全球变化知识资源中心

Soils of the Earth annually release ~60 Gt of carbon (C) to the atmosphere through soil-surface CO₂ efflux (F_S; ‘soil respiration’), dwarfing CO₂ emissions from fossil fuel combustion by a factor of seven⁶. This large C flux is approximately balanced by the flux of C entering soils through TBCF (the sum of C flux to below ground to support root production and respiration, root exudates, herbivory and symbionts) and litterfall⁷. Given the importance of SOC in the global C cycle, the effects of warming on the balance of inputs and losses will have a large impact on the net sink strength of the terrestrial biosphere². Efforts to quantify underlying processes, however, have been inadequate for projecting the effects of warming on terrestrial C balance⁴. For example, warming seems to be increasing global F_S (ref. 1), but how much, if any, of this increase is derived from accelerated SOC decomposition remains poorly quantified. Although many studies have documented short-term (annual to decadal) increases in SOC decomposition with warming, these responses often are ephemeral⁸, in part because of various acclimation processes including reduced substrate supply, microbial adjustments at cellular and community levels, and changes in litter and soil-C quality^{4, 8}. Extrapolating short-term results to long-term (centennial to millennial) responses is further complicated by observations that gross and net primary production also increase with warming^{9, 10, 11}, with a corresponding increase in the amount of C sent below ground by plants¹¹. Finally, SOC studies have failed to show changes in stock size with warming, with precipitation seeming to exert a much stronger influence on SOC storage than temperature¹².

To address these critical knowledge gaps, we tested two hypotheses on the potential response of SOC storage to long-term, whole-ecosystem warming. The first posits that warming increases the turnover rate for SOC, which drives the often-observed increase in F_S. This implies that the current capacity of the world’s forests to retain SOC will decline with warming if increased inputs do not keep pace with accelerated decomposition of SOC. Further, increased detrital production could stimulate SOC decomposition^{13, 14, 15} and accelerate net SOC loss. Our second hypothesis posits that warming-related increases in primary production drive higher F_S through elevated above-ground and below-ground carbon inputs¹¹, and their subsequent conversion to CO₂. With our second hypothesis, there need not be warming-driven increase in the turnover of older SOC as decomposition of the increased inputs can explain increased F_S. And although increased C inputs can stimulate SOC decomposition, thereby reducing storage, warmer temperatures can also accelerate processes of SOC formation¹⁶, with one potential outcome being no net change in SOC storage. These two hypotheses are conceptually straightforward, but tests have been lacking because of the logistical and technical difficulties associated with whole-stand warming and the tracking of below-ground C inputs. So far, results from artificial warming experiments, MAT gradient studies and ex situ incubation studies have been conflicting^{4, 17}.

We directly tested our hypotheses about the response of SOC storage to warming by using a whole-ecosystem study in tropical montane wet forest arrayed across a highly constrained 5.2 °C MAT gradient⁵. This MAT gradient represents a critical advance over previous gradient studies because the various factors that can affect ecosystem processes other than temperature are held constant⁵, including: soils (all Acrudoxic Hydrudands in four closely related soil series); parent material (all tephra-derived substrate of similar type and age); moisture (constant plant available soil moisture); vegetation (>85% of stand basal area across the MAT gradient is composed of one canopy and one mid-storey species); and long-term disturbance history (late-stage aggrading forests). To further constrain this gradient and minimize disturbance effects, we selected plots that represent maximum biomass for a given MAT (ref. 5).

We previously reported that F_S increased linearly and positively with MAT along this gradient⁵. Here we report that both TBCF and litterfall, representing most detrital C inputs to soil, also increase linearly and positively with MAT (Fig. 1), in line with cross-site global analyses of the response of TBCF to rising temperature¹¹. We then combined quantification of SOC stocks by depth (0–10 cm, 10–30 cm, 30–50 cm and 50–91.5 cm) with radiocarbon-based mean residence time (MRT) estimates for bulk SOC across the MAT gradient. Strikingly, radiocarbon-based estimates of MRT revealed no relationship between SOC MRT and MAT for any depth (Fig. 2a–d). Similarly, MAT had no effect on radiocarbon-based estimates of turnover for four SOC fractions (soluble, light, intermediate and heavy) in 0–10 cm depth soils, those soils most likely to show a response to MAT (Supplementary Information and Supplementary Figs A1 and A2). In addition, neither SOC stocks at any depth (Fig. 2e–h) nor SOC fraction size for surface soils (Supplementary Information and Supplementary Figs A1 and A2) varied with MAT. As radiocarbon-based turnover rate for bulk SOC is a strong predictor of the turnover rate for acid-insoluble SOC (Supplementary Information and Supplementary Fig. A3), we conclude that across our gradient, temperature has little detectable influence on turnover of bulk SOC, on the size and turnover of even labile SOC fractions with high MRT, or on the size and turnover of the most stable C fractions in mineral soil. Lending further support for our second hypothesis, SOC turnover estimated from stock and MRT measurements for bulk SOC represents <5% of F_S, or ~0.39 Mg C ha⁻¹ yr⁻¹. These numbers may underestimate the actual contribution of SOC decomposition to F_S because bulk SOC MRT may not accurately capture the dynamics of the more rapidly cycling SOC pools. Relying on fraction size and MRT for SOC in 0–10 cm soils, we calculated a total flux of 0.40 Mg C ha⁻¹ yr⁻¹. As this depth contributes >75% of SOC-derived F_S for the bulk SOC calculations, even a doubling of our estimated SOC flux translates to <10% of F_S being derived from decomposition of SOC that is older than 1 yr, in line with independent estimates for aggrading tropical forests¹⁸.

a–c, Annual flux rates for: litterfall (a), TBCF (b) and soil-surface CO₂ efflux (c) in tropical montane wet forest in Hawaii all showed strong linear increases with rising MAT (n = 9).

Site description.

This research took place on the eastern flank of Mauna Kea Volcano, Hawaii Island, within the Hawaii Experimental Tropical Forest and the Hakalau Forest National Wildlife Refuge. Forests are characterized as closed-canopy, Metrosideros polymorpha-dominated tropical montane wet forest. Soils are all Acrudoxic Hydrudands and are derived from tephra ash deposits from Mauna Kea volcanism. The underlying Pleistocene-aged flow is dominated by hawaiite and mugearite⁵. The nearly constant and old age (>10,000 yr) of SOC in the deepest soil layer supports a constant substrate age across the MAT gradient.

Plot selection.

To minimize disturbance history effects, repeat airborne light detection and ranging (LiDAR) measurements of forest structure were used to select seven sites at each of six target elevations, where each site represents the maximum above-ground biomass present at a given elevation. LiDAR-based information at a 1.12 m resolution was acquired with the Carnegie Airborne Observatory²⁸ (CAO) to quantify mean tree height across each elevation band specified on a single substrate type and age (Supplementary Information). For the two coolest sites, LiDAR data were not available, and so traditional inventory techniques were used to identify two high-biomass stands across a 4 km² area of forest growing on the appropriate geology and soils⁵.

Stock and flux measurements.

We measured F_S and litterfall in each of the nine 20 × 20 m plots located across a 5.2 °C MAT gradient (ref. 5). We used a mass-balance-based approach to estimate TBCF (ref. 18), which is defined as the annual total of C flux to below ground for the production and maintenance of roots, mycorrhizae and other symbionts, and C released as root exudates, herbivory or biomass turnover. As this C must be respired or stored, TBCF can be estimated as: F_S − litterfall + Δ [C_S + C_F + C_R], where C_S represents mineral soil C, C_F represents forest floor C and C_R represents live root C (ref. 18). We measured F_S monthly using previously described methods⁵. We measured litterfall monthly in 8 permanently installed 0.174 m⁻² collectors per plot, from which litter was collected, oven-dried and weighed using standard methods¹⁸. Both sets of flux measurements were conducted between April 2009 and March 2010. We assumed that annual change in soil C was negligible on the basis of our radiocarbon analyses and findings for adjacent but more disturbed sites¹⁸. We also assumed that erosion and leaching losses of C were minor components of TBCF at our sites¹⁸ and so were not measured. On the basis of previous results¹⁸, we assumed that 10% of TBCF was allocated to coarse root growth. Although relevant for TBCF accounting purposes, any errors associated with this assumption would have a minor influence on overall TBCF estimates¹⁸, and no effect on our SOC and F_S estimates because coarse root C is long-lived. From previous work on error distribution in TBCF calculations¹⁸, we anticipate that error propagation in calculating TBCF is negligible. Soil temperature and moisture were recorded at the location and time of measurement using temperature probes and loggers⁵. Detailed repeat measurements across MAT showed no diurnal variation in soil-surface CO₂ efflux, and so were not used to construct annual soil-surface CO₂ efflux budgets⁵.

We measured forest floor mass across the gradient to understand litterfall decomposition rates through collections of all recognizable plant material (litter layer C) at eight 0.174 m⁻² quadrats per plot. These samples were dried to constant weight and analysed for [C] (Costech Elemental Combustion System). Mineral-associated SOC (Mg C ha⁻¹) was estimated across the gradient in three cores per plot (or five cores if coefficient of variation was >25% based on the original 3 cores) to 91.5 cm using a 5.75-cm-diameter soil core with plastic sleeves, from which both C content as above and bulk density were determined¹⁸. We analysed bulk density and percentage of C from four depth increments (0–10 cm, 10–30 cm, 30–50 cm and 50–91.5 cm) to determine C stocks.

Soil samples were physically separated into soluble, light, intermediate and heavy fractions using a sequential density fractionation method relying on progressively denser solutions of sodium poly-tungstate to isolate soil C fractions²⁹. Sub-samples of bulk soil and soil fractions were ground to <150 μm mesh size for elemental and isotope analyses. Radiocarbon analyses of all bulk soils were completed at the Centre for Accelerator Mass Spectrometry at Lawrence Livermore National Laboratory and fraction-based radiocarbon analyses were completed at the ¹⁴CHRONO Centre for Climate, the Environment, and Chronology at Queen’s University Belfast (details provided in Supplementary Information). SRO Al concentration was determined by hydroxylamine hydrochloride hydrochloric acid extraction method combined with 16 h of shaking³⁰. Linear regression and diagnostic analyses for data conformance to assumptions were performed in SigmaPlot (Version 11.0).

1. Bond-Lamberty, B. & Thomson, A. Temperature-associated increases in the global soil respiration record. Nature 464, 579–582 (2010).
   • CAS
   • ISI
   • PubMed
   • Article
2. Davidson, E. & Janssens, I. Temperature sensitivity of soil carbon decomposition and feedbacks to climate change. Nature 440, 165–173 (2006).
   • CAS
   • PubMed
   • Article
3. Kuzyakov, Y. Sources of CO₂ efflux from soil and review of partitioning methods. Soil Biol. Biochem. 38, 425–448 (2006).
   • CAS
   • Article
   URL: