Rhizodeposition stimulated by elevated CO2 in a semiarid grassland


Author for correspondence: Elise Pendall Tel: +1 307 766 6293 Fax: +1 307 766 2851 Email: Pendall@uwyo.edu


  • • Rhizodeposition, or the addition of C from roots to soil C pools, is expected to increase if net primary production is stimulated and some excess C is allocated below-ground. We investigated the effects of 5 yrs of elevated CO2 on below-ground C dynamics in a native, C3–C4 grassland ecosystem in Colorado, USA.
  • • Cylinder harvests following each growing season and monolith excavation at the end of the experiment provided data on root biomass, root C : N ratios, and root and soil δ13C values. We applied an isotopic mixing model to quantify new soil C inputs on elevated and ambient CO2 treatments.
  • • Root biomass increased by 23% and root C : N ratios increased by 26% after 5 yrs of elevated CO2. Species-specific differences were found in root residence times, which ranged from 6 to 8 yrs.
  • • Rhizodeposition was roughly doubled in elevated compared with ambient CO2 chambers, at 83 ± 16 versus 35 ± 9 g C m−2 yr−1 over the last 4 yrs of the experiment (t-test, P = 0.006). Net C sequestration will depend on how decomposition rates are altered by elevated CO2.


Elevated CO2 stimulates carbon cycling in many ecosystems, but how much additional carbon may be stored in long-lived pools such as soils or wood has been difficult to assess. Perennial grassland soils may store carbon at higher rates when exposed to double CO2 concentrations (Jastrow et al., 2000), although decomposition of soil organic matter (SOM) has been shown to increase (Pendall et al., 2003). Because up to 80% of the biomass and at least 50% of net primary production (NPP) can occur below-ground in grasslands, changes in rhizodeposition will have a large impact on C cyling in these ecosystems (Milchunas & Lauenroth, 2001).

Rhizodeposition is soil C derived from turnover of fine roots, root hairs and mycorrhizas, secretion of soluble root exudates, and turnover of rhizosphere-associated microbial biomass. This C increment is generally too small to measure on an annual basis using total C analysis. However, in experiments where growing plants and pre-existing SOM have contrasting δ13C values, rhizodeposition can be quantified as the portion of soil C derived from the newly labelled plant C, whether that label results from a C3–C4 plant community shift, or from an imposed 13CO2 source (Paterson et al., 1997; Leavitt et al., 2001). Rhizodeposition may account for up to 40% of assimilated C (Paterson et al., 1997), although much of this is rapidly metabolized by the rhizosphere microbial community (Kuzyakov, 2002). The remaining C inputs to SOM can be quantified using isotope techniques.

The fate of C allocated below-ground under elevated CO2 will depend on its decomposability, which is partly related to tissue N concentration. Root N concentration has been shown to decrease by 10–25% under elevated CO2, as C accumulation dilutes N pools (Curtis et al., 1990; Berntson & Bazzaz, 1997; King et al., 1997; Cotrufo et al., 1998; Rogers et al., 1999; Pregitzer et al., 2000). In perennial grasslands like the N-limited shortgrass steppe, autumn senescence of leaves results in the translocation of N to perennial root systems for storage (Charley, 1977). For these reasons, the C : N ratios of below-ground biomass are of interest when evaluating the potential for altered C cycling under elevated CO2.

In a Colorado shortgrass steppe open-top chamber experiment, twice-ambient atmospheric CO2 concentrations stimulated NPP above-ground by 15–35% over 5 yrs (Morgan et al., 2001, 2004). Our goal was to evaluate the changes in root biomass and C : N ratios over 5 yrs, and to quantify the input rate of new soil C, or rhizodeposition, in both ambient and elevated CO2 treatments.


Experimental conditions

An elevated CO2 experiment was conducted in the shortgrass steppe region of north-eastern Colorado, at the USDA-ARS Central Plains Experimental Range (CPER; latitude 40°40′ N, longitude 104°45′ W), about 55 km northeast of Fort Collins. The most abundant species at the study site were the C4 grass, Bouteloua gracilis (H.B.K.) Lag. (blue grama), and the C3 grasses Stipa comata Trin and Rupr. (needle-and-thread grass) and Pascopyrum smithii (Rydb.) A. Love (western wheatgrass). Root biomass (including crowns) is responsible for up to 70% of NPP in this ecosystem (Milchunas & Lauenroth, 2001). Soils were weakly developed fine sandy loams in the Remmit series (Ustollic Camborthids), and were carbonate-free above 40 cm depth.

From 1997 to 2001, open-top chambers (OTC; 4.5 m diameter) were used to evaluate the effects of CO2 on the shortgrass steppe ecosystem, with 3 replicate chambers at ambient (AC, 360 ± 20 ppmv) and elevated (EC, 720 ± 20 ppmv) CO2. Three nonchambered (NC) plots of the same area allowed evaluation of chamber effects. Chambers were placed on the plots before growth started in late March or early April, and removed at the end of the growing season in late October. Blowers with ambient or elevated CO2 ran continuously. The experimental and chamber design was described in detail by Morgan et al. (2001).

An isotopic disequilibrium between growing plants and SOM of 4–5 on AC and NC plots may have been caused by increased C3 biomass following a reduction in grazing at our site (Milchunas et al., 1988). The disequilibrium might also have been partly due to fractionation by mycorrhizas (Henn & Chapela, 2000) or microbial decomposition (Mary et al., 1992). The tank gas used to double the atmospheric CO2 concentration had an approximate δ13C value of −40 during the experiment, which produced air in the elevated chambers of −24.7 ± 1.4, which compared with background air δ13C values of −8.1 ± 0.2 (Pendall et al., 2003). The isotopic disequilibrium imposed by the EC treatment was thus about 16.

Root and crown biomass and C : N ratios

Eighteen months before the start of the experiment (November, 1995), ten 20-cm diameter steel cylinders, 60-cm long, were driven into the soil in each experimental plot to within 1-cm of the soil surface. No litter or water was collected in the cylinders, and above-ground vegetation growing in the cylinders appeared no different from outside the cylinders. At the end of each growing season (1997–2001), two cylinders were removed from each plot, cut open, and soil and roots were removed quantitatively at the following depth intervals: 0–5, 5–10, 10–20, 20–30, 30–40 and 40–60 cm. Weight and gravel content were determined for each interval for bulk density estimation. Most roots were removed immediately after sampling, with additional root picking done by hand on dried subsamples. The same technicians performed the root picking in all 5 yrs of the experiment. Crowns were clipped from the roots at the soil surface, and above-ground biomass was removed. Roots and crowns were pooled by cylinder and not identified to species. Subsamples of roots and crowns were ashed at 450°C to determine the ash-free biomass. Additionally, subsamples were analysed in an elemental analyser for C and N determination (Europa Scientific ANCA-NT System Solid/Liquid Preparation Module, PDZ Europa Ltd., Crewe, UK); precision was 0.1% for both elements. Repeated measures analysis of variance (ANOVAR) with CO2 treatment and depth as fixed effects and year as the repeated measure was used to examine changes in root biomass and C : N ratios over the 5-yr experiment. Fisher's protected least significant difference was used for posthoc mean comparison.

At the end of the experiment, 9 root monoliths were excavated by backhoe (one per plot). One monolith, 20 × 100 cm and 75 cm deep, was removed from each experimental plot from the area that had been reserved for above-ground biomass harvests, and so had remained undisturbed below-ground for the duration. We assumed that experimental removal of the above-ground biomass at the end of each growing season was similar to what would have been removed by herbivory or fire in this ecosystem, and that below-ground herbivory was similar among treatments. For each monolith, entire root systems of several individual plants were separated, so that the isotopic composition of roots with depth could be determined for the three main species (S. comata, P. smithii, and B. gracilis). This was the only time root samples were collected by species; annual cylinder samples pooled all species. δ13C value of roots from the 3 main species were determined by elemental analyser-mass spectrometry (EA-MS; Europa Scientific 20–20 Stable Isotope Analyser). A two-part mixing model (Eqn 1, below) was used to estimate the new C added to roots of each species in the elevated CO2 chambers, using the average δ13C value of C3 or C4 plants from EC plots over the 5 yrs for the ‘new’ end-member, and NC root δ13C values for the ‘old’ end members for each species. The proportion of new root C added over 5 yrs was simply divided by 5 to estimate the annual proportional input rate, the inverse of which is the residence time. This approach assumed steady-state biomass and did not account for storage of pre-label nonstructural carbohydrates (Luo, 2003).

Rhizodeposition or ‘new’ soil carbon

Soil samples from the cylinders were picked by hand to remove visible roots, and then ground to a fine powder for analysis by EA-MS for C content and δ13C value, with precision of 0.1% and 0.2, respectively. Soils were carbonate-free to a depth of 40 cm, and we therefore did not acidify the samples. The fraction of C that has been contributed by rhizodeposition (Fnew) was determined using the two-part mixing model described by Balesdent et al. (1988):

δ13CSOC = Fnew13Cnew) + (1 –Fnew)(δ13Cold)(Eqn 1)

where δ13CSOC is δ13C of soil organic C (SOC) at a given depth interval in AC or EC soils, δ13Cnew refers to the ‘new’ end-member (defined below), and δ13Cold is the δ13C value of nonchambered SOC from the corresponding depth interval. To solve for the fraction (proportion) of new C in any layer, Equation (1) can be rearranged to:

Fnew = (δ13CSOC–δ13Cold)/(δ13Cnew–δ13Cold)(Eqn 2)

The ‘new’δ13C end member was determined as follows. Above-ground biomass clipping of 1.5 m2 subplots was done annually at peak green biomass (July). Leaves and stems were separated by species, dried, ground and analysed for analysis of C content and δ13C value by EA-MS. The leaf δ13C values were weighted by above-ground biomass amounts, after adding 1.4 to the leaf values to account for the average offset between leaves and roots in NC plots, for the final ‘new’ C end member signatures (δ13Cnew; Table 3). The three main species are associated with endomycorrhizas, making it impossible to distinguish C contributions from roots versus mycorrhizas. July harvest samples were assumed to represent the proportion of C3 and C4 species for the entire growing season; leaf samples harvested at the end of the season (peak total biomass) were not easily separated by species. The three dominant species maintain similar shoot : root ratios at ambient and elevated CO2, and we therefore assumed that C3 : C4 ratios in roots were the same as in above-ground biomass (Morgan et al., 1994). We also assumed that δ13Cnew was constant within a season and with depth in the profile. Periodic δ13C measurements of S. comata on NC plots through the dry 2001 growing season suggested that seasonal δ13C dynamics were minimal (standard error < 0.2, n = 8). C inputs from rhizodeposition can also include turnover of mycorrhizas; in the short grass steppe, all root symbionts were endomycorrhizas, and therefore our root analyses of C : N and δ13C included the mycorrhizal community.

Table 3.  Isotopic composition of C3 and C4 plants, the relative percent of C3 biomass, and the δ13C values of the ‘new’ C end-member
 C4C3% C3%‘New’δ13C
  1. se of the mean in parentheses.

AC−16.1 (0.5)−26.5 (0.7)49 (6)−19.8 (1.3)
EC−29.5 (0.4)−40.1 (0.6)61 (3)−34.6 (1.2)
NC−15.4 (0.4)−26.6 (0.8)57 (7) 
AC−14.7 (0.4)−26.9 (1.7)70 (2)−21.8 (2.0)
EC−30.1 (0.1)−41.3 (2.8)79 (8)−37.5 (3.0)
NC−15.1 (1.4)−27.1 (3.7)72 (15) 
AC−15.4 (0.2)−25.2 (0.9)72 (3)−21.1 (0.6)
EC−33.2 (0.8)−42.4 (1.9)81 (4)−39.3 (2.1)
NC−15.4 (0.3)−26.5 (0.5)67 (7) 
AC−13.4 (0.9)−24.0 (2.4)63 (5)−18.7 (2.8)
EC−28.3 (1.0)−32.8 (1.7)77 (4)−30.4 (2.2)
NC−15.2 (1.0)−26.0 (0.9)58 (7) 
AC−16.0 (0.1)−26.9 (0.6)53 (2)−20.4 (1.1)
EC−29.4 (0.7)−38.1 (4.1)64 (9)−33.6 (4.3)
NC−16.4 (0.1)−25.2 (0.9)50 (12) 

Fnew was multiplied by the total mass of C in each horizon or depth increment to calculate rhizodeposition going into each horizon on a mass basis using bulk density estimates. Whole-profile rhizodeposition was calculated by summing the depth increments. Uncertainties in estimating rhizodeposition were determined by accounting for variability in all of the components, including δ13CSOC, δ13Cold, and δ13Cnew, following a first-order Taylor series approach (Phillips & Gregg, 2001). Uncertainty estimates for δ13Cnew included standard errors of percentage biomass and δ13C values of C3 and C4 grasses, as well as the standard error for the average root δ13C. We further corrected for covariance of C content and δ13C values among soil depth intervals when calculating the standard error for the whole profile (Donald L. Phillips, personal communication).


Root biomass and C : N ratios

The root monoliths excavated after the 5-yr experiment appeared to contain more biomass in the elevated CO2 than ambient CO2 chambers, but the trend was not significant. High variability of root biomass within each treatment was observed, despite that each monolith contained ∼250 kg of soil. On average, root biomass per unit area in monolith samples was not significantly different from that in cylinders harvested in 2001.

In cylinders, which were excavated at the end of each growing season, root biomass was significantly increased by elevated CO2 (P = 0.006) and showed significant interannual variability (P < 0.0001; Table 1; Fig. 1a). After increasing from 1997 to 1998 (P = 0.008), root biomass declined in 1999 (P < 0.0001) and again in 2000 (P = 0.028). We investigated the marginally significant year by CO2 interaction (P = 0.066; Table 1) and found that in 1998, root biomass increased under elevated CO2 (EC > AC, P = 0.001) but was reduced by the presence of open-top chambers (NC > AC, P = 0.013). The enhancement of root biomass by EC was therefore stronger than the reduction caused by the chamber effect. Both the CO2 and chamber effects went away in 1999 and 2000 (Fig. 1a). In 2001, significant CO2 effects reappeared (EC > AC, P = 0.014), and the chamber effect diminished below significant levels. Crown biomass is clearly an important portion of the ecosystem C in the shortgrass steppe, and there was a tendency (NS) for higher amounts in elevated than ambient CO2 treatments (Fig. 1b). Interannual variability in crown biomass did not appear to follow any trend. After 5 yrs of exposure to the OTC experimental treatments, the depth distribution of root biomass was not significantly altered (Fig. 2).

Table 1.  Results of repeated measures analysis of variance for root biomass (ash corrected), for 1997–2001, with CO2 and Depth as main effects and Year as the repeated measure
CO2 2  0.0063
Depth 5< 0.0001
CO2 × Depth10  0.6095
Year 4< 0.0001
Year × CO2 8  0.0661
Year × Depth20< 0.0001
Year × CO2× Depth40  0.3535
Figure 1.

Root and crown biomass (ash-corrected) in nonchambered (NC), ambient CO2 (AC) and elevated CO2 (EC) chambers over the 5-yr experiment. Asterisks show years when EC was significantly different from AC (P < 0.05). Error bars show se of the mean of the total biomass. (a) Total root biomass to 60-cm depth. (b) Crown biomass (note different scale). No treatment differences were found in crown biomass.

Figure 2.

Root biomass depth profiles from 2001, after 5 yrs of treatments. Nonchambered (NC), ambient CO2 (AC) and elevated CO2 (EC) chambers.

Root C : N ratios were significantly altered by the CO2 treatments (P = 0.002), with significant interannual variability (P = 0.0008) and CO2 by year interaction (P = 0.0001; Table 2; Fig. 3). In 1998, the main effect was an increase in C : N ratios owing to the presence of open-top chambers (AC > NC, P = 0.021 and EC > NC, P = 0.012), with no CO2 effect. In 1999, the chamber effect was overwhelmed by the elevated CO2 effect, with EC > AC (P = 0.0024) and AC not different from NC. In 2000, the CO2 treatment grew stronger, with EC > AC (P = 0.0001) but a significant chamber effect, opposite to that of 1998, was seen (AC < NC, P = 0.011). After 5 yrs of experimental treatment, in 2001, the maximum treatment effect on C : N ratios in roots was observed, with EC > AC (P < 0.0001), and AC < NC (P = 0.014). A significant interaction between CO2 treatment and depth (P = 0.004) only became apparent in 2001; C : N in EC roots stayed relatively higher at deeper depths, in contrast to the pattern of increasing, then decreasing C : N ratios in NC roots (Fig. 4).

Table 2.  Results of repeated measures analysis of variance for root C : N ratios (ash corrected), for 1998–2001, with CO2 and Depth as main effects and Year as the repeated measure
CO2 2  0.002
Depth 5< 0.0001
CO2× Depth10  0.4367
Year 3  0.0008
Year × CO2 6  0.0001
Year × Depth15  0.0007
Year × CO2× Depth30  0.2703
Figure 3.

Average root C : N ratios in the last 4 yrs of the experiment; average weighted by mass of soil in each depth increment. Nonchambered (NC), ambient CO2 (AC) and elevated CO2 (EC) chambers. Asterisks show years when EC was significantly different from AC (P < 0.05). Error bars were calculated as the se of the mean.

Figure 4.

Depth profile of root C : N ratios in 2001, after 5 yrs of experimental treatment. Nonchambered (NC), ambient CO2 (AC) and elevated CO2 (EC) chambers.

δ13C of roots by species

We noted significant effects of CO2 treatment and plant part (leaves, crowns, roots at 0–20, 20–45 and 45–75 cm) on δ13C values of each species harvested at the end of the experiment. In general, leaves had more negative values than roots or crowns (Fig. 5). This effect was most noticeable in the EC treatment, because root and crown biomass accumulate over several years, whereas leaf biomass represents the current year's growth. Root δ13C values (averaged from 0 to 45 cm) of the most abundant C4 grass at the short grass steppe site, B. gracilis, were 1 higher than leaves in NC plots, 2 higher than leaves in AC plots, and 3.9 higher than leaves in EC plots (Fig. 5a). Root δ13C values of the C3, P. smithii, were 0.5 higher than leaves in NC and AC plots, and 6.4 higher than leaves in EC plots (Fig. 5b). Root δ13C values of the C3, S. comata, were 1.6 higher than leaves on NC plots, 1.4 higher than leaves on AC plots, and 3.4 higher than leaves on EC plots (Fig. 5c). Overall roots averaged 1.4 ± 0.9 heavier than leaves under ambient and nonchambered conditions, and this offset was used to estimate ‘new’ C input end member values (Table 3).

Figure 5.

δ13C values in leaves, crowns, and roots at 3 depths after 5 yrs of experimental treatments in the three most abundant grasses. AC, ambient CO2 chambers; NC, nonchambered control plots; EC, elevated CO2 chambers. (a) δ13C in the C4 grass, Bouteloua gracilis (BOGR). (b) δ13C in the C3 grass, Pascopyrum smithii (PASM). (c) δ13C in the C3 grass, Stipa comata (STCO).

Differences in root δ13C values between EC and NC monoliths were used to estimate the residence time for each grass species in the EC treatment (applying Equation (2) to the root values). After 5 yrs of elevated CO2, new roots made up 84% of S. comata, 63% of P. smithii, and 80% of B. gracilis roots in top 20 cm, corresponding to root residence times of approx. 6 yrs for S. comata and B. gracilis and approx. 8 yrs for P. smithii. In the 20–45 cm increment, 93% of S. comata roots were added during the 5-yr experiment (residence time 5.4 yrs) but similar proportions as in the top 20 cm were added for the other species.


The δ13C of ‘new’ C added annually to soils to 40-cm depth was estimated from the δ13C values of C3 and C4 plants and their proportional cover (Table 3). The ‘new’ end-members in EC plots had very low δ13C values because the tank gas added to the chambers to double the CO2 concentration was depleted in 13C (Pendall et al., 2003), and also because the proportion of C3 biomass increased over the experiment (Morgan et al., 2004). Most of the interannual variability in the average new end-member value resulted from changes in C3 versus C4 plant biomass in all treatments; this proportion varied spatially and decreased in the last 2 yrs because of extremely dry conditions (Table 3). The average δ13C of C3 and C4 plants in AC plots was slightly higher in 2000 than in the other years because a pulse-labelling experiment was conducted to evaluate short-term changes in exudation (Pendall et al., unpublished data). A similar pattern was seen in EC plots because of an unintentional change in tank gas isotopic composition for a few weeks at the beginning of the 2000 growing season. Because 2000 was a very dry growing season, especially in spring, plant growth was limited and only a small portion of the total C was affected by these variations. Nonetheless, we accounted for the interannual shifts in isotopic composition of plant material and proportional biomass when calculating rhizodeposition.

Significant changes in the total mass of soil carbon attributable to elevated CO2 were not found in this experiment (see Appendix). We therefore relied on changes in the stable C isotopic composition of SOM to improve the detection of new C inputs. In EC soils, C was added to all depth intervals every year, and in AC soils, detectable C was added in at least two depth intervals every year (Appendix). When summed up over the top 40 cm, in 1997, rhizodeposition contributed nearly 200 g C m−2 in both AC and EC treatments, and from 1998 to 2001, between 65 and 100 g C m−2 yr−1 was added to EC soils, and 26–40 g C m−2 yr−1 was added to AC soils (Fig. 6).

Figure 6.

Annual rates of rhizodeposition, or ‘new’ soil C inputs, in the top 40 cm of soil over the 5-yr experiment. Error bars represent cumulative errors in isotopic composition of end members and mixtures (Phillips & Gregg, 2001), and include corrections for spatial covariance among values at different depths in the soil (see the Materials and Methods section).


Root biomass and C : N ratios

Elevated CO2 stimulated above-ground biomass production in the short grass steppe OTC experiment by an average of 33% over the 5 yrs, and a chamber effect which lengthened the growing season stimulated above-ground biomass production in AC relative to NC plots by a similar amount (Morgan et al., 2004). After 5 yrs, we found that elevated CO2 stimulated root biomass by 23% in the top 60 cm (cylinder data). Root monoliths excavated to 75-cm depth and sampled at coarser depth intervals showed a trend of increased biomass under elevated CO2 (NS). Elevated CO2 has often stimulated root biomass production, especially in grasslands (Fitter et al., 1997; Jastrow et al., 2000; Higgins et al., 2002), but has sometimes had no effect or even reduced root biomass (Kandeler et al., 1998; Arnone et al., 2000).

The presence or absence of root biomass stimulation, and depth patterns of biomass enhancement are probably related to changing resource availability under elevated CO2. In our experiment, root biomass increased at all depths, but especially in the 5–10 cm increment. Root biomass increased in the top 6-cm of a calcareous grassland but decreased below 12 cm-depth under elevated CO2 (Arnone et al., 2000). Soil water content in the short grass steppe OTC experiment increased significantly, especially lower in the profile (Nelson et al., 2004). When soil resource availability is increased by elevated CO2, root production might be expected to decrease (Kandeler et al., 1998); however, increased water availability was probably responsible for a portion of the biomass enhancement both above- and below-ground in this water-limited ecosystem (Morgan et al., 2001, 2004).

Because the roots of most short grass steppe species are relatively long-lived (5–7 yrs based on 14C labelling; Milchunas & Lauenroth, 2001), we expected the response of our system to be slow. We did not expect to see the strong effect in 1998, after only 2 yrs of treatment, and we do not have a good explanation for the decreasing root biomass in the last 3 yrs of the experiment. A severe drought began in 2000 and lasted through 2001, but this would not explain the decrease beginning in 1999. Minirhizotron and root-ingrowth core data from this experiment should reveal details of inter- and intraseasonal dynamics in root production and its possible relationship to climate patterns (D. Milchunas, unpublished data). We believe that the insertion of the steel cylinders did not have a long-term effect on productivity because root biomass in cylinders in 2001 was similar to that in monoliths.

Decreased N concentrations are associated with lower root mortality or turnover rates, and thus may eventually reduce decomposition rates (reviewed in Pendall et al., 2004a). After 5 yrs of elevated CO2 on the short grass steppe, root C : N ratios increased by 26% when averaged through the whole 60-cm profile, but the increase was only about 6% in the top 10 cm, and about 33% between 10 and 40-cm depth. This depth by CO2 interaction appeared only in 2001, suggesting that the changing plant community composition, driven by elevated CO2 (Morgan et al., 2004), was differentially affecting rooting depth. It seems likely that different rhizosphere microbial communities will evolve at different depths, driven by depth-dependent quality of rhizodeposits, in turn affecting C turnover and nutrient cycling.

Root C : N ratios were not significantly altered at the end of an 8-yr-long tallgrass prairie OTC experiment, but particulate organic matter (POM) C : N in the top 5 cm did increase under elevated CO2 (Jastrow et al., 2000). The C : N ratios of POM, derived mainly from roots in experimental ryegrass swards (Lolium perenne), only increased under elevated CO2 if soil N availability was low (Loiseau & Soussana, 1999). Decomposition rates were slowed as the lower quality litter entered older SOM fractions. Older SOM decomposition rates were stimulated by elevated CO2 in the third and fourth years of the short grass steppe OTC experiment; apparently the differences in C : N we observed were insufficient to suppress decomposition rates at that point in the experiment (Pendall et al., 2003). However, the lower root N concentrations we observed may help explain lower specific root respiration in 1999 (Fitter et al., 1997; Pendall et al., 2003).

Root residence times estimated from changing isotopic composition must be interpreted cautiously, but our estimates in the EC treatment are similar to radiocarbon labelling estimates of 5–7 yrs on bulk roots in a nearby shortgrass steppe study under ambient CO2 conditions (Milchunas & Lauenroth, 2001). Isotope dilution by pretreatment nonstructural carbohydrates may cause root longevity to be overestimated, which may have been a factor in our long-lived root systems. We emphasize that our estimates came from the EC treatment only, and root longevity may be underestimated because root production was higher than in AC or NC treatments (Luo, 2003). However, the increased root C : N ratios on EC plots may counteract the effect of increased production by slowing root mortality. The species-specific response we noted, with S. comata having the shortest root residence time, likely reflects stimulated root growth in tandem with the above-ground biomass response of this species to elevated CO2; the other species in this study were not stimulated (Morgan et al., 2004). Large differences in root longevity between loblolly pine (P. taeda) and sweetgum (Liquidambar styraciflua) growing under elevated CO2 also point to the need for careful assessment of species effects in global change research (Matamala et al., 2003).

Rhizodeposition (‘new’ soil C inputs) and implications for altered soil C storage

Our isotopic approach defined rhizodeposition to be derived from recent photosynthate, fixed within each growing season. The isotopically labelled C accumulated over the 5 yrs experiment, but we evaluated inputs for each year. Root exudates, turnover of fine roots, and root crowns produced within the growing season may all contribute to rhizodeposition as measured by stable isotopes. In our study, endomycorrhizas were inherently included in the analysis of root C and N, and may have been partly responsible for the isotopic offset between leaves and roots. Pulse labelling studies may find most rhizodeposition allocated to root exudation, but continuous labelling studies such as ours are likely to quantify C pools that span a range of qualities with different decomposition dynamics (Paterson et al., 1997). As with most below-ground process studies, quantitative and qualitative responses of rhizodeposition to elevated CO2 are method-dependent (Pendall et al., 2004a).

In the initial year of experimental treatment in a Swiss grassland, new C inputs (determined from a continuous 13C label) were 210 g m2 yr−1 in the top 10-cm of soil (Niklaus et al., 2001). Initially rapid rates of new soil C inputs were also found in a Free-Air CO2 Enrichment (FACE) study on L. perenne and Trifolium repens pastures (Van Kessel et al., 2000). In both these studies, relatively high input rates were very nearly balanced by root turnover after the first year, because root residence times were < 2 yrs. In our study, the high input rates during the first year may have resulted from reduced soil moisture inside the chambers, which was subsequently remedied. This chamber effect in 1997 likely resulted in increased root mortality in both chambered treatments, some of which was distinguished by stable isotopes as rhizodeposition. Possibly, insertion of the cylinders contributed to variable root biomass and rhizodeposition dynamics in the first year or two. Nonetheless, elevated CO2 increased rates of rhizodeposition for the last 4 yrs of our study; rhizodeposition was roughly doubled in EC compared to AC treatments, averaging 83 ± 16 versus 35 ± 9 g C m−2 yr−1 over the last 4 yrs (t-test, P = 0.006).

Although the current study does not specifically address the fate of the rhizodeposits added under elevated CO2, widening C : N ratios of root biomass suggest that root turnover and decomposition rates will decrease, especially at depth. In a short-term microcosm study with L. perenne, radiocarbon labelling showed significantly increased rhizodeposition under elevated CO2, and the proportion of this substrate incorporated into microbial biomass was reduced, indicating that microbes preferred utilizing older SOM with higher C : N ratios (Paterson et al., 1999). Cardon et al. (2001) suggested that labile rhizodeposits were preferred by microbes over older SOM, at least when ample N was available, in an annual grassland elevated CO2 experiment with mesocosms and a transplanted C4 soil. In the shortgrass steppe OTC experiment, N appeared to be limiting enough to microbes that ‘mining’ and mineralization of older SOM occurred, at least during the first 3–4 yrs (Pendall et al., 2003).

Whether stimulation of rhizodeposition by elevated CO2 will result in a net accumulation of C in soils depends on whether pre-existing SOM is simply replaced by the new inputs. In the shortgrass steppe OTC experiment, stable isotopes provided constraints on the new C inputs as well as old C losses in both EC and AC treatments, and showed that net ecosystem production (NEP) did not change in 1999, a relatively wet year, as inputs balanced losses (Pendall et al., 2004b). This grassland appears to maintain a C balance close to steady state under elevated CO2, as decomposition increased nearly as much as rhizodeposition, which explains why total soil C pools did not change over 5 yrs of treatments.

Advantages and limitations of stable isotope techniques

Although the 13C approach has considerable utility and potential for quantifying C fluxes from plants, C inputs, and turnover rates of SOM (Leavitt et al., 1994, 2001; Paterson et al., 1997), certain assumptions must be met for appropriate interpretation of the results. For instance, the end-members must be well characterized, and their δ13C values should remain reasonably constant over space and time. In our case, we characterized spatial and temporal variations in the ‘new’ end-member, adjusted for a small offset between above- and below-ground δ13C values, and also measured ‘old’ end-member values from NC soils annually. The main limitation for this study was partitioning new C additions on the AC plots, because isotopic end-members were only about 4 apart; no new additions were detectable in nearly 20% of the cases (see Appendix).

To our knowledge, this is the first paper using δ13C to calculate soil C inputs which has accounted for all measurable sources of error; we used a Taylor-series approach to accumulate the errors in the end-members and the mixtures (Phillips & Gregg, 2001), and furthermore explicitly accounted for covariance among soil depths in δ13C and C content values. The main uncertainties in our ‘new’ end-member δ13C values stemmed from unknown contributions of leaf, root, and crown biomass to the new C pools, and how this varied with depth in the soil. Additional unknown uncertainties may result from differential decomposition dynamics of C3 and C4 plants, which would affect the δ13C of litter as it decomposes (presumably this may include roots; Wedin et al., 1995). Finally, root and leaf biomass is mainly structural, which may have a different δ13C value than metabolic C, while rhizodeposition is considered by some to include mainly metabolic C (Paterson et al., 1997). Our approach using δ13C values from bulk soil C and plant tissues included a range of compounds of varying quality. The accuracy of the 13C method will be improved as we learn more about the physiochemical nature of C inputs to soil and how they change isotopically during decomposition.


This research was supported by the Biological and Environmental Research Program (BER), US Department of Energy, through the Great Plains Regional Center of the National Institute for Global Environmental Change (NIGEC) under Cooperative Agreement DE-FC03–90ER61010; by Terrestrial Ecology and Climate Change Initiative, grant NSF-IBN-9524068; USDA/NRICGP-98–134; and from NSF DEB-9708596; with base support from USDA/ARS. We thank Donald Phillips for discussions regarding error analysis; and Dan LeCain, Mary Smith, Jim Nelson, and Susan Crookall for help in the field and the laboratory. The manuscript benefited from comments from three anonymous reviewers.


Bulk soil C content and isotopic composition, and new C input rate

Trt, yearDepth (cm)Total soil C (gm−2)δ13C ()C inputs (gm−2)
AC 0–5 617 (61)−17.7 (0.9) 72 (56)
AC 5–10 520 (38)−16.5 (0.8)  0
AC10–20 842 (24)−15.5 (0.1) 30 (9)
AC20–30 909 (29)−15.1 (0.1) 53 (14)
AC30–40 737 (66)−14.6 (0.3) 33 (14)
EC 0–5 645 (129)−18.8 (0.2) 49 (5)
EC 5–10 540 (90)−17.8 (0.1) 37 (3)
EC10–20 857 (115)−16.2 (0.2) 36 (2)
EC20–30 864 (61)−15.8 (0.3) 45 (3)
EC30–40 702 (88)−15.2 (0.4) 29 (2)
NC 0–5 545 (83)−17.4 (1.0) 
NC 5–10 526 (51)−16.6 (0.7) 
NC10–20 881 (95)−15.4 (0.3) 
NC20–30 816 (105)−14.8 (0.4) 
NC30–40 666 (110)−14.4 (0.4) 
AC 0–5 527 (44)−17.7 (0.5)  0
AC 5–10 569 (60)−16.5 (0.5)  0
AC10–20 882 (24)−15.2 (0.6)  0
AC20–30 837 (44)−14.9 (0.6) 23 (7)
AC30–40 778 (121)−15.1 (0.4) 65 (18)
EC 0–5 542 (74)−19.5 (0.5) 42 (7)
EC 5–10 561 (32)−18.4 (0.3) 39 (6)
EC10–20 906 (18)−16.1 (0.6) 29 (4)
EC20–30 899 (54)−15.7 (0.3) 36 (5)
EC30–40 817 (102)−14.8 (0.2) 10 (1)
NC 0–5 571 (115)−17.8 (0.9) 
NC 5–10 591 (66)−16.9 (0.3) 
NC10–20 869 (19)−15.4 (0.4) 
NC20–30 914 (130)−14.7 (0.2) 
NC30–40 698 (97)−14.5 (0.2) 
AC 0–5 488 (44)−18.4 (0.3) 18 (4)
AC 5–10 510 (39)−17.0 (0.3)  0
AC10–20 872 (39)−15.7 (0.1) 25 (3)
AC20–30 877 (47)−14.9 (0.2) 35 (3)
AC30–40 857 (118)−15.1 (0.3)  0
EC 0–5 522 (138)−19.9 (0.3) 41 (4)
EC 5–10 470 (110)−18.8 (0.3) 35 (3)
EC10–20 891 (62)−16.8 (0.1) 46 (4)
EC20–30 908 (55)−15.9 (0.2) 46 (4)
EC30–40 798 (102)−16.0 (0.3) 29 (3)
NC 0–5 540 (93)−18.3 (0.2) 
NC 5–10 596 (89)−17.2 (0.5) 
NC10–20 924 (45)−15.6 (0.4) 
NC20–30 963 (113)−14.6 (0.2) 
NC30–40 824 (113)−15.1 (0.4) 
AC 0–5 618 (103)−18.3 (0.1)  0
AC 5–10 596 (52)−17.1 (0.3) 28 (9)
AC10–20 962 (28)−15.9 (0.2) 51 (9)
AC20–301005 (101)−15.3 (0.1) 43 (4)
AC30–40 891 (141)−14.9 (0.2) 44 (3)
EC 0–5 616 (112)−20.8 (0.4)119 (8)
EC 5–10 581 (66)−18.9 (0.1) 79 (3)
EC10–20 912 (53)−17.2 (0.2) 92 (3)
EC20–30 910 (37)−16.3 (0.2) 66 (2)
EC30–40 784 (117)−15.8 (0.1) 55 (1)
NC 0–5 573 (47)−18.5 (0.6) 
NC 5–10 573 (61)−17.0 (0.4) 
NC10–20 900 (60)−15.8 (0.4) 
NC20–30 917 (60)−15.2 (0.3) 
NC30–40 810 (100)−14.7 (0.2) 
AC 0–5 603 (123)−18.1 (0.3)  0
AC 5–10 574 (33)−17.1 (0.4) 53 (7)
AC10–201056 (108)−16.2 (0.4) 84 (10)
AC20–301039 (79)−15.4 (0.4) 14 (1)
AC30–401008 (152)−15.0 (0.4)  0
EC 0–5 593 (47)−21.0 (0.3)106 (7)
EC 5–10 589 (32)−19.7 (0.1)101 (5)
EC10–201022 (67)−17.6 (0.2)102 (3)
EC20–30 947 (71)−16.9 (0.2) 82 (2)
EC30–40 797 (110)−16.3 (0.3) 46 (1)
NC 0–5 758 (71)−18.2 (0.6) 
NC 5–10 630 (57)−16.8 (0.3) 
NC10–201090 (59)−15.8 (0.4) 
NC20–301037 (101)−15.3 (0.3) 
NC30–40 909 (87)−15.3 (0.5)