Stem girdling affects the quantity of CO2 transported in xylem as well as CO2 efflux from soil

Authors


Summary

  • There is recent clear evidence that an important fraction of root-respired CO2 is transported upward in the transpiration stream in tree stems rather than fluxing to the soil. In this study, we aimed to quantify the contribution of root-respired CO2 to both soil CO2 efflux and xylem CO2 transport by manipulating the autotrophic component of belowground respiration.
  • We compared soil CO2 efflux and the flux of root-respired CO2 transported in the transpiration stream in girdled and nongirdled 9-yr-old oak trees (Quercus robur) to assess the impact of a change in the autotrophic component of belowground respiration on both CO2 fluxes.
  • Stem girdling decreased xylem CO2 concentration, indicating that belowground respiration contributes to the aboveground transport of internal CO2. Girdling also decreased soil CO2 efflux.
  • These results confirmed that root respiration contributes to xylem CO2 transport and that failure to account for this flux results in inaccurate estimates of belowground respiration when efflux-based methods are used. This research adds to the growing body of evidence that efflux-based measurements of belowground respiration underestimate autotrophic contributions.

Introduction

In forests, soil CO2 efflux (Esoil) contributes between 30 and 80% of total ecosystem respiration (Goulden et al., 1996; Davidson et al., 2006), thereby representing the second largest carbon flux after photosynthesis (Davidson et al., 2002; Bond-Lamberty et al., 2004; Subke et al., 2011). The different sources of belowground respiration contributing to Esoil can be divided into those originating from living roots, their mycorrhizal fungal symbionts, and rhizosphere micro-organisms (the autotrophic component (Ra)) and those from the decomposition of dead organic matter in the bulk soil (the heterotrophic component (Rh)) (Edwards et al., 1970; Bowden et al., 1993). Accurate estimates of both Ra and Rh are needed to better understand the dynamics in Esoil, which is a crucial component in modeling the carbon cycle in forests (Scott-Denton et al., 2006).

By removing a circumferential band of bark and phloem from a tree stem during girdling, the downward transport of photosynthates to roots and associated rhizomicrobial organisms is interrupted (De Schepper et al., 2010), which reduces Ra (Kuzyakov & Gavrichkova, 2010), while water and nutrient transport in the upward direction in the xylem can continue (Högberg et al., 2001; Frey et al., 2006). In various forest ecosystems, Esoil was reduced by girdling by between 24 and 65% relative to Esoil measured in nongirdled control plots (see table 1 in Högberg et al., 2009), illustrating the importance of the contribution of Ra to forest Esoil. However, these estimates were potentially biased by root death (Bhupinderpal-Singh et al., 2003) or increased use of starch (Högberg et al., 2001; Frey et al., 2006) after girdling.

Previous studies of Esoil might also have underestimated Ra because the upward transport of root-respired CO2 with the transpiration stream was not considered. With simultaneous measurements of sap flow and CO2 concentration ([CO2]%) in stems and Esoil in an eastern cottonwood (Populus deltoides) stand, Aubrey & Teskey (2009) estimated that twice the amount of CO2 derived from Ra was transported internally via the xylem compared with that which diffused into the soil atmosphere. This internally transported CO2 contributes to high internal [CO2] in tree stems (range < 1–26%; see table 1 in Teskey et al., 2008), which affects efflux from stems, and is not accounted for with conventional efflux-based measurements of belowground respiration. Grossiord et al. (2012) observed similar internal transport of CO2 derived from Ra in Eucalyptus PF1 (clone 1-41), but in a smaller quantity than observed by Aubrey & Teskey (2009). Nevertheless, the xylem transport of root-respired CO2 raises important questions about our understanding of respiration (Hanson & Gunderson, 2009), which is potentially misestimated in studies that neglect to account for it (Aubrey & Teskey, 2009).

To account for the contribution of Ra to both xylem CO2 transport and Esoil, we made measurements of sap flow, internal [CO2] and Esoil at high temporal resolution in girdled and nongirdled oak (Quercus robur) trees. Stem girdling was used to manipulate the contribution of Ra to belowground respiration. We addressed the following hypotheses: (1) root-respired CO2 in Q. robur trees does not solely diffuse to the soil atmosphere; rather, a portion dissolves in xylem sap and is transported upward in the transpiration stream; and (2) estimates of belowground respiration based on conventional Esoil measurements underestimate Ra, in particular at high sap flow during the daytime.

Materials and Methods

Study site

The study site was a managed 9-yr-old oak (Quercus robur L.) plantation located in Zwijnaarde technologiepark, which is a research facility of Ghent University near Gent, Belgium. The soil is sandy loam with fine roots concentrated in the organic and mineral horizon (maximum depth: 30 cm) (Vande Walle et al., 2007). No understory growth was present throughout the entire plantation and therefore contributions to the autotrophic component of belowground respiration (Ra) came only from oak roots, fungal symbionts, and associated micro-organisms in the rhizosphere.

Two plots of 90 m² were established within the plantation, each containing six trees. In each plot, three trees were randomly selected for measurement. Average diameter (± SD) at stem base and breast height was 11.5 ± 0.9 and 6.7 ± 0.6 cm, respectively, in the first (nongirdled) plot, hereafter referred to as the control treatment, and 11.1 ± 0.3 and 6.1 ± 0.7 cm in the second (girdled) plot, hereafter referred to as the girdling treatment (Table 1). We focused on one control and one girdling plot, because we had limited equipment for the detailed continuous measurements of soil CO2 efflux (Esoil; mg C m−2 h−1), internal CO2 concentration ([CO2];%), sap flow rate (Fs l h−1), and environmental parameters. These measurements at high temporal resolution were essential to understand short-term responses in belowground dynamics after stem girdling. The day before girdling, six undisturbed soil samples (0–30 cm depth) were taken per plot to determine soil bulk density and an additional eight soil subsamples per plot were taken with a soil corer and mixed to a composite sample to determine pH, organic carbon and soil texture for each plot. Analyses of soil bulk density and other soil properties were performed as described by D'Haene et al. (2008). In addition, near the end of September (23 September 2011; day of year (DOY) 266), soil samples were taken at 0–30 cm depth at two locations within 1 m of each tree from which roots were collected and dried to constant weight to determine average root biomass density per plot.

Table 1. Stem diameter at base (0.1 m) and breast height (1.3 m) and distance from the plot center for three Quercus robur trees selected for measurement in the control and girdling treatment plots
TreatmentTreeDiameter at 0.1 m (cm)Diameter at 1.3 m (cm)Distance from center (m)
Control111.06.65.1
210.96.14.2
312.67.43.1
Girdling111.15.23.5
211.26.64.7
311.16.44.1

In the center of each plot, soil temperature and soil moisture were recorded at 7.5 and 22.5 cm with a type-T thermocouple (Omega, Amstelveen, the Netherlands) and soil moisture probe (SM-300, Delta-T, Cambridge, UK), respectively. Air temperature (type-T thermocouple; Omega, Amstelveen, the Netherlands) and relative humidity (Hygroclip; Rotronic AG, Bassersdorf, Switzerland) were measured on site and used to calculate vapor pressure deficit (VPD). All data were recorded at 1-min intervals with a datalogger (CR1000; Campbell Scientific, Logan, UT, USA). Atmospheric pressure was obtained from a weather station located 7 km from the site.

Girdling treatment

On 5 August 2011 (DOY 217), the girdling treatment was applied to the six trees in the girdled plot by carefully removing a 5.5-cm-wide circumferential band of bark and phloem at a height of 50 cm from the stem without damaging the xylem tissue (Fig. 1). The exposed xylem was covered with clear plastic film to prevent tissue dehydration and to allow visual inspection for formation of new bark tissue, which was removed when it occurred. In addition to the trees in the girdled plot, four trees in close proximity to the selected trees but outside the plot were girdled to avoid possible edge effects from ingrowing roots of nongirdled trees. The trees in the control plot were left intact.

Figure 1.

Schematic of the experimental set-up for six Quercus robur trees, indicating the distances and positions of the stem girdle (three trees were girdled) and equipment installed on the stem and within the soil area occupied by the roots (Aroot) to measure the flux of root-respired CO2 transported in the transpiration stream and soil CO2 efflux (Esoil). The black and grey chambers represent the measurement of Esoil with automated and manual chambers, respectively. Displayed chambers were located at 40 cm from the stem. NDIR, nondispersive infrared CO2 sensor; TDP-10, thermal dissipation probe. The stem thermocouple installed 3 cm above the NDIR sensor and the manual chambers installed 70 cm from the stem for Esoil measurements are not shown.

Soil CO2 efflux measurements

For automated measurements of Esoil a PVC chamber (20 cm diameter and 15 cm height; design adapted from Suleau et al. (2009) was inserted 2 cm into the soil 40 cm from the stem base of each tree (Fig. 1). A pump (KNF Neuberger GmbH, Freiburg, Germany) was used to circulate air from the chambers to an infrared gas analyzer (GMP 343; Vaisala Inc., Helsinki, Finland) via a gas multiplexer system. At the start of a measurement cycle, the collar top of the first chamber was opened in half-open position for 1 min to allow flushing of the tubes before the chamber was closed for 4 min to determine the increase in [CO2]. Chamber [CO2] was recorded at 12-s intervals with a datalogger (CR1000x; Campbell Scientific). Finally, the collar top was opened. The remaining five chambers were measured sequentially in a similar manner to complete the measurement cycle. One measurement cycle lasted 30 min and measurements were made alternately in chambers in the control and girdling treatments. Based on the increase in [CO2] in the chambers over time, Esoil was calculated and analyzed as described by Savage et al. (2008). The most linear section in increase in [CO2] over time was identified and the rate of increase (slope) over time was calculated, but only accepted if the coefficient of determination of the linear regression (i.e. R²) was > 0.90. Slope estimates were scaled by chamber cross-sectional area and corrected for air temperature and atmospheric pressure to yield the CO2 efflux rate, according to Savage et al. (2008).

Automated measurements of Esoil were complemented with manual measurements (EGM-1 analyzer connected to a SRC-1 chamber; PP Systems, Amesbury, MA, USA) to account for variation in Esoil within the rooting area among the different cardinal directions. These measurements were performed on three additional PVC collars (diameter 20 cm; height 12 cm) installed 40 cm from each stem (Fig. 1). Later in the growing season (from DOY 231 (19 August 2011) onwards) we performed additional manual measurements with collars installed 70 cm from each stem to account for spatial variation in Esoil. Manual data were collected at 3-d intervals. Both manual and automated measurements continued for 1.5 months after girdling.

Xylem transport of root-respired CO2

The six selected trees in the control and girdled plots (three per plot) were instrumented to measure the transport of root-respired CO2 via the transpiration stream (Ft; mg C h−1) simultaneously with Esoil. We quantified Ft based on measurements of Fs and the concentration of dissolved CO2 in the xylem ([CO2*]; mM) as described by Aubrey & Teskey (2009):

display math(Eqn 1)

(a, the atomic weight of carbon.)

Xylem gaseous [CO2] was measured in situ by inserting nondispersive infrared (NDIR) CO2 sensors (model GMM221; Vaisala Inc.) in the base of each stem 5 cm above the soil (Fig. 1). Stem temperature was recorded in all trees with a type-T thermocouple (Omega), installed at 3 cm from the NDIR sensor. We assumed that xylem [CO2] measured at the stem base represented the [CO2] of xylem sap entering the stem from belowground. Xylem [CO2] was corrected for stem temperature and atmospheric pressure. As a consequence of technical problems with the sensors, data for the control trees were limited to the period from 30 July 2011 (DOY 211) until 13 August 2011 (DOY 225). [CO2*] was calculated using Henry's law coefficients from xylem [CO2], stem temperature and biweekly measurements of xylem sap pH (McGuire & Teskey, 2002; Erda et al., 2013). To obtain xylem pH, sap was expressed from excised twigs of nearby nongirdled trees with a pressure chamber (PMS Instruments, Corvallis, OR, USA). The expressed sap was transferred with a pipette to a solid-state pH microsensor connected to a pH meter (Hi 9124; Hanna Instruments Ltd, Bedfordshire, UK). The mean pH (± SD) of biweekly samples from five trees during the measurement period was 6.8 ± 0.1. Fs was determined by scaling sap flow density (l cm−2 h−1) with sapwood area (cm²). Sap flow density was measured using thermal dissipation probes (model TDP-10; Dynamax Inc., Houston, TX, USA) installed in each stem at 35 cm height with a vertical needle separation of 40 mm (Fig. 1). Zero flow was calculated based on the mean temperature difference between the needles from 03:00 to 05:00 h, assuming no or very limited nocturnal transpiration. Additional trees from the site were used to calibrate the sap flow sensors using a Mariotte-based verification system (Steppe et al., 2010) and generate site-specific calibration parameters for oak, as recommended by Bush et al. (2010), Steppe et al. (2010), and Sun et al. (2011). Stem [CO2], temperature, and sap flow density were recorded with a datalogger (CR1000; Campbell Scientific) at 1-min intervals.

Because Esoil is expressed on a m² area basis, Ft was scaled (Ft−s; mg C m−2 h−1) with the soil area occupied by the roots (Aroot; m²):

display math(Eqn 2)

To estimate the soil area occupied by the roots, we excavated the entire root system of three additional trees at the same site with similar dimensions as the measured trees. Based on the radii of Aroot for the three additional trees (0.8, 1.2 and 0.9 m), we estimated average Aroot as equal to a circular area around the trees with radius 1 m (3.14 m²), as shown in Fig. 1. Manual and automated Esoil point measurements performed within this area were assumed to represent average Esoil of Aroot.

Estimation of the autotrophic component of belowground respiration

We estimated Ra using two methods. The first estimate of Ra was made according to the approach used in previous girdling studies (Ra−conv), where Ra is assumed to be equal to the difference in average Esoil between the control (Esoil,c) and girdling (Esoil,g) treatments, hereafter referred to as the conventional approach:

display math(Eqn 3)

The second estimate of Ra (Ra-new) additionally accounts for internal transport of root-respired CO2 by including the difference in average Ft−s between the control (Ft−s,c) and girdling (Ft−s,g) treatments, hereafter referred to as the new approach:

display math(Eqn 4)

Soluble sugars and starch concentration of fine roots

We girdled three additional trees at the site for sampling fine roots for sugar and starch analysis. These samples were collected on the girdled trees and three nearby nongirdled trees on the day of girdling and 11 (16 August 2011 (DOY 228)) and 40 d after girdling (14 September 2011 (DOY 237)). The sampling dates were selected to obtain data concurrent with the flux measurements and to assess the long-term impact of girdling on fine-root soluble sugar and starch content. For sample harvesting, we carefully uncovered part of the root system and excised c. 3 mg of fine roots, according to Regier et al. (2010). Samples were immediately frozen in liquid nitrogen, transported to the laboratory and stored in a freezer at −18°C until analyzed for glucose, fructose, sucrose, and starch as described by De Schepper et al. (2012). Our intention was to use the soluble sugar and starch data to reveal the relative effect of girdling on fine-root sugar content and to identify the potential sources of root-respired CO2. As for this purpose no absolute fine-root soluble sugar and starch concentrations (shown in Supporting Information Table S1) are needed, we normalized concentrations for each tree by dividing the post-treatment concentrations by the concentration on the day of girdling, facilitating inter-treatment comparison and accounting for variation among trees.

Statistical analysis

We compared daily totals of continuously measured Ft and Esoil for both treatments using repeated measures multifactorial analysis of variance (ANOVA). Treatment (= 2, control and girdled) and date (= 14) were treated as fixed factors, while the individual tree (= 3 per treatment) was considered as a random factor. Similarly, automated and manual measurements performed at 40 cm were compared, but with fewer temporal measurements (= 5). Finally, manual Esoil measurements performed at 40 and 70 cm were compared, based on the measurements made later in the growing season (date; = 10). Relative sugar and starch concentrations were analyzed using a similar ANOVA model, but with fewer temporal measurements (date; = 3) and soluble sugar type (= 4; fructose, glucose, sucrose and starch) as an additional fixed factor. Akaike's information criterion corrected for small sample sizes (AICc) was used to determine the covariance structure that best estimated the correlation among individual trees over time. All analyses were performed using the mixed model procedure (PROC MIXED) of sas (Version 9.3; SAS Inc., Cary, NC, USA) with α = 0.05.

Results

Impact of tree girdling on soil CO2 efflux and xylem CO2 transport

The soil properties of the control and girdling treatment plots were similar (Table 2). Automated measurements showed that soil CO2 efflux (Esoil) before girdling was similar in the control and the girdling treatment plots (P = 0.95), with an average (± SD) of 162.5 ± 18.4 and 162.6 ± 20.9 mg C m−2 h−1, respectively (Fig. 2a). Following girdling, automated measurements showed a strong decrease in Esoil in the girdling treatment, while Esoil in the control treatment remained stable over time. Within 5 d, girdling significantly reduced Esoil (± SD) (= 0.02) by 21.8 ± 3.7% relative to the control treatment (Fig. 2a). This pronounced initial decrease of Esoil in the girdling treatment was followed by a slower decrease, such that, 25 d after girdling, Esoil (± SD) in the girdling treatment was 34.9 ± 4.3% lower than in the control treatment. The effect of girdling on Esoil was confirmed by manual measurements of Esoil 40 cm from the stem base (Fig. 2a). For manual measurements, significant differences in Esoil between the control and girdling plots were observed for the last two measurement dates (P = 0.01). Manual measurements of Esoil tended to be slightly higher than automated measurements, but no significant differences were observed between the manual and automated measurements for the control (= 0.92) or the girdling treatment (= 0.51). In addition, the manual Esoil measurements made at 70 cm from the stem base suggested that measurements performed at 40 cm were representative of Esoil in the rooting areas of the trees in the control and girdling treatments.

Table 2. Soil properties (± SD) of the control and girdling treatment plots established in a Quercus robur plantation measured to a depth of 30 cm
TreatmentRoot biomass density (g m−2)pHOrganic carbon (%)Soil bulk density (g cm−3)
  1. pH and organic carbon were determined for one composite sample per treatment, while root biomass density (= 8) and soil bulk density (= 8) were averaged per treatment.

Control286.7 ± 40.15.11.81.3 ± 0.1
Girdling287.8 ± 59.64.72.01.2 ± 0.2
Figure 2.

(a) Soil CO2 efflux (Esoil) measured with automated and manual chambers located 40 cm from the base of three Quercus robur trees in the control treatment plot (automated, = 3; manual, = 9) and three Quercus robur trees in the girdling treatment plot (automated, = 3; manual, = 9) (see Fig. 1). (b) Stem xylem CO2 concentration ([CO2]) at 5 cm above ground level of trees in the control treatment plot (= 3) and girdled treatment plot (= 3) (see Fig. 1). The vertical dashed line indicates the time of girdling. The shaded areas and bars represent standard deviation. For manual measurements, only upper or lower bars are presented to improve clarity. DOY, day of year.

Before girdling, xylem CO2 concentration, [CO2], measured at the stem base was similar in the control and girdling treatments, averaging (± SD) 10.7 ± 1.2% and 10.1 ± 1.4%, respectively (Fig. 2b). Girdling reduced average xylem [CO2] by approximately one-fifth (i.e. by 21.4 ± 1.1%) in the girdling treatment compared with the control treatment within 5 d after girdling.

Estimation of xylem CO2 transport in control and girdled trees

Before girdling, sap flow rate (Fs) was slightly higher in the girdled plot relative to the control plot, with 6-d averages (± SD) of 0.36 ± 0.12 and 0.29 ± 0.10 l h−1, respectively. After girdling, the three measured trees responded differently to the treatment: a pronounced decrease in Fs sap flow was observed in one tree, while sap flow remained similar to pre-girdling rates in the other two trees.

The decrease in xylem [CO2] in response to girdling (Fig. 2b) influenced the magnitude of root-respired CO2 transported with the transpiration stream (Ft). After girdling, Ft scaled with the soil area occupied by the roots (Ft-s) (Fig. 3c) was lower compared with the control treatment, particularly on days with the highest sap flow. On these days, average daily total Ft-s in the control treatment was significantly higher than in the girdling treatment (= 0.03) and at peak sap flow, average maximal Ft-s in the control treatment (44.7 ± 0.9 mg C m−2 h−1) more than doubled the average maximal Ft-s in the girdling treatment (16.5 ± 2.4 mg C m−2 h−1).

Figure 3.

Average sap flow rate (Fs) in (a) control and (b) girdled Quercus robur trees measured with TDP-10 sap flow sensors (see Fig. 1), and (c) xylem CO2 transport scaled with the soil area occupied by the roots (Ft-s) in control and girdled trees. The vertical dashed line indicates the time of girdling. The shaded areas represent standard deviation. DOY, day of year.

Estimation of the autotrophic component of belowground respiration

For estimating the autotrophic component of belowground respiration (Ra), we used data beginning 2 d after girdling. Diel variations were less pronounced when Ra was calculated with the conventional approach (Ra−conv) compared with estimates obtained with the new approach (Ra−new) (Fig. 4a). During the daytime, an important fraction of root-respired CO2 was transported upward in the xylem, which was not accounted for when Ra was estimated according to the conventional approach. Especially on the days with highest sap flow (DOY 222–223; Fig. 4a), large differences were observed between Ra−conv and Ra−new. Ra−conv decreased in the middle of the day, while Ra−new peaked during the day, when the highest sap flow occurred (Fig. 4a). We expressed Ra estimates relative to belowground respiration estimated according to the conventional and new approaches (Fig. 4b). According to the conventional approach, Esoil,c is an estimate of total belowground respiration. Averaged over 5 d, we (± SD) estimated that 27.0 ± 4.5% of belowground respiration was derived from Ra. With the new approach, Esoil,c Ft−s,c was used to estimate belowground respiration and we found that on average 32.8 ± 8.1% of belowground respiration was contributed by the autotrophic component (Fig. 4b). Larger differences were observed on the days with highest Fs (DOY 222–223). On these days, at peak sap flow (from 12 to16 h), 25.0 ± 2.7% and 45.4 ± 1.5% of belowground respiration was derived from Ra calculated by the conventional and new approaches, respectively. During nighttime hours, small differences were observed in Ra estimates between the two approaches, because low nocturnal sap flow rates (i.e. 1–2% of the 24-h flux) had the potential to transport only a small quantity of root-respired CO2.

Figure 4.

(a) Estimates of the autotrophic component of belowground respiration (Ra) calculated by the conventional approach based solely on measurements of soil CO2 efflux (Ra−conv), or calculated by the new approach, based on the combination of measurements of soil CO2 efflux and the flux of root-respired CO2 transported in the transpiration stream (Ra−new). (b) Contribution of Ra to belowground respiration, estimated according to the conventional approach and the new approach. Symbols represent 4-h averages. DOY, day of year.

Analysis of fine-root samples

Relative to pre-girdling values, the girdling treatment had a significant overall effect on the soluble sugar and starch concentrations of fine roots compared with the control treatment (= 0.002) (Fig. 5). At 11 d after girdling, relative starch (Fig. 5a), sucrose (Fig. 5b) and fructose (Fig. 5c) concentrations decreased in fine roots of the girdled trees, while relative soluble sugar and starch concentrations increased in control trees. Relative concentrations were significantly different between girdled and control trees for glucose (Fig. 5d, P = 0.02) and sucrose (= 0.04). At 40 d after girdling, relative average (± SD) sucrose and starch concentrations further decreased in girdled trees, to values 40.2 ± 2.7% and 14.4 ± 4.8% lower than observed before girdling, respectively, while fructose and glucose concentrations returned to the same level and increased relative to the day of girdling, respectively. In the control trees, soluble sugar and starch concentrations were higher relative to the day of girdling, with the largest increases in average starch (± SD) (44.8 ± 2.4%) and sucrose (43.3 ± 11.1%), the latter being significantly higher compared with the girdled trees (= 0.036).

Figure 5.

Concentration of (a) starch, (b) sucrose, (c) fructose, and (d) glucose relative to concentration on the day of girdling in the fine roots of control (grey bars) and girdled (black bars) Quercus robur trees at 11 d (16 August 2011) and 40 d (14 September 2011) after girdling. Bars represent the mean of three trees per treatment and four samples per tree. The horizontal dotted line represents the relative concentration of fine-root soluble sugars and starch on the day of girdling. Error bars represent + SD. Significant differences (< 0.05) per date between the two treatments are indicated by different letters.

Discussion

This study provides evidence that the autotrophic component of belowground respiration (Ra) contributes to root-respired CO2 transported in the transpiration stream (Ft). Root-respired CO2 either diffuses from the root surface into the soil atmosphere, thereby contributing to soil CO2 efflux (Esoil), or dissolves in xylem sap and is transported upward in the tree with the transpiration stream. Girdling reduced both Esoil and Ft, suggesting that efflux-based methods underestimate the contribution of Ra to belowground respiration.

The strong reduction in Esoil after stem girdling has been described in detail in previous studies (Högberg et al., 2001, 2009; Johnsen et al., 2007; review by Kuzyakov & Gavrichkova, 2010; Subke et al., 2011). Högberg et al. (2001) reported decreases of up to 37% in Esoil within 5 d after girdling Pinus sylvestris trees, while Subke et al. (2011) observed a similar decrease (35%) for Tsuga heterophylla trees within 2 wk after girdling. Data from our high-frequency Esoil measurements agree with these previous results. We found that girdling affected Esoil after 2 d and reduced Esoil after 5 and 25 d by 22 and 35%, respectively, relative to Esoil measured in the control treatment.

Previous studies exclusively used the reduction in Esoil induced by girdling to quantify the contribution of Ra to belowground respiration, based on the general belief that all root-respired CO2 diffuses into the soil and subsequently into the atmosphere. However, recent studies have suggested a large-magnitude upward flux of dissolved CO2 in xylem of trees derived from Ra (Aubrey & Teskey, 2009; Grossiord et al., 2012). Based on measurements of xylem [CO2] and sap flow at the stem base of Populus deltoides trees, Aubrey & Teskey (2009) found that the internal flux of root-respired CO2 in xylem rivaled the flux that contributed to Esoil. Additionally, they calculated that only a small fraction of Ft resulted from the uptake of CO2 dissolved in the soil solution. In our study, we also simultaneously measured the internal flux of CO2 transported in the xylem of the stems to quantify Ra. Girdling reduced xylem [CO2] at the stem base by 21% relative to xylem [CO2] of control trees within 5 d, which was similar to the reduction observed for Esoil. This result agrees with previous observations and supports our first hypothesis that a fraction of CO2 derived from Ra is transported in the transpiration stream.

Simultaneous high-frequency measurements of Ft and Esoil in girdling and control treatment plots allowed us to re-assess our understanding of Ra. When we accounted for the internal transport of root-respired CO2, Ra dynamics were more pronounced during the daytime and indicated a greater contribution of Ra to belowground respiration than previously reported. In particular, on days with high sap flow, a substantial quantity of root-respired CO2 was transported upward with the transpiration stream, especially at peak sap flow, resulting in a large underestimation of Ra when calculated by the conventional efflux-based method (second hypothesis). Similarly, Grossiord et al. (2012) reported for Eucalyptus PF1 (clone 1-41) that the largest underestimation (24%) of the contribution of autotrophic sources to belowground respiration was observed between 11:00 and 15:00 h, which was associated with peak sap flow.

Results of girdling studies may be prone to biases related to the interruption of assimilate flow to belowground tissues, which is assumed to reduce Ra. First, girdling leads to an accelerated use of stored carbohydrates in roots (Högberg et al., 2001, 2009; Olsson et al., 2005), that is, starch in oak trees (Barbaroux & Breda, 2002; Maunoury-Danger et al., 2010). In this case, Ra would be underestimated because tree roots would still be respiring after girdling. Girdling also leads to enhanced decomposition of roots and their associated mycorrhizas by soil heterotrophs (Högberg et al., 2001; Bhupinderpal-Singh et al., 2003; Ekberg et al., 2007), resulting in greater fungal abundance and a possible shift in the fungal community from symbiotic to saprophytic fungi in the long term (Subke et al., 2004). These rhizosphere effects could confound the appearance of reduced Ra in girdled plots (Kuzyakov & Larionova, 2006), leading to underestimation of Ra. However, enhanced decomposition of roots and associated mycorrhizas by heterotrophs and changes in fungal abundance and community structure have been observed to occur at multi-season (Högberg et al., 2001; Subke et al., 2004; Ekberg et al., 2007) and multi-year (Ekberg et al., 2007) scales. Therefore, we assume that these effects were negligible during our relatively short experimental period. After girdling, we observed a relative decrease in the fine-root starch concentration of the girdled trees relative to concentrations at the day of girdling (by 11% 40 d after girdling). At the same time, starch in fine roots of control trees increased (i.e. by 45% 40 d after girdling), probably as a result of late-season loading to maintain winter carbon reserves, as has been observed in previous studies (Barbaroux et al., 2003; Regier et al., 2010 and references therein). Increased use of starch in roots in response to girdling was observed in other field studies and has been suggested to lead to conservative estimates of Ra (Högberg et al., 2001; Frey et al., 2006). In addition, a fraction of the metabolic demand related to starch usage after girdling might be related to starch remobilization (Rodgers et al., 1995; Jordan & Habib, 1996), which suggests that the increased glucose concentrations we observed after girdling (Fig. 5d) could have resulted from hydrolysis of starch (Maunoury-Danger et al., 2010). Therefore, in our study, estimates of Ra contributions to Ft and Esoil in the girdled plot might be lower because the increased use of stored carbohydrates potentially fueled Ra after girdling. There was also potentially an effect of starch mobilization from larger roots to fine roots, as observed by Aubrey et al. (2012) in a canopy scorching experiment. All these different processes contribute to the large standard deviations in fine-root sugar concentrations observed among trees and might explain why the reduction in starch was not significant in girdled trees as compared with control trees. In spite of increased starch use, both xylem sap [CO2] and Esoil decreased rapidly in response to girdling, presumably as a result of the interruption of the translocation of photosynthates to the root system. Sucrose is commonly recognized as the main translocated sugar (Zimmerman, 1957; Rolland et al., 2006) and has a different nature compared with the other sugars. We observed a significant 40% reduction in sucrose concentration in fine roots in response to girdling. This result confirms previous findings that an important fraction of Ra may be dependent on the rapid turnover of recent photosynthate within roots (Epron et al., 2011, 2012 and references therein). Thus, autotrophic metabolic activity that contributes to both Ft and Esoil is probably fueled by a combination of recent photosynthate and stored carbohydrates, as has been observed in other root respiration studies (Lynch et al., 2013). Because root-respired CO2 both diffuses into the soil and is transported in the transpiration stream (Aubrey & Teskey, 2009) regardless of whether it is derived from recent photosynthate or stored carbohydrates, we expect that the biases caused by girdling-related increases in the use of reserves were similar for both fluxes.

Secondly, girdling may affect whole-tree sap flow (Domec & Pruyn, 2008; De Schepper et al., 2010), inducing different rates of water and nutrient uptake between girdled and nongirdled trees. By removing the bark, the outer xylem tissue is exposed to air and might dry out and become nonfunctional for water transport. Moreover, a reduction in stomatal conductance of leaves after girdling related to feedback inhibition of photosynthesis (see De Schepper et al., 2010 and references therein) could reduce sap flow. During our study, we tried to avoid desiccation of exposed xylem tissues by wrapping the girdled stem section in plastic film. However, sap flow in girdled trees was reduced by 20% on average relative to control trees, which was similar to the reduction in sap flow observed in a canopy scorching study by Aubrey et al. (2012). The decrease in sap flow probably affected the estimate of Ra according to both the conventional (Ra-conv) and new (Ra-new) approaches. As a consequence of a reduction in sap flow, a fraction of root-respired CO2 might have contributed to Esoil instead of being transported with the transpiration stream (Aubrey et al., 2012). In addition, previous studies have shown that internal [CO2] builds up in stems when sap flow decreases, because the transpiration stream is not removing respired CO2 from the tissues (Teskey & McGuire, 2002; McGuire et al., 2007; Saveyn et al., 2007b); a similar phenomenon may occur in roots, which would also contribute to increased efflux as a result of the larger concentration gradient.

Because root-respired CO2 both diffuses into the soil atmosphere and dissolves in the transpiration stream, the two fluxes must be investigated simultaneously to fully understand root metabolism dynamics. For instance, we observed non-temperature-related depressions in Ra-conv at high Fs when upward transport of root-respired CO2 with the transpiration stream was large (DOY 222 and 223; Fig. 4). Daytime depressions in CO2 efflux from roots related to the rate of transpiration have been observed for other species (Bekku et al., 2009, 2011) and Subke et al. (2009) observed unexpected decreases in daytime Esoil related to depressions in the autotrophic component of Esoil. However, these studies lacked measurements of Ft to fully elucidate the observed dynamics in Esoil and the tight coupling between the two fluxes. Daytime depressions in CO2 efflux from roots could be similar to what has been observed for stems, where daytime depressions in stem CO2 efflux were related to increased internal transport of CO2 away from the site of respiration as a result of increased sap flow (Teskey & McGuire, 2002; Gansert & Burgdorf, 2005; McGuire et al., 2007) or changes in stem turgor pressure (Saveyn et al., 2007a).

We have shown that efflux-based studies may underestimate the actual contribution of Ra to belowground respiration, as has been suggested previously (Aubrey & Teskey, 2009; Grossiord et al., 2012). Especially at high sap flow rates Fs, a substantial amount of root-respired CO2 is transported upward with the transpiration stream, resulting in large underestimations in belowground respiration when Ra is based solely on Esoil measurements. Therefore, future studies on the contribution of Ra to belowground respiration should use approaches for measuring CO2 fluxes that include the internal transport of respired CO2 (Bloemen et al., 2013; Trumbore et al., 2013) and consider potential factors that control the flux of root-respired CO2 with the transpiration stream, such as sap flow. These measurements are crucial for improving the accuracy of estimates of Ra and our ability to understand belowground respiration dynamics from a mechanistic point of view.

Acknowledgements

The authors wish to thank P. Deman and G. Favyts of the Laboratory of Plant Ecology for their enthusiastic technical support, Prof Dr L. Vanhaecke of the Department of Veterinary Public Health and Food Safety for the use of the analytical equipment for the soluble sugars and starch analysis of the fine-root samples, Dr S. Sleutel from the Department of Soil Management for help with the analysis of the soil samples, and Prof M. Aubinet and his group for information on the design of Esoil chambers. This project was supported by a starting grant from the Special Research Fund (BOF) of Ghent University to K.S.

Ancillary