How has soil carbon stock changed over recent decades?

Soil is the largest stock of carbon (C) in the terrestrial biosphere, so even slight changes in soil C stock may induce significant fluctuations in the atmospheric C dioxide (CO2) concentration. Early coupled C-climate models predicted that positive C-climate feedback would be triggered due to the acceleration of C release to the atmosphere under future climate warming (Cox et al., 2000). However, due to the omission of key microbial components and biogeochemical mechanisms in these models (Wieder et al., 2013), these predictions remain controversial, because soil C dynamics is still highly uncertain among results simulated by 11 Earth system models (ESMs) involved in CMIP5 (Ciais et al., 2013). Likewise, experimental evidence is also contradictory, revealing increasing, decreasing, or nonsignificant changes among individual experiments (Lu et al., 2013). Given the very mixed results from both modelling and experimental studies, we present a global synthesis of soil C changes to evaluate a central tendency. To gain a comprehensive understanding of current soil C dynamics, we compiled actual measurements from nearly 100 individual cases without land-use changes (Table S1). To increase data comparability, the original measurements were converted to soil C content in the top 15 cm using the depth function developed by Jobbágy & Jackson (2000). We then calculated the relative rate of soil C changes to make comparisons among various studies with different response variables (i.e. either soil C concentration or soil C density). Our results revealed that soil C content exhibited an overall increase across the synthesized studies, with an average accumulation rate of 0.19% yr−1 (Fig. 1). However, the rate of soil C changes showed large variability among individual cases, ranging from −1.67% yr−1 to 4.09% yr−1. Soil C content significantly increased in 43.1%, was stable in 31.4%, and decreased in 25.5% of all the synthesized cases, respectively. In terms of ecosystem types, significant increases occurred in forests and grasslands, but slight decreases were observed in croplands (Fig. 2a–c). Soil C content tended to increase across 50.0% of forest studies and 48.1% of grassland studies, respectively. The C accumulation in forest soils could be associated with enhanced biomass C sequestration under environmental changes (elevated CO2, nitrogen deposition, and climate warming) and forest management (reduced harvest and increased afforestation) (Pan et al., 2011). The C enrichment in grassland soils may be due to the higher manure inputs following an increase in cattle and sheep density over recent years (Meersmans et al., 2009). In contrast, soil C content exhibited a decreasing tendency across 45.7% of cropland studies. This reduction could be explained by the changes in land management including reduced application of farmyard/animal manure, increased residue removal by changing technology, and deeper ploughing due to increases in tractor power (Smith et al., 2007). When grouped by the methodology, a tendency towards higher significant soil C gain was observed in studies based on regional comparisons than in studies based on long-term experiments and repeated soil samplings (Fig. 2d–f). Such a discrepancy may be partly explained by the baseline effects, that is the larger the initial stock size is, the higher the soil C decrease or the lower the soil C increase (Bellamy et al., 2005). In this study, the initial stock size was estimated at 3.45 kg C m−2 among studies based on regional comparisons, significantly lower than that in studies on the basis of repeated soil samplings (4.74 kg C m−2) and long-term experiments (4.08 kg C m−2). In addition, the larger sample size involved in regional comparisons may also increase the possibility of detecting soil C changes. Soil C dynamics also depended on the time interval since the original assessment (Fig. 2g–i). The sampling interval of 20- to 30-yr period corresponded to the highest C accumulation rate of 0.46% yr−1. In contrast, stable soil C dynamics were observed among the majority studies over shorter sampling intervals. Shorter sampling intervals may be insufficient for detecting soil C changes. To our surprise, a limited C increment at a rate of 0.07% yr−1 also occurred over longer sampling intervals. This phenomenon could be explained by the larger number of studies available for croplands, but less for forests during longer sampling intervals than those during 20- to 30-yr sampling interval. The disproportionate variations in soil C dynamics between croplands and forests could then decrease the apparent rate of soil C accumulation over longer time periods. Taken together, these results demonstrate that most soils have experienced significant increases or neutral changes in C stock, suggesting that climate change has so far not resulted in soil C loss, and would not be expected to trigger strong positive C-climate feedback from terrestrial ecosystems. More importantly, the tendency for increasing C content in natural ecosystems, together with the tendency in managed ecosystems for a slight decrease, implies that changes in land management could obscure the prediction of soil C dynamics. To improve model prediction, more intensive measurements over longer sampling intervals are urgently needed, and models are needed to disentangle climatic and management-driven effects. This work was supported by the National Basic Research Program of China on Global Change (2014CB954001 and 2015CB954201), National Natural Science Foundation of China (31322011, 31400364 and 41371213), and the Thousand Young Talents Program. Appendix S1. Characteristics of individual studies included in this synthesis, including ln response ratios (lnR) and relative change rate (% yr−1) within each study. Appendix S2. The literature list from which soil organic carbon concentration or organic carbon density data during the two sampling periods were extracted. Please note: The publisher is not responsible for the content or functionality of any supporting information supplied by the authors. Any queries (other than missing content) should be directed to the corresponding author for the article.