全球变化知识资源中心

Old soil carbon (C) respired to the atmosphere as a result of permafrost thaw has the potential to become a large positive feedback to climate change. As permafrost thaws, quantifying old soil contributions to ecosystem respiration (R_eco) and understanding how these contributions change with warming is necessary to estimate the size of this positive feedback. We used naturally occurring C isotopes (δ¹³C and Δ¹⁴C) to partition R_eco into plant, young soil and old soil sources in a subarctic air and soil warming experiment over three years. We found that old soil contributions to R_eco increased with soil temperature and R_eco flux. However, the increase in the soil warming treatment was smaller than expected because experimentally warming the soils increased plant contributions to R_eco by 30%. On the basis of these data, an increase in mean annual temperature from −5 to 0 °C will increase old soil C losses from moist acidic tundra by 35–55 g C m⁻² during the growing season. The largest losses will probably occur where the plant response to warming is minimal.

Soils and sediments in the northern circumpolar permafrost zone store two times the amount of C as is in our atmosphere (~1,672 Pg; ref. 1) because frozen soil conditions have protected organic C from decomposition for hundreds to thousands of years^{2, 3}. The temperature increases (5–9 °C) predicted for high northern latitudes over the next century⁴ will make much of the world’s permafrost vulnerable to thaw. Models predict that 6–29% of permafrost will be lost for each degree of warming⁵. Model simulations also indicate that permafrost thaw has already occurred due to active layer thickening at a rate of 10 cm per decade in some areas⁶; such thaw has been documented in Alaska⁷, Greenland⁸ and Siberia⁹. Concurrent with permafrost thaw is C loss as these previously frozen soils are exposed to increased microbial activity. To better predict the size of the permafrost C feedback, the vulnerability of old soil C to thaw and warming needs to be quantified¹⁰.

Globally, respiration increases with greater mean annual temperatures¹¹. Numerous warming experiments in permafrost ecosystems have found R_eco increases with warming^{12, 13, 14, 15, 16}. Despite many studies on the response of R_eco to warming, it remains unclear how much of that response is due to increased microbial activity, which leads to losses of the critical old soil C pool. Both plant (autotrophic; R_A) and microbial (heterotrophic; R_H) respiration increase with warming^{13, 17, 18}; thus old soil C contributions to R_eco cannot be quantified by measuring land–atmosphere fluxes alone^{19, 20}.

Whether R_A or R_H drives the increase in R_eco determines whether or not a large positive feedback to climate change is occurring or has the potential to occur. Increased plant respiration is not a positive feedback to climate change because, on the timescales over which climate change is occurring, R_A is generally balanced by production, whereas R_H is not. This imbalance is extreme in permafrost ecosystems where soil C has been accumulating since the start of the Holocene epoch², or even during the Pleistocene, as in unglaciated regions of Siberia and Alaska²¹. The potential effect of increased R_H on atmospheric CO₂ levels is much greater than that of increased R_A because the soil C pool is orders of magnitude larger than the plant C pool in permafrost ecosystems.

Natural tracers partition R_eco into auto- and heterotrophic sources with minimal disturbance to the ecosystem. δ¹³C or Δ¹⁴C have often been used separately to estimate source contributions to respiration^{14, 22, 23, 24}. However, using both C isotopes is more powerful because it allows R_eco to be partitioned into more sources more accurately than with a single isotope²⁵. Natural abundance δ¹³C and Δ¹⁴C separate sources based on different principles. δ¹³C separates respiration sources by means of biological fractionation²⁶, and Δ¹⁴C separates respiration sources by age—on millennial timescales as a result of radioactive decay, and on annual to decadal timescales using the bomb enrichment of atmospheric ¹⁴C (ref. 24).

Here, we used natural abundance δ¹³C and Δ¹⁴C to partition R_eco at CiPEHR (Carbon in Permafrost Experimental Heating Research), a warming experiment in Alaskan subarctic tundra^{15, 27}. CiPEHR is unique among tundra warming experiments because it warms deep soil, causing permafrost thaw without confounding effects such as altered growing season length or water inputs. CiPEHR uses snow fences (with the excess snow removed each spring) to insulate soils during the winter (soil warming), which results in soils that are 1.5 °C warmer than the control during the growing season and increases thaw depths by 10% (ref. 15). Open-top chambers warm air in the summer (air warming) by about 1 °C, but do not affect soil temperatures¹⁵. After three years of experimental soil warming at CiPEHR, R_eco increased up to 57% relative to controls¹⁵. To investigate the contribution of old soil C loss to this R_eco increase, we partitioned R_eco into autotrophic (both aboveground and belowground plant structures), young surface soil (post-1963 bomb peak; 0–15 cm), and old deeper soil (15–75 cm) sources. This is the first study to present process-level relationships of how R_eco sources respond to soil temperature and experimental warming in a permafrost ecosystem.

Ecosystem respiration Δ¹⁴C became significantly more depleted with increases in R_eco (Fig. 1 and Table 1) and in depth-averaged soil temperature (hereafter called soil temperature; Fig. 2a and Table 1). The depletion in Δ¹⁴C with increased temperature was more pronounced in the control than in the experimental warming treatments (Fig. 2a and Table 1). Ecosystem respiration Δ¹⁴C was not affected by water table depth or soil moisture in our moist tundra; these variables were not significant predictors in the regression model. In contrast, wetting events caused a significant enrichment in soil Δ¹⁴CO₂ in semi-desert tundra²⁸.

The points are the data and the solid line is the prediction of the linear regression (see Table 1).

Overall, experimentally warming tundra caused an increase in R_eco that is driven by both plant and old soil respiration. The overall increase was driven more by autotrophs than heterotrophs, as demonstrated by the ratio of R_A to R_H, which is greater in the warming treatments than in the control (Table 1). The increase in autotrophic respiration is because plants are fixing more C as a result of the warming treatment such that the warming treatments are at present a growing season sink of 102 g C m⁻² (ref. 15). This increase in plant productivity is masking the critical loss of old soil C that has the potential to become a long-term feedback to climate change. It is only by using isotopes that we have been able to detect and quantify this old soil C loss.

At present, the growing season loss of old soil C in the soil warming treatments is being compensated for by C taken up by increased plant growth and stored in biomass because the site is a growing season C sink¹⁵. However, because the soil C pool (60 kg m⁻²; ref. 2) at our site is several orders of magnitude larger than the plant C pool (0.45–0.63 kg m⁻²; ref. 32), old soil C losses have the potential to eventually surpass gains in plant biomass. Much depends on how future plant productivity will respond to rising temperatures, changing soil moisture, and increased atmospheric CO₂. One prediction for mesic tundra found that plant production would continue to outpace R_H for the next century⁴⁰. However, at our experiment, the tipping point may have already been reached. This study examined only growing season R_eco dynamics, but soil warming increased winter R_eco by 50%, offsetting increased growing season C uptake¹⁵. Wintertime respiration increases are probably driven by old soil C losses because deep soils freeze last in the autumn, remain warmer than surface soils during the winter, and may form taliks (a layer of unfrozen soil between the permafrost and seasonally frozen soil)²². Old soil C losses due to warmer soils will probably continue into the winter and could lead to a positive climate change feedback even when the growing season C sink is increasing.

Site description.

CiPEHR is located near Eight Mile Lake (EML, 63° 52′ 59′′ N, 149° 13′ 32′′ W) in Healy, Alaska. The vegetation consists of moist acidic tussock tundra dominated by Eriophorum vaginatum and includes the graminoid Carex bigelowii, dwarf shrubs Vaccinium uliginosum, V. vitis-idaea, Betula nana, Rhododendron subarcticum, Rubus chamaemorus and Empetrum nigrum, and various mosses and lichens³². The soils are histels and consist of 0.3–0.5 m of organic soil atop a mixture of mineral loess deposits and glacial till. Mean annual temperature is −2.3 °C and the active layer thaws to about 60 cm deep during the growing season¹⁵. Permafrost temperatures in the area are monitored via a borehole and have been increasing over the past several decades⁷.

The soil warming (SW) and air warming (AW) treatments are set up in a factorial design: control, SW, AW and soil + air warming (SW + AW). In previous papers^{15, 27, 32}, soil warming was referred to as winter warming and air warming was referred to as summer warming. The air warming was achieved passively with open-top chambers (OTCs, 60 × 60 cm), and the soil warming was achieved with snowfences (8 m long by 1.5 m tall). The snowfences created snowdrifts over one metre deep, extending 10 m back from the fence that insulated soils throughout the winter⁴¹. At the end of each winter, excess snow was shovelled off the SW treatment so as not to add additional water or delay snowmelt. There were six replicate snowfences distributed evenly among three blocks. The SW treatment and SW control plots were the north and south sides of each fence, respectively. Each SW treatment and control plot contained both AW treatment and AW control plots.

Soil environment.

In all plots, the soil temperature at three depths (5, 20, 40 cm) and the soil volumetric water content (VWC) averaged over the top 20 cm were recorded every half hour during the growing season¹⁵. Soil temperature was measured using constantan-copper thermocouples, and VWC was measured using site-calibrated Campbell CS616 water content reflectometer probes. Thaw depth (the depth from the soil surface to the frozen soil) was sampled at three points in each plot using a metal probe pushed into the soil until it hit frozen ground. Throughout the growing season, water table depth (WTD) was measured three times per week from water wells installed in each SW treatment and control plot (12 in total; ref. 15).

Ecosystem respiration.

To measure the δ¹³C and Δ¹⁴C of R_eco, we installed 24 PVC collars (25.4 cm diameter  ×  10 cm deep) 6–7 cm into the soil, one per each of the four SW/AW treatment combinations at each of the six fences. We used previously published methods to sample R_ecoδ¹³C and Δ¹⁴C (ref. 20). In brief, 10 l dark chambers were fitted onto the collars encompassing aboveground biomass. Ecosystem respiration δ¹³C was analysed using the Keeling plot method, wherein CO₂ was collected into septa-capped glass vials (Exetainers, Labco Limited) every 2–3 min for a total of seven samples per collar, while an infrared gas analyzer (Licor-820, LICOR) simultaneously measured p_CO2. The samples and a set of field standards (−10.46‰; Oztech Trading Corporation) with similar p_CO2 were sent back to the University of Florida to be run on a GasBench II coupled to a Finnigan Delta Plus XL stable isotope ratio mass spectrometer (precision ±0.2‰, n = 215). δ¹³C changes due to travel and holding time were corrected using the field standards. R_ecoΔ¹⁴C was collected by pumping CO₂ through a molecular sieve (Alltech 13x, Alltech Associates) for 15 min. Before the collection, the chamber headspace was scrubbed for 45 min with soda lime while maintaining ambient p_CO2 to remove atmospheric contamination. The molecular sieves were baked at 625 °C to desorb CO₂ (ref. 42), which was purified using liquid N₂ on a vacuum line and reduced to graphite by Fe reduction in H₂ (ref. 43). The graphite was sent to the UC Irvine W.M. Keck Carbon Cycle Accelerator Mass Spectrometry (AMS) Laboratory for radiocarbon analysis (precision ±2.3‰, n = 102). Δ¹⁴C data were reported at the same δ¹³C value to correct for mass-dependent fractionation effects. Δ¹⁴C data were corrected for atmospheric contamination using each chamber’s δ¹³C data in a two-pool (atmospheric and R_eco) mixing model²⁰. R_eco isotopes were sampled at the beginning of July and in mid-September 2009, and in mid-August 2010 and 2011. Sampling occurred only under calm conditions and in the morning to control for potential diurnal variation. R_eco flux was measured from autochambers (60 × 60 cm each) adjacent to the PVC collars every 1.5 h throughout the growing season with a Licor-820 infrared gas analyzer (LICOR)^{15, 27}.

Source respiration.

Short-term incubations were used to measure the δ¹³C and Δ¹⁴C of autotrophic and heterotrophic source respiration²⁰. We collected aboveground (AG) and belowground (BG) plant material to measure R_A isotopes by cutting 5 × 10 cm blocks of tundra down to the frozen soil from each treatment at each fence. Samples from the same treatment were combined by block, for a total of three AG and three BG replicates per treatment. We clipped all live AG plant material from each block and placed it in foil-covered Mason jars (0.24 l). Belowground roots and rhizomes (>1 mm diameter) were collected from the thawed soil, rinsed, and shaken dry before being put into t