Feedback to climate warming from the carbon balance of terrestrial ecosystems depends critically on the temperature sensitivity of soil organic carbon (SOC) decomposition. Still, the temperature sensitivity is not known for the majority of the SOC, which is tens or hundreds of years old. This old fraction is paradoxically concluded to be more, less, or equally sensitive compared to the younger fraction. Here, we present results that explain these inconsistencies. We show that the temperature sensitivity of decomposition increases remarkably from the youngest annually cycling fraction (Q10 < 2) to a decadally cycling one (Q10 = 4.2–6.9) but decreases again to a centennially cycling fraction (Q10 = 2.4–2.8) in boreal forest soil. Compared to the method used for current global estimates (temperature sensitivity of all SOC equal to that of the total heterotrophic soil respiration), the soils studied will lose 30–45% more carbon in response to climate warming during the next few decades, if there is no change in carbon input. Carbon input, derivative of plant productivity, would have to increase by 100–120%, as compared to the earlier estimated 70–80%, in order to compensate for the accelerated decomposition.
Decomposition of organic material in soil is a process where soil organisms favor fresh, easily decomposable carbon compounds. A continuous inflow of new organic matter to soil results thus in a pool of soil organic carbon (SOC), which comprises a mixture of SOC age classes, with recalcitrant substances in the majority. The future carbon balance of terrestrial ecosystems is critically dependent on the effects of climate change on decomposition of this large, older SOC fraction (Cox et al. 2000, Friedlingstein et al. 2006).
We observed remarkable differences in the temperature sensitivity of decomposition between age classes of SOC in two typical podzol-type upland soils of boreal forests by measuring 14C activity of SOC fractions and heterotrophic soil respiration. In order to detect these differences, we had to incubate our soil samples for 18 months under favorable conditions to strengthen the decomposition signal from the older fractions sufficiently.
Materials and Methods
Individual 10-dm3 samples were taken from the organic and two mineral soil layers (0–15 and 15–30 cm) at two study sites located in southern Finland in 2005 (61°48′ N, 24°19′ E, 150–180 m above sea level, annual mean temperature 2.9°C, yearly precipitation 709 mm; see Kähkönen et al.  for other general information on study sites nearby ours). The podzolic soils studied had characteristically an approximately 10 cm thick organic layer (soil organic carbon [SOC] 45% m/m) of decaying litter on top of mineral soil (SOC 0.7–3.2%, mass/mass). Chemical characteristics of litter input to soil at the sites are described in the Appendix.
The soil samples were passed through a 2.8-mm (mineral soil) or 4-mm (organic layer, where the larger sieve was more practicable for the field-moist samples) sieve to remove larger roots. The samples were incubated in the laboratory at 25°C and 50% air humidity for 18 months correcting for moisture loss once a week. During this incubation period, respiration of our organic soil layer samples decreased by 84% or 85%, that of 0–15 cm mineral soil layer samples by 53% or 70%, and that of 15–30 cm mineral soil layer samples by 32% or 36% (see Appendix). These figures indicate a decreasing share of labile SOC toward a greater depth in the soils and a reduction in the amounts of these labile fractions during the incubation. Subsamples were put aside every threes months during the incubation period and analyzed for CO2 production rate and SOC content. SOC in the mineral soil samples was fractionated into particulate organic matter (POM; >63 μm, <1.85 g/cm3), mineral-associated organic matter (MOM; <63 μm), and sand-associated OM (>63 μm, >1.85 g/cm3) (Cambardella and Elliott 1992, Balesdent et al. 1998, Gale and Cambardella 2000). Also C dissolved in water during the fractionation process was measured (dissolved organic carbon [DOC], <0.45 μm). After the 18-month incubation at 25°C, 14C activity was measured from the CO2 produced at 8° and 25°C; the POM and MOM fractions and bulk organic layer samples were measured using accelerator mass spectroscopy (AMS).
CO2 was collected for the 14C measurements by incubating 500–2000 g of the soil samples, which had been incubated at 25°C for 18 months, in 5-L bottles first at 8°C and then at 25°C. The temperature range used is comparable to annual and sometimes diurnal variation in the surface layer of these soils (Kähkönen et al. 2002). The same samples were used at both temperatures to ensure that the quality of SOC was similar at each measurement temperature. The soil samples were allowed to settle at the measurement temperature for a week before starting the CO2 collection. The collection times were varied from a few days for the organic layer samples to over a month for the lower mineral soil layers in order to obtain at least 3 mg carbon for the measurements. The quality of SOC changed little as a result of decomposition during the CO2 collection; the amount of carbon collected represented 0.02–0.4% of the cumulative respiration during the entire 18-month incubation depending on soil sample. The CO2 was collected into a molecular sieve, released at 500°C temperature, purified, measured for δ13C value, and eventually graphitized. The AMS measurements were performed on the graphite targets.
To measure 14C in the SOC fractions, the bulk organic layer samples and the POM and MOM fractions of the mineral soil samples were combusted at 550°C for 10 h in closed ampoules with an excess of CuO. The produced CO2 was purified, measured for δ13C value and graphitized for the AMS measurements. All the AMS results were expressed as pMC (percentage modern carbon; Stuiver and Polach 1977). The term “activity” is used throughout the paper to describe the amount of radioactive 14C in the soil samples.
Soil respiration was modeled as R(T) = aebT, where R(T) is respiration at temperature T, a and b are parameters, and e is the base of the natural log, and its temperature sensitivity was estimated as Q10 = [R(25°C)/R(8°C)]10/17. The contributions of the younger and older SOC fractions to the CO2 produced at 8° and 25°C were calculated by soil layer using mass balance equations of respiration (Rtot= Ro+ Ry) and 14C activity (AtotRtot = AoRo + AyRy). The MOM and POM fractions were combined to form the older SOC fraction in mineral soil, because the mean residence times (MRTs) of these fractions were close to each other. The 14C activity (Ao) of this combined fraction was calculated as a mass-weighted mean. The 14C activity of the younger SOC fraction in mineral soil was calculated as
where Ro(25°C) was estimated from the mass loss of the combined POM and MOM fractions during the 18-month incubation (Appendix). To obtain a probability distribution of Ay (Appendix), we sampled the values of all the parameters in this equation from their probability distributions applying the following limitations: (1) Ay < 200 pMC (highest atmospheric 14C activity recorded), (2) Ay ≥ (25°C), (3) Ay ≥ 107 pMC (atmospheric activity in 2005), and (4) 0 < Ro < Rtot.
The 14C activity of the younger fraction in the organic soil layers was assumed to be the same as that of recent litter (Gaudinski et al. 2000). This activity was taken to be similar to the atmospheric activity three years before the soil sampling at the pine site and six or seven years before the sampling at the spruce site. Needles represent about one-third of the total litter input to the soils, and these delays account for different residence times of needles in pine and spruce trees at forest sites studied (Liski et al. 2006). The 14C activity of the older SOC fraction was assumed to be similar to the measurements of the bulk organic layer.
The contribution of the younger SOC fraction to the total CO2 production at temperatures (T) 8° and 25°C was calculated as
The temperature sensitivity of the younger SOC fraction was estimated as
Corresponding equations were used to determine the temperature sensitivity of the older SOC fraction.
where (t) is the measured or modeled 14C activity of a SOC fraction, Mi is the annual new SOC input, p is the year of soil sampling (2005), b is the year of starting the calculation (10 000 years BP in our case), is the atmospheric 14C activity in year i (Hua and Barbetti 2004, Levin and Kromer 2004, Reimer et al. 2004), and λ is the radioactive decay rate of 14C (1/8268 yr−1). We assumed that the productivity of the forest sites studied had not changed over time and thus applied a constant Mi. According to Eq. 4, this means that the values of (t) were not dependent on Mi. We applied a 3–4 year lag in the values at the pine site and a 6–7 year lag at the spruce site to account for a mean delay in litterfall compared to carbon fixation from the atmosphere. These lag values represent means of similar pine and spruce forests in southern Finland. They were calculated by weighing the turnover rates of litter components by their contributions to the total litter production (Liski et al. 2006). At the pine site, (1) foliage, (2) branches plus coarse roots, (3) fine roots, and (4) stem and (5) ground vegetation represented 21%, 9%, 45%, 5%, and 20% of the total litter production and had turnover rates equal to 6, 33, 1, 190, and 3 years, respectively. At the spruce site, these values were 20%, 8%, 49%, 6%, and 17% and 10, 80, 1, 370, and 3 years, respectively. When there were two solutions for the MRT values, we discarded one as unrealistically high or low considering CO2 production rates from these soils (our own laboratory measurements [see Appendix] or field measurements [Kähkönen et al. 2002]).
The reliability of the Q10, MRT, and estimates were assessed by conducting a Monte Carlo analysis using MATLAB (version 6.5, release 13; MathWorks, Natick, Massachusetts, USA). The probability densities of parameters a and b in the equation R(T) = aebT were sampled using a Markov chain Monte Carlo method (Hastings 1970). The other parameter values were drawn randomly from normal distributions with means and standard deviations determined from our measurements and measurement errors, respectively. These distributions show that our measurements were precise enough to detect significant differences in the Q10 values between the SOC fractions (Fig. 1). Consistency of the results between the two soils studied is another sign of an adequate measurement quality.
Results and Discussion
Qualitative differences in temperature sensitivity of decomposition between SOC fractions
We detected the temperature sensitivity of decomposition of the soil organic carbon (SOC) age classes by measuring the 14C activity of CO2 respired by the soil samples at 8° and 25°C after the 18-month incubation (Table 1). A difference in this activity between the temperatures revealed that various age classes of SOC had diverse temperature sensitivities, because 14C activity of SOC varies with age. SOC taken up from the atmosphere through photosynthesis before 1950 has activities lower than 100% modern carbon (pMC) because of radioactive decay. Younger SOC has higher activities as a consequence of nuclear bomb tests; the highest atmospheric pMC value was equal to 200 in 1964. The pMC value of recent SOC is near that of the present atmosphere: 107 pMC in 2005 when our soil samples were taken.
Table 1. Properties of soil organic carbon (SOC) at the two boreal forest sites located in southern Finland in 2005.
The 14C activities of CO2 respired by the organic layer samples were lower than the activities of the bulk SOC, but both were above the present atmospheric level (Table 1). The level of these activities means that both the bulk SOC and the decomposed SOC were dominantly of post-1950 origin. The lower activity of CO2 indicates that it originated mostly from decomposition of SOC age classes that were younger than the bulk SOC on average. In addition, the 14C activity of respired CO2 increased with the increased measurement temperature. This indicates a larger contribution of older, bomb-14C-enriched SOC to the respiration at the higher temperature: in other words, a higher temperature sensitivity of decomposition of the older, but still post-1950 fraction.
In the mineral soil layers, the 14C activities of mineral-associated organic matter (MOM) and particulate organic matter (POM) were lower than the activity of the present atmosphere, with the exception of the upper mineral soil layer at one of the sites (Table 1). As experimentally separated SOC fractions are always mixtures of different age classes (von Lützow et al. 2007), these activities, even the one slightly above the present atmospheric value, mean that each of these fractions contained a remarkable share of pre-1950 SOC (cf. e.g., Gaudinski et al. 2000). The 14C activities of CO2 respired by these samples were higher than the activities of POM or MOM. The level of these activities above the atmospheric level in 1950 indicates that a part of the respiration originated from post-1950 SOC. At the Scots pine site, the activities of CO2 were below the atmospheric level in 2005. This indicates that a part of the respiration came from pre-1950 SOC. At the Norway spruce site, the 14C activities of CO2, like all other 14C activities, were higher than at the Scots pine site. The higher level of the activities is explained by a longer life span of spruce needles (7–8 years) compared to pine needles (3–4 years) and consequently a higher bomb-14C-enrichment in litterfall. After accounting for this difference, a part of the respiration must have originated from pre-1950 SOC also at the Norway spruce site. At both sites and in each mineral soil layer, the 14C activity of CO2 was higher at the higher temperature. This means that the contribution of post-1950 SOC to respiration increased with the increased temperature. Thus, the decades old post-1950 SOC carbon was more temperature sensitive than the older pre-1950 SOC in the mineral soil layers.
These 14C activity measurements show that temperature sensitivity of decomposition was not similar across age classes of SOC carbon. The current models making this assumption give therefore biased estimates of the climate change effects on decomposition of SOC and consequent feedback to global warming. To estimate this bias at our study sites, we derived quantitative estimates for the temperature sensitivity of the SOC fractions based on our measurements.
Quantitative differences in temperature sensitivity of decomposition between SOC fractions
We calculated, first, the proportional contributions of the younger and older SOC fractions to CO2 production at 8° and 25°C in each soil layer. Then, to express the temperature sensitivity of decomposition of these fractions in terms of commonly used Q10 values (proportional growth in respiration for any 10°C increase in temperature), we assumed that respiration increased exponentially between these temperatures. In addition, we calculated mean residence times (MRT) of the fractions to characterize their mean stability. To describe uncertainty in these calculated results, we carried out a Monte Carlo analysis taking into account error in each affecting measurement and express the results as probability distributions. These distributions show that our measurements and laboratory analyses were precise enough to detect statistically significant differences in temperature sensitivity between the SOC fractions that had different MRT. It is worth noting that these error estimates do not account for sampling error.
Decomposition of the SOC fraction, which had MRT equal to about a decade, appeared highly temperature sensitive in both the organic layer and the 0–15 cm mineral soil layer (Fig. 1). The median Q10 estimates for this fraction ranged from 4.2 to 6.9. These values were substantially higher than those for the total heterotrophic respiration, which varied relatively little from 2.7 to 3.2 between all the soil layers. The decadally cycling SOC fraction was the most temperature sensitive one also in the 15–30 cm mineral soil layer (median Q10 from 3.1 to 3.7), but not quite as sensitive as in the upper layers. The oldest SOC fraction of the mineral soil layers (MRT = 140–490 years) and the youngest of the organic soil layers (MRT = 1–2 years) were least sensitive to temperature, with the median Q10 estimates between 2.4 and 2.8 for the former and <2 for the latter.
These results agree completely with a recent hypothesis about the temperature sensitivity of SOC decomposition (Davidson and Janssens 2006). Like this hypothesis suggests, the temperature sensitivity increased with an increasing MRT of the SOC fractions unless some other mechanism could start to prevent the fraction from decomposing according to its chemical characteristics. In the soils studied, the temperature sensitivity increased from the very youngest SOC fraction of the organic layer to the fraction, which had a MRT equal to about a decade independent of soil layer. However, in the deeper mineral soil layers, this decadally cycling fraction was somewhat less temperature sensitive than in the upper mineral soil and organic layers. A possible reason for this is a lack of labile substrates required for co-metabolic decomposition of recalcitrant materials such as lignin (Amelung et al. 1999, Fontaine et al. 2007). Most of the still older centennially cycling SOC fraction was not especially temperature sensitive. Most of this fraction was bonded to minerals (MOM), but a notable share was also found as POM (Table 1). Bonding of SOC to minerals is known to prevent decomposition (Six et al. 2002), and thus could have decreased the temperature sensitivity of this fraction. Another reason for the low temperature sensitivity of the centennially cycling fraction could be the lack of labile carbon substrates, which we expected also to have limited decomposition of the decadally cycling fraction in the deeper mineral soil layer. The relative importance of these mechanisms cannot be deduced from the data and require further studies. Nevertheless, an increased supply of labile substrates has been observed to accelerate decomposition of stable SOC fractions considerably (Fontaine et al. 2007).
Our measurements explain earlier influential results, which have been considered as paradoxical until now. Fang et al. (2005) concluded that temperature sensitivity of decomposition is similar for young and old SOC fractions because the temperature sensitivity of the total heterotrophic soil respiration did not change with decreased quality of SOC over their six-month incubation experiment. Our 14C measurements show that, despite relatively similar Q10 values for total heterotrophic soil respiration, different SOC fractions may still have remarkably different temperature sensitivities. Knorr et al. (2005) estimated on a theoretical basis the oldest SOC fractions to be highly temperature sensitive because these fractions have very complicated structures. However, most of the oldest fractions commonly occur associated with soil minerals, which can protect them from decomposition. Microbes in deeper mineral soil layers may also be so deprived that they cannot decompose recalcitrant materials without an addition of an easy energy source such as glucose or cellulose (Fontaine et al. 2007). These are possible reasons for why hundreds-of-years-old carbon found in mineral soil has not appeared to be particularly temperature sensitive in different soils (Liski et al. 1999, Giardina and Ryan 2000). Understanding that these earlier results are not contradictory at all is crucial to the development of our knowledge on the temperature sensitivity of SOC. It makes it possible to target future research more efficiently to the mechanisms that control the stability of SOC and to the changes in these controlling factors in response to changing climate, instead of sorting out the confusion about the earlier results.
Our results show that the total heterotrophic soil respiration gives a biased picture of the sensitivity of SOC decomposition to temperature changes. This is important because the current estimates for the effects of climate warming on SOC are based on such measurements (Cox et al. 2000, Jones et al. 2005).
Warming effects on SOC
Using the fraction-specific median Q10 values instead of the value of the total heterotrophic soil respiration (Fig. 1), assuming no change in carbon input and comparing steady states, the organic layers of our study sites would lose 30–45% more carbon in response to a predicted 5.1°C warming (Jylhä et al. 2004). This is because most of the SOC in the layer is actually much more temperature sensitive than the total heterotrophic respiration indicates (Fig. 1). The mineral soil layers, on the other hand, would lose 4–17% less, because decomposition of the major centennially cycling fraction is more tolerant of temperature. These estimates are sensitive to the division of the total SOC between the different temperature sensitivity fractions. Based on the loss of SOC during the first three months of our incubation experiment, we assumed that the annually cycling fraction represented 5% of the organic layer, whereas the rest of the material in the organic layers had MRT determined for the decadally cycling fraction (Table 1). Based on the loss of SOC during the entire 18 months of incubation, we approximated that 5–15% of the mineral soil carbon consisted of the decadally cycling fraction. These percentages are comparable to earlier estimates for similar soils nearby (Liski et al. 1998). During the next few decades, the response of these soils to climate warming will be dominated by the highly temperature-sensitive decadally cycling SOC fraction, whereas the more stable and less sensitive SOC pool of mineral soil will change more slowly.
If the SOC pools were to remain unchanged despite a 5.1°C warming, the annual heterotrophic respiration would increase by 100–120%, from 0.47 to 0.95 kg/m2 at the spruce site and from 0.36 to 0.79 kg/m2 at the pine site. If carbon input were to compensate for the accelerated decomposition, it would therefore need to increase by a similar factor. Based on the temperature sensitivity of the total heterotrophic soil respiration, one would incorrectly think that increases already by 70–80% would be adequate. The error is critical. The productivity of boreal forests is estimated to increase by 20–60% in response to a similar warming and a corresponding growth in atmospheric CO2 concentration (Bergh et al. 2003). Losses of soil carbon and a positive feedback to climate warming seem inevitable after accounting for the different temperature sensitivities of various SOC fractions. We estimated the current respiration rates by dividing the SOC fractions by their MRTs. The values are 20–27% smaller than estimates based on CO2-production measurements at the sites (Kähkönen et al. 2002), which is appropriate considering the differences in the approaches and the spatial variability inside the sites. The future respiration rates were estimated by applying the fraction-specific Q10 values.
This study points out a serious weakness in the current estimates of climate change effects on soil organic carbon (SOC). Applying the same temperature sensitivity of decomposition for all SOC fractions gives a substantially biased picture of future SOC cycle. The pool of decadally cycling SOC, which was highly temperature sensitive at our study sites, may be globally as large as the present atmospheric carbon pool (Jones et al. 2005). If this pool is commonly highly susceptible to warming, soil will provide a much stronger positive feedback to global warming than the current estimates indicate. To improve the reliability of future climate predictions, it is necessary, first, to study temperature sensitivity of SOC fractions in different soils and, second, to improve the soil carbon modules of Earth System Models according to the results.
This work was financially supported by the Academy of Finland (project 107253) and the Nessling foundation (project “Soil Carbon Cycling in Earth System Models”). We thank M. Sandberg and A.-M. Forss for technical assistance, R. Mäkipää and S. Palmroth for comments on the manuscript, and T. Kurtén for proofreading.
Description of litter chemical characteristics, SOC changes over the incubation period, respiration rate of the combined POM and MOM fraction, and the 14C activity of the younger SOC fraction in mineral soil (Ecological Archives E091-028-A1).