are deposited at Figshare (10.6084/m9.figshare.28348283)⁶⁰.

Code availability

The code used in this study is available at Figshare (10.6084/m9.figshare.28348283).

Competing interests

The authors declare no competing interests.

Supplementary information

The online version contains supplementary material available at 10.1038/s41467-025-65633-y.