Discussion
Our study showed how plant guild proportions and their pairwise interactions drove SOC at regional scale, and additionally to other environmental factors including climate, bedrock topography or livestock management (Table 1). Grasses, forbs and legumes were the most determinant plant guilds for SOC in Pyrenean grasslands. Legumes had an enhancement effect on SOC storage but only at moderate proportions (Fig. 1). When legume proportions reached intermediate values, SOC started to decrease. This effect was sharper at increasing grass proportions and more moderate at increasing forb proportions (Fig. 2).
SOC storage represents a complex equilibrium between primary production (inputs) and organic matter decomposition (outputs) that depends on multiple environmental factors including climate, soil texture and nutrients, and land management (Jenny, 1941; Schlesinger, 1977; Jackson et al., 2017). Our results demonstrate that some guilds at low proportions can have positive effects on SOC, but also that those enhancement effects can disappear or shift when the proportions among guilds are modified. In our DIM (Table 1 & Fig. 1), the critical role of legumes on SOC is not surprising, since their symbiosis with rhizobia bacteria allow them to fix N2 from the atmosphere (McGrath et al. 2014), having large effects on N availability and supply (Zanetti et al. 1997; Spehn et al. 2002; Scherer-Lorenzenet al. 2003). These effects on soil N may alter SOC inputs and outputs among other ecosystem processes (Hector et al. 1999; Fornara & Tilman 2008).
However, what is original in our work is the changing role of legumes on SOC storage depending on plant diversity and plant guild interactions. Former studies have reported positive, neutral and negative effects of legumes on SOC (Steinbeiss et al. 2008; Lange et al. 2015; Wu et al. 2017). Results from our DIM are novel because they suggest that those effects could be dependent on legume biomass proportion, as found in the case of Pyrenean natural grasslands. Nyfeler et al. (2009, 2011) found similar responses of plant N and yield to legume proportions in an experimental work, which can be behind the SOC-legume relationship we found. At low proportions (0-20%), an increase in legumes enhanced noticeably the positive effect on SOC (Figs. 1 & 2), which could be classified as a keystone effect (Mills 1993). Positive effects of legumes on SOC may have three different non-excluding explanations according to Zhao et al. (2014). Firstly, legumes would promote an increase in primary productivity of plant communities through increased N availability which would lead to an increase in SOC (Wu et al. 2017). Secondly, positive effects of legumes on SOC can be attributed to low C/N ratios of legume residues (i.e. litter, root exudates…) more similar to soil microorganisms and soil organic matter than other plant residuals(Jensen et al.2012). Substrates with low C/N ratios can reduce microbial N acquisition and increase their carbon use efficiency, facilitating humification processes and plant residues decomposition into soil organic matter (Spohn et al. 2016). Thirdly, N inputs through legumes may inhibit the production of oxidative enzymes to degrade the more recalcitrant compounds, leading to reduced ecosystem CO2emissions and C losses (De Deyn et al. 2011; Spohn et al.2016).
On the other hand, in our study, when legume proportions were high, they had a negative effect on SOC according to our model (Figs 1 & 2). Negative effects of legumes on SOC stocks have been attributed to a decrease in community root biomass (Lange et al. 2015). Recent work by Prommer et al. (2019) suggested that the allocation of carbon to roots is less necessary at the community level when legumes are present, due to the increase of available N they fix. Furthermore, Henneron et al. (2019) found that the resource acquisitive strategy of legumes , with high photosynthetic activity and capacity and high root metabolic activity, exudation and death, may enhance soil microbial activity, depleting SOC stocks (i.e. rhizosphere priming effect; Kuzyakov, 2002). This explanation is also suggested by the findings in Ibañez et al. (2020), where high efficiency in C capture does not translate into high biomass in the grassland, We also suggest that, at some point, pathways for positive effects of legumes on SOC may be reversed. For instance, due to legumes´ strategy of producing a more nutrient-rich and short-lived biomass than other plant guilds (Craine et al. 2002) high legume proportions may involve less total biomass in the long term, and low litter C/N ratios could lead to higher mineralization rates and less carbon storage (Orwin et al. 2010). Indeed, this fits with findings by Ibañez et al. (2020) in grasslands in the Pyrenees, where higher legume proportions are associated with lower yield. Moreover, at high legume proportions, the additional N supply provided by N2 fixation may be inhibited because of the reduction of plant competition for soil N (Nyfeler et al. 2011), hence the mechanisms which may enhance SOC stocks at low legume proportions could be inhibited at high legume proportions.
Additionally, our model pointed out that the SOC levels were enhanced in grass-dominated grasslands (Figs. 1 & 2). This is an interesting finding since synergetic effects between grasses and legumes on SOC have been described for C4 grasses and not for C3 grasses (Fornara & Tilman 2008; Yang et al. 2019), which is the main type of grass in the Pyrenees (Still et al. 2003). However, positive synergistic effects between grasses and legumes affecting N fixation and yield are well known (Kirwan et al. 2007; Nyfeler et al. 2011; Rasmussen et al. 2012; Schipanski & Drinkwater 2012; Ribaset al. 2015; Suter et al. 2015). Pirhofer-Walzl et al. (2012) suggested that grasses could be especially good receivers of legume-derived Ndue to their fibrous root systems, which provide larger root surface and superficiality to grasses. Additionally, N2 fixation activity of legumes could be promoted by an increased demand for soil N on the whole ecosystem, as a consequence of an enhancement of grass root systems driven by the supply of atmospheric N2 fixed by legumes and transferred by litter, dead roots and exudates (Nyfeler et al. 2011). That enhancement of grass root systems also would allow grasses to acquire more N from non-symbiotic sources (Nyfeler et al. 2011; Suter et al.2015). Other mechanisms like optimization of the non-symbiotic N consumption by spatial (different root depths) and temporal (different growing seasons) niche partitioning could be behind legume-grass synergy on SOC (Van Ruijven & Berendse 2005; Mueller et al. 2013). Grasses could also enhance SOC stocks since they are more adapted and tolerate grazing in a higher degree than forbs because of their evolutionary history linked to big ungulates (Coughenour 1985), and consequently they may present high regrowth rates and productivity in grasslands (Ganjurjav et al. 2019).
Most of the previous studies addressing plant guild effects on SOC of grasslands where carried out at local scales and/or employing experimental assemblages (Fornara & Tilman 2008; Prommer et al.2019), not natural ecosystems as considered here. Moreover, most of these studies only considered the effect of plant guild richness and the presence or absence of the different plant guilds on SOC (Lange et al. 2015; Wu et al. 2017), although guild proportions effects have been described for other ecosystem functions like yield (Kirwanet al. 2007; Finn et al. 2013; Ribas et al. 2015). In contrast, what our model suggests is that: plant guild effects vary depending on their mass proportion in plant communities; some guilds, even in small proportions, can greatly enhance ecosystem function; those effects can be reversed at increasing proportions; and those observed effects stand over a wide range of grasslands and environmental conditions at regional scale. Hence, in addition to direct effects, shifts in plant guilds will have clear consequences on the rest of the biota (i.e. other plants, microbes), triggering a set of cascading effects in the ecosystem (Loranger-Merciris et al. 2006; Cornwellet al. 2008; De Deyn et al. 2008). We postulate that legume proportion determines the predominance of some processes over others, leading to SOC accumulation at moderate legume proportions and to SOC depletion at high legume proportions.
Our results are also relevant for functional ecology since they illustrate the power and usefulness of plant guilds to study keystone effects and interactions in ecosystems. In the last decades PFTs, or plant guild approaches have been described as inferior methods in comparison with continuous trait indexes (Mason et al. 2005; Funket al. 2017). However, these last methods are focused on testing concrete hypothesis, like the mass-ratio (Grime 1998; Diaz et al.2007) or niche complementarity (Villéger et al. 2008; de Belloet al. 2016) hypotheses. Conversely, modelling of plant guild effects using DIM allowed us to detect critical functional phenomena like keystone legume effects (Spehn et al. 2002) or guild interactions effects (Fry et al. 2014) on soil organic carbon storage. Legume effects in this study worked contrary to the mass-ratio hypothesis. Indeed, small proportions of legumes produced high effects on ecosystem function, which disappeared at high legume proportions.
To conclude, our DIM revealed that SOC storage in the Pyrenees not only depends on regional, landscape and local scale factors including climate and topography but on the contribution of the different plant functional guilds and interactions between the guilds, although we did not found interactions between plant guilds and other environmental predictors. In particular, legumes had a complex effect as they enhanced SOC stocks at low proportions and minimize them at high proportions. The magnitude of those effects depended on the relative composition of other guilds in the grassland. Legume and grass proportions had interactive effects on SOC. SOC maximums were found at low-moderate legume proportions in grass-dominated grasslands. Different effects of the ability of legumes for fixing atmospheric N and their high nutrient acquisition strategy probably were behind this pattern. Our results stress the importance of the keystone role of the N2 fixation rate of legumes on SOC stocks in natural grasslands and provide a strong argument for species diversity conservation efforts under climate change conditions. In addition, our findings can facilitate the elaboration of regional and local strategies to ameliorate the soil capacity to absorb carbon (Conant et al., 2017), contributing to the global effort to preserve terrestrial carbon pools. However, natural grasslands are not as tractable as agronomic systems, and direct management of plant guilds is not possible. Hence, further research about how herbivores affect plant guilds at broad scales is needed, as it is possibly the only way to manage plant guild proportions in natural grasslands.