Response of an understory plant community to elevated [CO2] depends on differential responses of dominant invasive species and is mediated by soil water availability


  • R. Travis Belote,

    Corresponding author
    1. Department of Ecology and Evolutionary Biology, University of Tennessee, 569 Dabney Hall, Knoxville, TN
    2. 7996, USA;
    Search for more papers by this author
  • Jake F. Weltzin,

    1. Department of Ecology and Evolutionary Biology, University of Tennessee, 569 Dabney Hall, Knoxville, TN
    Search for more papers by this author
  • Richard J. Norby

    1. Environmental Sciences Division, Oak Ridge National Laboratory, Building 1059, PO Box 2008, Oak Ridge, TN 37831; Present address: Engineering-environmental Management, Inc., 1510 Canal Court, Suite 2000, Littleton, CO 80120, USA
    Search for more papers by this author

Author for correspondence: R. Travis Belote Tel: +1 303 721 9219 Fax: +1 303 721 9202 Email:


  • • Rising atmospheric CO2 concentrations are likely to have direct effects on terrestrial ecosystems. Here, we describe effects of elevated concentrations of CO2 on an understory plant community in terms of production and community composition.
  • • In 2001 and 2002 total and species-specific above-ground net primary productivity (ANPP) were estimated by harvesting above-ground biomass within an understory community receiving ambient [CO2] and elevated [CO2] at Oak Ridge National Laboratory's free-air carbon dioxide enrichment (FACE) facility.
  • • During a wet year, community composition differed between plots receiving ambient [CO2] and elevated [CO2], but total ANPP did not differ. By contrast, during a drier year, community composition did not differ, but total ANPP was greater in elevated than ambient [CO2] plots. These patterns were driven by the response of two codominant species, Lonicera japonica and Microstegium vimineum, both considered invasive species in the south-eastern United States. The ANPP of L. japonica was consistently greater under elevated [CO2], whereas the response of M. vimineum to CO2 enrichment differed between years and mediated total community response.
  • • These data suggest that community and species responses to a future, CO2-enriched atmosphere may be mediated by other environmental factors and will depend on individual species responses.


It is well known that the concentration of carbon dioxide ([CO2]) in the atmosphere is increasing and may double this century from preindustrial levels (Houghton et al., 2001). Elevated [CO2] is likely to have direct effects on vegetation (Poorter & Navas, 2003), while also causing changes in climate, which may affect patterns and processes of plant communities. To date, controlled CO2-enrichment studies demonstrate that elevated [CO2] enhances the growth of most plant species grown in monoculture, especially those using the C3 photosynthetic pathway (Poorter et al., 1996; Poorter & Navas, 2003). However, they might not provide realistic information on how plants will respond in natural communities, where the availability of resources is spatially and temporally heterogeneous, and where species interact (Körner & Bazzaz, 1996; Körner, 2000).

With the advent of open-top chamber and free-air carbon dioxide enrichment (FACE) facilities, more studies are beginning to investigate responses of plant communities to elevated [CO2] in more realistic settings (Potvin & Vasseur, 1997; Vasseur & Potvin, 1998; Norton et al., 1999; Niklaus et al., 2001; Shaw et al., 2002). Community responses to CO2 enrichment have generally been described in terms of total production (Koch & Mooney, 1996) or changes in community composition (Körner & Bazzaz, 1996). Increases in the productivity of communities in response to CO2 enrichment depend on species composition (Niklaus et al., 2001; Reich et al., 2001) and interspecific interactions (Stewart & Potvin, 1996; Dukes, 2002). However, elevated [CO2] can alter plant community composition even when total productivity is unaffected (Roy et al., 1996; Norton et al., 1999; Niklaus et al., 2001). Compositional responses to elevated [CO2] have included shifts in species abundance (Vasseur & Potvin, 1998), increased evenness and diversity of communities (Potvin & Vasseur, 1997; Leadley et al., 1999; Niklaus et al., 2001), decreased diversity (Zavaleta et al., 2003), and increased success of invasive species (Dukes & Mooney, 1999; Smith et al., 2000; Weltzin et al., 2003).

However, mixed and sometimes contradictory results have been observed in these studies under variations of resource availability, including light (Bazzaz & Miao, 1993), water (Owensby et al., 1993) and nitrogen (Roy et al., 1996; Cannell & Thornley, 1998). Specifically, resource limitations (e.g. water or nitrogen) typically attenuate the response of plants to elevated [CO2] (Poorter & Pérez-Soba, 2001), but may accentuate differences between different plant species or functional groups growing in natural communities. Owensby et al. (1993, 1999) found that elevated [CO2] increased biomass of a C4-dominated community during dry years. Smith et al. (2000) determined that red brome (Bromus madritensis ssp. rubens), a nonnative annual grass, increased its density and production of above-ground biomass and seeds under elevated [CO2] during a wet, El Niño year. Alternatively, elevated [CO2] may actually diminish the response of plants to other environmental factors (Shaw et al., 2002), although mechanisms contributing to this response are unclear. It is clear, however, that our ability to predict the future effects of elevated [CO2] on species composition, diversity, and invasibility is limited (Körner, 2000).

The present study was designed to determine the response of an in situ understory plant community to elevated [CO2]. We examined community and species responses in 2001 and 2002 to elevated [CO2] in the understory of the FACE facility at Oak Ridge National Environmental Research Park, TN, USA (Norby et al., 2002). We predicted that total production would be greater in plots receiving elevated [CO2] than plots receiving ambient [CO2], because C3 species were common in the understory. In addition, because of the variety of growth forms and functional groups (e.g. C3 herbaceous and woody dicots and monocots, and C4 monocots), we predicted that community composition would change in response to CO2-enrichment. Specifically, we predicted that production of dominant C3 species would be greater under elevated [CO2], because photosynthesis of C3 plants is limited by current [CO2]. Because C4 photosynthesis is saturated at current [CO2], we predicted that the production of C4 species would not differ between ambient and elevated [CO2]. Finally, the community was codominated by several nonnative invasive plant species, so it presented a unique opportunity to investigate whether changes in [CO2] would facilitate the success of invasive plants.

Materials and Methods

Site description

Research was conducted at the FACE facility, Oak Ridge National Environmental Research Park, Oak Ridge, TN, USA (35°54′ N; 84°20′ W). The research site is a planted sweetgum (Liquidambar styraciflua L.) monoculture established in 1988 on an old terrace of the Clinch River (elevation 230 m above sea level). The sweetgum trees were approximately 17 m tall in 2002, with a closed canopy that reduced the light in the understory 70–95% during the growing season (Belote, 2003; Norby et al., 2003). The soil, classified as an Aquic Hapludult, has a silty clay loam texture and is moderately well drained and slightly acidic (water pH approx. 5.5–6.0) (Soil Conservation Service, 1967; van Miegroet et al., 1994). Precipitation is generally evenly distributed throughout the year with an annual mean of 1322 mm. The mean annual temperature at the site is 13.9°C. Additional details about the physical and biological characteristics of the site are described in Norby et al. (2001, 2002).

The understory was sparse in 1997 when the FACE plots were laid out, but by the growing season of 2000 plant cover in the understory was continuous and codominated by two nonnative invasive plant species, Nepal grass (Microstegium vimineum (Trin.) A. Camus) and Japanese honeysuckle (Lonicera japonica Thunb.). Microstegium vimineum, a shade-tolerant C4 annual grass, was first reported in Tennessee in 1917 and since then has spread throughout most of the eastern USA (Fairbrothers & Gray, 1972). Lonicera japonica, a C3 evergreen woody vine, was introduced in 1806 and has become naturalized throughout the south-eastern USA (Leatherman, 1955). Other understory taxa at the FACE site include small clumps or scattered individuals of blackberry (Rubus spp.), goldenrod (Solidago canadensis L.), and box elder seedlings (Acer negundo L.), and about 25 other herbaceous and woody species.

Experimental design

Free-air CO2 enrichment (FACE) technology applies elevated [CO2] to natural systems with minimal effects on light, temperature and precipitation (Hendrey et al., 1999). In 1998, five 25-m diameter plots consisting of two [CO2] treatments were established in the sweetgum plantation (Norby et al., 2002). Four plots were surrounded by 24 vertical vent pipes spaced 3.3 m apart suspended from 12 aluminum towers. Two of the plots received elevated [CO2] (target = 565 p.p.m.) delivered to the vent pipes by blowers, while two control plots received ambient [CO2]. Mean [CO2] in elevated plots was 548 p.p.m. and 552 p.p.m. in 2001 and 2002, respectively. One ambient [CO2] plot with no vent pipes or other infrastructure served as a control for the presence of the [CO2] delivery apparatus (Norby et al., 2002). The [CO2] treatment was initiated in April 1998 and was maintained each year from April to November. Night-time fumigation was discontinued in 2001 because it interfered with soil respiration measurements.

Sampling methods

In March 2001, we randomly distributed four 0.5-m2 subplots within each of the five plots. For the 2002 season, we relocated each subplot to new random locations. At each subplot in each year, we determined foliar cover (%) and above-ground net primary productivity (ANPP; g m−2 yr−1) for all plant species. Cover was estimated in mid-May and late August to capture seasonal differences in plant community composition; maximum intra-annual species foliar cover was used in analyses. In early September, we determined ANPP within each subplot by clipping all individual herbaceous species at ground level. The ANPP of woody perennials (e.g. L. japonica) was determined by marking new shoots in early spring and clipping the subsequent seasonal growth during the destructive harvest. All plant tissue samples were oven-dried at 65°C to constant mass, which was determined by repeatedly weighing biomass samples. We summed all species within each year to determine total understory ANPP. The Shannon–Weaver (H″) diversity index (Shannon & Weaver, 1949), species richness (S), and evenness (E) were calculated for each subplot based on ANPP.

Six time–domain reflectometry (TDR; Soil Moisture Equipment Corp., Santa Barbara, CA, USA) probes were permanently installed within each 25-m diameter plot to determine volumetric water content (%; VWC) in the top 20 cm of soil. The VWC was recorded eight and ten times during the growing season in 2001 and 2002, respectively. We measured photosynthetically active radiation (PAR; µmol m−2 s−1) 1 m above subplots between 11:00 hours and 13:00 hours on clear days using a handheld line integrating ceptometer (AccuPAR; Decagon Devices, Inc., Pullman, WA, USA) twice in each of the 2001 and 2002 growing seasons.

Statistical analysis

We analysed total and species ANPP and biomass, species cover, H′, S, E and PAR for effect of [CO2] treatment with an unbalanced completely randomized design with sampling (CRDS), where each subplot was considered a sample within the plots (Filion et al., 2000). Data were analyzed with a mixed model analysis of variance (anova; procedure MIXED; SAS Institute, 1999) with the model:

yijk = µ + CO2 treatmenti + Rep(CO2)ij + subplot(Rep(CO2))ijk,

(µ is the overall mean; [CO2] treatment is a fixed effect; plot replicate is the random effect; subplots are the residual error that explain the measured dependent variable, yijk) (Filion et al., 1999). To minimize the number of statistical tests, we conducted species-specific comparisons of ANPP and cover for only the five dominant species (present in at least 25% of the subplots and with > 8% cover across treatments and years).

Data were tested for normality and homogeneity of variance with the Shapiro–Wilk W-statistic and Levene's test, respectively (Levene, 1960; Shapiro & Wilk, 1965). Continuous and proportional data that did not meet these assumptions were log-transformed or arcsine square-root-transformed before analysis, respectively. We excluded one outlying observation from the 2001 dataset.

We used principal components analysis (PCA) to analyse and display vegetation data based on maximum foliar cover within subplots. Species present in less than 5% of the subplots were removed from the dataset before ordination, which left 29 and 22 species for ordination in 2001 and 2002, respectively. The correlation matrix was used because of the high diversity of variances among species within the understory (Johnson, 1998). Because distortion problems can occur with PCA (Beals, 1984), we also examined the data using detrended correspondence analysis (DCA). However, the results of the DCA analyses were similar to PCA so we focused our interpretation on the latter. We conducted Pearson correlations of axes I and II scores with annual means of PAR and VWC, and maximum foliar cover of dominant species, within year, to determine environmental variables and species accounting for differences in the plant community, respectively.

The VWC measurements from the six locations within each plot were averaged for each sampling date. We analysed VWC for main and interactive effects of [CO2] and time (i.e. day of year, DOY) with a repeated measures multivariate analysis of variance (manovar; Pillai's Trace in SAS procedure GLM; SAS Institute, 1999).


Community and species production

Total understory ANPP in 2001 did not differ between [CO2] treatments (P = 0.97), but in 2002 it was 27% greater under elevated [CO2] than ambient [CO2] (P = 0.14; Fig. 1). In both years of this study, L. japonica comprised at least 24% of total community ANPP. The productivity of L. japonica was 3.3 and 2.5 times greater in plots that received elevated [CO2] than ambient [CO2] in 2001 and 2002, respectively (Fig. 1; Table 1). In 2001, cover of L. japonica was 3.0 times greater under elevated than ambient [CO2], but in 2002 the L. japonica cover did not differ between treatments (Table 1).

Figure 1.

Contribution of Microstegium vimineum, Lonicera japonica, and other species to total understory community ANPP (g m−2 yr−1) in plots receiving ambient [CO2] and elevated [CO2] treatments at the sweetgum free-air CO2 enrichment (FACE) site in a wet year, 2001 (a) and dry year, 2002 (b). Standard errors and P-values for M. vimineum and L. japonica are given in Table 1; the P-values for total understory production can be found in the Results section.

Table 1.  Above-ground net primary productivity (ANPP) and cover of five dominant (mean cover > 8%) understory plants in free-air CO2 enrichment plots receiving ambient [CO2] and elevated [CO2] in a wet year (2001) and dry year (2002)
SpeciesYear[CO2] TreatmentANPP (g m−2 yr−1)PCover (%)P
  1. Data are means ± SE; n = 3 for ambient [CO2] and n = 2 for elevated [CO2].

Acer negundo2001Ambient  4 ± 2 11 ± 6 
Elevated  2 ± 30.7812 ± 80.96
2002Ambient  1 ± 1  8 ± 4 
Elevated  1 ± 10.95 4 ± 50.70
Lonicera japonica2001Ambient 11 ± 4 21 ± 4 
Elevated 36 ± 50.0363 ± 50.008
2002Ambient 17 ± 7 37 ± 12 
Elevated 42 ± 90.1155 ± 150.42
Microstegium vimineum2001Ambient121 ± 11 85 ± 5 
Elevated 72 ± 150.0767 ± 60.09
2002Ambient 78 ± 14 81 ± 8 
Elevated 70 ± 170.7460 ± 100.19
Rubus spp.2001Ambient  1 ± 1  3 ± 3 
Elevated  3 ± 10.4216 ± 40.08
2002Ambient  1 ± 2  3 ± 5 
Elevated  5 ± 20.5919 ± 60.07
Solidago canadensis2001Ambient  1 ± 3  2 ± 4 
Elevated 18 ± 40.4322 ± 50.02
2002Ambient  0 ± 1  4 ± 3 
Elevated  3 ± 10.76 5 ± 40.79

Microstegium vimineum accounted for at least 64% of total production in both years of the study. The productivity of M. vimineum was 68% greater under ambient than elevated [CO2] in 2001, but did not differ between treatments in 2002 (Fig. 1; Table 1). The M. vimineum cover was 44% greater under ambient than under elevated [CO2] in 2001, but did not differ in 2002.

The productivity of Rubus did not differ between [CO2] treatments in either year, but Rubus cover was greater under elevated than ambient [CO2] in both years (Table 1). The productivity of S. canadensis in ambient and elevated [CO2] plots did not differ in either year. Solidago canadensis cover was greater under elevated than ambient [CO2] in 2001, but not in 2002. Production and cover of A. negundo did not differ between treatments in either years (Table 1). Foliar covers of subdominant species in the understory given are in Table 2.

Table 2.  Maximum cover of understory taxa in plots receiving ambient [CO2] and elevated [CO2] at the sweetgum free-air CO2 enrichment site in 2001 and 2002
Species2001 (%) AmbientElevated2002 (%) AmbientElevated
  1. s, seedling. Data are means ± SE; n = 3 for ambient [CO2] and n = 2 for elevated [CO2].

Acer saccharum (s) 1.5 ± 1.50.2 ± 0.2 3.8 ± 3.10
Allium sp. 3.1 ± 3.10.9 ± 0.500
Asplenium platyneuron 0.1 ± 0.1000
Aster dumosus 00.5 ± 0.300
Bignonia capreolata 0.6 ± 0.302.1 ± 1.50
Carex laevivaginata 0.1 ± 0.11.4 ± 1.000
Cerastium glomeratum 000.1 ± 0.10.1 ± 0.1
Clematis virginiana 1.5 ± 1.5000
Fraxinus pennsylvanica (s)16.5 ± 7.103.8 ± 3.12.2 ± 2.2
Galium aparine 0.6 ± 0.32.0 ± 1.000
Geum canadense 1.6 ± 1.40.4 ± 0.200
Juncus tenuis 0.1 ± 0.10.2 ± 0.20.6 ± 0.30.4 ± 0.4
Juniperus virginiana (s) 01.1 ± 1.100
Lespedeza cuneata 02.5 ± 2.500
Lobelia sp. 00.2 ± 0.200
Myotosis macrosperma 0.4 ± 0.30.4 ± 0.200.1 ± 0.1
Oxalis stricta 0.2 ± 0.10.2 ± 0.200
Panicum clandestinum 0.1 ± 0.15.4 ± 5.400
Panicum sp. 0.1 ± 0.10.5 ± 0.50.6 ± 0.60.8 ± 0.5
Parthenocissus quinquefolia 0.3 ± 0.30.5 ± 0.50.3 ± 0.26.9 ± 4.9
Potentilla simplex 3.5 ± 1.50.2 ± 0.22.6 ± 1.52.3 ± 1.2
Prunus serotina (s) 3.3 ± 1.90.2 ± 0.200
Quercus velutina (s) 1.5 ± 1.5000
Sanicula sp. 0.5 ± 0.32.7 ± 2.40.7 ± 0.60.4 ± 0.4
Solidago erecta 03.6 ± 2.500
Solidago sp. 01.1 ± 1.104.8 ± 2.8
Solidago speciosa 04.1 ± 2.30.1 ± 0.10.8 ± 0.5
Taraxacum officinale 0.2 ± 0.1000
Toxicodendron quercifolia 0.3 ± 0.13.0 ± 2.51.5 ± 1.55.7 ± 4.6
Ulmus sp. 0.1 ± 0.10.5 ± 0.50.4 ± 0.40.1 ± 0.1
Vitis sp. 1.5 ± 1.50.7 ± 0.30.3 ± 0.30.4 ± 0.4

Community structure and diversity

Principal components analysis reflected differences in species composition and abundance under ambient and elevated [CO2] in 2001, but the dissimilarity between treatments was less pronounced in 2002 (Fig. 2). In 2001, axis I and II explained 15% and 11% of the variation, respectively. Axis I scores were negatively correlated with cover of L. japonica, Rubus spp. and S. canadensis (r = −0.43, P = 0.07) and positively correlated with cover of M. vimineum (r = 0.61, P = 0.006).

Figure 2.

Principal components analysis ordination of quadrat scores of the understory plant community receiving ambient [CO2] (open circles) and elevated [CO2] (closed circles) at the sweetgum free-air CO2 enrichment (FACE) site in a wet year, 2001 (a) and dry year 2002 (b).

In 2002, axes I and II explained 14% and 13% of variation, respectively (Fig. 2). Axis I scores were positively correlated with L. japonica (r = 0.52, P = 0.02) and negatively correlated with M. vimineum (r = −0.50, P = 0.02) in 2002. Axis II scores were negatively (r = −0.50, P = 0.03) and positively (r = 0.67, P = 0.001) correlated with A. negundo and S. canadensis, respectively. In both years, neither mean annual PAR nor VWC were correlated with axis I or II quadrat scores (P = 0.24).

In 2001, species richness (S) did not differ between [CO2] treatments, but species evenness (E) and diversity (H′) of species production were greater under elevated than ambient [CO2] (Table 3). In 2002, S, E and H′ did not differ between [CO2] treatments.

Table 3.  Richness, evenness, and Shannon–Weaver's diversity index (H′) of subplots, calculated based on above-ground net primary productivity of all species in plots receiving ambient [CO2] and elevated [CO2])
Year[CO2] TreatmentRichnessPEvennessPHP
  1. Data are means ± SE; n = 3 for ambient [CO2] and n = 2 for elevated [CO2].

2001Ambient5.4 ± 0.9 0.35 ± 0.04 0.6 ± 0.1 
Elevated7.0 ± 1.20.380.54 ± ± 0.10.09
2002Ambient4.2 ± 0.5 0.47 ± 0.05 0.7 ± 0.1 
Elevated5.0 ± 0.60.370.56 ± 0.060.330.9 ± 0.10.27

Soil moisture and photosynthetically active radiation

Soil volumetric water content did not differ between plots under ambient [CO2] and elevated [CO2] in 2001 or 2002 (P = 0.24). The VWC varied throughout the year, depending on precipitation, but the pattern differed substantially between years (Fig. 3). In 2001, VWC was relatively constant throughout June and July, but increased in August (DOY effect, P = 0.0007), whereas in 2002 VWC peaked in May and declined throughout the growing season (DOY, P = 0.0001). In both years, patterns of soil moisture between treatments did not vary through time (CO2 × DOY, P = 0.51). Photosynthetically active radiation (µmol m−2 s−1) above subplots did not differ between [CO2] treatments on any sampling date in 2001 and 2002 (P = 0.34) (Table 4).

Figure 3.

Soil volumetric water content (VWC; %) in plots receiving ambient [CO2] (open circles) and elevated [CO2] (closed circles) throughout the growing seasons of 2001 (a) and 2002 (b). Broken horizontal lines in both figures represent mean VWC during the 2001 growing season to emphasize differences in patterns of VWC.

Table 4.  Photosynthetically active radiation at the top of the sweetgum canopy and at 1 m above subplots in plots receiving ambient [CO2] and elevated [CO2]
DateTop of tree canopy (µmol m−2 s−1)CO2 Treatment (µmol m−2 s−1)
  1. Data are means ± SE; n = 3 for ambient [CO2] and n = 2 for elevated [CO2]. Photosynthetically active radiation (PAR) did not differ between CO2 treatments on any date (P = 0.34).

Aug. 20011451 ± 32 80 ± 20 78 ± 25
Sept. 20011316 ± 24 83 ± 24102 ± 30
May 20021593 ± 77180 ± 24214 ± 29
June 20021672 ± 14167 ± 46 71 ± 56


Results indicate that ecosystem and community responses to future increases in [CO2] will depend on responses of individual species, as well as abiotic environmental factors (e.g. soil moisture). Consistent with other work that observed opposing responses of different plant species to elevated [CO2] (Norton et al., 1999), opposite responses of L. japonica and M. vimineum masked potential differences in total understory production but accentuated compositional (e.g. species composition and diversity) differences during the wet year (2001). Specifically, elevated [CO2] caused a threefold increase in production of L. japonica in both the wet and dry years, whereas M. vimineum varied between years. During the wet year, M. vimineum produced twice as much biomass in ambient [CO2] plots than elevated [CO2] plots, but did not differ between [CO2] treatments in the dry year. These results confirm research in other studies of elevated [CO2] on communities that found that community responses to elevated [CO2] are often unpredictable, in part because of the availability of other resources (Owensby et al., 1993, 1999; Smith et al., 2000).

Heterogeneity of resources in space and time is an important determinant of species composition and production (Tilman, 1982). However, our ability to predict responses of natural systems to elevated [CO2] is limited in that resource availability may either enhance or dampen the effects of elevated [CO2]. For example, in some communities, positive species responses to elevated [CO2] were observed only when availability of water was high (Smith et al., 2000). In other systems, community responses to CO2 enrichment occurred only when the availability of water was limited (Owensby et al., 1999). The photosynthetic pathway of the dominant species may explain the contradictory results. Specifically, C3 species may positively respond to elevated [CO2] by increasing the acquisition of carbon only when water resources are abundant (Huxman & Smith, 2001). By contrast, photosynthesis rates of C4 species are usually CO2-saturated at current [CO2] (Ghannoum et al., 2000), and may only benefit from elevated [CO2] through increased water-use efficiency during dry years (Clark et al., 1999) especially when growing in a community setting (Owensby et al., 1999). Contradictory results, for reasons not yet understood, can occur for other resources, such as light (Bazzaz & Miao, 1993; Poorter & Pérez-Soba, 2001) or nitrogen (Roy et al., 1996; Cannell & Thornley, 1998). Recently, Shaw et al. (2002) suggested that elevated [CO2] might actually diminish the otherwise positive effects of water, nitrogen and warming on California grassland production. The mechanisms driving these patterns are not fully understood, but may include differential species responses to availability of resources (Reich et al., 2001), spatial or temporal variation in resource availability (Tilman, 1982), nitrogen immobilization by soil microbes (Morgan, 2002), or species interactions (Arp et al., 1993; Belote, 2003) both above and below ground. It is clear that more long-term studies with multiple factors in naturalistic settings are needed to better understand potential effects of increasing atmospheric [CO2] on communities (Körner, 2000; Morgan, 2002).

The differential response of species to elevated [CO2] in each year also were responsible for observed differences in composition and diversity under elevated [CO2]. Based on PCA ordinations, opposing responses of the dominant species accentuated compositional differences during the wet year (2001). During the drier year (2002) differences in composition were not as pronounced. Variation in compositional shifts under elevated [CO2] caused by differential responses of species between years is not uncommon in community-level CO2-enrichment investigations (Vasseur & Potvin, 1998; Norton et al., 1999; Niklaus et al., 2001; Marissink & Hansson, 2002).

Differences in diversity between the [CO2] treatments in 2001 and 2002 were driven by the response of M. vimineum. Diversity was higher in elevated [CO2] plots in 2001, as a result of differences in species evenness, which was driven by the decreased production and dominance of M. vimineum. Similarly, elevated [CO2] has altered species diversity in other systems by increasing evenness (i.e. decreasing dominance) within communities by favoring certain species or functional groups (Potvin & Vasseur, 1997; Leadley et al., 1999; Niklaus et al., 2001). However, a recent report suggests that in some systems species, diversity may decrease under elevated [CO2] through species interactions, where small-stature annual forbs are excluded by dominant grasses (Zavaleta et al., 2003).

Species interactions may have been important mediators of plant community responses to elevated [CO2] in the understory community. Others have reported interspecific competition as an important factor mediating community responses to elevated [CO2], especially when certain species benefit from elevated [CO2] to the detriment of other co-occurring species (Owensby et al., 1993; Leadley et al., 1999; Zavaleta et al., 2003). In the present study, elevated [CO2] may have favored L. japonica to the detriment of M. vimineum, which coincides with the pattern in other systems where CO2-enrichment favors C3 species over C4 species (Reynolds, 1996). While L. japonica responded similarly during both the wet (2001) and dry (2002) growing seasons, the response of M. vimineum to variation in soil water availability, possible competition with L. japonica, and potential water savings under elevated [CO2] are not clear. We need to conduct more experiments to determine the relationship between elevated [CO2] and limited availability of water on the potential interactions between L. japonica and M. vimineum.

Response of invasive species to elevated [CO2]

An increase in atmospheric [CO2] may increase the success of nonnative invasive plants by directly enhancing their growth or by altering the availability of resources (Dukes & Mooney, 1999; Smith et al., 2000; Weltzin et al., 2003). However, studies examining the response of invasive plants to elevated [CO2] are rarely conducted in natural communities (Dukes, 2000). This study provides evidence that L. japonica responds positively to CO2 enrichment, not only in monoculture (Sasek & Strain, 1991), but also in natural communities. Thus, as [CO2] continues to rise, L. japonica may become more abundant and pose additional threats to native ecosystems.

Microstegium vimineum is also considered a problematic invasive plant (Redman, 1995), but it responded differently between years to elevated [CO2], possibly as a result of variations in soil moisture. Moreover, interannual variation in production of M. vimineum was an important driver of understory composition and diversity. Thus, abundance and impact of M. vimineum in a future, CO2-enriched atmosphere may be mediated by predicted changes in precipitation patterns and soil water availability (Houghton et al., 2001).

This study demonstrates that community responses to elevated [CO2] are dependent on the responses of individual species. However, other environmental factors, such as water availability, are important mediators of community and species responses to CO2 enrichment. Interannual variation in the production of M. vimineum affected total understory production, community composition, and diversity. Thus, opposing species responses masked potential differences in total understory production but accentuated differences in diversity during the year when soil water availability was high. These data, coupled with previous studies (Sasek & Strain, 1991), suggest that certain invasive species may become more abundant as [CO2] continues to increase. However, increases in [CO2] are likely to affect patterns of precipitation and temperature, so responses of invasive species and their recipient communities to increased [CO2] are likely to be mediated by subsequent changes in other environmental factors.


Melissa Moriarty, Erin Herndon, Philip Allen and Sara Jawdy provided field help with many aspects of the project. Jeff Dukes, Lou Gross and Dan Simberloff provided comments that improved earlier drafts. Chris Fleming, Eugene Wofford and Larry Pounds helped identify plant species; Arnold Saxton and Ann Reed provided statistical advice; Paul Hanson provided soil moisture data. Research at Oak Ridge National Laboratory (ORNL) was funded in part by the US Department of Energy Office of Science, Biological and Environmental Research Program and by the Department of Ecology and Evolutionary Biology, University of Tennessee, Knoxville. Oak Ridge National Laboratory is managed by UT-Battelle, LLC, for the US Department of Energy under contract DE-AC05-00-OR22725.