Distinct patterns in the diurnal and seasonal variability in four components of soil respiration in a temperate forest under free-air CO 2 enrichment

Introduction Conclusions References

different nature and age to the atmosphere, with root/rhizosphere respiration often assumed to be the dominant and variable one. Rates of R S vary greatly in time and space and the mechanisms underlying this temporal variability, or the R S components responsible for it, are poorly understood. It is often assumed the R s and its components are under abiotic control at almost all time scales. In this study, we used the ecosys-10 tem 13 C tracer at the Duke Forest Free Air CO 2 Enrichment site to separate forest R S into four components: root/rhizosphere respiration (R R ), litter decomposition (R L ), and decomposition of soil organic matter (SOM)of two age classes -up to 8 years old and SOM older than 8 years. We then examined and found that diurnal and seasonal variability in the components of R s occurred at different magnitudes and directions than 15 total R S . Soil respiration was generally dominated by R SOM during the growing season (44% of daytime R S ), especially at night. The contribution of heterotrophic respiration (R SOM and R L ) to R S was not constant during the growing season, indicating that the seasonal variability seen in R R alone cannot explain the seasonal variability in R S . Although there was no diurnal variability in R S , there were significant compensatory 20 differences in the contribution of individual R S components to daytime and nighttime rates. The average contribution of R SOM to R S was greater at night (54%) than during the day (44%) whereas the average contribution of R R to total R S was ∼30% during the day and ∼34% during the night. In contrast, R L constituted 26% of R S during the day and only 12% at night. Interestingly, the decomposition of C older than 8 years

Introduction
Terrestrial ecosystems exchange large amounts of C with the atmosphere through the processes of photosynthesis and ecosystem respiration (R E ). Annually, the difference 15 between these large fluxes determines the extent of C storage in the terrestrial biosphere and small imbalances between these fluxes can lead to significant variation in atmospheric CO 2 concentration. The role of ecosystems as a long-term sink or source for atmospheric C thus depends on the effects and feedbacks of changing environmental conditions on photosynthesis and the components of R E . The potential responses involving different substrates within the soil system (Taneva et al., 2006) and includes respiration by live roots, root-associated microorganisms, and microbial decomposition of root exudates (collectively referred to as root/rhizosphere respiration, R R ), as well as from heterotrophic respiration (R H ) associated with the decomposition of root and leaf litter, and other soil organic matter (SOM) pools of different ages. Ecosystem ex-10 posure to elevated [CO 2 ] has been shown to lead to enhanced R S rates initially (Zak et al., 2000;King et al., 2004;Bernhardt et al., 2006;Taneva et al., 2006). However, it remains unclear whether these changes are the result of increased R R , R H , or a combination of both (Gonzalez-Meler and Taneva, 2005;Subke et al., 2006). Because individual components of R S return soil carbon of different age back to 15 the atmosphere, a shift in their relative contributions to total R S with environmental changes, will impact the residence time of soil C and, therefore, atmospheric CO 2 levels, by affecting the C sink strength of soils. For instance, atmospheric CO 2 enrichment may cause increases in belowground plant biomass production (Hungate et al., 1997;Edwards and Norby, 1999;Matamala and Schlesinger, 2000;Pregitzer et al., 2000 Discussion Paper | Discussion Paper | Discussion Paper | Discussion Paper | C, with the potential to affect atmospheric [CO 2 ]. Therefore, a critical element in understanding the significance of increased net primary production (NPP) and R S rates for C storage under elevated [CO 2 ] is identifying the origin of soil-respired carbon and its biotic and abiotic controls. Partitioning R S into its components is inherently difficult and a variety of methods 5 have been applied to the separation of R R from R H (Hanson et al., 2000;Subke et al., 2006). The average contribution of R R to total R S in temperate forests has been estimated to be ∼45%, with a range of 10 to 90% (Hanson et al., 2000;Bond-Lamberty et al., 2004). The proportion of R R has been shown to be related to annual R S rates and may not be constant across temporal or spatial scales (Bond-Lamberty et al., 2004;Subke et al., 2006;Kuzyakov and Gavrichkova, 2010), challenging the use of a single annual value for R R /R S in terrestrial C cycle models. An emerging pattern from R E partitioning studies is that photosynthesis exerts a strong influence on R S on diel and seasonal time scales (Hogberg et al., 2001;Bowling et al., 2002;Tang et al., 2005;Kuzyakov and Gavrichkova, 2010). Trueman and Gonzalez-Meler (2005) showed that 15 both autotrophic and the heterotrophic oxidation of soil C pools older than 4 years were both influenced by changes in plant activity. These observations suggest that there are complex interactive effects between R S components that may operate at different time scales, involving several soil C pools that may differ in chemical composition and soil residence time (Heath et al., 2005). The interactive effects of biotic and abiotic 20 variables on R S are unclear. Temperature-and moisture-dependent models are widely used for predicting the response of terrestrial ecosystems to changing environmental conditions (Lloyd and Taylor, 1994;Reichstein et al., 2003;Luo et al., 2007). Individual components of R S , however, can often be independently affected by other abiotic or biotic variables, as nents of R S can often be confounded with the temperature-and moisture-dependent functions observed to explain variations in R S at seasonal time scales, potentially leading to limitations in our mechanistic predictions of ecosystem C budgets (Liu et al., 2006).
In this study, we used the long-term 13 C tracer at the Duke Forest Free Air CO 2 10 Enrichment (FACE) experiment (Chapel Hill, NC, USA) to partition growing season R S into the contributions of root/rhizosphere respiration, litter decomposition, and decomposition of two pools of root-free SOM in situ after 8 years of elevated CO 2 exposure. Stable isotope labeling techniques have been used successfully to partition R S into some of its components (e.g. Andrews et al., 1999;Matamala et al., 2003;Pendall et (Hendrey et al., 1999). Continuous fumigation of all plots began on 27 August 1996, when the trees were 15 years old. CO 2 fumigation is switched off when temperatures are below 5 • C 5 and when sustained wind speed exceeds 5 m s −1 . Starting 16 December 2002, fumigation was reduced to daytime only. The loblolly pine plantation was established in 1983, with 3-year-old seedlings planted at 2 × 2.4 m spacing. Through natural regeneration, a number of hardwood species have become established in the understory, the most abundant of which are 10 Acer rubrum, Liquidambar styraciflua, Liriodendron tulipifera, Ulmus alata, and Cercis canadensis. Soils at the site are clay-rich, low fertility Ultic Alfisols, derived from igneous rock, with a pH of ∼5. Fine roots are found mostly in the upper 20 cm of the soil profile (Matamala and Schlesinger, 2000). Mean annual temperature is 15.5 • C and mean annual precipitation is 1140 mm.

Ecosystem 13 C tracer
The CO 2 used for fumigation at FACTS-1 is strongly depleted in 13  new needles and fine roots produced under FACE have a δ 13 C of −41.8 ± 0.3‰ and −39.2 ± 0.8‰ compared to δ 13 C of −29.9 ± 0.2‰ and −27.6 ± 0.2‰ at ambient conditions, respectively (Matamala et al., 2003;L. Taneva, unpublished data, 2003). The fumigated forest plots have been exposed to a continuous ecosystem 13 C label since the beginning of the CO 2 treatment in 1996 and, through its incorporation into plant Introduction  (Andrews et al., 1999;Schlesinger and Lichter, 2001;Taneva et al., 2006).

Growing season soil respiration
During the 2003 and 2004 growing seasons, soil respiration rates were measured with a field-portable infrared gas analyzer (IRGA; LiCor 6400-09, Lincoln, Nebraska, USA) 5 at 12 PVC collars, randomly placed within each FACE plot. Soil collars were permanently inserted 3 cm into the mineral soil and were open to rainfall and litterfall, except during measurements. In May 2004, four additional soil collars were installed in each FACE plot, where the litter layer was completely removed down to the mineral soil and a thin layer of inert fiber glass was placed over the soil, in order to reproduce the CO 2 10 diffusivity and moisture content of the removed litter. Soil respiration rates were measured monthly during the growing season of the forest (May-October), both during the day (12:00-14:00 EST) and at night (22:00-00:00 EST). Measurements were made at the times previously determined to capture most of the variability in soil respiration rates (data not shown). The six FACE plots were grouped into three blocks, each in- 15 cluding one treatment and one control plot. The measurement time in each plot was ∼1 h and, therefore, only one block was measured a day, in order to avoid temporal variability of soil respiration rates during time of measurement. Measurements in all three blocks were carried out on days with comparable environmental conditions and were usually completed within 5-6 days.

Sample collection and stable isotope analysis of soil-respired CO 2
During the 2003 and 2004 growing seasons, soil-respired CO 2 samples were collected monthly from collars with and without litter layer, both during the day and at night, 24 h after soil respiration measurements were made (see above). Gas samples were collected from a LiCor 6400-09 soil chamber into evacuated 150-ml glass flasks, Introduction   -Meler, 2005;Moore et al., 2009). The CO 2 concentration of each sample was measured at the time of sample collection. Eight gas samples from collars containing litter, and four samples from collars where the litter layer was excluded, were collected from each FACE plot at each sampling time. Gas samples were collected when the CO 2 concentration differed by at least 80 ppm from that of previous samples. Sam-5 ples were collected from different collars to avoid alterations of convective patterns of CO 2 from soil to air and other recognized problems when collecting soil surface fluxes for building keeling plots (Phillips and Greg, 2001;Trueman and Gonzalez-Meler, 2005;Bowling et al., 2008;Kayler et al., 2010). Samples were shipped to the University of Illinois at Chicago for stable isotope analysis. In the laboratory, soil-respired CO 2 samples were purified by cryogenic extraction before they were analyzed for their stable C isotope composition with a Finnegan Delta Plus XL (Bremen, Germany) isotope ratio mass spectrometer. The δ 13 C of soil-respired CO 2 , in the absence of atmospheric air, was determined using Keeling Plot analyses (Pataki et al., 2003). The range in [CO 2 ] of samples used to construct Keeling Plots was at least 320 ppm. Keeling Plot 15 regressions with an r 2 value of <0.90 were excluded from further analysis.

Incubations for end-member determination
During the growing season of 2004 soil cores (0-10 cm) were collected from locations adjacent to collars after respiration and isotope measurements were done throughout the course of the experiment. Roots and leaf litter were removed from the soil immedi-20 ately after soil collection. Live fine roots were rinsed in distilled water of all attached soil and soil was removed from the litter layer by hand. In order to determine the δ 13 C of respired CO 2 , the litter layer, live roots, and root-free soil collected from each FACE plot were incubated separately in the dark in custom-designed PVC chambers with screw caps (400-ml chambers for soil and litter incubations and 150-ml chambers for root 25 incubation), using a method similar to that described in Taneva and Gonzalez-Meler (2008). The field incubation system consisted of a pump, a soda lime column placed before the incubation chamber, a desiccant column placed between the chamber and 2883 Introduction

Tables Figures
Back Close

Full Screen / Esc
Printer-friendly Version

Interactive Discussion
Discussion Paper | Discussion Paper | Discussion Paper | Discussion Paper | the glass flask, where the respired CO 2 was eventually collected, and an IRGA (LiCor 6262, Lincoln, Nebraska, USA). All components of the incubation system were connected to each other with Bev-A-Line ® tubing (1/4 outer diameter). Before sample incubation, the chamber, the 150-ml glass flask, and the line were flushed with CO 2free air by pumping dry ambient air through the soda lime column. The IRGA, placed 5 after the flask, was used to monitor the [CO 2 ] of the air in the incubation system. The air-tight chamber remained close with three-way valves (Swagelok, Solon, OH, USA) for an incubation period of 20-30 min, depending on respiration rate. After the incubation and prior to collecting the respired CO 2 from each sample, the incubation system was once again flushed with CO 2 -free air, bypassing the closed incubation chamber, to ensure the lines and flask were free of H 2 O and CO 2 . CO 2 -free air was then allowed to pass through the incubation chamber and the respired CO 2 was transferred through the desiccant column into the glass flask, reaching a maximum value between 400 and 1200 ppm. Flask samples were shipped to the University of Illinois at Chicago for analysis. These incubation experiments were also done at the ambient rings to account 15 for environmental variability in the isotopic composition of respired CO 2 from roots and litter that are independent from the addition of the post treatment isotope label in these pools. The δ 13 C value of respired CO 2 from roots, litter, and root-free soil from each plot (Table 1) was used in the partitioning of soil-respired CO 2 (see below). 20 Soil-respired CO 2 was first partitioned into C that was photosynthetically fixed since the beginning of CO 2 fumigation (referred to as "post-treatment" C) and C assimilated under ambient [CO 2 ] before fumigation started in 1996 (referred to as "pre-treatment" C), according to the following two end-member mixing equation: where δ 13 C Rs CO2 is the measured δ 13 C of soil-respired CO 2 at time t, δ 13 C post−tr is the end-member for post-treatment C, δ 13 C pre−tr is the end-member for pre-treatment C at time t and f represents the fraction of pre-treatment C in soil CO 2 at time t (Taneva et al., 2006). The δ 13 C pre−tr is determined by directly measuring δ 13 C Rs CO2 in the control plots 5 of the experiment at time t. This measured value incorporates respiration from both recalcitrant and labile soil C pools under ambient CO 2 conditions. Because the δ 13 C of recalcitrant soil C pools has little or no seasonal variation (Balesdent and Mariotti 1996), any seasonal variability in δ 13 C of soil-respired CO 2 in the control plots will be mainly due to differences in the signature of labile soil C pools (i.e. root/rhizosphere 10 respiration), reflecting, for instance, seasonal fluctuations in photosynthetic discrimination. Because the δ 13 C of the atmosphere in the CO 2 -enriched plots was changed by a constant value (E ) at the beginning of the experiment and because photosynthetic discrimination against 13 C is the same under ambient and elevated [CO 2 ] (due to lack of photosynthetic acclimation and conserved C i /C a between ambient and ele-15 vated [CO 2 ] plots (Ellsworth, 1999), the difference in δ 13 C of new photosynthate in the control and treatment plots also equals E . Therefore, the end-member for the δ 13 C of soil-respired CO 2 in the enriched plots (δ 13 C post−tr ) can be derived by subtracting E from the measured δ 13 C pre−tr and Eq. (1) can be rearranged as follows:

Mixing models and end-member determination
where E was found to be 12 ± 0.1‰.

Partitioning soil-respired CO 2 into its components
The litter exclusion experiments in 2004 enabled us to further partition soil-respired CO 2 into CO 2 originating from root/rhizosphere respiration (R R ), litter decomposition (R L ), and SOM decomposition (R SOM ). The contribution of R R , R SOM , and R L to total R S can then be express as follows: where the fraction of root-respired CO 2 (a) was determined from Eq. (4), b represents the fraction of soil-respired CO 2 produced in SOM decomposition, and the remaining CO 2 in R S , determined as (1−(a+b)), represents CO 2 produced in litter decomposition.

5
Because the ratio of R R to R SOM in the plots without litter is the same as that in plots with litter, the fractions of R R and R SOM in plots without litter (nl) can be expressed as follows: where a represents the fraction of root-respired CO 2 in R S , b is the fraction of SOM 10 decomposition in R S , δ 13 C nl CO2 is the δ 13 CO 2 from collars with no litter as determined from Keeling Plot analyses, δ 13 C root is the δ 13 C of root-respired CO 2 determined from root incubations, and δ 13 C SOM is the δ 13 C measured with litter-and root-free soil incubations (Table 1; incubations discussed above). To calculate the actual amount of C each of these components contribute to total R S , the fractional values of a, b, and 15 (1 − (a + b)) calculated over the growing season were multiplied by the measured R S rate at time t.
Assuming that the fraction of soil-respired CO 2 (i.e. in total R S ) derived from pretreatment C pools (i.e., C assimilated before 1996; Eq. 1) all originated from SOM decomposition, SOM decomposition was further partitioned into pre-treatment (oxida-20 tion of soil pools older than 8 years) and post-CO 2 treatment C decomposition. The contributions of C pre−tr and C post−tr were applied to the rate of SOM decomposition estimated from Eq. (3).
In addition, a sensitivity analyses was made to estimate the sensitivity of the calculated R R , R SOM and R L components of R S to endmember determinations. For this 25 sensitivity analyses we applied a ±1‰ to the root, SOM and litter respired CO 2 determinations form the incubation chambers to account for potential errors in gas collections 2886 Introduction and for potential rapid shift in substrates used for respiration in the isolated components. We performed this analysis for each month and time of day we calculated the root litter and SOM component partitioning of R S .

Soil temperature and moisture
Continuous soil temperature measurements were taken at 10 cm depth in each FACE 5 plot, using Siemens Type M 841/S1 thermistors (one per plot). Continuous soil moisture measurements were taken with a Campbell Scientific Model CS 615 probes (Logan, Utah, USA) consisting of two 30 cm long metal rods, over which each moisture measurement is integrated. Soil temperature and moisture measurements were taken every 5 or 30 s, averaged over 30 min intervals and automatically logged with Camp-10 bell 21X or 23X data loggers. The rate of total R S and each R S component was plotted against soil temperature and soil moisture at each measurement date and time (regressions not shown). The relationship between soil temperature and R S and its components was determined by fitting a second-order exponential growth function to the data, according to the equation 15 f = ae bx (Lloyd and Taylor, 1994). The relationship between R S and each R S component and soil moisture was determined by fitting linear functions to the data, according to the equation f = y 0 + ax (Orchard and Cook, 1983).

Statistical analyses
Temporal variability in R S and R S components was examined with mixed-effects re-

Seasonal and interannual variability of soil respiration (R S )
Soil respiration rates were not significantly stimulated by ecosystem exposure to elevated [CO 2 ] during 2003 (p > 0.5), but there was a significant CO 2 treatment effect on R S in 2004 (p < 0.03), when overall R S rates were higher (Fig. 1). Rates of R S were on 5 average 14% higher under elevated [CO 2 ] in 2004 relative to ambient CO 2 conditions. The magnitude of the CO 2 treatment effect on R S differed during the day and night and varied seasonally in the two years of measurement (Fig. 1) were between −5% (in August, p > 0.5) and 10% (in September, p > 0.5) higher than nighttime rates under ambient [CO 2 ]. In 2004, the enhancement of daytime R S rates in the treatment plots was between 9% (in August, p > 0.1) and 21% (in July, p < 0.002). Nighttime R S rates under elevated [CO 2 ] in 2004 were between 5% (in August, p > 0.4) and 17% (in July, p < 0.02) higher than R S rates in the control plots.

15
Rates of R S differed significantly in the two years of study (p < 0.0001); R S rates in 2004 were on average 16% higher than R S rates in 2003 (Fig. 1). Rates of R S in both treatment and control plots showed seasonal variability in both years of study (p < 0.0001) with higher R S in the middle of the growing season (Fig. 1).

Differences in nighttime and daytime R s 20
Daytime and nighttime R S rates were not significant different during the two years of measurements (p > 0.2) under either ambient or elevated [CO 2 ], with the exception of a significant CO 2 treatment x time interaction in 2003 (p < 0.04; Fig. 1 (Fig. 1).

Pre-Treatment and Post-Treatment C in R S in 2003 and 2004
In 2003, the contribution of post-treatment C (less than 8 years old) to daytime R S ranged from 58.6±8.3% in May to 87.5±5.1% in July (Fig. 2). At night, the contribution of post-treatment C to R S ranged from 56.2 ± 12.9% in June to 84.0 ± 3.8% in August.
In 2004, the daytime contribution of post-treatment C to R S was higher than in 2003 and ranged from 82.5 ± 9.1% in May to 89.9 ± 3.5% in July (Fig. 2). The contribution of post-treatment C to nighttime R S was less than during the day and ranged from 76.6 ± 9.1% in August to 84.8 ± 6.3% in May (Fig. 2).

Root/rhizosphere respiration (R R )
In 2004, R R had a seasonal average of 2.81 ± 0.50 µmol CO 2 m −2 s −1 during the day 15 and 3.04 ± 0.66 µmol CO 2 m −2 s −1 at night ( Fig. 3; Table 2). Overall daytime and nighttime R R rates were not significantly different (p > 0.03), despite significant differences in diel rates in August. Overall, R R rates were 8% lower during the day than at night (Table 2). Significantly higher rates of R R were observed in the middle of the season (July and August) relative to rates early or late in the season (June and September), 20 both during the day (36%; p < 0.001) and at night (39%; p < 0.0001). These differences were mostly due to much lower daytime R R rates in September and much higher nighttime rates in August, relative to the rest of the season (Table 2). Daytime and nighttime R R rates in the middle of the season were also significantly different from those early or late in the season (p < 0.0001). The average contribution of R R to total R S was 29.7 ± 5.3% during the day, ranging from 14.1 ± 4.4% in September to 36.8 ± 4.1% in June (Table 2). At night, the average contribution of R R to total R S was 33.7 ± 5.9%, ranging from 26.6 ± 4.5% in September to 51.4 ± 3.5% in August (Table 2).

5
In 2004, R L had a seasonal average of 2.56 ± 0.99 µmol CO 2 m −2 s −1 during the day and 1.16 ± 0.57 µmol CO 2 m −2 s −1 at night (Fig. 3; Table 2). Overall, daytime R L rates were significantly different from nighttime rates (p < 0.0001), despite non-significant differences in diel rates in June and July (Fig. 3). On average, daytime R L rates were 55% higher than nighttime R L rates. Neither daytime nor nighttime rates of R L showed seasonal variability (p > 0.2).
The average contribution of R L to total R S was 26.3 ± 10.5% during the day, ranging from 0% in June to 51.4±6.7 in September (Table 2). At night, the average contribution of R L to total R S was 12.6 ± 6.2%, ranging from 0% in June to 24.3 ± 6.2% in July (Table 2).  Table 2). There were significant differences in daytime and nighttime R SOM rates (p < 0.02), mostly because of significant differences between daytime and nighttime R SOM rates in September (p < 0.01; Fig. 3). 20 Overall, daytime R SOM rates were 14% lower than nighttime R SOM rates. Significantly higher rates of R SOM were observed early in the season (June and July) relative to later in the season (August and September) during the day (23%; p < 0.003), mostly because of high R SOM rates in June ( Fig. 3; Table 2); there were no significant differences between early-and late-season nighttime R SOM rates (p > 0.2). The average relative contribution of R SOM to total R S was 44.1±6.6% during the day, ranging from 34.5 ± 8.0% in September to 63.5 ± 8.4% in June (Table 2). At night, the average contribution of R SOM to total R S was 53.7 ± 5.8%, ranging from 44.7 ± 7.3% in August to 70.8 ± 9.7% in June (Table 2).

Post-treatment SOM decomposition
5 R SOM can be further distinguished between two age pools: pre-treatment C consisting of C fixed by the ecosystem prior to September 1996 (>8 years old), and posttreatment C, assimilated after fumigation began (<8 years old). In 2004, the seasonal average rate of post-treatment SOM decomposition was 2.83 ± 0.53 µmol CO 2 m −2 s −1 during the day and 2.89 ± 0.45 µmol CO 2 m −2 s −1 at night (Fig. 4). Overall 10 rates of post-treatment SOM-C decomposition did not significantly differ between day and night (p > 0.7; Fig. 4). Post-treatment SOM decomposition showed seasonal variability, with higher rates earlier in the season (June and July), both during the day (40%, p < 0.0001) and at night (34%, p < 0.001). The average contribution of posttreatment SOM decomposition to total R S was 30.4±6.8% during the day, ranging from 15 16.2 ± 5.4% in September to 48.6 ± 5.6% in June (Table 2). At night, the average contribution of post-treatment R SOM to total R S was 33.7 ± 7.5%, ranging from 22.5 ± 4.8% in August to 55.9 ± 6.5% in June (Table 2).

Pre-treatment SOM decomposition
In 2004, the seasonal average rate of pre-treatment SOM decomposition had a sea-20 sonal average of 1.28 ± 0.16 µmol CO 2 m −2 s −1 during the day and 1.81 ± 0.26 µmol CO 2 m −2 s −1 at night (Fig. 4). Unlike rates of post-treatment SOM decomposition, the rates of pre-treatment SOM decomposition differed significantly between day and night (p < 0.0001), despite non-significant differences in June (p > 0.4), with nighttime rates 29% higher than daytime rates ( decomposition, rates of pre-treatment SOM decomposition earlier in the season (June and July) were lower than rates later in the season (August and September), both during the day (25%, p < 0.005) and at night (36%, p < 0.0001). The average relative contribution of pre-treatment SOM decomposition to total R S was 13.7 ± 1.9% during the day, ranging from 9.7 ± 4.7% in July to 18.5 ± 5.1% in 5 September (Table 2). At night, the average contribution of pre-treatment SOM decomposition to total R S was 20.0 ± 2.2%, ranging from 14.9 ± 6.3% in June to 24.8 ± 5.3% in September (Table 2).

Sensitivity analyses of endmember determinations
It has been documented that the isotopic composition of respired CO 2 can shift relatively rapid with changes in substrates, diffusion or other factors (see Vargas et al., 2011). To account for this potential variation, we performed a sensitivity analysis for each month and time of day for which the component partitioning of R s was measured by calculating the effect of a 1‰ shift in the isotopic composition of respired CO 2 of any given endmember on the component partitioning results of R S . We found that for 15 every 1‰ change in the isotopic composition of end members respired-CO 2 , the R R component of R S varied up to 15% whereas the variation in the R SOM or R L contribution to R S was less than 6% on average.

Soil temperature and moisture
There were no significant differences in daytime and nighttime soil temperature or mois-20 ture for the measurement periods in this study (t-test, p > 0.05). None of the relationships between total R S or individual R S components and soil temperature or moisture were significant (regressions not shown).

Discussion
In this study, we documented that variations in individual components of R S do not always lead to measurable variations in overall R S efflux rates. We also report that diel differences in rates of R S components are not easily explained by passive temperature and moisture controls, and that biotic controls of R H may be important in determining 5 rates of SOM oxidation. While this is not the first study to separate R S into more than two components (Sulzman et al., 2005;Cisneros-Dozal et al., 2006;Subke et al., 2011), to our knowledge, this is among the first reports of the diel and seasonal changes in the contribution of several R S components to total growing season efflux rates under field conditions. Our results suggest that using R S as an integrator of 10 fast and slow soil biological activity and their responses to environmental change may be too coarse; individual components contributing to soil CO 2 efflux could respond differently to the same variables. Understanding the sources of soil CO 2 efflux and its dependent biotic and abiotic controls are important in elucidating the environmental effects on R S rates at different time scales. 15 By combining the ecosystem 13 C tracing of fumigation CO 2 (as in Taneva et al., 2006) with incubations of litter, roots, and root-free soil, we were able to partition R S further into rhizosphere-, litter-, and two pools of soil-derived C in situ during the day and at night throughout the growing season of a warm-temperature forest exposed to elevated atmospheric [CO 2 ] after 8 years of 13 C application. Due to the lack of a 13 C tracer in the 20 control plots at the Duke Forest FACE site, we were only able to study R S components under ecosystem exposure to elevated [CO 2 ] and a comparison of the contributions of different R S components under ambient and elevated CO 2 conditions was not possible. Several studies have reported increased R S rates under elevated [CO 2 ] (Zak et al., 2000;King et al., 2004;Bernhardt et al., 2006;Taneva et al., 2006) and our results 25 are in agreement with these reports (Fig. 1). Even though the stimulation of R S by elevated [CO 2 ] in this study was not always significant, we recognize that our results are constrained by enhanced R S rates in the treatment plots. By taking advantage of the ecosystem 13 C tracer, however, we were able to examine if and how the temporal dynamics of R S components translate into temporal variability of total R S .

Soil respiration and its components
We studied the daytime and nighttime differences and seasonal dynamics of four R S components in a temperate forest exposed to elevated [CO 2 ] and found that the pres-5 ence of absence of temporal variability of total R S rate could not be attributed to variability in the rate of any single R S component. The seasonal contribution of R R to total R S in this study ranged from 14% to 37% during the day, in the lower end of the annual range of 20-84% reported for temperate coniferous forests (Subke et al., 2006). Our estimate of the contribution of R R to R S is also lower than the estimates of 55% for 10 the beginning of the FACTS-1 experiment in September 1997 (Andrews et al., 1999) or that of 48% derived from a processed-based Q 10 approach (Hamilton et al., 2002), both calculated from midday rates of R S . Furthermore, most of the studies reviewed by Subke et al. (2006), report the fraction of R R in R S on an annual basis, while our results are from the growing season only when rates of R S are at their maximum. Interestingly, 15 Subke et al. (2006) report a relative increase in the fraction of R R with increasing R S rate, suggesting that R R may dominate R S during the growing season, whereas R H may be the relative dominant during the dormant season when total R S rates are low. The proportion of R R in R S found in this study is consistent with the relatively low levels of root productivity and turnover in this pine forest as compared to other temperate 20 forests (Matamala and Schlesinger, 2000;Matamala et al., 2003). Heterotrophic respiration was the dominant component of growing season R S in this forest, constituting 63 to 86% of daytime R S rates (Table 2), within the reported range of 16-80% for the contribution of R H to R S in temperate coniferous forests (Subke et al., 2006) turnover time, and turnover controls. In this study, we partitioned R H into litter decomposition and SOM decomposition and found that the contribution of R SOM to daytime R S rates ranged from 35 to 64% (Table 2). Litter decomposition in this study constituted from 0 to 51% of total R S . Cisneroz-Dozal et al. (2006) found the contribution of litter decomposition to total R S to range from 1 to 42% of R S during the growing season of 5 a temperate deciduous forest and they attribute this variability to forest floor moisture content, in agreement with other studies (Hanson et al., 2003;Goulden et al., 2004;Lee et al., 2004). No significant relationships between soil moisture/temperature and R L were found during the day or at night in this study, suggesting other possible controls on R L in this forest (see Malcom et al., 2009). Although root litter decomposition was not directly accounted for in this study, its contribution to SOM decomposition is expected to be low due to low root productivity and slow root turnover rates in this forest.

Temporal variability in total R S and R S components
In this forest, R S rates varied during the growing season, with higher R S rates in the 15 middle of the season (July and August; Fig. 1). While greater rates of R R in July and August were correlated with increased R S , the contribution of R H (R SOM and R L ) to R S was not constant during the growing season (Table 2), indicating that seasonal variability in R R alone cannot explain the seasonal variability in R S . Notably, when R S rates were at their highest, the rates of both R R and R SOM increased. These results indicate 20 that increases in R S rates are not always solely driven by higher root and rhizosphere activity, as seen in other studies (e.g. Hogberg et al., 2001). Different soil C pools may interact to produce observed rates of R S and measurements of soil CO 2 efflux alone cannot account for the variability of and interactions between R S components (Trueman and Gonzalez-Meler, 2005 and night, although the magnitude and direction of the difference in rates of each component varied throughout the growing season. The contribution of R R to total R S was greater at night later in the season than during the day ( Table 2). The observed diel differences in R R are likely the result of the diurnal variability in the allocation of photosynthetic C to roots (Trueman and Gonzalez-Meler, 2005;Tang et al., 2005) and not 5 necessarily to changes in soil temperature and moisture; daytime and nighttime values of soil temperature and moisture were not significantly different (Table 3). Because the diel differences in R R were not always significant in this study, the time lag between photosynthesis and R R may not be constant during the growing season of this forest (but see Stoy et al., 2007). 10 The average contribution of R SOM to R S was greater at night than during the day (Table 2). Since total R S rate did not differ between day and night, it is reasonable to expect that nighttime decreases in R R due to the absence of photosynthesis may be compensated for by increases in nighttime R SOM . Root/rhizosphere respiration was not always lower at night, however, and lower R R did not always translate to higher R SOM 15 ( Fig. 3 and Table 2). Whether these variations were independent or the result of more complex interactions needs further study. Interestingly, the decomposition of C older than 8 years (R pre−tr ) showed more pronounced and consistent diel differences than any other R S component, with nighttime rates on average 29% higher than daytime rates. In contrast, the decomposition of more recent, post-treatment C (R post−tr ) did not 20 differ between day and night, suggesting that the variability in R SOM appear to be due to changes in the decomposition of older C, rather than the decomposition of recently added SOM. Furthermore, the decomposition of older C was always higher at night and increased consistently during the growing season. These results indicate that oxidation of older decadal soil pools, or components within these C pools, may be able to respond to short-term biotic and/or abiotic changes. If decomposition of decadal SOM pools can rapidly respond to ecosystem exposure to environmental change, the implications for atmospheric [CO 2 ] could be substantial. The presence of unpredictable diel patterns in the rates of R S components, with no changes in soil temperature or moisture, suggests that primary and secondary responses of decomposers to changes in soil conditions exist. Because R S components may have different seasonal and diel patterns from total R S , extrapolation of daytime measurements of R S to monthly or annual scales or application of growing season Q 10 5 values to annual R S may introduce a bias in long-term ecosystem C budgets. In our study, decomposition of SOM, particularly pre-fumigation SOM (R pre−tr ), was the only R S component that exhibited consistently higher contribution to R S at night, which increased towards the end of the growing season, despite no significant differences in intrinsic decomposition kinetics between C pre−tr and C post−tr at FACTS-1 (Taneva and 10 Gonzalez-Meler 2005) and other studies (Trueman et al., 2009). These results suggest that the oxidation of decadal soil C pools may be affected by short-term environmental or biotic controls that may result from interactions between plant and decomposer activity.

BGD
Indirect evidence has shown that R R may be a possible driver of R S variability (Hog- 15 berg et al., 2001;Janssens et al., 2001;Bond-Lamberty et al., 2004;Subke et al., 2006). In this study, variability in R R alone was insufficient in explaining the seasonal and diel variability of R S because temporal changes in other R S components compensated for changes in R R and the decomposition of older soil C pools constituted a substantial fraction of total R S during the growing season of this forest (Table 2). Our 20 results indicate that plant activity may exert a direct and/or indirect control over R S through cascading effects on other R S components beyond R R . Plant activity has been previously linked to greater rates of SOM decomposition (Kuzyakov and Cheng, 2001;Kuzyakov, 2002;Subke et al., 2004;Trueman and Gonzalez-Meler, 2005) and an increasing number of studies have indicated that R S components are not independent 25 of each other, but have interactive effects on R S . These studies indicate that predicted increases in above-and belowground NPP with elevated [CO 2 ] may not necessarily translate into greater soil C storage, as increases in plant activity may simultaneously increase the decomposition of recent and older C in forests (Hoosbeek et al., 2004; Subke et al., 2004;Sulzman et al., 2005;Trueman and Gonzalez-Meler, 2005). Despite the importance of potential priming of SOM decomposition by enhanced plant activity with changing environmental conditions, mechanisms of priming remain poorly understood.
In summary, the results from these experiments show that the lack of diel changes in 5 total R S cannot be interpreted as a sign that source components within R S do not vary. Conversely, because the diel changes in the four components of R S we measured was not consistent, the seasonal variation seen in R S for this forest cannot be attributed to proportional variation within these components. Our results also suggest that there are interactions between components of R S at both diel and seasonal time scales.
Although the nature of these interactions could not be elucidated here, caution should be employed when applying temperature-and moisture-dependent functions to R S , as soil organisms and roots are likely to actively modulate their activity rather than passively respond to abiotic factors.  Table 1. The 13 C signature of respired CO 2 from roots, root-free soil organic matter, and aboveground litter from control and treatment plots at FACTS-1. The average values listed here were derived from incubations (see methods) and site-and time-specific measurements were used as end-member values in the partitioning of soil-respired CO 2 (Eqs. 3 and 4). Average values are expressed in per mil ± standard error (n = 3).  Table 2. Absolute and relative contribution of soil respiration components to total CO 2 efflux during the day and at night in treatment plots at FACTS-1. Respiration rates are expressed in µmol CO 2 m −2 s −1 and show average values (n = 3) ± standard error.