In previous sections, we discussed the capacity of ECM and N2‐fixing plants to acquire additional N under eCO2, which feeds back on plant productivity. Both N‐acquisition through SOM decomposition (outputs) and productivity (inputs) affect soil C storage. Meta‐analyses show that, indeed, eCO2 increases belowground C inputs through enhanced fine‐root production by 44% (Dieleman et al ., 2010) and rhizodeposition by 37.9% (Nie et al ., 2013). Although greater inputs of root‐derived C may increase soil C storage, much of the C that is released to the soil can also stimulate microbes to accelerate SOM decay and N release via ‘priming effects’ (Cheng et al ., 2014; Finzi et al ., 2015). Indeed, meta‐analyses have shown that increases in soil C inputs under eCO2 are offset by loses (Hungate et al ., 2009; van Groenigen et al ., 2014). These studies, however, did not account for potential differential effects among symbiotic types. The quantification of priming effects has important implications on the magnitude of the terrestrial CO2 sink, but these effects are difficult to measure and model (Georgiou et al ., 2015).
How do changes in N availability under eCO2 affect soil C storage?
We found a pattern of changes in soil C storage across N‐acquisition strategies, with eCO2 generally stimulating soil C losses in ECM, and soil C storage in AM systems under low N availability. The marginally significant relationship between soil C storage and (Fig. 4; P =0.0503), however, highlights that other factors beyond are at play.
Enhanced N‐mining activity in ECM under eCO2 involves CO2 release through heterotrophic respiration, minimizing net accumulation of soil C with eCO2 (Fig. 4). For example, the large CO2 fertilization effect on ANPP in Duke FACE (ECM) (McCarthy et al ., 2010) was likely driven by increased allocation to ECM fungi (Drake et al ., 2011) and root exudation (Phillips et al ., 2011), which stimulated microbial activity and SOM decomposition (priming), increasing N availability to plants (see also Cheng et al ., 2014). This, however, was accompanied by increased soil respiration (Oishi et al ., 2014), reducing soil C content (Fig. 4). In the Populus tremuloides (ECM) community from the Aspen FACE experiment, eCO2 increased litter inputs, but also decreased soil C content (Fig. 4), suggesting strong stimulation in SOM decomposition (Talhelm et al ., 2014). Likewise in the Florida OTC experiment, eCO2 increased plant productivity of scrub oaks (ECM) under low N availability (Fig. 3) through enhanced N‐mineralization (Langley et al ., 2009), but the stimulation of SOM decomposition yielded no effect on C storage at the ecosystem level (Hungate et al ., 2013).
By contrast, several AM‐ecosystems under low N have shown limited eCO2‐effects on N‐mineralization and plant productivity, together with significant increases in soil C content. For example, the lack of a significant eCO2 effect on biomass after 10 yr in the Nevada Desert FACE (AM) (Newingham et al ., 2013) was accompanied by a significantly positive effect on soil C content (Evans et al ., 2014), with increased fungal activity (Jin & Evans, 2010), but not fine‐root inputs (Ferguson & Nowak, 2011) – suggesting Ctransfer as the main driver of this effect (Jin & Evans, 2010). The same pattern of smaller than average biomass responses but soil C accumulation was observed, for example, in an AM‐forest ecosystem at ORNL (Iversen et al ., 2012), an AM‐grassland ecosystem in Australia (Pendall et al ., 2011) and a shortgrass steppe in the US (Pendall & King, 2007), accompanied by a doubling in rhizodeposition (Pendall et al ., 2004).
Other AM ecosystems, however, do not follow this pattern. In the SwissFACE experiment, neither the AM grass Lolium perenne nor the N2‐fixer Trifolium repens showed an increase in soil C storage after 10 yr of eCO2 (van Kessel et al ., 2006), despite a positive effect on photosynthesis (Ainsworth et al ., 2003) and a lack of N‐mineralization and ANPP response under low N availability (Schneider et al ., 2004). eCO2 did not increase soil C content at GiFACE either (Lenhart et al ., 2016), but the presence of legumes may have contributed to an increase in the allocation of Ctransfer to N2‐fixation, rather than soil C stabilization, which would explain the strong increase in abundance of legume species from c . 1% at the beginning of the experiment to 10% in later years, together with an increasingly positive overall effect on plant biomass (Andresen et al ., 2017). A certain degree of CO2‐driven enhancement of N‐mineralization in grasslands might also follow from increased soil water (e.g. Pendall et al ., 2003).
Although there have been reports of AM plants accelerating litter decomposition under eCO2 (Cheng et al ., 2012), there is little evidence that AM plants can increase the decay of SOM under eCO2, particularly in low N soils. Thus, CO2‐induced priming effects in AM systems are likely to be more short‐lived relative to those occurring in ECM‐dominated ecosystems (Sulman et al ., 2017).
An intermediate situation might be found for N2‐fixers (Fig. 4), which can obtain (additional) N from the atmosphere. eCO2 generally increases growth in legumes (Fig. 3; Ainsworth & Long, 2005), and thus likely also enhances soil C inputs, but whether SOM decomposition offsets additional inputs is uncertain. For example, eCO2 increased C inputs through biomass and productivity (Fig. 3) in a grassland FACE experiment with N2‐fixers in New Zealand. But eCO2 also increased N‐mineralization (Rütting et al ., 2010) and N availability (Newton et al ., 2010), yielding a modest increase in soil C storage (Ross et al ., 2013; Fig. 4). Various factors are probably at play to determine the balance between inputs and outputs, including species composition, litter quality, climate and nutrient and water availability.
The eCO2 effects on soil C under high N availability do not appear to follow a clear pattern in this dataset (Fig. 4). Meta‐analyses show that N‐fertilization may increase the positive effects of eCO2 on soil respiration further (Zhou et al ., 2016), but the effect of N has been shown to be negative in trees (Janssens et al ., 2010), and positive in grasslands and croplands (Zhou et al ., 2014). Whether this variability indicate different effects of N‐fertilization among N‐acquisition strategies or plant functional types remains to be disentangled.
These differences in the sign and magnitude of the effects of eCO2 on N‐mineralization, priming and soil C storage across symbiotic types might explain the large variability and nonsignificance of these effects found in several meta‐analyses (de Graaff et al ., 2006; Hungate et al ., 2009; van Groenigen et al ., 2014). The reasons for these different patterns among symbiotic types, however, remain elusive. Recent empirical observations and model analyses suggest that labile litter (low C : N) is quickly assimilated by microbes, and this microbial necromass contributes to the formation of stable SOM in greater proportion than recalcitrant litter (high C : N), which decomposes slowly (Knicker, 2011; Castellano et al ., 2015; Cotrufo et al ., 2015). On the other hand, the stabilization of labile litter in SOM should protect plant material, constraining the eCO2‐driven priming effect (Sulman et al ., 2014, 2017). Thus, recalcitrant litter should be more easily primed provided that it is ‘unprotected’. A recent meta‐analysis showed that, overall, AM trees produce litter that is significantly more labile than ECM trees (Lin et al ., 2017). Therefore, AM litter may be more easily stabilized by microbes, protecting new C from priming, whereas recalcitrant ECM litter may be more susceptible to priming, stimulating N‐mineralization and N availability. This would explain the limited CO2‐driven priming observed in some AM experiments, together with increased soil C content in AM‐low N systems.
Evidence from eCO2 experiments suggests that mycorrhizal status plays a key role in determining the sign of the eCO2 effect on soil C storage. Under low N availability, some AM‐ and ECM‐dominated ecosystems show opposite patterns. In some AM‐dominated ecosystems, eCO2‐driven priming is more limited than in ECM‐dominated ecosystems, which results in lower C losses in the former. By contrast, many ECM systems show strong priming effect and N‐acquisition in response to eCO2. This mechanism, however, enhances SOM decomposition and may thus partially offset the increase in biomass storage and limit CO2 sequestration at the ecosystem level. The result is a C‐allocation shift in AM vs ECM ecosystems, which may result in enhanced soil‐C gains in AM and enhanced biomass‐C gains in ECM. It is, however, the final balance between the (changes in) C inputs and outputs that eventually determines whether soil C storage increases, decreases or remains unaltered.