Delayed recovery of soil respiration after wetting of dry soil further reduces C losses from a Norway spruce forest soil



[1] This experiment investigated the effects of prolonged summer drought on soil respiration (SR) in a mountainous Norway spruce forest in south Germany. On three manipulation plots we excluded summer throughfall in the years of 2006/2007 and measured SR fluxes in comparison to three control plots. Using radiocarbon measurements we quantified the contribution of rhizosphere (RR) and heterotrophic respiration (HR) to total SR. In both manipulation years, mean CO2 emissions (±SE) from the throughfall exclusion (TE) plots were smaller than from the control plots with 5.7 t C ha−1 (±0.3) compared to 6.7 t C ha−1 (±0.2) in 2006 and 5.9 t C ha−1 (±0.3) compared to 7.0 t C ha−1 (±0.4) in 2007. Under control conditions, CO2 originated mainly from HR (60–95% of SR). Prolonged drought reduced HR, whereas RR was not affected or even increased slightly. Reduction of CO2 emissions on the TE plots was found up to 6 weeks after differences in matric potential conditions disappeared, possibly either because water repellency inhibited homogeneous rewetting of the organic horizons or because of severe damage to the microbial population. No evidence was found for the release of new, formerly protected substrates by preceding drought. Continuous measurements in 2008 (no manipulation) did not reveal increased CO2 emissions on the TE plots that could compensate for the reduction during the years 2006/2007. Based on our results, we postulate a negative feedback between increased frequency and magnitude of summer droughts and SR in Norway spruce stands.

1. Introduction

[2] Drought is one of the most common types of environmental stress that soil organisms experience. According to model simulations we face a globally increasing likelihood of severe drought periods in central Europe that will cause irregular and extreme water stress for soil organisms [Intergovernmental Panel on Climate Change (IPCC), 2007]. Soils, in turn, are globally important carbon (C) pools, containing more than twice as much C as vegetation or the atmosphere [Schlesinger and Andrews, 2000]. Soil respiration (SR) is the largest CO2 flux from the terrestrial biosphere to the atmosphere. Among the various ecosystems studied, boreal and temperate forests have been identified as particularly important due to their large C pools [Lal, 2003]. At present, many European forest ecosystems act as C sinks due to a fine balance between net primary production and respiration losses [Valentini et al., 2000]. Respiration has been identified as the main determinant of this C balance. This raises the question how changes in the global water cycle will affect soil respiration. The possibility that climate change is being reinforced by increased carbon dioxide emissions from soils due to altered boundary conditions emphasizes the necessity to improve our understanding of climate change feedbacks on soil carbon processes.

[3] The CO2 respired from soils originates from various sources. [cf. Kuzyakov, 2006]. Functionally, SR can be divided into respiration by autotrophs (plants) and by heterotrophs (microorganisms). However, the roots of most plants form intimate associations with myccorhizal fungi, supplying them with C in form of photosynthates [Högberg and Read, 2006]. Nonmyccorhizal microorganisms within the rhizosphere can also receive significant amounts of C from plants, mainly low molecular weight organic compounds [van Hees et al., 2005], that are deposited from roots into the rhizosphere, predominantly as root exudates [Nguyen, 2003]. Thus, the substrates fueling the respiration of these ‘heterotrophic’ microorganisms are inseparable from the substrates used by plant roots [Högberg and Read, 2006], and their respiration shows a close temporal coupling to tree canopy photosynthesis [Högberg et al., 2008]. Thus, SR components are often operationally defined as CO2 evolving from respiration of roots, myccorhizal fungi and root-associated microorganisms (summarized as rhizosphere respiration, hereafter RR), and CO2 evolving from the decomposition of soil organic matter (SOM) by decomposing soil microorganisms (called heterotrophic respiration, hereafter HR) [van Hees et al., 2005]. The external factors controlling these distinct sources of SR differ, so it can be expected that changing boundary conditions like, e.g., droughts can result in altered contributions of these sources to total SR: Whereas microorganisms, e.g., are dependent on water in their immediate surroundings, trees, even shallow rooting Picea abies (L.) Karst. can lift up water from deeper horizons and redistribute it [Nadezhdina et al., 2006].

[4] A variety of methods has been developed to separate the contribution of RR and HR to SR (reviewed, e.g., by Kuzyakov [2006]). Using radiocarbon (14C) produced during atmospheric testing of thermonuclear weapons during the early 1960s as a ‘conservative’ tracer is one method that has been applied repeatedly in the past [Dörr and Münnich, 1986; Gaudinski et al., 2000; Trumbore, 2000; Schuur and Trumbore, 2006]. Before 1950, the atmospheric abundance of 14C was relatively constant for hundreds to thousands of years (“prebomb 14C”). Atmospheric testing of nuclear weapons nearly doubled the atmospheric abundance of 14C in the 1960s (‘bomb-14C’). After atmospheric test were banned in 1964, atmospheric 14C levels decreased due to uptake of excess 14C into oceanic and terrestrial pools and due to emission of 14C-free CO2 from the combustion of fossil fuels. The radiocarbon signature (Δ14C) of the atmosphere propagates into plant biomass via photosynthetic assimilation of atmospheric CO2, thus effectively labeling each years biomass with a specific isotopic signature. The annual rate of decline during the last decades was greater than the precision of 14C measurements, so we are able to determine the year of C fixation during the last 40 years. As described above, RR is fueled mainly by recent photosynthates, thus having a Δ14C close to that of the atmosphere in the same year. The CO2 evolving from HR, however, originates from the decomposition of SOM that was formed several years ago. Thus, CO2 respired in HR generally contains higher amounts of bomb-14C and consequently has a significantly higher Δ14C than RR. The contribution of two isotopically different sources to a mixture can easily be calculated using isotope mixing models [Amundson et al., 1998; Phillips and Gregg, 2001].

[5] The effects of drought on the components of SR still are not fully understood in forest ecosystems. As C losses originating from the decomposition of SOM have been identified as more important with regard of the CO2-driven greenhouse effect [Kuzyakov, 2006], we will concentrate here on summarizing the state of knowledge of effects of drought on HR. Generally, it is understood that SR in well drained soils is reduced during periods of drought. Liquid films become smaller and substrate accessibility for microorganisms via diffusion decreases [Voroney, 2007]. Furthermore, soil organisms have to respond to reduced soil water potentials. Strategies for surviving drought periods include the formation of dormant spores [Chen and Alexander, 1973] or the accumulation of compatible solutes [Harris, 1981; Schimel et al., 2007]. Accumulation of compatible solutes demands high amounts of C and nitrogen (N) therefore further reducing SR. As soon as the dry soil is rewetted, these restrictions are lifted and SR regenerates quickly, sometimes within minutes [Borken et al., 2003].

[6] There are two fundamentally different mechanisms to explain the effects of drying-wetting events on soil microbial processes affecting C mineralization: The “microbial stress” mechanism versus the “substrate supply” mechanism [cf. Xiang et al., 2008]. The concepts behind the microbial stress mechanism are based on microbial drought tolerance physiology. As mentioned above, accumulation of compatible solutes is one way for cells to survive periods of drought. Rewetting of the soil, resulting in a sudden increase of water potential, forces cells to dispose of these compatible solutes or risk cell rupture due to massive water uptake. The no longer needed compatible solutes are set free and can now easily be mineralized by microorganisms. The microbial stress mechanism postulates that the fast increase of SR during rewetting is mainly due to the mineralization of these compatible solutes, i.e., substrates that were already available to the microorganisms, but where not mineralized because they were needed otherwise. The microbial stress mechanism bears no potential of creating new substrates. Instead, Xiang et al. [2008] postulated that it should result in a loss of microbial biomass thereby reducing the metabolic capability of the microbial pool.

[7] This concept contradicts findings by Birch [1958, 1959], who was the first to recognize that the cumulative CO2 loss from soil that was subjected to drying-rewetting events can be higher than the losses from the same soil under constantly moist conditions. His findings indicate that drying-wetting events can mobilize substrates that were not available before. This priming effect has recently been named the ‘Birch effect’ [Jarvis et al., 2007], and can be best explained by the “substrate supply mechanism” [Xiang et al., 2008]. It assumes that physical processes during rewetting of dry soil like, e.g., aggregate disruption, organic matter redistribution or desorption result in the destabilization of formerly physically protected SOM pools and enhance labile C availability in soils. However, the occurrence of the Birch effect does not seem to be a universal phenomenon [cf. Borken and Matzner, 2009]. Whereas some authors reported increased C losses due to drying-rewetting [Seneviratne and Wild, 1985; Xiang et al., 2008], others did not [Degens and Sparling, 1995; Muhr et al., 2008].

[8] As 14C is an indicator of age, both on decadal timescales (due to bomb-14C) and on centennial to millennial timescales (due to radioactive decay), it offers the possibility of detecting changes in the predominant sources of respired CO2, as long as this change is accompanied by a change of the age of the substrate. Physically protected substrates that are made bioavailable by drying-rewetting presumably are older than more recent C sources. Under this assumption, the creation of new substrates from physically protected pools should result in measurable changes of the Δ14C of respired CO2.

[9] This experiment was designed to investigate the effect of prolonged summer drought followed by natural or artificial rewetting of the soil on SR underneath a Norway spruce stand (Picea abies (L.) Karst.). We excluded summer throughfall in two subsequent years (2006/2007) and measured one additional year without manipulation (2008) to test the following hypotheses: (1) SR is reduced during drought; (2) rewetting leads to a fast recovery of SR; (3) no additional substrates are mineralized by drying-rewetting, so that in the end (4) the cumulative CO2 losses are reduced by drying-rewetting compared to control conditions; (5) HR is more strongly reduced by drought than RR; (6) reduced C losses in a dry year and the assumed substrate accumulation is compensated for by increased C losses in subsequent years.

2. Materials and Methods

2.1. Site Description

[10] The research site Coulissenhieb II is located in a mature Norway spruce forest (Picea abies, (L.) Karst., mean age 145 years) in the Fichtelgebirge in Southern Germany (50°08′N, 11°52′E) at an elevation of 770 m a.s.l. Mean annual air temperature is 5.3°C and the mean annual precipitation (1971–2000) ranges around 1160 mm [Gerstberger et al., 2004]. The understory vegetation is dominated by Calamagrostis villosa (Chaix ex Vill), Deschampsia flexuosa (L.), Vaccinium myrtillus (L.) and Oxalis acetosella (L.). The soil is classified as a Haplic Podzol with a sandy to loamy texture according to the FAO soil classification [IUSS Working Group WRB, 2006]. The mor-like forest floor has a thickness of 6–10 cm and is composed of Oi, Oe and Oa horizons (Table 1). The pH (CaCl2) value of the soil ranges around 3.3 in the Oa horizon and increases with depth to around 4.2 in the Bw and C horizon. Carbon contents decrease with depth, ranging around 40–50% in the Oi and the Oe and less than 1% in the C horizon [from Schulze et al., 2009].

Table 1. Mean Thickness of the Horizons, Bulk Density, pH, Carbon Content, and Bulk Radiocarbon Signature of a Podzol Soil From a Norway Spruce Stand in the Fichtelgebirgea
HorizonThickness (cm)BD (g cm−3)pH (CaCl2)C (%)Δ14C (‰)
  • a

    BD, bulk density. Values represent mean values from nine soil profiles (radiocarbon signature: three soil profiles). Numbers in parentheses give the standard deviation of the mean (changed from Schulze et al. [2009]).

Oi (±SD)2.10.07 45.8113.8
(0.1)(0.00) (0.9)(8.0)
Oe (±SD)2.20.15 42.1161.5
(0.2)(0.02) (6.3)(16.0)
Oa (±SD)
Ea (±SD)5.20.603.48.323.0
Bsh (±SD)5.30.753.66.0−13.8
Bs (±SD)11.40.793.83.6−63.2
Bv (±SD)−145.4

[11] A storm event on 18 January 2007 severely damaged the research site, considerably thinning out the forest. All results beyond this date may therefore be subject to disturbance caused by the storm.

2.2. Experimental Design

[12] In the summer of 2005, three control and three throughfall exclusion (TE) plots were established, each covering an area of 20 m × 20 m. We chose one representative plot from each group for soil moisture measurements in the organic layer. Three ECH2O EC-20 soil moisture probes (Decagon Devices, WA, USA) per plot were installed within the Oa horizon and logged automatically every hour. The ECH2O probes were calibrated specifically for the Oa horizon of our site to calculate volumetric water contents from the mV signal. To translate volumetric water contents into matric potentials, we used the van Genuchten model [van Genuchten, 1980]:

equation image

with θ (ψ) = volumetric water content as a function of suction power [m3 m−3]; ψ = suction power [hPa]; θr = residual soil moisture [m3 m−3]; θs = saturation water content [m3 m−3]; α, n, m = van Genuchten equation parameters with m = 1 − 1/n. The necessary parameters (θr = 0.000 m3 m−3, θs = 0.860 m3 m−3, α = 0.163, n = 1.209, m = 0.173) were determined on soil from our site by Zuber [2008].

[13] Soil moisture in 20 cm mineral soil depth was measured on all six plots by two to four custom-built, calibrated tensiometers per study plot and automatically logged hourly. Soil temperature in the organic layer was measured on every plot by custom-built temperature data loggers. Throughfall was measured by custom-built rainwater collectors that were emptied regularly.

[14] Control plots were used to asses the natural dynamic of all measured parameters without any experimental disturbance. The TE plots were equipped with a wood structure that was covered with transparent plastic sheets during the manipulation periods to exclude throughfall on the entire plot area (400 m2). Roofs were built beneath the forest canopy, about 2.5 to 3 m above the forest floor. Rainwater falling on the roofs during the TE period was channeled through rain gutters and water pipes over a distance of ca. 35 m before it could soak into the ground outside the plots. By trenching the TE plots down to a depth of approximately 0.4 m lateral water inflow or uptake by roots was reduced. Roofs on the TE plots were closed from 22 June to 8 August 2006 and from 2 July to 13 August 2007. In 2006, 67 mm of throughfall were excluded, compared to 121 mm in 2007. In the year 2006 we irrigated the TE plots for 2 days with a total of 67 mm artificial throughfall solution via a sprinkler system at the very beginning of the posttreatment period. By irrigating the plots, we guaranteed that the total amount of throughfall was the same on TE and control plots, so we only changed the precipitation pattern. Based on the results in 2006 and the enormous logistic effort, we decided to omit irrigation in 2007. The manipulation in 2007 therefore was a combination of prolonged summer drought and a reduction of total annual throughfall.

[15] The period before closure of the roofs, when control and TE plots both received the same amount of throughfall, will be addressed as the pretreatment period here, the manipulation period, when roofs were closed on the TE plots will be called the TE period. The rest of the year, beginning with the reopening of the roofs will be addressed as posttreatment period here.

2.3. Measurement of CO2 Fluxes

[16] In each of the six plots, three plastic collars with a length of 20 cm and an inner diameter of 49.5 cm were installed permanently for SR measurements. The collars were driven 5 cm into the forest floor several months before the first measurements. Positions of the collars were chosen randomly on the plots. For gas measurements, the collars were manually closed with a plastic lid and connected to a portable infrared gas analyzer (IRGA, Li-820 from Li-Cor Biosciences GmbH). Air was circulated in this closed system by a pump at a constant flow rate of 0.5 l min−1 and the CO2 concentration inside the chamber was logged every 10 s for a period of 3–5 min. A linear regression was performed on the increasing CO2 to determine a flux rate, which was corrected for atmospheric pressure and chamber air temperature. Measurements were conducted simultaneously on the control and the TE plots, and always between 9:00 A.M. and 12:00 noon. For more details on the method, see Borken et al. [2006].

2.4. Radiocarbon Signature of SR and of Its Components (HR, RR)

[17] We simultaneously measured the radiocarbon signature of total soil respiration (Δ14CSR) and of its two operationally defined belowground components (heterotrophic respiration: Δ14CHR, rhizosphere respiration: Δ14CRR) on seven dates in the two manipulation years (Table 2). There were two exceptions: (1) On 8 June 2006 we failed to measure Δ14CHR; (2) on 3 August 2006 we did not measure Δ14CRR, but instead used the value measured on 16 August 06 for both of these dates. All data from one of the measurement dates (19 June 2007) was discarded due to extreme variation. We cannot quantify to which extent the Δ14C data of the year 2007 is influenced by the damage caused by the storm in January 2007, and to which extent this storm might create differences between control and TE plots.

Table 2. Mean Δ14C of SR and Its Operationally Defined Components, HR and RR, as Used for the Partitioning of SRa
DateMean Δ14C (‰) ± SE
  • a

    Samples dominated by prebomb 14C (see Figure 2) have not been included in this calculation. Note that during the first four measurement dates, the atmospheric Δ14C was used as a proxy for Δ14CRR. Data from an additional measurement date (19 June 2007) were discarded due to extreme variation. Sampling was incomplete at the very first measurement date, so some data were not determined (nd).

Control Plots
8 Jun 200691.4 ± 2.2nd50.6 ± nd
3 Aug 200678.7 ± 4.992.5 ± 1.553.0 ± 1.4
16 Aug 200681.0 ± 2.486.3 ± 7.353.0 ± 1.4
15 Mar 200785.8 ± 3.188.0 ± 5.839.7 ± 1.8
9 Aug 200790.2 ± 4.198.0 ± 6.755.2 ± 0.2
16 Oct 200785.1 ± 5.295.6 ± 4.051.6 ± 1.1
TE Plots
8 Jun 200688.0 ± 2.4nd50.6 ± nd
3 Aug 200673.1 ± 2.5100.1 ± 13.653.0 ± 1.4
16 Aug 200678.6 ± 3.087.3 ± 4.753.0 ± 1.4
15 Mar 200774.7 ± 5.090.7 ± 5.139.7 ± 1.8
9 Aug 200773.0 ± 3.6100.5 ± 4.755.2 ± 0.2
16 Oct 200774.9 ± 1.4101.3 ± 3.251.6 ± 1.1

[18] For measuring Δ14CSR, we used the same chambers as for SR measurements. Two of these chambers on each plot (resulting in a total of six replicates per treatment) were closed and then flushed with CO2-free synthetic air for 90 min at a moderate flow rate of 1.5 l min−1 thereby effectively flushing the respiration chambers with an amount of gas equal to at least three times the chamber volume. Following flushing, the respiration chambers were sealed and left until the CO2 concentration inside the chambers reached at least 1500 ppmv. Incubation time depended on CO2 flux rates at the sampling day. Evacuated stainless steel sampling cylinders (2 l) were connected to the respiration chambers and slowly filled with gas from inside the chamber. Via mass-flow controllers the cylinders were connected to a high-vacuum extraction line at the University of Bayreuth. CO2 was cryogenically purified and converted to graphite targets using the modified sealed tube zinc reduction method described by Xu et al. [2007]. Graphite targets were analyzed by the Keck Carbon Cycle AMS facility at UC Irvine, USA with a precision of 2–3‰. Radiocarbon data are expressed as Δ14C, which is the per mil deviation from the 14C/12C ratio of oxalic acid standard in 1950. The sample 14C/12C ratio has been corrected to a δ13C value of −25‰ to account for any mass-dependent fractionation effects [Stuiver and Polach, 1977].

[19] Between the beginning of the experiment in 2006 and March 2007, the radiocarbon signature of rhizosphere respiration (Δ14CRR) was not measured directly. Instead we measured atmospheric Δ14C, assuming a close temporal coupling between assimilation and rhizosphere respiration [Högberg et al., 2008]. Evacuated sampling cylinders were filled with air at 2 m above ground and CO2 was cryogenically purified. Using the atmospheric Δ14C as a proxy for Δ14CRR caused two problems: (1) Some of the atmospheric samples had unusually low Δ14C, most likely resulting from airborne fossil fuel emissions; (2) although RR is closely coupled to photosynthetic assimilation, Schuur and Trumbore [2006] reported differences between the Δ14C of the atmosphere and CO2 respired by roots, indicating that part of the C respired by roots originated from pools with turnover times >1 year. Due to these problems, in 2007 we passed on to directly measuring the Δ14C of CO2 respired by live roots. Roots were excavated from the forest floor at three locations nearby our field plots at three dates in 2007 (Table 2). Excavated roots were washed with deionized water to remove soil particles clinging to the roots and then transferred to gastight incubation containers. Prior to the incubation the containers were flushed with CO2-free synthetic air to remove all atmospheric CO2 and then incubated for 1–2 days in the laboratory at a constant temperature of +15°C. Gas samples again were taken by connecting an evacuated sampling cylinder that was then opened slowly to take in gas from the incubation container.

[20] The Δ14C of HR was determined by incubating root-free soil cores from the uppermost 25–30 cm of our field plots including the Oi, Oe, Oa, Ea, Bsh and part of the Bs horizon. After soil was stored at +5°C for four to six weeks, roots were manually removed and the disturbed soil from each soil core was transferred to one incubation container. After flushing the incubation containers with CO2-free synthetic air to remove atmospheric CO2, the soil was incubated for 1–2 days at a constant temperature of +15°C. Following previous work by Dioumaeva et al. [2002] and more recent experiments by Czimczik and Trumbore [2007], temperature does not affect the Δ14C of evolving CO2. Gas from the containers was sampled using stainless steel containers and processed like described before.

[21] Total SR was partitioned into RR and HR components using a mass balance approach described by Gaudinski et al. [2000] and Schuur and Trumbore [2006] using the following equations:

equation image
equation image

where FCO2,SR is the CO2 flux of total soil respiration, FCO2,HR the flux of heterotrophic respiration and FCO2,RR the flux of rhizosphere respiration, all in mmol m−2 h−1. Δ14CSR is the radiocarbon signature of the total soil respiration, Δ14CHR of the heterotrophic respiration and Δ14CRR of the rhizosphere respiration, all given in ‰. This mass balance approach was designed for samples that are dominated by bomb-14C [Trumbore, 2000]. In this study, samples were regarded as predominantly consisting of bomb-14C when they had Δ14C values between a maximum of ca. 850‰ (atmospheric Δ14C peak around 1964) and a minimum of ca. 50‰ (atmospheric Δ14C around 2007). Samples with Δ14C significantly below that range were regarded as containing significant amount of prebomb 14C and were excluded from the partitioning calculation.

2.5. Data Analysis

[22] For each measurement day, we averaged the individual SR fluxes of the nine chambers per treatment (control, TE) to calculate a mean SR flux. Data were analyzed using STATISTICA 6.1. Differences in SR fluxes between the treatments were tested using the nonparametric Mann-Whitney U-Test. We compared values of the same treatment at different dates using the Tukey HSD test.

[23] We calculated the cumulative CO2 emissions from each individual chamber by linear interpolation of the measured SR fluxes between adjoining measurement dates. For statistical analysis, we calculated the mean cumulative C emissions for the nine control chambers and the nine TE chambers and applied the nonparametric Mann-Whitney U-Test.

[24] For partitioning, we calculated mean values for Δ14CSR (n = 6), Δ14CHR (n = 3), and Δ14CRR (n = 3). Source partitioning was calculated following Phillips and Gregg [2001], accounting for variability in the isotopic signatures of both the sources (Δ14CRR, Δ14CHR) and the mixture (Δ14CSR). When multiplying the calculated source proportions with mean SR fluxes, standard errors were calculated accounting for error propagation.

3. Results

3.1. Throughfall, Soil Moisture, and Soil Temperature

[25] The year 2006 was dry, with only 868 mm of throughfall, followed by a relatively wet year (2007) with 1152 mm and another dry year (2008) with 924 mm (Figure 1a). During the manipulation period, 67 mm of throughfall were excluded in 2006, compared to 121 mm in 2007 (no manipulation in 2008). Matric potential measured beneath the Oa horizon varied between values of pF 2 (moist) and pF 5–6 (very dry) in the observed years (Figure 1b). Driest conditions were found during the summer of 2006, when a period of natural drought hit both the control and the TE plots. A period of low matric potentials in the winter of 2008 is attributed to freezing of soil water due to deeply penetrating soil frost. Differences between control and TE plots were found only during summer time: In the summer of 2006, control plots were naturally dry with pF values around 5 compared to even drier conditions on the TE plots (pF ca. 6). In 2007, no water stress was observed on the control plots in summer (pF ca. 3), whereas the TE plots experienced a moderate drought (pF ca. 4). Although matric potentials regenerated quickly after natural rewetting of the TE plots in 2007, the volumetric water content of the organic horizon on the TE plots remained significantly reduced until winter (data not shown). It should be emphasized that the measurements represent the matric potential at the transition between organic and mineral horizon and that differences between TE and control plots within the organic horizons or even on top of the forest floor are expected to be more pronounced.

Figure 1.

(a) Throughfall, (b and c) matric potential beneath the Oa horizon and in 20 cm mineral soil depth, (d) soil temperature 10 cm below the forest floor surface, and (e) mean (n = 9) soil respiration (±SE) on the control and the throughfall exclusion (TE) plots from September 2005 to November 2008. Shaded areas in Figures 1b–1e label periods when the roofs were closed on the TE plots. Throughfall that was excluded by the roofs during that period is presented in white in Figure 1a. Arrows in Figure 1e mark dates when Δ14C samples were taken for partitioning of soil respiration.

[26] Matric potential in 20 cm mineral soil depth was slightly affected by TE in both manipulation years, resulting in more negative matric potentials on the TE plots (Figure 1c). In the naturally dry year of 2006, TE resulted in a minimum matric potential of at least −650 hPa (tensiometers failed at this point; pF 2.8) compared to ca. −400 hPa (pF 2.6) on the control plots. In the wet summer of 2007, matric potential on the control plots ranged around −50 hPa (pF 1.7), compared to a minimum soil matric potential of −200 hPa (pF 2.3) on the TE plots. In 2008, TE and control plots showed the exact same minimum soil matric potential of ca. −300 hPa (pF 2.5). Soil temperature was not affected by the manipulation and therefore followed the same seasonal dynamic on both the control and the TE plots (Figure 1d).

3.2. Soil Respiration

[27] In the two manipulation years, significantly less CO2 was emitted from the TE plots compared to the control plots (p < 0.05; see Table 3). These differences on an annual basis resulted from reduced CO2 emissions both during the TE period as well as the posttreatment period (Figure 1e and Table 3). No significant differences were found during the pretreatment period, prior to the closure of the roofs. In the year 2006, 6.7 t C ha−1 (±0.2) were emitted from the control plots and only 5.7 t C ha−1 (±0.3) from the TE plots (mean values from nine measurement chambers±SE), resulting in a difference of ca. 0.9 t ha−1. Of this total difference of 0.9 t ha−1, 50% (or approximately 0.5 t C ha−1) can be explained by decreased CO2 emissions during the TE period itself. Another 44% of the total reduction (approximately 0.4 t ha−1) can be explained by reduced CO2 emissions during the posttreatment period.

Table 3. Cumulative C Emissions From the Control and the TE Plotsa
YearPeriodCumulative C Emissions (t C ha−1)DifferenceStatistics p
DescriptionDurationMean ± SEMean ± SEAbsolutePercent of Annual
  • a

    Cumulative C emissions were calculated on an annual basis for the years 2006–2008, and also for the individual periods (pretreatment, TE, and posttreatment periods) in the two manipulation years (2006, 2007). Mean values were calculated from the data of nine individual SR chambers (n = 9) on each manipulation type (control, TE). We calculated differences between the control and the TE plots on an annual basis. By also calculating differences for each individual period, we quantified the relative contribution of differences in a certain period to the total differences of that year. Mean values of control and TE plots were statistically compared by the Mann-Whitney U-Test to detect differences in cumulative C emissions during a certain period. Asterisks indicate values of p < 0.05 are considered statistically different.

2006total1 year6.7 ± 0.25.7 ± 0.30.91000.031*
pretreatment172 days2.1 ± 0.12.1 ±
TE47 days1.4 ± 0.11.0 ± 0.10.5500.002*
posttreatment146 days3.1 ± 0.12.7 ± 0.10.4440.047*
2007total1 year7.0 ± 0.45.9 ± 0.31.21000.019*
pretreatment182 days3.0 ± 0.22.9 ± 0.10.2130.825
TE42 days1.6 ± 0.11.0 ± 0.10.6540.001*
posttreatment141 days2.4 ± 0.12.0 ± 0.10.4330.031*
2008total1 year6.7 ± 0.46.7 ± 0.30.01000.965

[28] In 2007, 7.0 t C ha−1 (±0.4) were emitted from the control plots, compared to only 5.9 t C ha−1 (±0.3) from the TE plots, so total emissions from the TE plots in 2007 were smaller by approximately 1.1 t C ha−1. Again, most of this difference (54% or approximately 0.6 t C ha−1) can be explained by reduced emissions during the TE period itself. Reduced CO2 emissions during the post treatment period explain 33% or approximately 0.4 t C ha−1.

[29] In the year 2008, no manipulation was carried out. Mean total CO2 emissions from the control and the TE plots were identical in this year, both ranging around 6.7 t C ha−1.

3.3. Dynamics of Δ14C of SR

[30] During the pretreatment period in the year 2006 (8 June 2006), we measured little variation in the Δ14CSR on all plots, and we found no significant differences between TE and control plots with mean values (n = 6, ±SE) of 88.0‰ (±2.4) and 91.4‰ (±2.2), respectively (Figure 2). We found no samples that were dominated by old prebomb 14C (i.e., all samples were above atmospheric Δ14C). During the first TE manipulation period (3 August 2006), the variation of measured Δ14C values increased within both treatment groups, and Δ14C values beneath contemporary atmosphere occurred, indicating the influence of prebomb carbon. Variation of measured Δ14CSR decreased considerably as soon as soils were rewet again during the posttreatment period (16 August 2006), and mean values on TE and control plots revealed no significant differences with 78.6‰ (±3.0) and 81.0‰ (±2.4), respectively.

Figure 2.

Radiocarbon signature of total soil respiration (Δ14CSR) and of heterotrophic (root-free) respiration (Δ14CHR) on the control and the throughfall exclusion (TE) plots on different sampling dates during 2006 and 2007. The dashed line at 50‰ represents the mean radiocarbon signature of atmospheric CO2 in the measurement years. A Δ14C beneath this line indicates that a substantial amount of the CO2 of the sample originates from the mineralization of old substrate (prebomb). These samples were not applicable for partitioning calculations.

[31] In the year 2007, premanipulation measurements of Δ14CSR (15 March 2007) revealed little variation for the control plots with mean Δ14CSR of 85.8‰ (±3.1), but high variation for the TE plots and the occurrence of CO2 with prebomb Δ14C in one of the samples. During the TE period, Δ14CSR measurements for both TE and control plots showed a smaller variation and no samples with obvious prebomb influence were found. Mean values of TE and control plots during this second manipulation period differed significantly with 73.0‰ (±3.6) and 90.2‰ (±4.1), respectively. Differences decreased during rewetting with Δ14CSR values around 74.9‰ (±1.4) on TE and 85.1‰ (±5.2) on control plots.

3.4. Partitioning of SR

[32] The contribution of HR and RR to SR was calculated on the basis of the mean Δ14C of SR, HR, and RR like presented in Table 2. Absolute fluxes of HR and RR are shown in Figure 3. Partitioning revealed that SR on the control plots was dominated by HR on all measurement dates, with RR contribution to total SR ranging only between 5 and 40%. Absolute RR emissions were consistently small throughout both years in the control plots, except for the summer of 2006, when RR emissions ranged between 2 and 3 mmol CO2 m−2 h−1. Absolute HR emissions in the control plots followed the seasonal trend of SR.

Figure 3.

Following Phillips and Gregg [2001], we calculated the absolute amount of CO2 emissions (±SE) originating from heterotrophic (HR) and rhizosphere (RR) respiration on the control and throughfall exclusion (TE) plots on several dates in 2006 and 2007.

[33] Natural heterogeneity made it difficult to find differences between the control and TE plots (Figure 3). On all measurement dates, the CO2 emissions attributed to HR were smaller on the TE plots than on the control plots. These differences were predominantly small and statistically not significant. However, during both manipulation periods the differences became much more pronounced. On 3 August 2006, HR on the TE plots was only 33% of the control, and on 9 August 2007, it even decreased to 23% of the control. It should be mentioned, that we also measured significantly (p < 0.05) smaller SR fluxes from the TE plots for both of these dates, whereas SR did not differ at any of the other 14C sampling dates. In contrast to HR, RR tended to be increased on the TE plots at most measurement dates, but differences never were significant.

4. Discussion

[34] Exclusion of summer throughfall effectively reduced the total CO2 emissions from this Norway spruce forest soil in both manipulation years. We ascribe this reduction to reduced water availability indicated by matric potential differences between control and TE plots. More than 50% of the reduction were explained by reduced fluxes during the manipulation period itself. However, a major proportion of the remaining reduction was due to continuously reduced fluxes during the posttreatment period. In both years, CO2 emissions on the TE plots remained reduced until six to seven weeks after the reopening of the roofs, although matric potential differences disappeared a few weeks earlier. Based on 14C data, we attribute the reduction of total CO2 emissions mainly to reduced respiration by heterotrophic soil organisms. It has already been reported that for this soil C losses via DOC leaching are insignificant compared to CO2 emissions, even during drought [Hentschel et al., 2007; Muhr et al., 2008]. The observed reduction of CO2 emissions therefore can be seen as equivalent to a reduction of gross soil C losses at this stand.

[35] During the manipulation periods, the observed reduction of CO2 emissions can be explained best by reduced matric potential in the organic and the uppermost mineral horizons. In the naturally dry summer of 2006, matric potentials indicate drought stress on both the control and the TE plots. Nevertheless, exclusion of summer throughfall led to measurable differences between control and TE plots with matric potentials being one order of magnitude more negative on the TE plots than on the control plots. The soil moisture data shown here reflect the conditions 8 cm below the surface. More pronounced differences between control and TE plots can be expected in the top of the forest floor, because small precipitation events (like documented for this period) likely led to a temporary increase of soil moisture and therefore SR in the control plots, whereas the TE plots were constantly dry throughout the whole manipulation period. On the other hand, when the moisture sensors indicated drought stress this close to the transition of organic and mineral horizons, we can assume that at least the uppermost centimeters of the mineral horizon were subjected to drought stress as well. This assumption is supported by the matric potential measurements in 20 cm mineral soil depth, which revealed that the manipulation produced measurable matric potential differences even in this depth.

[36] After we removed the roofs on the TE plots, it took several weeks until the matric potentials regenerated back to the level of the control plots. In 2007, this could be due to the fact that we did not irrigate the TE plots to compensate for the excluded amount of throughfall. However, we observed the same phenomenon during the posttreatment period in 2006, despite irrigation. This indicates that the delayed regeneration of the matric potential was independent of irrigation. A reason to explain this slow regeneration of matric potential could be water repellency and preferential flow patterns in the organic layer of this soil as reported by C. Bogner et al. (Investigating flow mechanisms in a forest soil by mixed-effects modeling, submitted to European Journal of Soil Science, 2009). It is well known that water repellency is increased under conifer stands [Doerr et al., 2000]. Drying can even further increase water repellency of soils [Dekker and Ritsema, 1996]. Consequently, CO2 emissions remained smaller on the TE plots as long as matric potentials had not recovered.

[37] However, we found that differences in SR rates persisted much longer than differences in matric potential. A possible explanation for this observation addresses the regeneration of the microbial population. Although we do not have direct measurements of microbial biomass, our Δ14C data indicated significantly reduced HR rates under dry conditions. To demonstrate the severity of the drought experienced by the microorganisms, we compared our data with data from a parallel experiment at the same site. In this experiment, the contribution of organic horizons to total SR was quantified by completely removing the organic horizons. This removal reduced SR rates in summer by maximal 40–45% (T. Froitzheim, personal communication, 2009), indicating that CO2 originating from the mineral horizons accounts for a high proportion of the SR emissions. However, with our TE manipulation, we achieved a reduction of SR rates of up to 60% on several measurement dates. We therefore assume that drought stress not only affected the organic horizons, but also the respiratory emissions from the uppermost mineral horizons. In previous laboratory experiments with soil columns from the same site, Muhr et al. [2008] tested the reaction of organic and mineral horizons to intensive drying. They reported that rewetting could completely and quickly restore the CO2 emissions from dry organic horizons back to the level of a continuously moist control, regardless of drought intensity (J. Muhr et al., Drying-rewetting events reduce C and N losses from a Norway spruce forest floor, submitted to Soil Biology and Biochemistry, 2009). In contrast to this, rewetting of dry mineral horizons resulted in an incomplete regeneration, as CO2 emissions remained below the level of a continuously moist control even after rewetting. Thus, when conditions become dry enough not only to affect the organic horizons but also the mineral horizons, it can be expected that rewetting will result in an incomplete restoration of SR fluxes. A possible explanation for this phenomenon could be that the microbial population inhabiting the mineral horizons is less adapted to drought. This idea is supported by findings by Fritze et al. [2000], who described that in the organic horizons of typical Podzol profiles under coniferous forest fungi are relatively abundant, whereas in the uppermost mineral horizon the relative abundance of bacteria is increasing. Fungi, in turn, are known to be more drought tolerant than bacteria [Griffin, 1981; Voroney, 2007]. To investigate the relative abundance of fungi and bacteria at the Coulissenhieb II site, A. Schmitt et al. (Organic matter quality of a forest soil subjected to repeated drying and different rewetting intensities, submitted to European Journal of Soil Science, 2009) analyzed phospholipid fatty acid (PLFA) patterns in undisturbed soil cores. Their results confirm that the ratio of fungal to bacterial PLFAs decreases from the organic to the mineral horizon at our research site. We therefore postulate that SR rates can be reduced persistently by drying because (1) water repellency and preferential flow patterns hinder fast and complete rewetting, and (2) a part of the microbial population can be severely damaged and regenerates slowly.

[38] Reduced CO2 emissions during periods of drought are not surprising: it has long been described that drought reduces the activity of soil microorganisms and therefore SR [Kieft et al., 1987; Degens and Sparling, 1995; Borken et al., 2006]. However, during the last years an increasing number of experiments indicated that drought can also trigger the release of formerly protected, unavailable substrates that can result in a net increase of annual CO2 emissions even if fluxes are significantly reduced during the actual drought period, a phenomenon recently referred to as the ‘Birch effect’ [Miller et al., 2005; Jarvis et al., 2007; Xiang et al., 2008]. In this experiment, CO2 emissions from TE plots exposed to summer drought never exceeded emissions from the control plots. One might argue that due to the nature of our measurements, we might have missed a short-lasting increase of CO2 emissions. In the year 2006, e.g., we measured SR a few hours before we started irrigation and then again 36 h after the beginning of irrigation. We found no increase in SR between those two measurement dates. It still is possible, though, that a pulse with a duration less than 36 h occurred. However, we would like to emphasize that annual CO2 emissions on the TE plots were reduced by 0.9 t C ha−1 compared to the control plots. Even when we missed a short-lasting pulse, it would have to come close to 0.9 t C ha−1 (or about 210 mmol m−2 h−1 on average) to compensate for this reduction. Laboratory measurements on soil from this site with high temporal resolution do not indicate that such an enormous SR pulse is likely to occur [Muhr et al., 2008].

[39] Furthermore, our findings are in agreement with nearly all other field experiments simulating prolonged drought periods we know of, all reporting reduced C losses due to drought [cf. Borken and Matzner, 2009]. The only field experiment that reported an increase of C losses during a drying-rewetting manipulation [Borken et al., 1999] attributed this increase to artificial rewetting during an extremely warm period, when control plots received less water and probably were water limited. So far, increasing C losses from soils due to preceding drought have mainly been reported from laboratory experiments with mineral soil [cf. Borken and Matzner, 2009, and references therein]. In the majority of these experiments, the soil has been sieved. This represents a major disturbance, creates new surfaces and can possibly facilitate the release of formerly protected substrates. Furthermore, it changes the physical characteristic of the soil, and it has been reported that rewetting of disturbed soil occurs faster than of undisturbed soil [Schjønning et al., 1999]. Delayed rewetting due to water repellency and preferential flow patterns, e.g., is not to be expected in disturbed soil.

[40] Reduced C losses in one year do not necessarily have to result in increased C sequestration. If losses are reduced because of reduced mineralization, material that was not metabolized during the dry year might simply be mineralized in the subsequent year, resulting in increased C losses that compensate for the dry year. However, measurements of SR during 2008 do not indicate that reduced C losses in 2006 and 2007 are in any way compensated for by increased C losses in 2008. One possible explanation might be reduced metabolic capacity of the microbial community like postulated as a possible effect of drying-rewetting in the ‘microbial stress’ mechanism [Xiang et al., 2008]. Alternatively, drying might result in stabilization and thus C sequestration of substrates. Based on our results, we so far only can show that substrates not used in dry years are not immediately metabolized in a subsequent wet year.

[41] Even though decreased SR rates on the TE plots contradict the idea of protected substrates being released, Δ14CSR data indicate the contribution of old prebomb carbon to SR mainly on the TE plots and mainly during periods of drought, thus indicating a shift in the quality of the predominant substrate. Without additional information, this might be interpreted as the release of formerly protected substrate, which would be expected to be older. However, in this experiment a second explanation is much more likely: CO2 respired in SR originates from both organic and mineral horizons. Under normal conditions, most of the CO2 emitted in SR is originating in the organic horizons and the uppermost mineral horizon. Bulk Δ14C of these horizons is clearly dominated by postbomb material (see Table 1). As soon as CO2 emissions from organic horizons and uppermost mineral horizons decrease due to drought, the relative contribution of CO2 originating from deeper mineral horizons increases. This will consequently lead to a change in the measured Δ14C like observed in this experiment.

[42] The calculation of the source contribution of HR and RR to SR is based on a variety of parameters (see equations (2) and (3)), all of which are prone to statistical error due to natural heterogeneity. Interpretation of the results therefore has to be done very critically. However, there are two very pronounced results: (1) The contribution of HR to SR on the control plots is always bigger than the contribution of RR; (2) the CO2 emissions from control and TE plots are always about the same size, except during the TE periods in both manipulation years: During this period, both SR and HR become considerably smaller on the TE plots than on the control plots, while RR remains unaffected.

[43] The overall observed dominance of HR over RR might have had several reasons. An explanation might be the thickness of the organic horizons and the high amount of substrate that is stored there. Another explanation might be the disturbance resulting from the chamber installation depth (5 cm). Wang et al. [2005] discussed that the insertion of SR chambers into the soil cuts off superficial fine roots and thereby reduces RR. It is impossible for us to quantify this effect in our experiment, so quantitative conclusions about RR have to be considered with care. However, qualitative conclusions concerning differences in RR between control and TE plots and the reaction of RR below 5 cm depth still should be valid. Further, seasonal periods of low RR contribution to SR were observed from spruce forest soil in other studies as well [e.g., Schindlbacher et al., 2009].

[44] During the TE periods, SR on the TE plots was significantly smaller than on the control plots. Partitioning revealed that this reduction can be explained with a corresponding reduction of HR. At the same time, RR was not affected by the TE manipulation. We therefore can conclude that drying negatively affects heterotrophic soil microorganisms, whereas roots and root-associated microorganisms seem to be able to withstand the drought stress caused by TE manipulation.

[45] Summarizing, we conclude that the exclusion of summer throughfall leads to a significant reduction of SR mainly in the organic and uppermost mineral horizons during the exclusion period. HR is affected stronger than RR. Regeneration of SR takes several weeks, most likely due to a combination of water repellency and microbial casualties. The manipulation effect is strong enough to significantly reduce annual C losses. So far, no evidence has been found that metabolization of unused substrates can lead to a compensation of these reduced C losses either during rewetting of the dry soil or during the following years.

5. Conclusion

[46] Prolonged summer droughts are likely to lead to a significant reduction of annual CO2 losses in this temperate Norway spruce forest. CO2 emissions are not only reduced during the actual drought period, due to either water repellency or serious damage in the microbial population it takes several weeks before they are restored back to control levels. Data on Δ14C indicate that a reduction of SOM mineralization in the organic horizon and the uppermost mineral horizon is mainly responsible for this reduction. No evidence has been found that preceding drought can release new, formerly protected substrate and thereby result in increased carbon losses from soils, like discussed in the context of the so-called “substrate supply” mechanism. So far, no evidence has been found that reduced C losses are compensated for by increased CO2 emissions in subsequent years. Based on our results and in face of the current climate change scenarios, we expect a negative feedback between increased frequency and magnitude of summer droughts and SR in Norway spruce stands.


[47] This research was financially supported by the program 562 “Soil processes under extreme meteorological boundary conditions” of the Deutsche Forschungsgemeinschaft (DFG). We would like to thank Sue Trumbore and Xiaomei Xu from the University of Irvine for invaluable help with the radiocarbon analysis. We are also very thankful to two anonymous reviewers for all their critical recommendations contributing to the improvement of this study.