Tree root and soil heterotrophic respiration as revealed by girdling of boreal Scots pine forest: extending observations beyond the first year

Authors


Peter Högberg. Fax: + 46 90 7867750; e-mail: Peter.Hogberg@sek.slu.se

ABSTRACT

Limitations in available techniques to separate autotrophic (root) and soil heterotrophic respiration have hampered the understanding of forest C cycling. The former is here defined as respiration by roots, their associated mycorrhizal fungi and other micro-organisms in the rhizosphere directly dependent on labile C compounds leaked from roots. In order to separate the autotrophic and heterotrophic components of soil respiration, all Scots pine trees in 900 m2 plots were girdled to instantaneously terminate the supply of current photosynthates from the tree canopy to roots. Högberg et al. (Nature 411, 789–792, 2001) reported that autotrophic activity contributed up to 56% of total soil respiration during the first summer of this experiment. They also found that mobilization of stored starch (and likely also sugars) in roots after girdling caused an increased apparent heterotrophic respiration on girdled plots. Herein a transient increase in the δ13C of soil CO2 efflux after girdling, thought to be due to decomposition of 13C-enriched ectomycorrhizal mycelium and root starch and sugar reserves, is reported. In the second year after girdling, when starch reserves of girdled tree roots were exhausted, calculated root respiration increased up to 65% of total soil CO2 efflux. It is suggested that this estimate of its contribution to soil respiration is more precise than the previous based on one year of observation. Heterotrophic respiration declined in response to a 20-day-long 6 °C decline in soil temperature during the second summer, whereas root respiration did not decline. This did not support the idea that root respiration should be more sensitive to variations in soil temperature. It is suggested that above-ground photosynthetic activity and allocation patterns of recent photosynthates to roots should be considered in models of responses of forest C balances to global climate change.

INTRODUCTION

The ecosystem C balance is the net result of photosynthesis (CO2 uptake) and ecosystem respiration (CO2 loss). The soil is the major site of ecosystem respiration, and globally soil respiration emits 68–76.5 Pg CO2-C year−1 (Raich & Schlesinger 1992; Raich & Potter 1995) to the atmosphere. A study of European forests showed that there is more inter-annual variability in forest ecosystem respiration than in photosynthesis (Valentini et al. 2000), which calls for a deeper understanding of controls on soil respiratory activity. Boreal forests contain a large proportion of the C stored in terrestrial ecosystems (Dixon et al. 1994), and the soil respiration from these forest ecosystems is mainly derived from ectomycorrhizal roots and heterotrophs. It is, however, difficult to separate the two soil components of respiration (Hanson et al. 2000). Furthermore, they may have different temperature sensitivities (Boone et al. 1998; Atkin, Edwards & Loveys 2000; Epron et al. 2001) and their relative contributions to the total soil CO2 efflux may vary with season (Hanson et al. 2000; Epron et al. 2001). Estimates of the contribution of root respiration to total soil CO2 efflux vary as widely as between 10 and 90% depending upon the type of ecosystem studied and the method used (Hanson et al. 2000).

The techniques employed to estimate the contributions of root respiration and heterotrophic decomposition of soil organic matter to overall soil CO2 efflux are: (1) integration of respiratory components (roots, litter, etc.) contributing to in situ forest soil CO2 efflux; (2) flux comparisons with and without root exclusions; and (3) the use of stable or radioactive isotopes (see review by Hanson et al. 2000). Most of these techniques tend to disturb the intimately linked processes by which C is allocated to fine roots, mycorrhizal symbionts and the wider soil community. Isotope studies have the advantage of limited soil and root disturbance over the other methods, but are costly. They are thus mostly conducted on small trees in more or less artificial settings, which may not encompass the spatial and temporal variability and other complexities found at larger scales in the field. An exception was the continuous 13C-labelling technique in a Free Air CO2 Enrichment (FACE) experiment of an intact forest ecosystem (Andrews et al. 1999). This represented a study of an undisturbed system, but used elevated concentrations of CO2, which is a future scenario, and was not representative of systems working at ambient CO2.

Recently, Högberg et al. (2001) presented an approach involving ‘stem-girdling’ all trees of large experimental plots of boreal Scots pines in northern Sweden. Girdling instantaneously terminates the flux of photosynthates from the tree canopy to roots and mycorrhizas. The treatment, however, initially allowed upward water transport through the xylem. This avoided artefacts associated with destructive techniques, such as trenching, which frequently alters moisture and other physical soil parameters and severs structural integrity of roots and fungal hyphae. Högberg et al. (2001) suggested that, during the year of girdling, at least 50% of the soil respiration in these boreal forests was derived from roots and mycorrizhas. They furthermore demonstrated that roots of girdled trees depleted their starch reserves and therefore suggested that the above estimates of root respiration were conservative. Högberg et al. (2001) also suggested that respiration is driven mainly by the seasonal pattern of supply of recent photosynthates to roots, which is highest at the end of the summer in northern coniferous forests (Hansen et al. 1997). However, conventionally temperature has been used as the major environmental driver of respiration in models of global climate change (McGuire et al. 1992; Melillo et al. 1993; Xiao et al. 1998). It is thus of some concern that root activity has been suggested to be more sensitive than heterotrophic activity to variations in soil temperature with Q10 values up to twice as high as those for heterotrophic activity (Boone et al. 1998; Epron et al. 2001). We found this latter suggestion intriguing and worthy of further study.

The δ13C signature of respired CO2 is determined by that of CO2 derived from respiration of live roots (including associated mycorrhizas) and the soil heterotrophs. As the relative contributions of these components vary, the carbon isotopic signature of respired CO2 could vary (Ekblad & Högberg 2001; Pendall et al. 2001; Bowling et al. 2002). In particular, in C3 plants, the enzyme RuBisCo contributes to a strong fractionation against 13C during photosynthesis, but this effect becomes less distinct as stomata close because of water stress (e.g. Farquhar, Ehleringer & Hubick 1989). Ekblad & Högberg (2001) thus showed that air relative humidity (ARH) 2–4 d before sampling of soil CO2 efflux plays an important role in determining the δ13C signature of soil-respired CO2. Their data from periods of low ARH also indicated that root respiration in a boreal forest may account for more than 50% of the total soil respiration. More recently, Bowling et al. (2002) reported a similar close link between the ecosystem respiration from coniferous forests and the vapour saturation deficit of air (which correlates with ARH) 5–10 d earlier. These findings suggest that both above- and below-ground biotic and abiotic conditions could determine seasonal variations in the 13C signature of soil-respired CO2 besides their effect on soil surface CO2 fluxes. The fact that ectomycorrhizal fungi are approximately 2‰ enriched in 13C relative to their host trees (Högberg et al. 1999) adds complexity to the interpretation of the causation of δ13C of soil-respired CO2.

The soil respiration measurements in the above-mentioned girdling experiment (Högberg et al. 2001) were continued for the second successive year. An additional observation was made in the third year anticipating high δ13C of root respiration in connection with high daily maximum vapour pressure deficit (VPDmax) of air. The main aims of the present study were: (1) to determine the fractional contribution of root respiration to soil respiration during the second year after girdling, namely when the girdled trees had their roots totally deprived of starch reserves; (2) to examine the influence of soil temperature on the dynamics of ectomycorrhizal root and heterotrophic respiration in view of the idea that the former should be more sensitive to variations in soil temperature than the latter; and (3) to investigate whether the girdling treatment affected the δ13C signature of respired CO2, and if this could in turn provide further information on the below-ground C dynamics.

Note that we use the term ectomycorrhizal roots because we want to emphasize that these roots also have a significant fungal component. We use the term ectomycorrhizal mycelium to denote the mycelium extending from these roots and to separate it from mycelium formed by other types of mycorrhiza or by saprotrophic fungi. The respiration by ectomycorrhizal roots and mycelium, together with that of other rhizosphere micro-organisms directly dependent on labile C compounds leaked from roots forms the autotrophic (tree root) respiration.

MATERIALS AND METHODS

Study site and experimental design

The experimental site is located at Åheden (64°14′ N, 19°46′ E, 175 m above sea level) in northern Sweden. Nine square plots of 900 m2 and with 120 ± 12 trees each arranged in three separate blocks (each block containing three plots) were established in the year 2000 in a naturally regenerated 45- to 55-year-old Scots pine (Pinus sylvestris L.) forest. There is a sparse understorey of the dwarf shrubs Vaccinium vitis-idaea L. and Calluna vulgaris L. The soil is a weakly podzolized sediment of sandy silt. The annual mean air temperature is +1.0 °C (1980–99). Snow usually covers the frozen ground from the end of October to mid-May Mean annual precipitation is approximately 600 mm (1980–99), of which half falls as rain and half as snow.

In early June 2000, the trees on three plots (one in each block) were girdled at 1.5 m above ground, by complete removal of bark, to the depth of the current xylem, over 0.3 m long sections around the circumferences of the stems. These plots were named early-girdled (EG) plots. In mid-August 2000, the trees on three other plots were girdled; these plots were named late-girdled (LG) plots. On both occasions, it took two persons 4 d to girdle all the trees (approximately 360). Three plots of ungirdled trees were used as controls throughout the experiment. Girdling was conducted at two times to detect any phenological effects on soil respiratory activity. There was no clear increase in needle litter-fall after girdling before the spring of 2002. However, at the end of July 2001, needles of trees on both EG and LG plots started to show some yellowing.

Meteorological data

Air temperature, ARH, soil temperature at 5 cm soil depth in the mineral soil, and daily accumulated precipitation were measured continuously (at 10 min resolution for air temperature and ARH, and 1 h for soil temperature) at a standard meteorological station in an open field 150 m from the centre of the experiment. Air temperature and ARH data were used to derive the daily VPDmax. During the year 2000, soil temperature was also measured continuously (at 30 min interval) in the centre of one plot of each treatment using loggers (type Titbits; Onset Computer Corporation, Bourne, MA, USA) installed below the superficial organic layer (approximately 2 cm deep), but above the mineral soil surface. During the year 2001, soil temperature below the superficial organic layer was measured in all the nine experimental plots (logger type Optic; Onset Computer Corporation). Soil moisture content was measured gravimetrically throughout the year 2000, during which no treatment effects on soil moisture were observed. Nor were any treatment effects on soil temperature observed during 2000–01.

Since both 2000 and 2001 were wet with the May–October precipitation exceeding the 20-year-mean by 25 and 36% (Fig. 1d), respectively (see also VPDmax data, Fig. 2c), we did not observe any drought-stress-induced high δ13C values of root respiration these years (see Fig. 2a & b). A warm and dry spring of 2002, however, provided such an opportunity, and we collected samples of the soil CO2 efflux on 13 June 2002, which was a week after a period of high VPDmax (Fig. 2c).

Figure 1.

Soil respiration (mg C m−2 h−1) in different tree-girdling treatments in Scots pine forest at Åheden in relation to sap flow (mm week−1) and meteorological data for the years 2000 (left, Högberg et al. 2001) and 2001 (right): (a) respiratory soil CO2 efflux from control (filled circles), EG (open triangles) and LG (open squares) plots; (b) calculated root respiration (filled triangles, obtained after subtraction of respiratory activity on the EG plots in 2000 and the mean of EG and LG plots in 2001, respectively, from that on control plots in the respective year); (c) trunk sap flow in the year 2000; and (d) open-field meteorological station data for daily precipitation (bars), daily averages of air temperature and soil temperature at 5 cm depth (presented for only 2000) in the mineral soil (see broken and thick solid lines, respectively), and average soil temperatures at approximately 2 cm depth below the superficial organic layer but above the mineral soil (thin solid line) in the experimental forest. Arrows indicates times of early and late girdling. A 6 °C temperature drop during the summer of 2001 is also indicated. Error bars are ± SE of means of three plots.

Figure 2.

The isotopic composition of respired CO2, δ13Cr (‰), in different tree-girdling treatments in Scots pine forest at Åheden in relation to daily maximum vapour pressure deficit (VPDmax, hPa) of air for the years 2000 (left), 2001 (middle) and until mid-June 2002 (right): (a) comparison of δ13Cr on control (filled circles) and EG (open triangles) plots; (b) comparison of δ13Cr on control (filled circles) and LG (open squares) plots; and (c) daily VPDmax of air. Error bars are ± SE of means of three plots.

Soil surface CO2 flux and 13C natural abundance of respired CO2

The CO2 concentration and isotopic ratio (δ13C) of each gas sample collected from the enclosure (on 10 occasions during June to October 2000, on eight occasions during June to October 2001, and once on 13 June 2002) was measured as described by Högberg & Ekblad (1996; see below) with the exception that δ13C of respired CO2 (δ13Cr) was estimated using the ‘Keeling plot’ approach (Keeling 1958; see below). The natural abundance of 13C is expressed, relative to the international standard Vienna Pee Dee Belemnite, as: δ13C = 1000 × (Rsample/Rstandard − 1) (‰), where R is the molar ratio 13C/12C.

Briefly, within 1 m of the centre of each plot, three opaque plastic cylinders (diameter 243 mm, height 135 mm), covered with a removable lids, were placed on the ground to create headspaces of approximately 6.1 L above the soil surface (Högberg & Ekblad 1996). The superficial layer of lichens, mosses and litter found on top of lichens and mosses were temporarily removed before their placement on the ground. In order to provide a tight seal between the soil surface and the enclosure, it was gently pushed approximately 2 mm into the superficial organic layer and a weight (approximately 2 kg) was placed on top of the lid. Starting 2 min after the placement of the enclosure on the ground, five 12 mL gas samples were removed by a syringe through a rubber membrane in the lid at 2 min intervals. At the two last sampling times during 2000, when soil respiration rates were low, the intervals between samplings were 4 min. In 2001, however, all the samplings were performed at intervals of 4 min after the initial 2 min waiting period. In 2002, the samplings were performed at intervals of 5 min and the first sampling done immediately after covering with the lid (i.e. at approximately 0 min). Gas samples were then transferred to pre-evacuated vials, which were analysed for CO2 concentration and δ13C on a gas purification module coupled on-line to a continuous flow isotope ratio mass spectrometer (CF-IRMS) (Högberg & Ekblad 1996). The rate of CO2 evolution from the soil was calculated by linear regression. Carbon isotopic ratios of respired CO2 (δ13Cr) were modelled by regressing the inverse of enclosure CO2 concentration against the corresponding measured δ13C (Keeling 1958; Sternberg 1989). The linear relationship describes the turbulent mixing of two major sources within the enclosure (atmospheric and respired CO2) and the intercept on the y-axis is the weighted average isotopic composition of respired CO2 of the biological sources (root and heterotrophic respiration). The standard deviation for δ13C based on analysis of replicated reference samples was < ± 0.3‰.

15N natural abundance in understorey plants

We did know that ectomycorrhizal fungi are some 2‰ enriched in δ13C relative to their host trees (Högberg et al. 1999), and anticipated that their mycelium would be decomposed after tree girdling (Högberg & Högberg 2002). Decomposing ectomycorrhizal mycelium should thus contribute to the δ13C of the soil CO2 efflux, and we wanted an additional indicator of this process. Ectomycorrhizal fungi are even more (5–10‰) enriched in 15N than their plant hosts (e.g. Högberg et al. 1996; Taylor et al. 1997). It is also known that ericoid mycorrhizal dwarf shrubs are able to take up N from dead mycelium (Kerley & Read 1998). We therefore sampled the understorey shrubs on 15 October 2001, 16 and 14 months after girdling of the EG and LG plots, respectively, assuming that their δ15N would be elevated on girdled plots if the shrubs had taken up N from decomposing ectomycorrhizal mycelium. Leaves of C. vulgaris and V. vitis-idaea were sampled near the border of the inner 100 m2 of each plot. The leaves were dried (70 °C, 24 h) and ground to a fine powder in a ball mill before analysis. Percentage N and δ15N were analysed using an on-line C and N analyser coupled to an IRMS (an ANCA-NT solid/liquid preparation module coupled to a model 20–20 IRMS; Europa Scientific Ltd. Crewe, UK, Ohlsson & Wallmark 1999). Results are expressed in δ15N (‰) deviations from the standard atmospheric N2: δ15C = 1000 × (Rsample/Rstandard − 1) (‰), where R denotes the ratio mass 29/mass 28. The standard deviation based on analysis of replicated samples for δ15N was < ±0.1‰.

Sap flow measurements

As a proxy for seasonal variation in canopy photosynthesis, we used data on sap flow (Fig. 1c), which was measured using the heat balance method (Cermak, Deml & Penka 1973; Kucera, Cermak & Penka 1977) with commercially available sap flow meters from EMS (Environmental Measuring Systems, Brno, Czech Republic). Measurements were made on one tree stem approximately 100 m from the girdling experiment, but in the same stand, using internal heating and internal sensing of temperature gradients. Tree-trunk sap flow was measured using two measuring channels on opposite sides of the trunk in order to account for flux variability along the stem circumference.

The water flux at tree level was extrapolated to stand level on the basis of stem circumference. This simplified scaling approach assumed a linear relationship between tree circumference and tree water use (Cienciala et al. 2002).

Statistical analysis

The soil respiration rate and 13C isotopic measurements for each treatment are based on means per experimental plot and presented as means of three plots ± standard error (SE). In calculations of the contribution of root respiration to total soil respiration, treatment mean values were used. Statistical analysis of the isotope data was performed with repeated measures analysis of variance (anova), using SPSS 10.0 (SPSS Inc., Chicago, IL, USA), to compare treatments effect (P < 0.05) on isotopic composition of respired CO2 over two growing season (2000–01). The block effect was excluded from the repeated measure anova, because it was insignificant. For comparison of the δ13Cr on the control and EG plots, the mean isotope data at 11 repeated times were used (i.e. from 27 June 2000–3 August 2001) and the δ13Cr data on the control plot on 21 August 2000 was omitted, because the corresponding δ13Cr data on the EG plot was not available. Similarly, for comparison between the control and LG plots, the mean isotope data at eight repeated times were used (i.e. from 2 September 2000–3 August 2001). The difference in δ13Cr on the control and girdled (data from both EG and LG) plots at the sampling time in 2002 (i.e. on 13 June) was tested at the 5% level of significance, using the two-tailed t-test and assuming equal variances.

RESULTS

Seasonal (2-year) pattern of mycorrhizal root and heterotrophic respiration

In 2000, measured soil respiration (the sum of respiration by mycorrhizal roots and heterotrophs) rates on the control plots ranged from 20 to 60 mg CO2-C m−2 h−1 early and late in the growing season to 150 mg CO2-C m−2 h−1 in mid August (Fig. 1a). The peak in soil respiration coincided with the seasonal peak in sap flow (Fig. 1c). Similarly, in 2001, the soil respiration rates on the control plot ranged from approximately 30 mg CO2-C m−2 h−1 early and late in the growing season to a maximum of 115 mg CO2-C m−2 h−1 in late August However, a decrease of about 30 mg CO2-C m−2 h−1 in soil respiration rate occurred between mid-July and early August before the peak in late August 2001. Simultaneously, there was a 20-day-long drop of approximately 6 °C in soil temperature, namely between mid-July and early August 2001 (Fig. 1d). The lower soil respiration in July–August 2001 as compared to during the same period in 2000 was correlated with the above-mentioned drop in soil temperature.

In the EG plots, soil respiration decreased within 5 d after girdling by 27% relative to that of control plots. This relative difference between the treatments was kept for a further 2 weeks, although soil respiration increased in both treatments. By this time, at the end of June, respiration in the EG plots had reached its seasonal maximum. Soil respiration on the control plots continued to increase for a further 7 weeks until mid-August 2000. This led to 52% lower respiration on the EG plots in comparison with the control plots, which clearly indicates that the increase on the control plots was due to root rather than heterotrophic activity. In the LG plots, soil respiration declined by 37% relative to the control plots within 5 d of treatment (Fig. 1a). This rapid decline in soil respiration continued, and 2 weeks after girdling the CO2 efflux was 56% lower than in the control plots. Thus, respiratory activity on the LG plots rapidly reached a level close to that on the EG plots (Fig. 1a).

In the next year (2001), the soil respiration on the EG and LG plots showed an almost identical seasonal pattern, although the EG plots showed slightly higher respiration rates, especially between early June and early August. During the early and later part of the season, the soil respiration rates on the girdled plots were at about the same level as that of the control plots. However, during relatively warmer periods, namely between end of June and mid-September (Fig. 1d), the soil respiratory activity on the EG and LG plots were 30–60% and 40–70% lower in comparison with the control plots, respectively, depending on time of the season. These data reveal that the relative contribution of roots to soil respiration is less during the early and later part of the growing season, when the air and soil temperatures are low. Thus, the contribution of roots to total soil respiration increased in the summer, and peaked during mid to late summer (Fig. 1b).

The levels of respiration by roots and associated mycorrhizas were calculated by taking the values of respiratory activity on the EG plots and the mean of EG and LG plots as proxies for heterotrophic respiration during 2000 and 2001, respectively, and then subtracting these values from the respiratory activity on the control plots in the respective year (Fig. 1b). Thus, earlier in the summer of 2000, calculated root respiration increased later than heterotrophic respiration, whereas during the autumnal decline (Fig. 1d) heterotrophic and root respiration were similar (Fig. 1b). Overall, the early stem-girdling resulted in approximately 40% decline in soil respiration during the 2000 growing season compared with that on the control plots (Fig. 3a). In 2001, the heterotrophic respiration showed marked variations indicating its sensitivity to changes in soil temperature, such as in connection with the 6 °C drop in soil temperature from mid-July to early August, as mentioned above. However, the calculated root respiration was seemingly not affected by this decrease in temperature. On a cumulative basis, the contribution of heterotrophic respiration (on the EG and LG plots) was about 50% to the total soil respiration (on the control plots) during the 2001 growing season (Fig. 3a). The cumulative respiration of both the growing seasons (Fig. 3b) indicates that on average the EG plots respired the equivalent of 45% of the total soil respiration (i.e. 5.5 Mg CO2-C ha−1 2-year−1) on the control plots.

Figure 3.

Cumulative soil respiration (g C m−2) in the EG (open triangles), LG (open squares) and control (filled circles) plots in Scots pine forest at Åheden: (a) cumulative respiration in each year (i.e. 2000 and 2001); and (b) cumulative respiration of both the years.

δ13C of heterotrophic respiration versus total soil respiration

In 2000, the δ13Cr on control plots showed some variations (Fig. 2a & b). However, in 2001, it did not show variations as those observed in 2000. The early-girdling treatment resulted in an increase (by up to 2‰) in δ13Cr on the EG plots compared to the control plots. Interestingly, this difference between δ13Cr on the EG and control plots remained more or less the same from 27 June 2000 (i.e. approximately 3 weeks after early girdling) to 3 August 2001 (repeated measure anova, P < 0.023) resulting in a similar seasonal pattern of δ13Cr on these plots in 2000 (Fig. 2a). The late-girdling treatment, however, caused a greater increase (up to 4‰) in the δ13Cr on the LG plots approximately 3 weeks after the treatment (i.e. 2 September) compared to the control plots, suggesting an influence of the time of season on the shift in δ13C of respired CO2 (Fig. 3b). This greater difference (repeated measure anova, P < 0.019) was observed up to early August 2001, but disappeared thereafter. Hence, there were no differences in the δ13C of soil respiration among EG, LG and control plots after early August in 2001.

Based on the observation by Ekblad & Högberg (2001), we expected that conditions of high air VPDmax would result in high δ13C of soil respiration in the control plots, but not in the girdled plots, where root activity should be greatly reduced. However, during the years 2000 and 2001, the daily VPDmax of air rarely increased above 20 hPa (Fig. 2c), and we observed the contrary pattern in δ13Cr, namely higher values on the girdled plots, as mentioned above. The warm and dry early June of 2002, however, resulted in the expected high δ13Cr on the control plots (−23.2‰) in connection with high VPDmax (approximately 39 hPa observed a week before the sampling on 13 June 2002) compared to the girdled plots (−24.9‰) (two-tailed t-test assuming equal variances, P < 0.004) (cf. Fig. 2a–c).

As stated above, we analysed δ15N of understorey dwarf shrubs to indicate that the ectomycorrhizal mycelium, which constitute a part of the functional tree ‘root’ system, was decomposed and its N taken up by the dwarf shrubs. We found that the δ15N values of understorey C. vulgaris and V. vitis-idaea on the girdled plots (average of both EG and LG) were enriched by 1.3‰, relative to the values on the control plots (Fig. 4). Similarly, the N contents of C. vulgaris and V. vitis-idaea were 22.0% higher in the girdled plots.

Figure 4.

N concentration (%) and 15N natural abundance (‰) of understorey plants (Vaccinium vitis-idaea L. and Calluna vulgaris L.) from control (filled circles), EG (open triangles) and LG (open squares) plots in Scots pine forest at Åheden.

DISCUSSION

During the first year of this experiment, Högberg et al. (2001) calculated that up to 56% of the soil respiration was derived from roots and their associated mycorrhizal fungi (Fig. 1b). However, this estimate of root respiration was pointed out to be conservative by Högberg et al. (2001), because root starch reserves were shown to be used up after girdling when the above-ground supply of C was curtailed. This was especially important during the early part of the growing season, when the initial starch reserves were larger (Högberg et al. 2001). Furthermore, the contribution from the respiration by understorey dwarf shrubs, and from the enhanced decomposition of starved roots and ectomycorrhizas by heterotrophs on the girdled plots should have led to underestimation of the proportion of soil respiration accounted for by the tree roots.

Högberg & Högberg (2002) showed a decline of on average 32% of soil microbial biomass in the girdled plots 1–3 months after girdling, and attributed this to a loss of biomass of extramatrical mycelium of ectomycorrhizal fungi. This decline was equivalent of approximately 60 kg C ha−1. Such a contribution to soil C could be respired by heterotrophs well within 2 weeks after girdling at the rate at which CO2-C was respired from both the EG and LG plots (Fig. 1a & b). As pointed out above, ectomycorrhizal fungi are highly (5–10‰) enriched in 15N in addition to their approximately 2‰ enrichment in 13C relative to their host trees. The clear increase in N% and δ15N in the ericaceous dwarf shrubs, thus, strongly suggests that the ectomycorrhizal fungal biomass N was turned over and made available to the dwarf shrubs. Simultaneously, and as suggested by the soil CO2 efflux data, the C from this 13C-enriched source was respired contributing to some of the increase in δ13C of the CO2 after girdling. Furthermore, most of the stored starch in roots of girdled trees was also metabolized towards the end of the 2000 growing season. Then, there was only 0.6 and 1.5% starch in fine roots from EG and LG plots, respectively, in comparison with 5.1% in the control (Högberg et al. 2001). Starch and sugars are, like the ectomycorrhizal mycelium, likely to be enriched in 13C relative to the structural C compounds (Brugnoli et al. 1988; Gleixner et al. 1993; Schmidt & Gleixner 1998). The stores of these compounds in fine and coarse roots of this stand should be in the order of 120 kg C ha−1 based on root biomass data from the site (Plamboeck, Grip & Nygren 1999). We suggest that respiration during metabolism of stored sugars and starch, and also during decomposition of ectomycorrhizal mycelium, is one explanation of the transient increase in δ13C of soil respiration after girdling.

Therefore, the second year (i.e. 2001) respiration measurements from the girdled plots, after the transient increase in CO2 efflux described above, should allow more precise estimates of the contribution from heterotrophic respiration. Those estimates would form a more correct basis for the calculation of the contribution by ectomycorrhizal roots to total soil respiration. Hence, based on the average CO2 being respired from the EG and LG plots in 2001, ectomycorrhizal roots contributed up to 65% of the total soil respiration on the control plots (Fig. 1a & b). This supports the previous suggestions by Högberg et al. (2001) that their estimate of root respiration in 2000 was conservative. On a cumulative basis, in 2001, half of total soil respiration was respired by heterotrophs and the other half by ectomycorrhizal roots (Fig. 2a), mainly because root respiration was low early and late in the season.

True heterotrophic activity (we submit that mycorrhizal fungi are an integral component of the root system of the trees) in the soil is supplied with substrate via above-ground litter-fall and root litter. Based on the idea that root growth and mortality is a major fate of photosynthate C allocated below-ground it has been widely held that decomposition of root litter C could be a larger contribution to soil respiration than root respiration (see Högberg, Nordgren & Ågren 2002 and references therein). The large and immediate losses of soil respiratory activity observed in this girdling experiment, however, clearly suggests that the major fate of photosynthate allocated below-ground is root respiration rather than root growth (Högberg et al. 2002). The girdling treatment will probably cause accelerated senescence of roots, but once the standing crop of roots is decomposed there will be no further cohorts of fine roots produced to feed the heterotrophs. The total root biomass in this forest stand (Plamboeck et al. 1999) is 2.35 ton ha−1 (dry matter), which is equivalent of 106 g C m−2. This corresponds to roughly 1–2 months of respiration during the summer, if all these roots were completely decomposed, which is unlikely in such a short time. Hence, even this major substrate to heterotrophs represents a small amount of C in comparison with the total soil respiratory activity. Changes in root litter production, and other processes, such as accelerated needle loss, and thereby increased litter-fall and higher soil temperatures, will confound the interpretation of contributions to soil respiration in the longer term in this kind of experiments. We have chosen to report the data from the two first years, because no treatment effect on soil temperature was observed, and also since there was no notable increase in litter-fall on the girdled plots during that period.

In northern coniferous forests, photosynthesis peaks in the middle of the summer (Linder & Troeng 1980; Troeng & Linder 1982), whereas the allocation of photosynthate-C to roots is thought to be at maximum in the later part of the season (Hansen et al. 1997). The more rapid decrease in soil respiration after late as opposed to early girdling supported this notion (Högberg et al. 2001; Fig. 1a; i.e. the year 2000 data). Moreover, the sap flow in 2000 evidently peaked in August, supporting the idea of a relatively late peak in photosynthesis. The data on soil respiration from 2001 do not contradict the inference on C allocation to roots made above, because the calculated root respiration exceeded heterotrophic respiration in early August to early September Evidently therefore modelling of root respiration should consider seasonal C allocation patterns.

Conventionally, however, models of global change use temperature as the driving variable for soil respiration. Such models are less sophisticated than those used by plant physiologists, which partition root respiration into components, notably maintenance and growth respiration. Nor do such simple models acknowledge differences in short- and long-term responses to changes in temperature (e.g. Larigauderie & Körner 1995; Atkin et al. 2000). Several authors (e.g. Boone et al. 1998; Epron et al. 2001) have recently suggested that root respiration responds more strongly to variations in soil temperature than does heterotrophic respiration. Their arguments are based on higher calculated Q10 values of root respiration as compared to heterotrophic respiration. Moreover, their Q10 values for root respiration are more than twice as high as those of around 2 commonly reported for plant respiration (Atkin et al. 2000). We find these large differences in Q10 between auto- and heterotrophs unlikely. First of all, much of the basic respiratory system is common to bacteria, fungi and plant roots, and the activity of roots, when measured in the field, includes that of their mycorrhizal fungi. Secondly, application of the concept of Q10 requires, in a strict sense, that only the temperature factor varies. But in studies in which Q10 was calculated based on field data from one or several vegetation periods, also many other factors varied; for example, photosynthesis and the allocation of C to the root system, which is a major driving force for root activity. As an example, up to 37% of soil respiration was lost in 5 d (and 56% in 2 weeks) after girdling in August (Högberg et al. 2001). Moreover, Ekblad & Högberg (2001) showed that photosynthates were transported from the canopy to roots of 20-m-tall trees within a few days. We suspect that the high Q10 values previously reported for root respiration is due to the fact that root respiration starts later but stops earlier during a growing season than does heterotrophic respiration (Fig. 1b). Root respiration thus shows a large seasonal variation during a period of a small variation in soil temperature.

We submit that the temperature anomaly of a 20-day-long 6 °C decline in soil temperature in the middle of the summer of 2001 provided a unique opportunity to examine the effect of temperature alone during a short period, when C allocation patterns should show no gross change. In line with our arguments, the heterotrophic activity responded with a drastic decline, but there was no decline of the calculated root respiration (Fig. 1b). However, the root respiration did not increase as fast as it did during the corresponding period in the previous summer, when no such temperature decline occurred. The data suggest that soil temperature does not have the major control on root respiration, provided (of course) that soil temperature is above the freezing point. In our view, heterotrophic respiration could very well be modelled as a function of temperature at mesic boreal sites. However, models of root respiration should consider above-ground plant photosynthetic activity (Craine, Wedin & Chapin 1998; Kuzyakov & Cheng 2001) and allocation patterns of recent photosynthates to roots (Högberg et al. 2001; Pendall et al. 2001). We do not want to imply that root temperature is unimportant globally, but that the C supply seems to be more important under these conditions. Several recent studies suggest that other factors may be as important or more important than soil temperature in governing rates of root respiration (Fitter et al. 1998; Giardina & Ryan 2002; Rey et al. 2002), thereby also questioning a universal role for soil temperature.

We also thought this experiment would be ideally suited to confirm the finding of Ekblad & Högberg (2001) that drought stress-induced high δ13C of photosynthates rapidly affects the δ13C of soil respiration. However, both 2000 and 2001 were relatively wet years (Figs 1d & 2c), and as discussed above, the metabolism of 13C-enriched mycorrhizal mycelium, soluble carbohydrates and starch in roots after girdling initially caused an increase in the δ13C of the soil CO2 efflux rather than a decrease. We had to wait until 2002 to get a period of dry and warm weather and its associated high VPDmax. At that time, the above-mentioned labile and 13C-enriched C sources were also depleted. Then we observed, as expected, that the girdled plots (with only heterotrophic respiration), had a significantly lower δ13C (i.e. up to 1.7‰) of the soil CO2 efflux in comparison with the control plots with their large contribution from root respiration (Fig. 2a & b).

CONCLUSIONS

We have shown that girdling of a boreal tree stand leads to a relatively small transient increase in the apparent soil heterotrophic respiration. Up to two-third of soil respiration after this transient flush was calculated to be respiration by roots, their ectomycorrhizal fungi and other organisms mainly using recent plant photosynthates. This large contribution of root respiration to soil activity provides a new perspective on forest C cycling. We would also like to question the idea that root respiration has a higher Q10 than soil heterotrophic respiration. Models of the root component of soil respiration should consider the importance of the flux of photosynthates to roots; the role played by temperature deserves further critical evaluation.

ACKNOWLEDGMENTS

We gratefully acknowledge funding by the EU (through the project FORCAST), the Swedish Natural Sciences Research Council and the Swedish National Energy Administration. Göte Moen and Leif Ohlsson carried out the girdling, and Anders Ohlsson and Håkan Wallmark analysed the gas and plant samples, respectively.

Received 27 December 2002; received in revised form 3 March 2003; accepted for publication 7 March 2003

Ancillary