Introduction

Soils represent the largest organic carbon reservoir in the terrestrial biosphere, globally exceeding the total carbon pool of vegetation and the atmosphere (Bhattacharya et al. 2016). Most studies estimate the global amount of soil organic carbon (SOC) to be around 1400–1500 Pg C in the top 100 cm (Scharlemann et al. 2014; Batjes 2016).

Among the ecosystem services they provide, soils play a key role in climate regulation through the sequestration of atmospheric carbon dioxide (CO2) and its storage as organic carbon (Jónsson and Davídsdóttir 2016).

The Sustainable Development Goals (SDGs) defined by the United Nations in 2015 include soil conservation as an urgent climate action to be taken, with the goal of preserving its multifunctional role.

Indeed, studies on the potential of soils to act as carbon sinks and mitigate climate change have increased in recent decades. Studies on alpine soils have also increased, as they are notoriously rich in organic matter (Zhu et al. 2019) and particularly susceptible to climate change, in terms of increased erosion, changes in microbial community structure and activity, and rising CO2 emissions (Garcia-Pausas et al. 2007; Sjögersten et al. 2011).

Soil carbon content depends on many factors. Most studies agree that vegetation is the environmental variable that mostly influences the quantity, quality, and distribution of SOC (Kumar et al. 2017). Indeed, vegetation, with the biomass it holds, is the main contributor to soil organic matter, and once processed by the soil biota in relation to abiotic factors, it determines the formation and accumulation of SOC (Kuznetsova 2022). Variations in vegetation composition can lead to large variations in the amount and type of soil organic matter (Drewnik et al. 2016). For example, in mountain regions, broadleaf forests, coniferous forests, and grasslands have different rates of SOC production, accumulation, and decomposition (Oueslati et al. 2013). In particular, the SOC stock of alpine grasslands and broadleaf forests is higher compared to alpine coniferous forest soils (Canedoli et al. 2020). Many authors have investigated the relationship between SOC and vegetation type, but only a few have worked at the level of the plant community (Ferré et al. 2020; Yimer et al. 2006), investigating the effect of the mosaic of plant communities on SOC content (Zhao et al. 2021).

However, SOC is also greatly affected by topographical (aspect, elevation, and slope), pedoclimatic and pedological factors (temperature, water content, chemical, physical, and biological characteristics of soil).

As regards the relationship between SOC stocks and soil types, the existing papers have mostly been carried out at small scale, giving great importance to land use and land cover (Mendoza-Vega et al. 2003; Munoz-Rojas et al. 2012; Wiesmeier et al. 2015). Only Conforti et al. (2016) have worked on a larger scale, in a nature reserve in Calabria.

Climatic conditions and, therefore, climate change, can also strongly influence the carbon cycle, alter carbon fluxes, and consequently affect the carbon storage capacity of soils. Several studies have investigated the response of SOC to warming (Li et al. 2021). Many of these have been conducted in the alpine regions of the Tibetan Plateau, considered one of the most sensitive areas to climate change (Chen et al. 2020). These studies have shown that both C and N cycles are sensitive to warming at the ecosystem level, but the observed changes in total C and N stocks are minimal, so that warming effects are likely to offset each other. In fact, on the one hand, warming increases plant biomass, resulting in higher inputs of C and N into the soil, and on the other hand, it stimulates soil respiration, leading to large losses of C as CO2 to the atmosphere. Contrary to these results, other authors have observed an overall loss of carbon, due to the predominant effect of warming on the microbial activity of decomposing organic matter (Hagedorn et al. 2010; Streit et al. 2014; Melillo et al. 2017).

Studies on SOC content and its relationships with environmental factors are particularly complex in mountain areas, which are characterized by high local variability in soil environment, topography, and microclimate. This variability influences the carbon cycle, induces changes in soil organic matter (SOM) quality and quantity (Canedoli et al. 2020; Ferré et al. 2020) and determines a high spatial heterogeneity of SOC (Zhu et al. 2019). Particularly, elevation, aspect, slope, temperature, and water content have been reported to be the main determinants of the great variability in the spatial patterns of mountain SOC (Hoffmann et al. 2014; Zhu et al. 2019).

Elevation is the most studied topographic variable, due to its close relationship with temperature and consequently with vegetation, which develops mainly along altitudinal bands. In most cases, SOC stock (Dorji et al. 2014; Zhang et al. 2021; Zhu et al. 2019), or SOC content (Parras-Alcántara et al. 2015; Choudhury et al. 2016), has shown a positive correlation with altitude, which strongly influences the mesoclimate and turnover of vegetation types, affecting organic matter decomposition rate and plant production (Zhu et al. 2019). Specifically, the increase in SOC with altitude is explained by the decrease in temperature, which limits the microbial activity and the decomposition rate of SOM, allowing the accumulation of SOC (Parras-Alcántara et al. 2015). However, over a certain elevation, the strong reduction of vegetation cover and the soil itself would cause a decrease in the SOC stock.

Aspect affects the amount of solar radiation incident on the ground and the microclimatic characteristics of an area. Some authors (Jendoubi et al. 2019; Zhu et al. 2019) have found that south-facing soils have a lower SOC content than north-facing soils, while others (Drewnik et al. 2016; Yimer et al. 2006) have observed the opposite trend.

With regard to the gradient of slopes, previous studies observed a strong negative correlation between SOC content and increasing slopes (Jendoubi et al. 2019; Yu et al. 2020). This is due to the higher intensity of runoff processes that characterize steeper slopes compared to flat areas, where the accumulation of material is favoured (Jendoubi et al. 2019).

However, since the effects of elevation, aspect, and slope are often strongly related to each other, it is difficult to disentangle their single contribution to the SOC stock.

To date, studies on the climate regulation services of mountain soils that assess the interaction of multiple factors in influencing C stocks are quite limited. Therefore, we present here the results of a study in a valley of the Central Alps (Italy) on the climate regulation services of alpine soils. The objectives of this work were: i) the investigation of the effects of soils, vegetation types and environmental parameters on the SOC stocks and other soil properties; and ii) the formulation of hypotheses on the effects of climate change on the evolution and biodiversity of plant communities and the SOC storage.

Materials and methods

Study area

The study area (Lat: 46.1090°, Lon: 10.5260°) is located in the Valle Adamé, one of the main glacial valleys of the Adamello Group in the Central Alps (Fig. 1), which extends with a sub-planar morphology and an N–S trend for about 8 km, starting from the glacial threshold at 2000 m a.s.l. that separates it from the Val Saviore below.

Fig. 1
figure 1

Study area with schematic indication of the altitudinal range to which plots belong (plots are divided by vegetation type)

Full size image

From a geologic point of view (Callegari and Dal Piaz 1973), the Adamello Group is made up of a large Cenozoic batholith surrounded by a band of older crystalline metamorphic rocks. The central massif, including the highest peak (Mt. Adamello, 3539 m), is mainly made up of tonalites, quartzdiorites, and granodiorites, emplaced during the Late Eocene–Late Oligocene alpine orogenesis. The summit area of the massif hosts the Adamello glacier, the largest in the Italian Alps (about 16 km2 in 2007; Baroni et al. 2007). During the Last Glacial Maximum, the Valle Adamé was covered by glaciers that shaped the local morphology, which was characterized by a succession of basins, steps, and glacial deposits (Baroni et al. 2014). The climate of the Valle Adamé is continental (endoalpine type), with cold winters and mild summer temperatures. The annual mean temperature at about 2200 m is 5.6 °C, and annual precipitation reaches 1350 mm in the upper Adamé Valley (Ceriani and Carelli 1999). Snow may persist locally in sheltered channels even during the late summer (Baroni et al. 2014).

Above the current tree line, the vegetation in the Valle Adamé mainly consists of Festuca luedii (Markgr.-Dann.) Foggi, Gr.Rossi, Parolo & Wallossek grasslands (mainly on the southern slopes), Nardus stricta L. pastures, and Rhododendron ferrugineum L. heatlands. Sites with long-lasting snow cover are characterized by alpine tundra (Salix herbacea L.) and by tall herbs (Adenostyles leucophylla (Willd.) Rchb.).

Soils are mostly acidic, rich in rock fragments, coarse-textured, and well-drained; the main soil types (IUSS Working Group WRB 2022) are Leptosols, Podzols, and Umbrisols (Sartori et al. 2005).

The study area is interested in the summer period by pasture. Dairy cattle and goats graze freely throughout the area, mainly using the pasture areas on the valley floor, which are a mosaic of fat, lean, wet and shrub pasture plots. The livestock load, evaluated in ABU (Adult Bovine Units) has not changed in the last 20 years (data of Lombardy alpine pasture census for 2000 and 2021; unpublished): 2/3 is represented by cattle and the remaining by goats.

Vegetation survey

To investigate the soil–vegetation relationship and how the SOC stock varied between the main plant communities, five vegetation types were selected according to the dominant species: 1. Salix herbacea (Salix), Nardus stricta (Nardus), Rhododendron ferrugineum (Rhododendron), Adenostyles leucophylla (Adenostyles), Festuca luedii (Festuca). The field survey took place in the second half of July 2019, when vegetation development was at its peak. For each vegetation type, five plots of about 30 × 30 m were selected along the valley in areas where the dominant species exceeded 50% cover (Fig. 1). In each plot, three 50 × 50 cm sampling points were randomly selected, approximately 20 m apart, where the floristic survey was carried out, to also verify that the vegetation types were well-defined and separated before proceeding with the soil sampling.

The nomenclature follows the updated checklists of the vascular flora of Italy (Bartolucci et al. 2018), continuously updated in The Portal to the Flora of Italy (https://dryades.units.it/floritaly/) (Martellos et al. 2020).

Soil sampling and laboratory analyses

For each vegetation type, three minipits were opened and described by horizon within each of the five plots near the floristic sampling points; for each plot, only the minipit considered most representative was sampled for taxonomic classification (IUSS Working Group WRB 2022). For the evaluation of the effects of vegetation, soil types and environmental factors on SOC stock and other soil properties, soil samples were collected from the 0–10 cm, 10–20 cm and 20–40 cm layers of the 75 georeferenced minipits. In some cases, the depth reached was less than 40 cm due to the presence of the rocky substrate near the soil surface. A total of 188 mineral soil samples were collected.

For each layer, the percentage of rock fragments (RF) was visually estimated, and volumetric samples (core diameter 5 cm; sample volume 100 cm3) were taken for the determination of bulk density (BD); the latter sampling was only carried out in the minipit considered to be the most representative, with a total of 75 core samples collected. For the BD determination, soil cores were oven dried at 105 °C for 24 h and weighed; in the presence of rock fragments, the soil volume and mass were proportionally reduced to obtain the BD of fine earth (< 2 mm diameter).

For those layers for which no core sample was obtained, site-specific empirical pedofunctions, one for each investigated layer (0–10 cm [1]; 10–20 cm [2]; 20–40 cm [3]), were used to estimate the BD value as a function of the SOC content (%):

$$y=1.0435\times {SOC}^{-0.255} \quad {r}^{2}=0.53,$$
(1)
$$y=-0.183 \times \mathrm{ln}\left(SOC\right)+1.0861 \quad {r}^{2}=0.79,$$
(2)
$$y=-0.185 \times \mathrm{ln}\left(SOC\right)+1.1358 \quad {r}^{2}=0.86.$$
(3)

Soil mineral samples were air dried, sieved (2 mm mesh) and analyzed to determine pH in H2O (pH; soil to water ratio of 1:2.5). A fraction of each sample was further crushed, sieved to a finer mesh (0.5 mm) and analyzed for the determination of organic carbon and total nitrogen contents by dry combustion using a Flash EA 1112 NC-Soil elemental analyzer (Thermo Fisher Scientific CN, Pittsburgh, USA). The SOC content (%) was converted to SOC stock (kg m−2) considering the soil BD and the percentage volume of rock fragments in each layer.

For each soil sampling point, the organic horizon (OL only) was also collected using a 33 × 33 cm frame, oven-dried at 70 °C for 48 h and then weighed. The organic carbon and nitrogen content of the organic samples was determined using the NC elemental analyzer and converted to content on an area basis (kg m−2).

The list and abbreviations of all the parameters considered are given in Table 1.

Table 1 Investigated soil properties and types, environmental variables, vegetation types and indexes and their abbreviations
Full size table

Sampling of aboveground and root biomass

One sample of epigean biomass and one of hypogean biomass were taken per plot, in correspondence with one of the three minipits, and the floristic survey area. The same 33 × 33 cm frame used for the sampling of the organic horizon was used for the sampling of aboveground biomass, cutting approximately 2 cm above the ground. The collected biomass was air-dried, then put in an oven at 80 °C, weighed and expressed as weight on an area basis (kg m−2).

A soil cube sample was taken to a depth of 10 cm (sample volume 1 dm3) for the determination of the root biomass. The extraction of roots from the cores was carried out by presoaking the sample in agitation overnight in a solution of EDTA (40 g L−1) to facilitate washing. After that, roots were washed, picked out, separated into three diameter classes (< 2 mm, very fine roots; 2–5 mm, fine roots; > 5 mm, coarse roots), and oven dried at 80 °C for 24 h to determine the dry masses of the root biomass, expressed as g dm−3.

The organic carbon and nitrogen contents of the aboveground and belowground samples were determined using the NC elemental analyzer; the C:N ratio for roots and aboveground biomass was then calculated.

Aboveground biomass and root samples were taken for all the vegetation types, and the ratio of aboveground to belowground biomass was calculated for each, with the only exception of Rhododendron, which is a shrub species also characterized by dense, abundant root systems and large roots, that do not allow for proper volumetric sampling.

Ancillary data

For each minipit location, elevation, aspect and slope were extracted from the Digital Terrain Model (DTM, 5 m resolution) using QGIS software (version 3.16).

Aspect was converted into Northness and Eastness Index (cosine and sine of the aspect expressed in radians, respectively).

In addition, the soil water content was measured at each minipit location in the 0–8 cm layer using a portable TDR system (IMKO Micromodultechnik, GmbH, Ettlingen, Germany) and then converted (using total soil porosity, totPor) to water-filled pore space (WFPS). The totPor was previously calculated from \(\frac{\mathrm{PD}-\mathrm{BD}}{\mathrm{PD}} ,\) where PD (g cm−3) is the soil particle density (\(\mathrm{PD}=2.65-0.02\times \mathrm{SOC}(\%)\times 1.724)\).

Statistical analyses

The similarity percentage procedure (SIMPER analysis) was performed to assess the species contribution across vegetation types using the average Bray–Curtis dissimilarity index (Table 2). Analyses of similarity (one-way ANOSIM; Bray–Curtis index; 10,000 permutations) were used on the full species data set to test for differences in species composition between the vegetation types before starting soil sampling.

Table 2 SIMPER analysis of the selected vegetation types. The overall dissimilarity according to the Bray–Curtis index was 91.72
Full size table

To evaluate the effects of vegetation, soil types and environmental factors on SOC stock and the other soil properties, a linear mixed effect model (LME) procedure was run in the R environment (version 4.2.2; R Core Team 2022) with soil type, vegetation type or environmental factors as fixed effects and “plot” and “soil type” (if not used as a fixed effect) as random effects, with plot nested within soil.

To identify differences in biodiversity between vegetation types, the same LME procedure used for soil properties was also performed, with vegetation type as a fixed effect and plot, nested within the soil type, as a random effect.

To analyse the relationships among plant species, soil properties (0–20 cm layer) and environmental variables, the canonical correspondence analysis (CCA) ordination method was used. The CCA was performed with the CANOCO software, version 4.5 (Ter Braak et al. 2002). Before CCA, correlation analysis (Spearman’s rho) was carried out to reduce the number of variables and noise in the variance components. One of each pair of highly correlated variables, with a correlation coefficient ≥ 0.75, was eliminated from the analysis (Perez-Riverol et al. 2017). During the CCA analysis, variables showing a variance inflection factor (VIF) higher than 10 were also removed. The significance of the variables was tested using permutation tests (N = 999).

Particularly at first, we performed a CCA, including the selected variables. Subsequently, to assess the effect of topographical variables on the variance of the other environmental variables (especially C stock), we carried out a partial-CCA considering elevation, slope and aspect as covariates.

Results

Vegetation

The selected vegetation types exhibited differences in species composition and relative abundance, as also highlighted by the ANOSIM analysis, using the Bray–Curtis index (mean rank within = 308.9; mean rank between = 1640; r 0.9592; p < 0.0001). All pairwise comparisons between groups exhibited significant differences (p < 0.01 for all comparisons after the Bonferroni correction). According to the SIMPER analysis (Table 2), the total dissimilarity among groups was 91.72 (Bray–Curtis index). Besides the dominant species characterizing the five vegetation types, the five species that mostly contributed to the dissimilarity among groups were (values above 3%): Imperatoria ostruthium, Mutellina adonidifolia, Scorzoneroides helvetica, Festuca rubra and Potentilla aurea.

The cumulative vegetation cover was the highest in the Adenostyles type and the lowest in the Nardus and Salix types (F = 11.3267; p < 0.0001; Fig. SM1).

Overall, the α-diversity among groups was similar, except for Festuca type that showed the lowest value of richness (F = 15.27; p < 0.0001) and Shannon index (F = 11.27; p < 0.0001; Fig. SM1). On the other hand, β-diversity according to the Bray–Curtis dissimilarity was the highest in Festuca and Salix types compared to the other three vegetation types (F = 5.703; p < 0.0038; Fig. SM1).

Soil

Soil types

Based on the WRB classification (IUSS Working Group WRB 2022), the main soil types were Leptosols, Umbrisols, Podzols and Fluvisols (Table SM1).

Leptosols, thin soils, developed on boulders on very steep slopes, showed a high rock fragment content and an umbric horizon (Umbric Skeletic Leptosols; 9–25 cm thick). They were mainly located on the NW-facing slopes, under the Adenostyles vegetation type, and to a lesser extent on the lower slopes at the junction with the valley floor, under the Rhododendron vegetation type.

On the SE-facing slopes, under the Festuca vegetation type, we found the deepest soils among those studied: Umbric Entic Podzols (45–47 cm thick) and Cambic Umbrisols (42–45 cm thick).

Podzols (Entic Podzol; 30–45 cm thick) and Umbrisols (30–40 cm thick with evidence of colluvium and a protospodic horizon), were also found under the Salix vegetation type.

On the lower slope and at the junction with the valley floor, where Nardus and Rhododendron types predominated, Dystric Fluvisols, Podzols and Umbrisols were found.

SOC stock and other soil properties by soil types

The average SOC stock in the 0–40 cm layer (SOCstock) was 11.7 ± 8.9 (mean ± SD) kg m−2 with Podzols and Umbrisols showing a significantly higher SOCstock than Fluvisols and Leptosols (Fig. 2; F3,47 = 6.07, p < 0.01, R2 = 0.80).

Fig. 2
figure 2

Soil carbon stock, pH, C:N ratio and rock fragments content of the 0–40 cm layer by soil types. For abbreviations refer to Table 1. Box plot legend: middle point: median, box: IQR range, whisker: non-outlier range

Full size image

SOCstock by soil type and overlying vegetation are reported in Table SM2.

In Podzols, the highest SOC stocks were found under Salix, but mainly under Festuca; Umbrisols with the highest SOCstock were found mainly under Nardus, Salix and Festuca. Among the Leptosols, which contained lower SOCstock than the other soil types due to their limited thickness, the richest ones were found under Festuca.

Statistically significant differences among soil types were found for the C:N ratio of the 0–40 cm layer (CN) (F3,47 = 6.36, p < 0.01, R2 = 0.64), which was lower in Fluvisols (mean ± SD: 11.7 ± 1.4) than in Leptosols (14.2 ± 1.6), Umbrisols (13.8 ± 1.3) and Podzols (15.2 ± 1.8), but not for pH (Fig. 2 and Table SM3). Among the investigated soil types, the thin Leptosols had the highest rock fragment content, statistically different from the other types in the 0–20 cm layer (F3,47 = 8.34, p < 0.001, R2 = 0.40).

Effects of geomorphology on SOC stock

The investigated plots were located in the upper part of the valley, between 2000 and 2350 m, mainly facing NW and SE and on slopes ranging from 4° to 41° (Table SM3).

Considering the effects of morphology on SOCstock (Table 3) the LME model showed that the stock increased with increasing altitude (F1,45 = 7.19, p < 0.05) and decreasing Northness Index (F1,45 = 6.44, p < 0.05), while it showed no relationship with Eastness Index (F1,45 = 0.34, p = 0.56) and slope (F1,45 = 1.94, p = 0.17).

Table 3 SOC stock (0–40 cm) relationships with morphological variables (slope, elevation and aspect expressed as Northness and Eastness Indexes)
Full size table

Plant–soil relationships

Vegetation was observed to influence the SOC stock mainly in the first investigated layer (SOCstock1) and partly in the second one (SOCstock2), while it had no effect in the third layer (SOCstock3; Table 4). Festuca and Adenostyles showed higher SOCstock1 compared to the other types; the SOCstock2 under Festuca was significantly higher than that of the other vegetation types, but not different from that of Salix, which only differed from that of Adenostyles.

Table 4 Linear mixed model results of SOC stock of each investigated layer; vegetation type was fitted as fixed factor, while “plot” and “soil type” as random effects, with “soil” nested within “plot”
Full size table

In this regard, it should be noted that, according to the field description and the laboratory determination (the latter only for the first layer), the biomass of the root system was mainly in the first layer, with differences among vegetation types, and remained high in the second layer only under Festuca and Salix (Table 5). The differences among the vegetation types were not only limited to the SOC stock of the 0–20 cm layer (SOCstock1–2: F4,18 = 12.14, p < 0.001, R2 = 0.80), but also concerned other soil properties (Fig. 3), such as the SOC stock and C:N ratio of OL horizon (SOCstockOL: F4,18 = 10.71, p < 0.001, R2 = 0.59; CNOL: F4,15 = 27.72, p < 0.001, R2 = 0.68), the C:N ratio and pH up to 20 cm depth (CN1–2: F4,18 = 3.21, p < 0.05, R2 = 0.63; pH1–2: F4,18 = 9.16, p < 0.001, R2 = 0.70), the WFPS (F4,18 = 13.05, p < 0.001, R2 = 0.57), the aboveground biomass and the ratio of aboveground biomass to total root biomass (AgB: F3,16 = 9.48, p < 0.001, R2 = 0.60; AbB:totRoots: F3,14 = 22.71, p < 0.001, R2 = 0.97).

Table 5 Total, coarse, fine and very fine root density (g dm−3) in the 0–10 cm layer for each vegetation type and field estimation of root amount for each investigated layer and vegetation; n.d.: not determined
Full size table
Fig. 3
figure 3

SOC stock and C:N ratio of organic horizon, SOC stock, C:N ratio and pH of 0–20 cm mineral layer, WFPS, aboveground biomass, ratio between aboveground and belowground biomass by vegetation types. For abbreviations of the variables, refer to Table 1. Box plot legend: middle point: median, box: IQR range, whisker: non-outlier range

Full size image

Regarding the organic horizon, the SOCstockOL under Festuca and Rhodedendron was higher than that of the other vegetation types, statistically different from Nardus and Salix; CNOL under Festuca was the highest, while under Salix and Adenostyles was the lowest; Rhododendron and Nardus types showed intermediate values of CN.

The first 20 cm of soil under Adenostyles had the highest pH1–2 (5.0 ± 0.3), statistically different from all the other vegetation types.

Concerning soil water content, Nardus and Salix prefer wetter soils, with WFPS values ranging between 60% and 70%, significantly different from those of the other vegetation types. Under Festuca, located on the south-facing slope, the topsoil was the driest.

Festuca and Adenostyles were also characterized by the highest aboveground biomass (mean ± SD: 0.51 ± 0.09 kg m−2 and 0.44 ± 0.28 kg m−2, respectively), while no difference in root biomass in the first 10 cm was found between vegetation types, so that the ratio between above- and belowground biomass was the highest in the case of Festuca and the lowest for Salix.

The CCA analyses generally resulted in high eigenvalues and high cumulative percent variances for the species data. In both the full model and that considering covariables, the first three CCA axes explained about 60% of the variance (Table 6; Fig. 4).

Table 6 CCA analysis of plant communities in relation to the considered soil properties (0–20 cm layer) and types, and environmental variables
Full size table
Fig. 4
figure 4

CCA analysis of plant communities in relation to a the considered soil properties (0–20 cm layer) and types and environmental variables b with topographic variables as covariates. For abbreviations of the variables, refer to Table 1. Ade_leu: Adenostyles leucophylla; Agr_rup: Agrostis rupestris All.; Alc_xan:: Alchemilla xanthochlora Rothm.; Ant_nip: Anthoxanthum nipponicum Honda; Bar_alp: Bartsia alpina L.; Bis_viv: Bistorta vivipara; Car_sem: Carex sempervirens; Cir_spi: Cirsium spinosissimum (L.) Scop.; Fes_lue: Festuca luedii; Fes_rub: Festuca rubra; Gen_aca: Gentiana acaulis L.; Geu_mon: Geum montanum L.; Hel_pra: Helictochloa praetutiana (Parl. ex Arcang.) Bartolucci, F.Conti, Peruzzi & Banfi; Hie_vil: Hieracium villosum Jacq.; Hom_alp: Homogyne alpina (L.) Cass.; Imp_ost: Imperatoria ostruthium; Sco_hel: Scorzoneroides helvetica; Leu_alp: Leucanthemopsis alpina; Lot_alp: Lotus corniculatus L. subsp. alpinus (DC.) Rothm.; Luz_alp: Luzula alpinopilosa (Chaix) Breistr.; Mut_ado: Mutellina adonidifolia; Myo_alp: Myosotis alpestris F.W.Schmidt; Nar_str: Nardus stricta; Phy_hed: Phyteuma hedraianthifolium Rich.Schulz; Poa_alp: Poa alpina L.; Pot_aur: Potentilla aurea; Pot_ere: Potentilla erecta (L.) Raeusch.; Ran_mon: Ranunculus montanus Willd.; Rho_fer: Rhododendron ferrugineum; Rho_ros: Rhodiola rosea L.; Sal_her: Salix herbacea; Vac_myr: Vaccinium myrtillus; Vac_uli: Vaccinium uliginosum; Ver_alb: Veratrum album L.; Vac_vit: Vaccinium vitis-idaea L.; Vio_bif: Viola biflora L

Full size image

In the full model, environmental variables that showed significance (p value < 0.05) and the highest levels of LambdaA score (above 0.15) were: slope (0.69), SOCstock1–2 (0.58), elevation (0.34), soil depth (0.17).

In the biplot of the first two axes (Fig. 4a), slope and elevation were positively correlated with the Festuca and Adenostyles types and negatively correlated with the other vegetation types.

Salix, Nardus and Rhododendron were close to each other and positively correlated with soil water content, while they were negatively correlated with elevation, slope, and rock fragments. Festuca and Adenostyles were isolated from the other vegetation types. Festuca vegetation was positively correlated with SOC stock, elevation, and slope, while it was inversely correlated with the Northness Index. Adenostyles type was positively correlated with pH, rock fragment content, Leptosols, slope and elevation.

In the model considering the covariates (Fig. 4b), the highest levels of LambdaA score (above 0.15) were: soil thickness (0.20) and Podzol (0.17). Soil thickness was positively correlated with Festuca and Nardus types and negatively correlated with Adenostyles type.

Considering the first two CCA axes, the different types of vegetation were closer together and overlapped, and they were distinguished primarily by the properties of the soils with which they were associated.

Discussion

Factors affecting SOC stock

Although the investigated alpine soils had limited thicknesses and a high rock fragment content, the average SOC stock down to a depth of 40 cm was 11.72 ± 8.97 kg m−2, higher than that found in the acidophilous grasslands (6.29 ± 2.31 kg m−2) of Gran Paradiso National Park (Italy; Canedoli et al. 2020).

The wide range of SOC stock variation is in part evidence of the effect of soil type: the highest average SOC was found in more developed soils, such as Umbrisols and Podzols, while Leptosols and Fluvisols showed low storage capacity because of their low pedogenetic development and the frequent disturbances (snow erosion, water erosion and burying) to which they are subjected due to their location on the slope. This is in line with some studies (Munoz-Rojas et al. 2012; Wiesmeier et al. 2015) that have examined the impact of land use on SOC stock and have also demonstrated the influence of soil development.

SOC values in the first 10 cm (5.2 ± 2.6 kg m−2) were comparable to those found in a sinkhole with highly heterogeneous soil and vegetation types in the upper Valchiavenna (Lombardy, Italy), at a height of about 1900 m, ranging between 3.3 and 6.5 kg m−2 (Ferré et al. 2020).

A declining trend with depth was observed, in line with previous studies (Mendoza-Vega et al. 2003; Munoz-Rojas et al. 2012); 74% of the SOC stock was sequestered in the upper 0–20 cm layer, while the remaining, equal to 3.0 kg m−2, was stored in the 20–40 cm layer, where rock fragments occupied about half of the soil volume. Below 40 cm depth, the rock fragments predominated, resulting in a negligible carbon stock.

Vegetation type was undoubtedly the factor that most influenced the topsoil organic carbon differentiating stocks, vertical distribution, and quality.

The vegetation types differentiated the organic horizons, with the Rhododendron type and Festuca stocking more SOC (0.18 kg m−2 and 0.26 kg m−2, respectively) than the other vegetation types. Both had thick litter layers with high CNOL (30.2 and 47.2 for the Rhododendron type and Festuca), indicating a slowdown in organic matter decomposition (Hobbie 2015; Steffen et al. 2007). In contrast, the organic horizons of Salix and Nardus were thinner or absent.

Regardless of the soil type, Festuca appears to contribute to SOC enrichment in the 20-cm topsoil. This is likely because of its high (above- and belowground) biomass, which is promoted by the high solar radiation on the SE-facing slope and SOM quality. From high values in the OL layer, the C:N ratio in the mineral soil decreased, but still remained quite high, indicating a type of SOM that is more resistant to mineralisation.

In terms of topsoil, SOC accumulation under the Salix vegetation was comparable to that under the Festuca type. The complex process of plant litter decomposition is regulated by the quality of the litter, but also by pedoclimatic factors (Coûteaux et al. 1995; Santa Regina and Tarazona 2001). In this regard, the Salix type is the typical vegetation of the nival valleys, where the snow cover persists for many months and moisture levels are high even in the absence of snow cover; this may have slowed down the SOC mineralization by limiting microbial activity due to the lack of oxygen and the lower temperatures caused by the higher soil water content. Salix had the lowest ratio of aboveground to belowground biomass. Previous authors have reported that in plants found in tundra ecosystems (in our case, alpine tundra), the belowground biomass is much higher than the aboveground biomass. Particularly, fine roots, rhizomes and belowground stems are present in a shallow layer of organic matter at the soil surface. This material accumulates because of reduced decomposition rates in cold, and sometimes saturated, soils (Iversen et al. 2015). This imbalance can be attributed to a different phenological development between above- and belowground parts in most tundra species, where a longer growing season (up to about 50%) can be observed in the belowground apparatus than aboveground (Blume‐Werry et al. 2016).

Soils under Adenostyles had the lowest SOC stocks (3.1 kg m−2) in the 0–20 cm layer. In this regard, although Adenostyles had almost twice the belowground biomass of the other vegetation types, it had a significant fraction of coarse and fine roots with lower turnover rates than very fine roots (Schneider et al. 2006). This suggests that, in addition to a SOC loss through erosion, due to the distribution of this vegetation on steep slopes, the scarcity of SOC compared to the other vegetation types may also be due to the characteristics of the organic matter inputs. As a further confirmation of this, based on the C:N values obtained, which are strictly related to the rate of SOM decomposition (Newbery and Gartlan 1996; Hodge et al. 1998), the fine roots of Adenostyles showed lower degradation rates than the other plant types studied.

Topographic and geomorphological factors were mainly found to indirectly influence the SOC stock, and directly act as discriminator of the plant community, similar to the findings of Zhu et al. (2019). In our study area, elevation, aspect and slope were determinants in shaping the spatial patterns of the plant types: Salix, Rhododendron and Nardus types were mostly found on the valley floor or connecting the valley floor and slopes, with a low slope gradient. Festuca and Adenostyles were mainly found at higher elevations on steep slopes, the former on the south-facing slope and the latter on the more rocky north-facing slope.

Despite the limited altitudinal range, the SOC stock increased with elevation, confirming what other authors have observed (Leifeld et al. 2009; Zhang et al. 2021). This in addition to the direct effect that elevation has on stock variability by affecting plant distributions, as in the case of Festuca, where soils had the highest observed SOC stock.

Hypotheses on the effects of climate change

Alpine ecosystems above treelines are particularly prone to global warming, since they are characterised by low temperatures (Körner 2003).

The SOC balance could be affected by the increase in temperature, which would lead to a change in the composition and extension of the plant community, altering the quality and quantity of organic matter input to the soil or the composition and activity of microbial communities, thus changing the C utilisation efficiency of soil microorganisms and the losses through mineralisation (Allison et al. 2010; Schindlbacher et al. 2011; Manzoni et al. 2012).

Long-term monitoring of vegetation in alpine environments indicates that rising temperatures observed over the last century have influenced the range and ecology of plant species (Parolo and Rossi 2008). In particular, alpine plants may respond to the rising temperatures predicted in the IPCC report (IPCC 2021) either by adapting their life cycles to the new environmental conditions or shifting their distribution range upwards. Particularly at higher elevations above the current tree line, shrub species (such as Rhododendron ferrugineum) are expected to expand as glaciers retreat (Gentili et al. 2020). Cannone et al. (2007), in their work on the upward shift of alpine vegetation due to climate change during the period 1953–2003, observed a marked increase in shrubs (Rhododendron ferrugineum and Kalmia procumbens (L.) Gift, Kron & Stevens ex Galasso, Banfi & F.Conti), at the expense of alpine grasslands. In the alpine belts, the abandonment of pastures has also led in the recent past and continues to lead to the recovery of Rhododendron ferrugineum shrublands (especially on northern and eastern aspects or in moist sites), confirming the great vitality and competitiveness of this community (Lasen 2017; Francon et al. 2021). Therefore, from the subalpine to the alpine belt, Rhododendron ferrugineum is a dominant shrub species functioning as an ecological indicator of the response of alpine ecosystems to global warming (Francon et al. 2020). In addition, in our study area, Rhododendron ferrugineum was abundant and had expanded over the last 20 years (observing the orthophotos) and it is expected to continue to expand in the short to medium term replacing other plant communities, such as Salix herbacea, associated with prolonged snow cover and relatively lower temperatures, first resulting in a decline in biodiversity.

The change in vegetation following the rise in temperature would also have an impact on the SOC stock of the investigated soils. The expansion of the Rhododendron type would imply a change in the quantity and quality of organic inputs; the results showed that its litter was among the most recalcitrant, thicker, and slower to decompose and transform into stable humus than that of the vegetation types that it would replace. Although we lack data on the Rhododendron roots, we can hypothesize a decrease in SOC input of hypogean origin in the short to medium term based on evidence that the mass of roots less than 2 mm in diameter, which have a higher turnover rate than other size classes (Schneider et al. 2006), is lower in shrubs than in herbaceous species (Pornaro et al. 2017).

The SOC stock balance is determined by organic carbon gains through litter and plant roots (Smith 2008), but also by losses due to mineralisation. In this regard, several field and laboratory warming experiments and observational studies across climatic gradients have largely shown that warming increases the rate of SOC decomposition to CO2 (Melillo et al. 2002; Conant et al. 2011), because it stimulates microorganism activity (Kirschbaum 2006).

As a result, it is expected that global warming will result in an increase in SOC loss in the investigated valley in the near future, due to enhanced SOC mineralization but also to a lower input of organic carbon due to the likely vegetation shift toward species with more recalcitrant organic matter, leading to a positive impact of soil carbon feedback on climate change.

Conclusion

We studied the soils of the upper Adamé Valley by investigating the relationships between SOC stocks and topographic factors, soil characteristics and vegetation communities.

The study improved our knowledge of alpine soils and their important role as sinks of SOC stocks. First, it revealed the importance of soil development in the differentiation of SOC stocks; topographic factors had mostly an indirect effect on SOC stocks, affecting the vegetation types that essentially determine the SOC spatial variability, quantity, and quality.

Based on the results, it can be hypothesized that soils will lose OC in the short to medium term in response to rising temperatures, and shorter snow permanence times, reducing the effectiveness of the climate regulation service they provide.