Does a decade of elevated [CO2] affect a desert perennial plant community?



  • Understanding the effects of elevated [CO2] on plant community structure is crucial to predicting ecosystem responses to global change. Early predictions suggested that productivity in deserts would increase via enhanced water-use efficiency under elevated [CO2], but the response of intact arid plant communities to elevated [CO2] is largely unknown.
  • We measured changes in perennial plant community characteristics (cover, species richness and diversity) after 10 yr of elevated [CO2] exposure in an intact Mojave Desert community at the Nevada Desert Free-Air CO2 Enrichment (FACE) Facility.
  • Contrary to expectations, total cover, species richness, and diversity were not affected by elevated [CO2]. Over the course of the experiment, elevated [CO2] had no effect on changes in cover of the evergreen C3 shrub, Larrea tridentata; alleviated decreases in cover of the C4 bunchgrass, Pleuraphis rigida; and slightly reduced the cover of C3 drought-deciduous shrubs.
  • Thus, we generally found no effect of elevated [CO2] on plant communities in this arid ecosystem. Extended drought, slow plant growth rates, and highly episodic germination and recruitment of new individuals explain the lack of strong perennial plant community shifts after a decade of elevated [CO2].


Elevated [CO2] enhances photosynthesis, water-use efficiency, and biomass accumulation in many species (Nowak et al., 2004a; Ainsworth & Long, 2005). However, the dynamics of plant communities can seldom be predicted from individual plant responses, and few studies have examined how intact plant communities respond to elevated [CO2]. Prediction of community responses might be simplified if functional groups responded in predictable ways to elevated [CO2] based on their underlying physiology and phenology. For example, C3 species are expected to realize greater benefits in productivity from elevated [CO2] than C4 species (Bowes, 1993), which would, in turn, alter the composition of mixed C3–C4 communities and thus ecosystem function (Polley et al., 2012). By contrast, we have little comparative information about how evergreen and deciduous species may respond to elevated [CO2].

Whether plants respond by species or as functional groups, it is likely that elevated [CO2] will modify resource limitations and potentially alter competitive interactions. Altered competition will affect the persistence and representation of particular species in a community, with effects on relative dominance, species richness, and species diversity. Previous work measuring diversity under elevated [CO2] has not yielded a consistent response; studies have reported that diversity increased (Niklaus et al., 2001; He et al., 2002), decreased (Zavaleta et al., 2003), or remained the same (Navas et al., 1995; Morgan et al., 2007). The variability in community responses suggests that initial species composition and higher-level interactions are key elements in elevated [CO2] community responses (He et al., 2002; Davis & Ainsworth, 2012). The complex interplay among species interactions (competition, facilitation, and herbivory; Reich et al., 2001; Nowak et al., 2004a), resource limitations (water and nutrient availability), and species or functional group presence and relative representation (Curtis & Wang, 1998; Bradley & Pregitzer, 2007) make studies of intact ecosystems the best predictors of community responses to elevated [CO2]. Therefore, there is a great need for in situ long-term studies to understand how individual species and plant communities may respond to rapidly changing [CO2] in the Earth's atmosphere.

Despite the clear value of community-level experiments, few studies have assessed the effects of elevated [CO2] on diversity parameters. At many free-air CO2 enrichment (FACE) sites, the long life span of perennials and the necessity of maintaining intact, unmanaged vegetation have precluded community-level analyses (Nowak et al., 2004a). Thus, most experiments evaluating community responses to elevated [CO2] have involved short-lived species in old fields, grasslands, and annual communities (Morgan et al., 2004a; Ramseier et al., 2005; Polley et al., 2012). There is a particular dearth of studies addressing effects of elevated [CO2] on aridland ecosystems, which represent substantial amounts of global land cover. Community responses to elevated [CO2] may be stronger in arid regions because desert plants may be particularly responsive to elevated [CO2] conditions as a result of improved water-use efficiency (Melillo et al., 1993). Species richness in deserts is often low because of environmental constraints (Kier et al., 2005) and could be further reduced with increasing [CO2] as a result of competition with invasive species (Smith et al., 2000). In addition, aridland plant community responses to elevated [CO2] may be complicated by episodic droughts common to desert regions, which could limit desert plants' ability to respond to elevated [CO2] (Morgan et al., 2004b; Jasoni et al., 2005; Housman et al., 2006). Previous research has found short-term, positive effects of elevated [CO2] on above-ground or below-ground plant growth in the Mojave Desert but only in wetter years (Housman et al., 2006; Ferguson & Nowak, 2011). However, elevated [CO2] did not affect perennial plant standing biomass both above and below ground at the end of the 10 yr FACE experiment (Newingham et al., 2013). Although there was no effect of elevated [CO2] on final standing biomass, it is possible that there were changes in plant community characteristics under elevated [CO2] over the course of the experiment.

We examined plant community changes in an undisturbed Mojave Desert perennial plant community exposed to 10 yr of elevated [CO2]. Plant community shifts were assessed by examining changes in cover, species richness, and species diversity. We predicted that elevated [CO2] would increase cover through higher canopy production of the dominant species, which would decrease species richness and diversity over time. We also predicted that C3 shrub species would respond more strongly to elevated [CO2] than a C4 bunchgrass, and that the dominant evergreen shrub, Larrea tridentata, would have a stronger response to elevated [CO2] than drought-deciduous shrubs and perennial grasses.

Materials and Methods

Experimental site

The Nevada Desert FACE Facility (NDFF) is located in southern Nevada on the Nevada National Security Site (formerly Nevada Test Site; 36°46′12.009″N, 115°57′54.173″W, 970 m). Mean annual precipitation is c. 146 mm, with most precipitation occurring as winter rainfall and episodic summer monsoonal rains. The NDFF consisted of nine circular plots, each 23 m in diameter, within native plant communities in the Mojave Desert. CO2 was distributed using FACE technology, and plots were spaced at least 100 m apart to avoid CO2 contamination in untreated plots. Three plots received continuous exposure to elevated [CO2] (550 μmol mol−1 target; ‘elevated’) except when air temperature fell below 4°C or wind speed (5 min average) exceeded 7 m s−1, resulting in an average daytime [CO2] of 544 μmol mol−1 and a 24 h average [CO2] of 513 μmol mol−1. Three plots received ambient air ([CO2] averaged 375 μmol mol−1 through the FACE infrastructure; ‘ambient’), and three control plots did not have the FACE infrastructure (‘nonblower controls’). Previous results suggest there is no difference between ambient [CO2] and nonblower control plots at the NDFF (Nowak et al., 2001). Suspended walkways with an attached sampling platform were used to eliminate disturbance of plants, biological soil crusts, and soils (Jordan et al., 1999). Carbon dioxide fumigation commenced in April 1997 and ended in June 2007. Owing to extreme drought conditions in the final year of the experiment, we irrigated all plots in March 2007 with c. 30 mm of water to stimulate perennial green-up before terminating the experiment. This was necessary to break drought dormancy and allow for the final experimental harvest.

Experimental plots were located in an intact L. tridentataAmbrosia dumosa desert scrub community, which has been closed to the public and livestock grazing for over 50 yr. The five most common perennial species included a C3 evergreen shrub (Larrea tridentata (DC.) Coville), C3 drought-deciduous shrubs (Ambrosia dumosa (A. Gray) Payne, Lycium andersonii A. Gray, and Lycium pallidum Miers), and a C4 bunchgrass (Pleuraphis rigida Thurb.). Other perennial shrubs, grasses, and forbs (‘other species’) comprised the remainder of the plant community. Perennial plants that were clearly dead were identified to species when possible, and unidentifiable dead individuals were placed in an unidentifiable dead plants category. Our study focused on long-term above-ground responses by perennial species as a proxy for community structure. While annual plant communities are an important component of this ecosystem, their presence or absence in response to fluctuating resources warrants separate consideration and is discussed elsewhere (Smith et al., 2013).

Measurements and calculations

We recorded the species, height (h), widest canopy diameter (c1), and the canopy diameter perpendicular to the widest diameter (c2) for each individual plant in every plot at the beginning of the experiment (June 1997–September 2000) and in 2007 just before the final experimental harvest. In the initial measurements, 7072 plants were measured; in 2007, 5780 plants were measured. There were differences in sampling date by treatment for the initial measurements; elevated [CO2] plots were measured between June 1997 and August 1998, ambient [CO2] plots were measured between September 1997 and February 1999, and nonblower control plots were measured between December 1997 and September 2000. We plotted the average sampling date for each plot against total cover, species richness, and Larrea cover. There was no trend associated with the average date of sampling, so we concluded that the sampling date differences at the beginning of the experiment did not bias our results.

The initial and final canopy cover for each individual was calculated as the area of an ellipse (π × 0.5c1 × 0.5c2) and then summed to estimate total cover per ring per species. After estimating their cover, the identifiable and unidentifiable dead plants were placed in a ‘dead plants’ category for analysis. Dead plants were not measured in three plots (two elevated, one ambient) during the initial sampling period, so the average cover of dead plants from the six other plots in the initial sampling period were used to replace the missing values. Absolute cover for the five dominant species, dead plants, and ‘other species’ was calculated, as well as the relative cover (cover for each species or category divided by total perennial plant cover in the plot). Species richness was defined as the total number of perennial species observed, while diversity was calculated from the cover data using the Shannon diversity index. Both species richness and diversity excluded dead plants.

Statistical analyses

The split-plot experimental design was identical for all dependent variables, and all data were analyzed using mixed-effects ANOVA. The [CO2] treatment was applied to entire plots. Each variable was sampled at two time points within each plot (initial and final). Therefore, all models contained [CO2] treatment as a fixed effect (tested over plot within [CO2] treatment), a fixed effect of time, and an interaction between [CO2] treatment and time (both tested over the interaction between time and plot). We also estimated the differences between final and initial values for each [CO2] treatment as a priori contrasts. A significant [CO2] treatment effect alone may indicate that there were pre-existing differences among plots. Therefore, a significant treatment × time interaction or differences in the a priori contrasts within each treatment across time were necessary to suggest elevated [CO2] effects. Analyses were conducted using SAS 9.2 (SAS Institute 2002–2008, Cary, NC, USA), and details unique to each analysis are included below.

Because we were particularly interested in how the five most common species individually responded to [CO2] treatments, we also analyzed the cover of each of the five most common species, dead plants, and ‘other species’ in mixed-effects ANOVAs individually as already described. Effects of time and [CO2] treatment on total cover, species richness, and species diversity were also analyzed in separate mixed-effects ANOVAs as already described. Species diversity was back-transformed to species number equivalents for descriptive purposes. A species number equivalent is the number of equally represented species that would yield the diversity index value (Jost, 2006).


During initial sampling, 30 species of perennial grasses, forbs, cacti, and shrubs were observed across all plots, while 23 species were observed in 2007 (Table 1). Total cover, species richness, and species diversity significantly decreased over the course of the experiment, but the decreases between the initial and final sampling were similar for all [CO2] treatments (Table 2; Fig. 1).

Table 1. Perennial plant species observed at the Nevada Desert Free-Air CO2 Enrichment (FACE) facility during initial (1997–2000) and final (2007) sampling
SpeciesCommon nameGrowth formInitialFinal
  1. A letter indicates that the species was present in a [CO2] treatment for the listed time period, where ‘E’ is for elevated [CO2], ‘A’ is for ambient [CO2], and ‘C’ is for the nonblower control.

Achnatherum hymenoides Indian ricegrassBunchgrassEACEA
Acamptopappus shockleyi Shockley's goldenheadSubshrubEACEAC
Ambrosia dumosa White burrobushShrubEACEAC
Androstephium breviflorum Pink funnel lillyForbE 
Argemone corymbosa Prickly poppyForbEA 
Astragalus tidestromii Tidestrom's milkvetchForbEACAC
Atriplex canescens Fourwing saltbushShrubAA
Baileya multiradiata Desert marigoldForbEACEAC
Delphinium parishii Desert larkspurForbAC 
Encelia virginensis Virgin River brittlebushShrubE 
Ephedra nevadensis Nevada jointfirShrubEACEAC
Eriogonum inflatum Desert trumpetForbEACE
Grayia spinosa Spiny hopsageShrubACEAC
Hymenoclea salsola CheeseweedSubshrubAA
Krameria erecta Littleleaf ratanyShrubEACEAC
Krascheninnikovia lanata WinterfatSubshrub/shrubEACEAC
Larrea tridentata CreosotebushShrubEACEAC
Lycium andersonii Anderson's wolfberryShrubEACEAC
Lycium pallidum Pale wolfberryShrubEACEAC
Mirabilis pudica Four o'clockForbEACAC
Opuntia basilaris Beavertail pricklypearCactusEACEAC
Opuntia echinocarpa Staghorn chollaCactusEACEAC
Opuntia ramosissima Pencil cactusCactusECEAC
Pleuraphis rigida Big galletaBunchgrassEACEAC
Polygala subspinosa Spiny milkwortSubshrub/shrubEACEAC
Psorothamnus fremonti Indigo bushShrubEACEAC
Sphaeralcea ambigua Desert globemallowForb/subshrubEAC 
Sphaeralcea grossulariifolia Gooseberry leaf globemallowForb/subshrubAC 
Stephanomeria pauciflora WirelettuceForb/subshrubEACEA
Thamnosma montana Turpentine-broomSubshrubA 
Table 2. ANOVA for total cover, species richness, and species diversity across treatment (nonblower control, ambient [CO2], and elevated [CO2]) and time
EffectdfCoverSpecies richnessSpecies diversity
  1. Degrees of freedom (df) represent numerator degrees of freedom; denominator degrees of freedom were 6 for all effects. Bold values are statistically significant at α = 0.05.

Time123.8 0.003 16.9 0.006 49.3 0.001
Treatment × time21.00.4101.00.4390.30.722
Figure 1.

Percentage cover, species richness, and diversity (expressed as species number equivalents) for all species treated with nonblower controls, ambient [CO2], and elevated [CO2] at initial (closed bars) and final (open bars) sample points (1997–2007). For percentage cover, species richness, and diversity, the only significant effect was time at α = 0.05. Values are mean ± SE.

We further explored the decrease in cover during the study by examining the five most common individual species. The relative cover of Ambrosia and Pleuraphis decreased while L. pallidum, L. andersonii, and Larrea increased over time; the ‘other species’ group did not change substantially during the study (Tables 3, 4; Fig. 2). There were treatment effects within these larger trends. Ambrosia dumosa relative cover decreased more in elevated [CO2] than in nonblower control and ambient [CO2] (Tables 3, 4; Fig. 2). Lycium andersonii and L. pallidum relative cover increased in nonblower control and ambient [CO2] plots over time but not under elevated [CO2] (Tables 3, 4; Fig. 2). By contrast, the ‘other species’ group had little change in cover for any treatment, and no differences were noted over time relative to [CO2] treatment (Tables 3, 4; Fig. 2). The absolute and relative cover of Larrea tended to increase during the study, but this increase occurred in all CO2 treatments, although it was greatest in nonblower control and elevated [CO2] plots (Tables 3, 4; Fig. 2). Pleuraphis relative cover decreased during the study (Tables 3, 4), but the relative cover of Pleuraphis decreased less in elevated [CO2] plots than in nonblower control or ambient [CO2] (Fig. 2). Overall, the patterns for the drought-deciduous shrubs and ‘other species’ grouped together suggest that elevated [CO2] negatively impacted their cover relative to Pleuraphis during the study (Fig. 2).

Table 3. ANOVA table for the effects of treatment (nonblower control, ambient [CO2], and elevated [CO2]) and time on the absolute (percentage of plot) and relative (each species relative to total plant cover) cover of the five most common species, dead plants, and other species
SpeciesEffectdfAbsolute coverRelative cover
  1. Degrees of freedom (df) represent numerator degrees of freedom; denominator degrees of freedom were 6 for all effects. Bold values are statistically significant at α = 0.05. Italicized values are statistically significant at α = 0.10.

Ambrosia dumosa Treatment22.70.1482.90.133
Time125.3 0.002 12.3 0.013
Treatment × time21.60.2761.20.369
Larrea tridentata Treatment20.80.5020.90.449
Time16.9 0.040 56.6 < 0.001
Treatment × time22.30.1801.70.267
Lycium andersonii Treatment21.30.3502.50.165
Time114.2 0.009 9.4 0.022
Treatment × time25.6 0.043 7.8 0.021
Lycium pallidum Treatment20.00.9630.30.747
Time10.40.55515.9 0.007
Treatment × time22.20.1881.70.265
Pleuraphis rigida Treatment22.00.2191.80.237
Time128.4 0.002 35.5 0.001
Treatment × time23.10.1172.90.134
Dead plantsTreatment22.80.1401.20.368
Time13.60.10612.3 0.013
Treatment × time20.30.7450.50.632
Other speciesTreatment20.30.7250.30.758
Time110.1 0.019 0.10.795
Treatment × time20.20.8060.30.746
Table 4. Table of initial and final absolute cover (%) and SEs of the five most common species, dead plants, and other species under nonblower control ambient [CO2], and elevated [CO2] treatments for initial and final sampling periods
TimeTreatment Ambrosia dumosa SE Larrea tridentata SE Lycium andersonii SE Lycium pallidum SE Pleuraphis rigida SEOther speciesSEDead plantsSE
Figure 2.

Change in relative cover for the five most common species – Ambrosia dumosa, Larrea tridentata, Lycium andersonii, Lycium pallidum, and Pleuraphis rigida – and other species over the study period treated with nonblower controls (closed bars), ambient [CO2] (open bars), and elevated [CO2] (hatched bars). The asterisk indicates statistically significant differences from initial to final biomass (1997–2007) within species and treatment at α = 0.05. Values are means ± SE.


We found that elevated [CO2] had no effect on total cover, species richness, and species diversity in our long-term desert FACE experiment. Other studies have found no plant community responses to elevated [CO2], although many have been shorter-term or in less natural ecosystems (Navas et al., 1995; Morgan et al., 2007; Davis & Ainsworth, 2012). Yet other studies have found that elevated [CO2] can negatively affect species richness, diversity, or both (Zavaleta et al., 2003; Dawes et al., 2011). These plant community responses to elevated [CO2] are likely mediated by precipitation and soil moisture (Niklaus & Körner, 2004). The NDFF was the most arid FACE site, representing an area with extremely low annual precipitation. Arid plant community responses to elevated [CO2] may be quite different from semiarid community responses as a result of extreme water (and potentially nutrient) limitations and highly fluctuating rates of productivity (Shaw et al., 2005) in these pulse-dominated systems. Notably, higher soil moisture has been observed under elevated [CO2] in semiarid ecosystems (Niklaus et al., 1998; Morgan et al., 2001), but there was no effect of elevated [CO2] on soil moisture at the NDFF over the long term (Nowak et al., 2004b).

Previous work during four consecutive years (1998–2001) at the NDFF revealed significant increases in photosynthetic rate (Naumburg et al., 2003) and productivity (Housman et al., 2006) with elevated [CO2] in wet years and small or no changes in normal or dry years for Larrea, Ambrosia, and Krameria erecta. Based on these short-term responses, we incorrectly predicted there would be an overall increase in long-term cover with elevated [CO2]. Long-term data from other FACE sites have exposed similar incongruities between short-term physiological data and long-term changes in cover or biomass (Navas et al., 1995). We propose that enhanced plant growth in elevated [CO2] during wet years was counteracted by plant death and substantive biomass loss through canopy dieback (Miriti et al., 2007; McAuliffe & Hamerlynck, 2010) in ensuing low rainfall years. Indeed, total cover decreased from initial to final sampling in all treatments, probably because initial sampling took place during higher than average rainfall conditions. Surprisingly, these changes were not influenced by [CO2].

Few clear-cut patterns have emerged when examining plant responses to elevated [CO2] by functional group (Poorter & Navas, 2003; Nowak et al., 2004a; Ainsworth & Long, 2005). We predicted that C3 species would respond more favorably than C4 species to elevated [CO2] based on improvements in WUE in C3 plants. However, there was no evidence that C3 species consistently benefited from elevated [CO2] to cause a shift in C3–C4 dominance. On the contrary, our results suggest that C3 deciduous shrubs may have reduced performance and the C4 bunchgrass relatively better performance in elevated [CO2] conditions in the Mojave Desert. Considering L. tridentata did not respond in the same way as the other C3 plants, this work adds to the body of literature suggesting that photosynthetic pathway is not a chief determinant of response to elevated [CO2] (Nowak et al., 2004a; Ainsworth & Long, 2005). Pleuraphis was the only C4 plant and bunchgrass that we examined, making comparative conclusions about this functional type and photosynthetic pathway untenable. Although the relative cover of Pleuraphis decreased over the study, this decrease was ameliorated under elevated [CO2], suggesting that canopy dieback and mortality were less severe. This contrasts with other studies, which have found that grasses have a competitive disadvantage under elevated [CO2] (Shaw et al., 2005).

The episodic germination and establishment of long-lived plants in the Mojave Desert limit the scope of the species diversity and species richness patterns that we report. Our initial measurements took place during a pronounced wet cycle, which temporarily increased species richness through the appearance of shrub seedlings. Housman et al. (2003) reported that neither germination nor mortality of shrub seedlings was ultimately influenced by elevated [CO2], although initial seedling survivorship of Larrea and Ambrosia was higher in elevated [CO2]. All seedlings died during the study period, which prohibited any increase in species richness. Definitive answers on the effects of elevated [CO2] on desert perennial plant recruitment must involve observations during rare periods when perennial plants establish in the Mojave Desert, which did not occur during the decade of our study.

Contrary to predictions, elevated [CO2] did not affect total cover, species richness, or species diversity, even though we found small effects of elevated [CO2] on the cover of individual species that could eventually alter the overall plant community over longer timescales. We propose that the lack of strong plant community responses to elevated [CO2] may be explained by the long-term steady state of this desert ecosystem, which has slow plant growth and rare perennial recruitment events. Indeed, perennial plant cover of c. 17% at the NDFF is a function of low precipitation and recurring drought, and elevated [CO2] may have little influence on this dynamic. It is likely that overall plant cover will remain low, particularly if precipitation in the region decreases in response to climate change (Seager et al., 2007). Our results provide the first information about the effects of elevated [CO2] on aridland perennial communities, which comprise a large and growing portion of global land cover.


Thanks to L. Fenstermaker and E. Knight for administrative and logistical support. A big thanks to S. Ferguson, N. Gehres, and S. Schmid, and many other harvesters. We thank B. Wissinger for manuscript assistance. We appreciate reviews by R.M. Callaway and three anonymous reviewers. Final harvest work was funded by the US Department of Energy Office of Science (DE-FG02-03ER63651) and earlier work by the National Science Foundation Ecosystem Studies Program (DEB-0212812).