•An unbiased partitioning of autotrophic and heterotrophic components of soil CO2 efflux is important to estimate forest carbon budgets and soil carbon sequestration. The contribution of autotrophic sources to soil CO2 efflux (FA) may be underestimated during the daytime as a result of internal transport of CO2 produced by root respiration through the transpiration stream.
•Here, we tested the hypothesis that carbon isotope composition of soil CO2 efflux (δFS) in a Eucalyptus plantation grown on a C4 soil is enriched during the daytime, which will indicate a decrease in FA during the periods of high transpiration.
•Mean δFS of soil CO2 efflux decreased to −25.7‰ during the night and increased to −24.7‰ between 11:00 and 15:00 h when the xylem sap flux density was at its maximum.
•Our results indicate a decrease in the contribution of root respiration to soil CO2 efflux during the day that may be interpreted as a departure of root-produced CO2 in the transpiration stream, leading to a 17% underestimation of autotrophic contribution to soil CO2 efflux on a daily timescale.
In forest ecosystems, a major part of gross primary production is allocated below ground and c. 60% of below-ground net primary production is used for root respiration (Litton et al., 2007), contributing to soil CO2 efflux (FS) in an almost similar proportion than heterotrophic sources (Raich & Schlesinger, 1992). This average value hides a large variability, the estimated contribution of the autotrophic sources (FA) to FS typically ranging from 20% to 80% (Epron, 2009). Besides differences among biomes (Subke et al., 2006) or related to age (Bond-Lamberty et al., 2004), methodological biases associated to the various techniques that were used to partition soil CO2 efflux might also account for this large variability (Hanson et al., 2000; Kuzyakov, 2006).
The most common approaches used in the field are based on soil CO2 efflux measurements (Bowden et al., 1993; Epron et al., 1999; Högberg et al., 2001; Lavigne et al., 2004; Ngao et al., 2007; Marsden et al., 2008) that are made during the daytime. It has been recently suggested that a large amount of CO2 produced by root respiration can be dissolved in the xylem sap and be moved upward into the stem via the transpiration stream (Teskey & McGuire, 2007). The authors indeed found a high CO2 concentration at the base of trunks and they hypothesized that part of the CO2 in xylem sap may derive directly from below ground and may contribute to trunk CO2 efflux. More recently, it was calculated that the amount of root-derived CO2 transported above ground was of the same order of magnitude as soil CO2 efflux and that two-thirds of the CO2 produced by autotrophic sources below ground did not diffuse into the soil and did not contribute to soil CO2 efflux (Aubrey & Teskey, 2009). This finding questions the validity of FS and FA values that have been estimated over several decades and our understanding of the global carbon budget.
The aim of this study was to characterize the influence of transpiration on the contribution of autotrophic and heterotrophic components of soil CO2 efflux in a native savannah afforested with Eucalyptus. The conversion of C4-type vegetation into C3-type vegetation is a powerful tool to study the contribution of the two sources to FS (Epron et al., 2009). We hypothesized that carbon isotope composition of soil CO2 efflux (δFS) is enriched during the daytime compared with the night-time, which will mean a decrease in FA during the periods of high transpiration.
Materials and Methods
The experiment was conducted in a 2 ha Eucalyptus stand near the village of Yanika in the coastal Congo (4.17°S, 11.39°E, 80 m elevation). The native vegetation was tropical grassland dominated by a C4 grass, Loudetia simplex (Nees) Hubb, that was burned each year at the beginning of the dry season. It was planted in June 2007 with the hybrid PF1 (clone l-41, Bouillet et al., 2002) at a spacing of 2.5 × 3.2 m.
The subequatorial climate is characterized by a high atmospheric humidity of c. 80% and a high air temperature (25°C) with low seasonal variations and a mean annual rainfall of 1200 mm with a marked dry season between May and September. Field measurements were made during 3 d in April 2011 during the transition period between the wet and the dry season (Table 1). The soils are Ferralic Arenosols (Laclau et al., 2000; Mareschal et al., 2011).
Table 1. Daily maximal air temperature (Tmax), daily minimum relative air humidity (RHmin) and daily cumulative global radiation (RG) 1 wk before measurements and during the 3 d of measurements (in bold) (the monthly average for April is also given)
RG (MJ m−2 d−1)
Sap flux measurement
The thermal dissipation method (Granier, 1985) was used to measure volumetric sap flux density on three trees with an average circumference of 35 cm at 1.3 m height above soil level. The two homemade temperature probes of the sensor were installed at 1.3 m above soil level with 10 cm vertical spacing and at 20 mm depth in the trunk. Data were collected every 5 s with a CR3000 datalogger (Campbell Scientific Inc, Loughborough, UK) and 10 min averages were stored. The volumetric sap flux density (JW, l dm−2 h−1) was calculated as follows:
where ΔT(0) is the maximal temperature difference during the night and ΔT(u) is the current temperature difference between the two temperature probes (Granier, 1985).
Soil CO2 efflux (Fs)
Soil CO2 efflux (FS) was measured using the Li 8100-103 soil chamber (Li-Cor Inc, Lincoln, NE, USA) combined with the Li 8100 infrared gas analyser. Three PVC collars (20 cm diameter) located at 50 cm of each selected tree and inserted to a depth of 2.5 cm into the soil were set up 24 h before the beginning of the measurements. Measurements were made every 4 h from 10:00 h on 25 April 2011 to 23:00 h on 27 April 2011 and the increase of CO2 in the chamber was recorded during 1 min and 30 s.
13C composition of soil CO2 efflux (δFS)
Immediately after measuring Fs, we collected gas samples in a 10 ml Exetainer glass vial (Labco Ltd, High Wycombe, UK) using a self-made sampling device consisting of a customized Plexiglas body (Ngao et al., 2005) connected between the output of the infrared gas analyser and the soil CO2 efflux chamber. Five air samples were collected at intervals of c. 50 ppm. The carbon isotope composition of CO2 within the vials () was measured within a week on a mass spectrometer (Delta S, ThermoFinnigan, Bremen, Germany) coupled to a gas purification device (Gas-Bench II, ThermoFinnigan) and expressed relative to the international Vienna Pee Dee Belemnite (VPDB) standard. was plotted against the inverse of CO2 concentration (Keeling, 1958) to estimate δFS.
13C composition of phloem sap exudates (δP)
One small disk of bark was collected just after the soil CO2 sampling at 1.3 m height on three trees in the vicinity of those selected for sap flux density measurements. They were immediately placed in 2 ml distilled water and left to incubate for 5 h at ambient temperature (Gessler et al., 2004). After removing the bark disk, the phloem extract solution was filtrated on a 0.2 μm nylon cartridge (Whatman, ref. 17 463 433), stored in a freezer and evaporated for 3 d at 45°C. The 13C composition of the dried extracts of phloem exudate (δP) was determined using mass spectrometry as already described.
13C composition of CO2 respired by incubated soil samples and of soil carbon
At the end of soil CO2 efflux measurements, one soil core (5 cm depth) was collected in the centre of each soil collar. The soil was sieved through a 2 mm mesh and root fragments were discarded. A root-free soil subsample of each core (c. 150 g) was then enclosed in a 250 ml glass flask and incubated at room temperature. A pump was used to circulate gas from the flask to an infrared gas analyser (Li840, Li-Cor). The CO2 initially inside the flask was removed using a soda lime trap. After 10 min, the soda lime trap was bypassed and the increase in CO2 concentration in the flask was recorded until reaching values > 400 ppmv. Air was then sampled in a 10 ml Exetainer glass vial and the isotope composition of CO2 in sampled air was measured as already described. After incubation, the soil was oven-dried, ground and analysed for carbon isotope composition.
Soil temperature, soil water content and global radiation
Soil temperature (TS) was measured with a home-made copper-constantan thermocouple inserted at a depth of 10 cm and recorded with the datalogger used for sap flux sensors. Soil volumetric water content in the 0–6 cm soil layer (θS) was measured simultaneously with FS using a Theta Probe (ML2X, Delta-T Device Ltd, Cambridge, UK). The probe was unfortunately moved at several occasions, thereby confounding spatial variability and temporal trends. Global radiation (RG, SP lite sensor, Campbell Scientific Inc, Loughborough, UK), air temperature and relative air humidity (HMP45C, Campbell Scientific Inc) were measured in a nearby savannah experimental site (Tchizalamou, 7 km from the plantation) every 30 s and 30-min averages were stored on a CR10X datalogger (Campbell Scientific Inc).
The analyses were performed using the statistical software R 2.13.0 (R Development Core Team, 2011). Mixed-effect models were used to estimate the effect of the time of the day as a fixed effect on FS and δFS with collars as a random effect and on δP with trees as a random effect. Contrasts were used to test relevant differences between the different times of the day when the overall model was significant (P <0.05). The linear correlations between two variables were tested for significance using Pearson’s correlation coefficients with or without introducing a time lag between both. Temporal autocorrelation (R) was calculated for δFS and δP:
where k is a discrete time lag increment of 4 h (k = 1 means a time lag of 4 h, k = 6 a time lag of 24 h).
A univariate model was fitted using nonlinear regression analysis for determining the Q10 of soil CO2 efflux.
The contribution of autotrophic sources to soil CO2 efflux (FA/FS) was calculated as
with δA and δH the isotope composition of CO2 released by autotrophic and heterotrophic sources, respectively. The CO2 flux from heterotrophic sources (FH) is the difference between FS and FA. The average carbon isotope composition of phloem exudate (δP) was used as a proxy of δA and the isotope composition of CO2 respired by incubated soil samples as a proxy of δH. The confidence interval of FA/FS was calculated from the standard deviation of both end members and of δFS using Isoerror 1.4 (Phillips & Gregg, 2001).
Sap flux density showed typical diurnal patterns with a maximum in the middle of the day (1.9 l dm−2 h−1) corresponding to the highest global radiation and decreased to 0 during the night (Fig. 1a). δFS followed the same pattern as sap flux density, ranging from –24.7‰ around midday to −25.7‰ in the middle of the night (Fig. 1a). Data collected at 03:00 h were statistically different from those collected at 11:00 and 15:00 h (P <0.01) and temporal autocorrelation showed a clear 24 h periodicity (Fig. 2). Daytime values of δFS (07:00–19:00 h) were strongly correlated with JW (R2 = 0.75, P <0.001), with lower JW and δFS on 26 April, which was cloudier than the two other days (Table 1). TS and δFS were not correlated.
δP was far more negative than δFS, ranging from −31.3‰ to −32.8‰ (Fig. 1b). Although δP varied during the measurement period, we did not detect any daily pattern, in contrast to δFS. However, a 16 h periodicity was observed (Fig. 2). No correlation was found between δFS and δP whatever the time lag we considered. Including δP in a bivariate model between δFS and JW did not explain more variance than did JW alone.
In contrast to δP, the isotope composition of soil carbon (−17.3 ± 0.4 ‰) was less negative than δFS, while the isotope composition of CO2 respired by incubated root-free soil samples (−22.7 ± 0.1 ‰) was intermediate between δFS and the isotope composition of soil carbon.
FS exhibited only weak daily variations, ranging from 2.1 to 2.8 μmol m−2 s−1 (Fig. 1c). No significant difference of FS was found between the different sampling times. Soil temperature ranged from 25.7°C after sunrise to 30.8°C after sunset. Despite the small range of variation, the daily changes in soil CO2 efflux were exponentially related to soil temperature (R2 = 0.57, P <0.01) with a Q10 of 1.6. Volumetric soil water content exhibited some erratic variations that probably reflected the confounding effect of spatial variability and a decreasing trend during the three measurement days after the last rain event (17 mm) ending 24 h before the start of the measurements.
FA accounted, on average, for 27% (0.63 μmol m−2 s−1) of FS with a confidence interval of ± 2%, but varied from a maximum of 32% at night (03:00 h) to a minimum of 23–25% during the day between 11:00 and 15:00 h (Fig. 1d). If we further assumed that FA estimated at night is not affected by transpiration and that root and microbial respiration should change with daily changes in soil temperature similarly, the expected FA values (not affected by transpiration) were above the measured values during the day (Fig. 1d). This suggests that, on average, 17% of root respiration might be diverted from soil CO2 efflux during the daytime, with a maximum of 24% between 11:00 and 15:00 h.
Our results confirm the tested hypothesis that δFS is enriched during the daytime in a Eucalyptus plantation grown on a C4 soil. The values of the carbon isotope composition of soil CO2 efflux and of soil carbon for a 4-yr-old plantation are consistent with those obtained on a nearby chronosequence. The isotope signature of soil organic matter ranged from –15.1‰ 18 months after afforestation of a C4-dominated grassland with the same Eucalyptus clone to −18.5‰ 60 months after afforestation (Epron et al., 2009). Those results bracket our values in a 48-month-old plantation. The difference between the isotope composition of soil carbon at 0–5 cm depth (−17.3‰) and the typical isotope composition of C4 grass found in this region (−12.8‰) indicates that a substantial amount of Eucalyptus-derived organic matter has already been mixed with the soil organic matter from the previous land use. Moreover, this additional pool of fresh soil organic matter is more readily available to microbial respiration, as indicated by the more negative isotope composition of CO2 respired by incubated root-free soil samples (−22.7‰). The carbon isotope composition of soil CO2 efflux was even more depleted in 13C as expected, because of the contribution of root respiration to soil CO2 efflux. δFS exhibited daily variations, characterized by 13C enrichment in the middle of the day that was not related to variation in soil temperature.
The range of observed values of δP (−31.3 to −32.8‰) is quite depleted in comparison to the results obtained on Eucalyptus delegatensis (−28.5‰ on average, Gessler et al., 2007). This probably reflects differences in water-use efficiency between temperate and tropical species and also the fact that our measurements were done during the transition between the rainy and the dry seasons when soil water content is not yet limiting stomatal conductance. Large seasonal variations in δP were indeed reported for Eucalyptus globulus (−24 to −30‰; Merchant et al., 2010). The observed variation in δP exhibited a 16 h periodicity which is not recovered in soil CO2 efflux that exhibited a clear 24 h periodicity. The lack of daily periodicity in the δP has also being reported for pine trees (Brandes et al., 2006; Kodama et al., 2008), for spruce (Betson et al., 2007) and for E. delegatensis (Gessler et al., 2007). It was ascribed to the mixing of sugars of different metabolic origins during phloem transport that weakens the signal recorded in the phloem exudate compared with the signal in leaf carbohydrates. In addition, root respiration might be mostly fuelled by stored carbohydrates in Eucalyptus trees (Binkley et al., 2006), which would explain why the temporal variations in δFS were not related to the temporal variations in δP. Similarly, isotope composition of coarse root respiration of beech and spruce exhibited less pronounced seasonal variations than did branches and trunks (Kuptz et al., 2011) and δFS was less coupled to photosynthetic fractionation than was respiration of above-ground compartments in pine (Wingate et al., 2010).
The observed daily variations in δFS were thus more likely a result of a change in the relative contribution of the different respiratory sources having different signatures in the soil. Daily variations of δFS have already been observed in a pine forest (Kodama et al., 2008) and a beech forest (Marron et al., 2009). In both cases, a change in the relative contribution of autotrophic and heterotrophic sources was also hypothesized. A difference in temperature sensitivity of autotrophic and heterotrophic sources (Epron et al., 2001) would not account for the daily variation in δFS because no correlation between TS and δFS was observed.
The daily change in δFS provides indirect experimental evidence of an internal transport of root-respired CO2 in the xylem sap, as suggested by Aubrey & Teskey (2009). While they concluded that flux through this pathway can represent twice the amount of root-respired CO2 that diffuses into the soil atmosphere in a poplar plantation, our results indicated that only 24% of autotrophic respiration would be diverted from soil CO2 efflux to xylem transport in the middle of the day. The discrepancy between the two studies may indicate that the effect is species-specific or site-specific, and that methods used to partition soil CO2 efflux are marred by uncertainties. Aubrey & Teskey (2009) did not measure the autotrophic contribution but based their estimate on a partition of 50% for each component of soil CO2 efflux. The limitations of the short-term sampling period used, and the effects of small sampling sizes on our calculations, have to be acknowledged.
Our estimation of the autotrophic contribution to soil CO2 efflux (27% on average) is prone to random errors related mainly to sampling issues and measurement precision which leads to a confidence interval ranging from 25 to 29%. But at least three systematic biases can also affect this estimation, and therefore the estimation of the fraction of FA that is diverted from soil CO2 efflux.
First, in our calculation, we assumed that root respiration has the same isotope composition as the average value of phloem sap exudate, while fractionation might occur during root respiration (Badeck et al., 2005; Bathellier et al., 2009) or respiration might be fuelled by stored carbohydrates that can be enriched or depleted in comparison to current carbohydrates (Gessler et al., 2007). The isotope composition of bulk fine root organic matter of the same clone averaged −30.8 ± 0.3 ‰ in nearby stands (Epron et al., 2009), c. 1.2‰ enriched compared with δP that we used as proxy for δA (−32.0 ± 0.6 ‰). But root respiration is known to be depleted compared with bulk organic matter (Bathellier et al., 2008). If we had assumed that root respiration was either 2‰ depleted or 2‰ enriched compared with δP, we would have estimated a lower or higher absolute value of FA, respectively (22 or 34%, respectively, of FS instead of 27%), but it would not have changed the fraction of autotrophic respiration that is diverted from soil CO2 efflux during the day (17%). We have assumed δA is constant throughout the day. If we had allowed δA to vary by 2‰ over the course of the day, the fraction of autotrophic respiration that is diverted from soil CO2 efflux during the daytime would have ranged from 14% (with δA peaking at midday) to 21% (with δA peaking at midnight).
Secondly, because disturbance may free more labile carbon, the incubated sieved soils likely release more depleted 13C than would intact soils. If we had assumed that the measured 13C composition of CO2 respired by incubated soil samples was 2‰ depleted (i.e. a δH of 20.7‰ instead of 22.7‰), we would have estimated a higher absolute value of FA (40% of FS instead of 27%) and a lower fraction of autotrophic respiration diverted from soil CO2 efflux during the daytime (10% instead of 17%).
And thirdly, our calculation assumed that soil CO2 concentration and soil CO2 efflux are at equilibrium and that apparent fractionation between the respiratory sources in the soil and the efflux of CO2 did not occur. This would not be the case if advective transport of CO2 within the soil occurred (Kayler et al., 2010) and it would have potentially confounded the mixing of heterotrophic and autotrophic components of soil CO2 efflux with effects resulting from physical isotopic fractionation related to transport processes. The fact that variations in FS and δFS are not at all correlated (R2 = 0.07), however, suggests that our assumption is reliable.
It has been demonstrated that roots can absorb dissolved organic carbon that is further transported above ground (Ford et al., 2007). In our case, such additional transport of below-ground CO2 to above ground would have damped the daily variation of δFS, leading to an underestimation of the fraction of FA that is not recovered in FS during the daytime. However, if a significant amount of FH was also diverted from soil CO2 efflux, we would have expected a decrease in soil CO2 efflux during the daytime, which was not observed.
Because the autotrophic sources only contributed to 27% of soil CO2 efflux in our 4-yr-old plantation, there is no visible decrease in soil CO2 efflux during the daytime. However, it may become more visible for older plantations, because the contribution of autotrophic sources to soil CO2 efflux is thought to increase with stand age in the nearby chronosequence (Nouvellon et al., 2012). Daily patterns of soil CO2 efflux that are not explained by variations in soil temperature, with peaks occurring during the night, have been reported (Kodama et al., 2008; Marron et al., 2009). If these daily patterns are related to the transport of root-respired CO2 into the transpiration stream, then soil CO2 efflux would underestimate the contribution of below-ground sources to total ecosystem respiration, and overestimate the contribution of above-ground respiration that would include a part of CO2 respired by roots that would not contribute to soil CO2 efflux. The magnitude of this internal transport is a key issue that should be carefully addressed in studies that aim to quantify below-ground carbon allocation and forest ecosystem carbon budgets. How seasonality, environmental constraints and phenology affect the upward transport of root-respired CO2 in different forest ecosystems are important, yet unsolved, questions.
We thank C. Plain and A. Granier for their help in designing the experiments; B. Caquet, B. Geneste and J-C. Mazoumbou for their help in the field; G. Sola who provided us with global radiation data; Christian Hossann (‘functional ecology’ platform of the IFR 110) for IRMS analyses; and Lydie-Stella Koutika for critical reading of the manuscript. C. G. was supported by an internal grant from UMR Eco&Sols.