General patterns of trait variation with increasing shrub density
This study provides the first trait-based approach of the effects of shrub encroachment in subalpine grasslands, testing the relevance of the plant economics spectrum (Reich 2014) to ecosystems comprising significant proportions of these woody species. Our results support general hypotheses on the effects of shrubs on community-level plant strategies and their implications for carbon and nutrient cycling (Myers-Smith et al. 2019; Wookey et al. 2009), thereby complementing evidence from arctic and sub-arctic (Crofts et al. 2018; Kemppinen et al. 2021), arid and semi-arid (Ambrosino et al. 2023; Guo et al. 2022b), subhumid (Vogel et al. 2022), alpine (Collins et al. 2020), and mediterranean (Garnier et al. 2004; Nunes et al. 2019) ecosystems.
The PCA (Fig. 3) showed that most of the variation in plant and soil parameters is explained by shrub biomass (Dim 1), and associated with a gradient of plant resource economics. Grassland plots were located at the more acquisitive side of the spectrum with softer tissues and higher nutrient concentrations (lower CWM DMC, lower CWM CN). Communities became more conservative along succession towards shrubland, with lower N concentration and greater investment in tougher tissues (higher CWM CN and CWM DMC). DMC and C concentration were higher in shrub stems compared with herbaceous leaves, which on the contrary had higher N concentration. We acknowledge that as biomass fractions of plant organs were used to calculate CWM of traits, correlation of plant traits with increasing shrub density were expected, especially for CWM DMC and CWM CN. These results are consistent with findings for shrub-encroached subalpine grasslands in the Pyrenees, where C concentration in shrub leaves was higher than in herbaceous leaves, and negatively correlated to N concentration (Urbina et al. 2020). They also match patterns of secondary succession elsewhere including Mediterranean shrublands (Navas et al. 2010). Contrary to the Pyrenees, we however did not find a high contribution of aboveground P to Dim 1, which suggests that although plant P concentrations might decrease with shrub encroachment (Urbina et al. 2020) and along some successional gradients (Tsujii et al. 2023), it may not be a core trait of the resource economics spectrum in our communities.
Soil properties showed a less strong association with the shrub density gradient than aboveground traits. The second axis of variation in the PCA (Dim 2) reflected shrub identity effects on CWM P, available nitrate and pH. Despite the fact that aboveground traits analysed here may be less strongly associated to soil parameters than belowground traits, which interact more directly with soil microenvironment (Spitzer et al. 2023), multiple studies have shown the effect of aboveground traits on belowground processes (Kahl et al. 2017; Sokol et al. 2022). Nevertheless, our analyses for the Stubai site showed that Specific Root Length and Root Tissue Density decreased with shrub encroachment (Tello-García et al. forthcoming). Patterns of association between above- and belowground traits as part of the plant economics spectrum are not consistent or fully understood, but overall coordination has been confirmed in floras relevant to our sites (de la Riva et al. 2016; Freschet et al. 2010; Pérez-Ramos et al. 2012), suggesting that aboveground plant traits are relevant indicators for quantifying ecosystem-level functional responses during shrub encroachment processes.
Linear changes of plant traits and soil properties in shrub encroached subalpine grasslands
Contrary to our expectations, we found no indication of nonlinear trends in community-level plant traits or soil processes with increasing shrub density. Instead, we observed linear responses of plant community traits to increasing shrub biomass. All our linear models for plant traits were significant with a high R2, with the exception of P concentration which, while still significant, showed a higher variance. These results suggest the functional response of plant communities to increasing shrub biomass in subalpine grasslands in the Alps might be gradual along gradients of encroachment, at least for low shrubs such as those considered here. Although soil properties showed a higher variance, we did not detect abrupt changes along the density gradient either. Soil C, C:N and SOM increased significantly and linearly with increasing shrub biomass. However, we measured concentrations of total soil SOM and C soil, as opposed to specific carbon fractions or phenolic compounds. While some studies have found nonlinear changes of soil organic C and lignin in shrub encroached temperate grasslands (Zhou et al. 2017), they included a narrower range of density (0 to 30%) and different functional groups (i.e. leguminous shrub C. microphylla) compared with our study. Instead, our results are consistent with linear increases of soil C found in subalpine grasslands with increasing cover of R. ferrugineum in the Eastern Italian Alps (Pornaro et al. 2017).
Nevertheless, our analysis across the two sites’ plots with deciduous Vaccinium species showed a qualitative shift in fungal communities taxonomic and functional composition at around 35% shrub cover (Mouhamadou et al. forthcoming), suggesting abrupt changes in soil microbial communities. Likewise, nonlinear decreases in the fungi:bacteria and Gram-:Gram + bacteria ratios were observed at levels as low as 10% shrub cover in inner Mongolia (Zhou et al. 2017). In contrast, no change in the fungi:bacteria ratio was detected in response to shrub removal in an Austrian subalpine grasslands (Broadbent et al. 2021). Lastly, we did not detect nonlinear relationships between soil properties and plant community traits, which suggests that qualitative shifts in microbial communities along a gradient of aboveground plant community traits might not translate to changed total soil carbon or nutrient availability.
Cascading effects of shrub encroachment on plant traits and soil properties
Our SEM supports our initial hypothesis that shrub encroachment has cascading effects on soil properties via plant traits (Fig. 6b). We found that the succession toward shrubland leads to plant communities with tougher and nutrient-poorer tissues (Fig. 7). These in turn resulted in increased SOM and soil CN, and decreased pH and P availability (Fig. 7), confirming our hypothesis that shrub encroachment has indirect effects on soil properties. Contrary to our hypothesis, these cascading effects did however not apply to soil N availability.
Plant trait responses to shrub encroachment
Our results support previous evidence of shrub encroachment in subalpine grasslands in the European mountains resulting in plant communities with tougher tissues and lower nutrient content (Grau et al. 2019; Seeber et al. 2022; Urbina et al. 2020). While some previous studies have used CN as a proxy for tissue quality, we found that CWM DMC as a weighted sum of LDMC and StDMC was also a good indicator. CWM DMC was highly and positively correlated not only with CN in the PCA, but also with CWM Lignin:N. Indeed, in a test of alternative models incorporating CWM Lignin:N, we found that these had a lower level of support compared with models including CWM DMC as an indicator of tissue quality (Fig S5).
This suggests that aboveground CWM DMC, an easily measurable trait, can be used as an indicator of the amount of recalcitrant compounds (especially lignin) in shrub encroached grassland communities, and thus of decomposability (Bumb et al. 2018; Freschet et al. 2012; Pietsch et al. 2014) and integration into soil organic matter (Cheng et al. 2023). Our findings add to previous research that focused on leaf traits only (Grau et al. 2019; Seeber et al. 2022; Urbina et al. 2020) by demonstrating the relevance of integrating stem traits (Kobayashi et al., 2024; Zuo et al., 2018) with approaches that consider biomass allocation to leaves vs. stems in woody plants (Guo et al., 2022a). Lastly, these findings confirm that considering allocation is important in community trait-based studies (Garnier et al. 2004) in which not only species dominance but also structural components are relevant to understand ecosystem functioning.
We also found that increasing shrub biomass led to plant communities with lower N and P concentration. Likewise, Urbina et al. (2020) found that plant N and P concentrations decreased along succession to shrubland. However, they showed that these were combined with an increase in their aboveground stocks, given that high shrub biomass per se overrides the dilution of nutrients in their aboveground biomass. Accordingly, despite the decrease of nutrient concentrations in the aboveground biomass, we also found and increase of total aboveground stocks of N and P with increasing shrub biomass (Fig S6 c-f). This increase in stocks was mainly driven by the stem biomass of shrub species, which emphasises once again the importance of biomass allocation to understand nutrient cycling and storage in these ecosystems.
Cascading effects on soil properties
We found that the increase of woody material, which has high CN, high DMC and high lignin (and lignin:N), led to increased SOM and soil CN and to decreased pH. These results are consistent with well-known trait effects on decomposition (Cornwell et al. 2008; Rosenfield et al. 2020, Zuo et al. 2018) and hence on soil carbon accumulation (De Deyn et al. 2008; Huys et al. 2022; Sokol et al. 2022). Recalcitrance of woody tissues results from their high fibre density (Bani et al. 2018) and their high concentration in polyphenols (i.e. tannins, lignin) (Kahl et al. 2017), compared to leaves. Accordingly, leaf and wood litter traits (i.e. lignin) have been shown to be independent within woody tree species, resulting in different effects on decomposition (Kobayashi et al. 2024). Lignin can impact decomposition and nutrient cycling in multiple ways. First, it can have direct effects on decomposer communities such as saprotrophic fungi or soil macrofauna (Sanghaw et al. 2023; Yang et al., 2024a; Zuo et al. 2014). Second, lignin can interact with soil minerals and be protected from microbial access, therefore influencing decomposition (Liao et al. 2022). Our model also showed decreasing soil pH with shrub encroachment, which was also reported in other studies (Collins et al. 2020; Eldridge et al. 2011), and might be linked to the leaching of organic acids into de soil (Swift et al. 1979; Whitford 1992).
Shrub encroachment decreased soil P availability as a result of the lower P concentration in plant communities, and despite the increase in total aboveground P stocks and soil P concentrations (Fig S6 c-d). These results highlight the importance of quality, and not only quantity, for nutrient release in the case of P, thus validating observed effects of plant functional traits on nutrient cycling in shrub encroached grasslands (Yang et al. 2022; Zhang et al., 2008a). Moreover, P cycling depends on other mechanisms such as fungal communities and their ability transform organic P into forms of P available for plants (Ferrol et al. 2019; L. Wang et al. 2023). These mechanisms can be affected by soil pH. It is well known that low soil pH suppresses microbial activity and can hence lower the nutrient use efficiency (Zhang et al., 2008b). We suspect that the immobilization of P in aboveground and belowground compartments, together with decreased pH, cause a decline in available P in shrub encroached plant communities.
Surprisingly, we found that decreases of N in plant communities did not cascade to decreases of soil N availability. We expected that the increase of woody biomass with high concentrations of lignin and tannins, which are known to inhibit N release (Rahman et al. 2013), would result in decreasing soil available N. Broadbent et al. (2022) observed decreased soil N availability in Austrian subalpine grasslands encroached by ericaceous species. However, seasonality, soil biota or microclimate might offset this effect. For instance, Crofts et al. (2018) found no differences in N availability in tundra ecosystems with increasing shrub density, potentially linked to a lower pulse of nutrients in summer (vs. spring), when the sampling was done. The symbiosis of shrubs with ericaceous mycorrhizae and ectomycorrhiza may also impact N cycling by immobilizing soil nutrients (Broadbent et al. 2022). Yet, we did not find an increase in the abundance of ericoid and ectomycorrhiza in the soil in the same experimental plots encroached by Vaccinium species (Mouhamadou et al., forthcoming). Lastly, microclimatic factors such soil moisture or temperature due to rainfall interception or shading by shrubs can affect N cycling (Aguirre et al. 2021). It is likely that the multifactorial determinism of N cycling, including microbial communities, dominant functional shrub species, and abiotic factors such as precipitation regimes and soil water content, made it difficult to observe decreasing trends of available N in this study.
At our sites, land use history with varying duration since mowing (Lautaret) and more intensive grazing (Stubai), along with idiosyncratic colonisation into initially homogenous grasslands (Lavorel et al. 2017) have resulted in varying shrub density. Our results based on our directional structural equation model show that this shrub density gradient then influences soil properties through plant-soil feedbacks (Bennett et al. 2017; Wardle et al. 2004). Furthermore, we cannot exclude that changed soil conditions (greater amounts of more recalcitrant organic matter, lower pH and P availability) may result in a feedforward loop further favouring shrub colonization through changes in microclimate and/or soil microbial communities (Myers-Smith et al. 2019; Wookey et al. 2009). For example, the presence of ericoid mycorrhiza fungi present in ericaceous species can outcompete the ubiquitous arbuscular mycorrhiza fungi in the obtention of nutrients from recalcitrant litter (Wurzburger and Hendrick, 2009) and limit N availability to non-ericoid mycorrhizal plants, including the herbaceous matrix (Northup et al. 1998; Ward et al. 2021). Such effects require experimental testing including long-term shrub removal experiments (Aguirre et al. 2024; Broadbent et al. 2021; Fanin et al. 2022; Urcelay et al. 2009).
Effects of shrub species identity
The functional type of the dominant species has been found to be one of the main drivers of the functional structure of alpine subshrub communities (Illa et al. 2017), and of functional effects of shrub encroachment globally (Eldridge et al. 2017). The effect of ericaceous species on soil properties has been well documented (Hattenschwiler 2000), and their negative effect on N availability shown in various ecosystems, including boreal forests (Fanin et al. 2022) and subalpine grassland (Broadbent et al. 2022). In this study, we found consistent changes in responses of soil properties and nutrient availability to shrub density for ericaceous species, whether deciduous or evergreen, across the two sites. In contrast, our second SEM incorporating J. communis as a source of variation confirmed that the effects of shrub encroachment on soil properties were species dependant (Fig. 4). Plots with high abundance of J. communis had soils with higher pH, lower concentrations of PO4 − 3 and NH4+ and higher concentrations of NO3− compared to plots dominated by ericaceous species or R. spinosa, while grassland plots did not significantly differ across corresponding communities. Responses to shrub density also differed in J. communis encroached communities, with limited changes in pH, SOM or soil CN. The higher litter persistence of J. communis (Illa et al. 2017) might explain the lower incorporation of plant material into soil organic matter and corresponding stability in C:N stoichiometry. Grau et al. (2019) found similar results in subalpine grasslands encroached by Juniperus sabina in south facing slopes of the Pyrenees, with no change in pH between young and mature shrublands. Consistent with these results, in our study J. communis dominated communities had lower P concentrations in plant tissues than other, and especially ericaceous dominated communities, reflecting lower P concentration in this species than in grassland or other shrub species. In the Pyrenees, this resulted in decreases in total aboveground and belowground P stocks in plots encroached by J. communis or J. sabina (Urbina et al. 2020). Sequestration of nutrients in the abundant biomass of shrubs and especially Juniper species (Illa et al. 2017) might explain the observed decrease of available P. Despite the non-significant change of N availability with increasing shrub density, we found higher concentrations of NO3− in plots encroached by J. communis compared with plots encroached by ericaceous species. The later also showed lower values of pH at high shrub densities (Fig. S3b), which can have negative impacts on N availability (Wang and Kuzyakov 2024), and thus explain the lower NO3− concentrations compared with plots encroached by J. communis.
Implications for the quantification of ecosystem services in shrub-encroached land
Shrub encroachment is expected to significantly modify societal values of grasslands through changes in ecosystem services (Bardgett et al. 2021). Quantification of these changes with trait-based models is required to assess their magnitude and project the effects of future climate and land use change scenarios (Barros et al. 2017; Kambach et al. 2024). Previous models were developed for herbaceous communities (Grigulis et al. 2013; Lavorel et al. 2011), and have been applied to project changes in ecosystem services in response to subalpine grassland management scenarios (Lamarque et al. 2014; Lavorel 2018; Schirpke et al. 2017). These models based on the plant economics spectrum and especially on traits reflecting plant size (height) and aboveground tissue composition (dry matter content, nutrient concentrations) could however not be applied to shrub-encroached communities until demonstrating the validity of these traits beyond grasslands. With this study, we show that this extrapolation is warranted.