Discussion
Using an isotopic ingrowth core technique, we quantified root-derived soil C accumulation (in the bulk soil and MAOM fraction) and root production across gradients of ECM-associated tree dominance in six temperate forests. We found that root-derived C accumulation does not mirror root production patterns, with greater root-derived C in AM-dominated plots yet greater root production in ECM-dominated plots (Fig. 2). We also recovered more root-derived inputs in the MAOM fraction in AM compared to ECM plots (Fig. 4). Finally, our results highlight the impressive magnitude of root-derived C inputs (199.5 ± 14.7 g C m-2 y-1), emphasizing the importance of adequately characterizing this plant-to-soil C flux in order to understand how tree community composition influences ecosystem C cycling (Fig. 3).
Given that root production did not predict root-derived soil C in our study, and that fungal production is typically greater in ECM plots (Clemmensen et al. 2013; Cheeke et al. 2017), greater root and/or fungal production most likely do not explain the greater root-derived C accumulation in AM plots. Importantly, our measure of root-derived soil C accounts for all root and fungal inputs plus rhizodeposition that persisted in soil after two years. Data quantifying rhizodeposition are less abundant, but previous work has found this flux can be greater in ECM stands (Yin et al. 2014) or similar between mycorrhizal types (Keller & Phillips 2019b). Finally, our two-pool mixing model accounts for changes in soil C within each core. This minimizes the effects of mycorrhizal type differences in priming on our estimates of root-derived C. Thus, it is unlikely that the observed variation in root-derived soil C accumulation across the mycorrhizal gradient is principally driven by mycorrhizal type differences in root production or rhizodeposition.
Instead, mycorrhizal type differences in root and fungal turnover between AM and ECM trees may contribute to the greater root-derived C in AM plots. AM plant tissues tend to decay more quickly than those of ECM plants (Keller & Phillips 2019a; See et al. 2019), and faster turnover rates of these tissues could result in greater total root-derived C inputs in AM plots when measured over multiple phenological cycles. Mycorrhizal type differences in root turnover may also explain, to some degree, the lack of relationship between root-derived C accumulation and root production (quantified in this study as root biomass recovered from ingrowth cores after two growing seasons). Likewise, turnover of AM fungi can exceed that of ECM fungi by an order of magnitude (Staddon et al. 2003; Tedersoo & Bahram 2019). Differences in fungal turnover rates between mycorrhizal types may be particularly important in driving root-derived soil C accumulation as fungal inputs to soil C have been shown to exceed that of both leaf and root litter (Godbold et al. 2006).
The greater recovery of root-derived C in AM soils also reflects mycorrhizal-associated differences in soil organic matter formation pathways. Whereas plant inputs to ECM soils tend to accumulate in organic horizons or particulate C pools, there is increasing evidence that AM systems transfer greater amounts of plant-derived C into mineral-associated forms (Cotrufo et al. 2019) and our results support this idea (Fig. 4). Faster decomposition of AM inputs leads to more microbial products which are important MAOM precursors (Cotrufoet al. 2013). To the extent that MAOM cycles slowly and protects C from microbial decomposers (Grandy and Neff 2008, Bradford et al. 2013, but see Jilling et al. 2018), this could explain the greater root-derived C accumulation in both the bulk soil and MAOM fraction in AM compared to ECM soils. Our ingrowth core method did control for edaphic differences (cores were filled with a uniform soil matrix across all plots and sites) and thus differences in microbial and soil C cycling dynamics driven by edaphic factors were minimized. However, soil C cycling is also driven by distinct plant and microbial traits which can promote (or reduce) soil C aggregation and stabilization (Cheng & Kuzyakov 2005; Schmidt et al. 2011). For example, AM fungi are known to produce an aggregate-promoting glycoprotein (Rillig 2004), while ECM fungi have greater oxidative enzyme capacity to destabilize organic matter (Shah et al. 2016). In this way, the observed inverse relationship between root-derived soil C and ECM dominance (Fig. 1a) may reflect differences between mycorrhizal types in their input chemistry, and in their capacity to destabilize soil organic matter to acquire nutrients.
While root-derived soil C accumulation in forests has been poorly quantified to date, our estimates are similar in magnitude to previous studies using the isotopic ingrowth core technique. Across all plots in our study, root-derived soil C averaged 199 g m-2y-1 (ranging from 59 to 500 g m-2y-1). In an ECM-dominated 130-year old forest, Martinez et al. (2016) estimated root-derived C inputs to be 303 g m-2 y-1, and Panzacchi et al. (2016) reported a similar rate (309 g m-2y-1) in a young mixed hardwood plantation. Moreover, mean root-derived C inputs were estimated to be ~40% and ~70% of ANPP at these sites, respectively. Across our six temperate forest sites, we found root-derived C to range from 74% (HF) to 157% (SERC) of ANPP. While our ANPP estimates (SI Fig. 1) may be conservative given that we excluded both small, understory trees (i.e. < 10 cm diameter, including individuals that grew into the >10 cm diameter class between measurement periods) and trees that died between measurement periods, annual root-derived C accumulation is still larger on average than leaf litter flux (Fig. 3; SI Fig. 2). This highlights the relative importance of belowground C inputs to soils and suggests that models that presume the primacy of leaf litter fluxes are drivers of soil C dynamics may be underestimating the importance of root dynamics.
Overall, our results suggest the magnitude of root-derived soil C inputs is large and can vary significantly across sites and mycorrhizal types. Importantly, we show direct evidence of distinct plant mycorrhizal type effects on soil C formation. Accurate predictions of ecosystem C cycling in ecosystem and land surface models depend on improved quantification of the belowground C flux from plants to soil C pools, and improved understanding of the factors that control soil C stabilization. Our results suggest that better estimates of root and fungal contributions to stable soil organic matter pools are clearly needed in order to better understand how plant species shifts affect ecosystem C cycling now and in the future.