Increasing CO2 accelerates root growth and enhances water acquisition during early stages of development in Larrea tridentata


Author for correspondence: D. Obrist Tel: +1 (775) 784 1484 Fax: +1 (775) 673 4583 Email:


  • • Stimulation of root growth under elevated CO2 has been hypothesized to enhance soil water uptake under water-limiting conditions. The objectives of this study were to quantify the effects of rising CO2 on root development and soil water uptake in Larrea tridentata and to quantify root proliferation into small water patches.
  • • Seedling communities of L. tridentata were grown in rhizotrons under controlled environmental conditions at three CO2 concentrations (280, 360, and 600 µl l−1). Patches of water were applied to small areas of the root systems in the rhizotrons and to L. tridentata shrubs in the field.
  • • Rising CO2 significantly stimulated root length production, but only in the lower half of the soil profile. Stimulation of root production led to faster depletion of soil water. Neither mature shrubs nor seedlings responded to water-enriched soil patches via root proliferation.
  • • The results of our study indicate that rising CO2 may accelerate seedling root growth in L. tridentata, could lead to proportionally greater investment of roots in deeper soil layers and may enhance water acquisition.


Increasing atmospheric CO2 stimulates root biomass and root length production in many plant species, even more than above-ground biomass or leaf area production (e.g. reviews by Norby, 1994; Rogers et al., 1999; Pritchard & Rogers, 2000; but see Bernacchi et al., 2000). Such stimulation can lead to increased soil exploration via (1) expansion of the rooting zone (Gifford, 1979; Baker et al., 1990; Berntson & Woodward, 1992; Rogers et al., 1992), (2) a more intense exploration of a given soil volume (Chaudhuri et al., 1986; Del Castillo et al., 1989) including enhanced proliferation of roots into resource-rich microsites (Arnone, 1997), or (3) more rapid root deployment in the soil (Chaudhuri et al., 1990; Arnone, 1997). Enhanced soil exploration via any, or all, of these mechanisms may increase nutrient acquisition (Barber, 1984; St. John et al., 1983; Rogers et al., 1992; Van Vuuren et al., 1996) and has been hypothesized to increase water capture (Tolley & Strain, 1985; Berntson & Woodward, 1992; Stulen & den Hertog, 1993). No data are available, however, that specifically relate the possible array of root growth responses caused by rising CO2 to changes in the pattern of soil water uptake. For desert plants, especially those in the seedling stage, CO2-induced increases in root system size, rate of root deployment, or root proliferation into ephemeral or spatially rare patches of moister soil – typical of deserts (Noy-Meir, 1973) – may increase water acquisition by these plants and enhance their survival.

The objectives of our study were (1) to quantify the effects of rising atmospheric CO2 on temporal and spatial patterns of root development in seedling communities of the dominant Mojave Desert shrub Larrea tridentata growing in rhizotrons in natural soil, (2) to determine whether an expected stimulation of root development (rate) and root system size with rising CO2 would lead to increased soil water uptake under water-limited conditions and (3) to quantify the extent to which rising CO2 promotes proliferation of L. tridentata seedling roots into small water-enriched soil patches. To accomplish these objectives, we grew plants in rhizotrons maintained under preindustrial (280 µl l−1), current (360 µl l−1), or future (600 µl l−1) CO2 concentrations. The rhizotrons permitted continuous, nondestructive quantification of entire root systems over time. Water patches were applied in small areas in these rhizotrons during part of the study. We also applied water patches around adult L. tridentata shrubs growing under present CO2 (360 µl l−1) levels in the field in the northern Mojave Desert to assess responses of mature plants and relate these to responses of seedlings grown in the rhizotrons.

Materials and Methods

Controlled environment: Rhizotron experiment

(1) Plant material and growth conditions Eighteen Plexiglas rhizotrons (30 cm wide × 3 cm thick × 40 cm high) were filled with 3.6 l of a 1 : 1 mixture (by volume) of topsoil collected from the northern Mojave Desert (next to the Nevada Desert FACE Facility, Jordan et al., 1999) and an inert, coarse, washed industrial outdoor blasting sand (Silica resources Inc., Marysville, WA, USA). Particle sized distribution of the mixed soil corresponded to that of a sandy soil. The sidewalls and bottom of the rhizotrons were covered with opaque aluminum fabric to minimize soil heating during the day and create a dark environment for roots. Initial volumetric soil water content in all rhizotrons was adjusted to 5%.

Thirty seeds of Larrea tridentata L. (Homan Brothers Seed Inc., Phoenix, AZ, USA) were placed in each of the 18 rhizotrons on November 22, 1999. Six rhizotrons were randomly placed in each of three naturally lit, environmentally controlled growth chambers (volume 2.9 m3) at the Desert Research Institute in Reno, Nevada, set at an angle 20° from vertical in order to allow root growth along the front Plexiglas window. CO2 levels in the three chambers were maintained at past (280 µl l−1), present (360 µl l−1), and future (600 µl l−1) CO2 concentrations, respectively. Rhizotrons and CO2 treatments were rotated between the three chambers every 4 wk to minimize chamber effects. Day and night temperatures were maintained at 15 and 30°C, respectively.

The number of plants that germinated in the rhizotrons was not affected by CO2 concentrations resulting in an average of 9.6 ± 0.8 plants per rhizotron. Rhizotrons were watered daily throughout the experiment with 20 ml of tap water added on the topsoil. Although this watering regime was equivalent to 811 mm water per year, high soil evaporation rates from the top surface resulted in very dry growing conditions in the rhizotrons.

(2) Application of water and water-plus-nitrogen enriched patches In order to assess proliferation of L. tridentata seedling roots into small, water-enriched soil patches, we marked three small zones in each of the rhizotrons in which we randomly assigned either (1) a patch of water, (2) a water-plus-nitrogen patch, or (3) a control that did not receive any additional resources. The purpose of the water-plus-nitrogen patch was to differentiate between direct effects of water addition and possible indirect effects of increased water availability via nutrients (e.g. increased nitrogen mineralization), and to assess potential root proliferation into these patches relative to patches of water alone. The three zones (water patch, water-plus-nitrogen patch, and control location) we established in each rhizotron were at equal depths within a rhizotron, but the depths of the zones were different between rhizotrons depending on the depth of the root system (no significant differences between CO2 treatments). Eighty days after seeding we applied the first patches of water (5 ml tap water) and water-plus-nitrogen (10 mm NH4NO3 in 5 ml tap water), injected through small holes in the front windows of the rhizotrons. These injections created visibly wet patches (c. 5.5 cm diameter) on both sides of the rhizotrons around the injection holes. Application of water and water-plus-nitrogen patches increased volumetric soil water content by c. 7% volumetric soil water content (e.g. from 5% to 12% volumetric soil water content). After patch applications, we covered the holes with transparent tape. After the first patch application on day 80, we renewed patches weekly for the next 4 wk. Since the two applied patches accounted for only 4% of the total rhizotron area, we did not expect these treatments to affect overall spatial and temporal root system responses to rising CO2 (Objective 1).

(3) Measurements of root development and soil water content All roots growing along the front window of the rhizotrons were traced on clear acetate sheets every 3 wk throughout the experiment using permanent markers. Whole root length in each rhizotron (in two separate depth layers: 0–20 cm and 20–40 cm) and maximum rooting depth were then measured using a digital plan measuring tool (Scale Master Classic, Calculated Industries Inc., Carson City, NV, USA) and expressed as root intensity (root length per unit surface area of the rhizotron). Root length in water patches, water-plus-nitrogen patches, and in control locations were traced before the first patch treatment, every week during the 5 wk of patch application, and 1 wk after the last patch application by tracing all roots growing in circles of 5.5 cm around the center of patch and control locations.

We harvested above- and below-ground biomass in all rhizotrons 178 d after seeding. First, roots from all patch and control locations were harvested by removing the front wall of the rhizotrons and cutting out the soil corresponding to patch and control locations using a sharp-edged 5.5 cm diameter aluminum can. The rest of the roots were harvested in two separated depth layers (0–20 cm; 20–40 cm). All roots were removed from the soil samples and washed, and root length density (RLD: cm roots l−1soil) was determined using a line intersect method (Böhm, 1979). We determined leaf area by analyzing scans of leaves using a digital analysis program (Image Pro Plus, SciMeasure Analytical Systems Inc., Atlanta, GA, USA). For determination of root dry mass, leaf dry mass, and total above- and below-ground dry mass, plant material was dried in a convection oven at 60°C for 2 wk. Since application of patches did not result in an increase in root length (see Results), root intensities and RLD in the patches were included in the analysis and in figures of overall rhizotron root parameters. We monitored soil water contents of each rhizotron by recording rhizotron weights, assuming plant biomass was small and negligible compared to changes in soil water.

Controlled environment: Pot watering experiment

To determine if growing conditions in the rhizotrons were limited by water, we performed an additional experiment in which L. tridentata seedlings were grown in nine pots (8 cm wide × 8 cm thick × 15 cm high) in each growth chamber next to the rhizotrons during the same time period. Compared with the rhizotrons, three of these pots received (1) equal amounts of water, (2) the equivalent of half the amount of water per unit of soil surface area (achieved by reducing irrigation frequency), or (3) the equivalent of twice the amount of water per unit of soil surface area (achieved by increasing irrigation frequency). At the end of the experiment, we measured plant biomass in these pots to assess the degree by which biomass production was limited by availability of soil water.

Field experiment

(1) Field site and growth conditions The field experiment to assess root proliferation into water patches of adult L. tridentata shrubs was performed adjacent to the Free-Air-CO2-Enrichment (FACE) facility in the northern Mojave Desert (36°49′ N, 115°55′ W, elevation a. 970 m). With average precipitation of 140 mm yr−1, this region is one of the driest habitats in North America (MacMahon & Wagner, 1985). Soils are classified as Aridisols derived from calcareous alluvium. Vegetation is characterized as a Larrea tridentata-Ambrosia dumosa-Lycium spp. desert scrub community (Jordan et al., 1999) in which L. tridentata is the dominant shrub covering nearly two-thirds of the total area. For further details about the study site, see Rundel & Gibson (1996) and Jordan et al. (1999).

(2) Application of water- and water-plus-nitrogen patches Three times in 1999 (March, May and October), patches of water and water-plus-nitrogen were applied to each of six individual mature L. tridentata shrubs at a distance of 20 cm from the stems. Six different plants were selected for each date. As in the rhizotron experiment described above, water-plus-nitrogen patches were applied to differentiate between direct and indirect (e.g. via increased nitrogen mineralization) effects of water on root proliferation. Water-enriched patches were created by dripping water onto the soil surface during a 2-wk period using a gravity-driven drip system consisting of a 9-l water bottle elevated on a cinder block (40 cm above ground surface) and connected to a vinyl tubing syringe and a needle. Water-plus-nitrogen patches were applied using the same drip system with grains (2 g) of nitrogen fertilizer (38-0-0, N-P-K) placed directly under the tip of syringe needle. All grains completely dissolved within the 2-wk drip period resulting in a concentration of c. 0.2 mg N ml−1 H2O (assuming no nitrogen volatilization). Additional patches applied in October and excavated during the 2-wk exposure period indicated that the drip system increased volumetric soil water content in the patches (c. 20 cm diameter and 50 cm depth) by about 10% volumetric soil water content (e.g. from 3 ± 0.5% to 13 ± 1.5% volumetric soil water content).

(3) Measurement of root biomass and RLD Two weeks after the application of the patch treatments, three soil cores (8 cm diameter; 35 cm depth) were removed around each of the six shrubs using a hand auger (one core in each of the two patch locations and one core at a control location). Because we were only interested in L. tridentata roots, we immediately removed roots that visibly belonged to annual forbs and grasses growing on top of the patches from the soil. The soil cores were then sieved, the remaining roots were removed from the soil, and root biomass (g l−1 soil) and RLD (cm roots l−1 soil, line intersect method, Böhm, 1979) were determined.

Statistical analysis

Time-repeated observations in the rhizotron experiment (overall root intensity, root intensity 0–20 cm, root intensity 20–40 cm) were analyzed using a repeated-measures anova (using Box's conservative epsilon) with CO2 as between-group independent (i.e. fixed) variable and Time (repeated measure) and interactions as within-group independent variables. The effects of patches on root intensity were analyzed using a similar repeated-measures anova with the additional between-group variable Patch and interactions. We first tested for differences between the two types of patches (i.e. water and water-plus-nitrogen). Since we did not detect any differences between the two types of patches, we tested for differences between control locations (n = 6 per CO2 treatment) and all patch locations (n = 12 per CO2 treatment). We analyzed the effects of CO2 on soil water contents for each date separately using Type III anovas with CO2 as independent variable. All observations at the end of the experiment (on day 178; above- and below-ground biomass, total biomass, root shoot ratio, root mass fraction, RLD, specific root length) also were analyzed using Type III anovas with the independent variable CO2. CO2 was treated as a continuous variable in all analyses (280 µl l−1, 360 µl l−1, 600 µl l−1). Data from the field experiment (root biomass and RLD) were analyzed using Type III anova with independent variables Patch (two treatments: no patch or patch location; no significant differences between water and water-plus-nitrogen patches), Time-of-the-year (three treatments), and interactions. All analyses were performed using Stata Version 6 (Stata Press, College Station, TX, USA). All error estimates in the text and error bars in the figures are standard errors of the means. Statistically significant differences are defined as P ≤ 0.05.


Controlled environment: Rhizotron experiment

(1) Effects of rising CO2 on temporal and spatial patterns of root system development Rising CO2 significantly stimulated root intensities (Fig. 1a; Table 1; in mm roots cm−2 rhizotron surface). The stimulatory effect of rising CO2 on root intensities, however, was not consistent throughout the duration of the experiment (Table 1). The strongest stimulation of root intensity occurred within the first 100 d of the experiment, with stimulation decreasing toward the end of the experiment. Relative stimulation of root intensities under future CO2 levels compared with present CO2 levels was 11% at the end of the experiment, while relative stimulation of root intensity under present compared with past CO2 concen-trations was much greater (+30%).

Figure 1.

Root development of Larrea tridentata seedlings in rhizotrons. (a) Temporal patterns of root intensities (means and standard errors, n = 6) measured in Larrea tridentata seedling communities grown in rhizotrons exposed to past (280 µl CO2 l−1, open symbols), present (360 µl CO2 l−1, shaded symbols), and future (600 µl CO2 l−1, closed symbols) CO2 concentrations. (b) Temporal patterns of root intensities measured in the upper 20 cm of the rhizotrons exposed to the three CO2 levels. (c) Temporal patterns of root intensities measured in the lower 20 cm (20–40 cm) of the rhizotrons exposed to the three CO2 levels. P-values are based on repeated-measures anovas (see Table 1).

Table 1.  Summary of statistical analyses (P-values from repeated measures anovas) evaluating the effects of CO2 (independent, continuous variable: 280, 360, and 600 µmol mol−1) and time (independent, repeated measures variable) on total root intensities, root intensities from 0- to 20 cm depth and from 20 to 40 cm depth of Larrea tridentata
Source of errorDependent variable
Root intensity totalRoot intensity (0–20 cm)Root intensity (20–40 cm)
  1. df, degrees of freedom; MS, mean squares; *, P < 0.05; **, P < 0.01.

Between-subject variable
CO2 119976.710.008**606.060.32713623.73< 0.001**
Residual16 2206.83 592.97   794.73 
Within-subject variable
Time 6  774.010.050*154.720.094  273.76  0.125
Time × CO2 6 1796.770.007**244.300.040*  774.84  0.016*
Residual88  180.61  48.19   103.34 

Increases in root intensity observed with increasing CO2 was not accompanied by increases in rooting depth (i.e. the deepest root in a rhizotron; P = 0.1338; data not shown) up to the time when roots in all treatments reached the bottom of the rhizotrons (c. 100 d after planting). Thus, the strong CO2-induced stimulation of root length production observed early in the seedling development (first 100 d) did not lead to increases in rooting depth even at a time when this would have been possible (i.e. before roots reached the bottom of the rhizotrons). In spite of the similarity of rooting depths among CO2 treatments, rising CO2 only led to increases in root intensities in the lower part of the rhizotrons (20–40 cm; Fig. 1c; Table 1), with root intensities in the upper 20 cm remaining unaltered by CO2 levels (Fig. 1b). No black roots were observed during the experiment indicating no or only minimal root mortality during the experiment.

As observed with nondestructive measurements (Fig. 1), total RLD (cm roots l−1 soil) measured destructively at the end of the experiment increased with increasing CO2 (Table 2), with stronger increases observed between past and present CO2 levels (+55%) than between present and future CO2 levels (+21%). Increases in RLD observed with rising CO2, however, were less pronounced than increases of root biomass, resulting in a marginally significant decrease in specific root length (root length per root mass (Table 2)).

Table 2.  Means and standard errors (n = 6) of biomass (dry weight), leaf area, and root morphological data at harvest 178 d after plant growth
 CO2 concentrations
280 µl l−1360 µl l−1600 µl l−1P
  1. P-values are based on anovas using CO2 as a continuous variable (i.e. df = 1; 280, 360, and 600 µL CO2 l−1).

Biomass (g)
 Total0.28 ± 0.040.37 ± 0.030.60 ± 0.05< 0.001
 Above-ground0.19 ± 0.020.22 ± 0.020.37 ± 0.04< 0.001
 Leaf0.10 ± 0.010.11 ± 0.010.20 ± 0.02< 0.001
 Root0.09 ± 0.020.15 ± 0.020.24 ± 0.04  0.003
Root : shoot ratio0.48 ± 0.070.69 ± 0.130.67 ± 0.13  0.367
Root mass fraction0.32 ± 0.040.39 ± 0.050.38 ± 0.04  0.388 (root mass/plant mass)
RLD (cm l−1 soil)    140 ± 32    214 ± 34   258 ± 43  0.054
Specific root length (m g−1)54.0 ± 9.652.1 ± 3.339.6 ± 3.7  0.054
Leaf area11.2 ± 1.2 9.7 ± 0.717.3 ± 1.3< 0.001

(2) Effects of rising CO2 on soil water use From the beginning of the experiment until the first weight determination of the rhizotrons (60 d after seeding), volumetric soil water content increased under all CO2 treatments from the initial value of 5 ± 0% to 6.5 ± 0.3% (Fig. 2). Thereafter, evapotranspiration in the rhizotrons exceeded the amount of water added, resulting in an exponential decrease in soil water content to 1.9 ± 0.1% at the end of the experiment. Rising CO2 resulted in a more rapid depletion of soil water from the rhizotrons, however, these differences were only significant at 88 and 96 d after seeding (88 d: P = 0.050; 96 d: P = 0.019) and thereafter disappeared. At the end of the experiment, volumetric soil water content in all CO2 treatments reached equal levels of 1.9% under all CO2 treatments.

Figure 2.

Volumetric soil water content (means and standard errors, n = 6) measured in rhizotrons exposed to past (280 µl CO2 l−1, open symbols), present (360 µl CO2 l−1, shaded symbols), and future (600 µl CO2 l−1, closed symbols) CO2 concentrations. Soil water content in rhizotrons was calculated by weight determination of rhizotrons (i.e. mean value for entire rhizotrons). Decline in soil water contents is an indicator for root water uptake. *, P ≤ 0.01; *, P ≤ 0.05 (based on anovas performed for each date).

(3) Effects of rising CO2 on root proliferation into localized water patches Before the first application of water and water-plus-nitrogen patches (80 d after seeding), root intensities in the control and two patch locations were higher than root intensities outside these locations. This was because we specifically chose locations for the application of patches where roots were present, while other areas in the rhizotrons were void of roots at that time. During the 6-wk period of patch exposure (five weekly patch applications), root intensities in the two patch locations and the control location increased significantly (Fig. 3a,b,c; Table 3). Rising CO2 also stimulated root production in all locations during this time. Compared with the control locations, however, application of water or water-plus-nitrogen patches did not stimulate root intensities (Table 3). Also, no patch effects were observed for absolute or relative increases in root intensities compared to the values observed before the first patch application (Table 3).

Figure 3.

Root development in patch and non-patch locations of Larrea tridentata seedlings in rhizotrons. (a) Temporal patterns of Larrea tridentata root intensities (means and standard errors, n = 6) in control patches (no resource enrichment, open symbols), water patches (shaded symbols) and water-plus-nitrogen patches (closed symbols) during a period of 6 wk and five patch applications (arrows) in rhizotrons exposed to past CO2 concentrations (280 µl CO2 l−1). (b) Temporal patterns of root intensities in patches exposed to present CO2 concentrations (360 µl CO2 l−1). (c) Temporal patterns of root intensities in patches exposed to future CO2 concentrations (600 µl CO2 l−1). All patch areas were 23.8 cm2. P-values are based on repeated measures anovas for each CO2 level (for combined analysis, see Table 3).

Table 3.  Summary of statistical analyses (P-values from repeated measures anovas) evaluating the effects of patch applications (two treatments; no patch [control] and patch [water or water-plus-nitrogen]), CO2 (independent, continuous variable: 280, 360, and 600 µmol mol−1) and time (independent, repeated measures variable) on root intensities, changes in root intensities in relation to the time immediately before first patch application (Δ root intensity absolute), and percent changes in root intensities in relation to the time before patch application (Δ root intensity relative)
Source of errorDependant variable
Root intensityΔ root intensity (absolute)Δ root intensity (relative)
  1. df, degrees of freedom; MS, mean squares; *, P < 0.05; **, P < 0.01.

Between-subject variable
Patch  11404.500.174 71.090.595  2618 0.745
CO2  15975.380.006**877.520.066 (*)301766 0.001**
Patch × CO2  1 665.130.347 19.410.781  1378 0.814
Residual 50 738.32 248.61  24558 
Within-subject variable
Time  5  583.82< 0.001**583.82< 0.001** 51286< 0.001**
Time × patch  5  38.030.315 38.03 0.315  3068 0.368
Time × CO2  5  94.450.116 94.45 0.116 18578 0.030*
Time × patch × CO2  5  35.880.329 35.88 0.329  1791 0.490
Residual250  36.85  36.85 3710 

(4) Overall seedling responses to rising atmospheric CO2 Rising CO2 levels significantly increased final plant biomass, above-ground biomass, leaf area and biomass, and below-ground biomass at the end of the experiment (Table 3). Total biomass was less stimulated between past and present CO2 levels (+32%) than between present and future CO2 levels (+62%). Below-ground biomass stimulation, however, was similar for the two CO2 steps (+66% and +60%, respectively). Despite these patterns, root shoot ratio and root mass fraction (root mass/plant mass) were not significantly affected by atmospheric CO2 level.

Controlled environment: Pot watering experiment

Seedling biomass production in the pots placed adjacent to the rhizotrons in the growth chambers increased with increasing water supply under all three CO2 treatments (Fig. 4). These increases were highly linear (r2 ≥ 0.99) over the range of watering applied to the pots (i.e. from approximately half to double the amount of water added to the rhizotron experiment per unit surface area). This indicated that biomass production of the plants growing in the rhizotrons was also likely limited by soil water availability in all three CO2 levels. The effects of rising CO2 on final plant biomass (per pot) were similar to those observed in the rhizotrons under the higher of the three watering treatments (i.e. more biomass with rising CO2), but not at the two very low watering regimes (no biomass effect). The reason for these differences was a decrease in germination success with rising CO2 concentrations at the two very low watering treatments, which compensated for larger individual plant biomass observed under rising CO2 concentrations. In the rhizotron experiment, no such differences in germination success were observed among CO2 treatments.

Figure 4.

Linear regression between amount of water added and biomass production (end of experiment) of Larrea tridentata plants growing in pots next to rhizotrons under past (280 µl CO2 l−1, r2 = 1.00, open symbols), present (360 µl CO2 l−1, r2 = 1.00, shaded symbols), and future (600 µl CO2 l−1, r2 = 0.99, closed symbols) CO2 concentrations. Amount of watering was adjusted by increasing or decreasing watering frequency. P-values for all regressions were < 0.01.

Field experiment

Patch application around mature L. tridentata shrubs in the upper 40 cm of the soil did not affect RLD after exposure to these patches for two weeks (Fig. 5; P = 0.226) in all three periods when patches were applied (May, May, or October 1999). RLD ranged from 25.8 ± 4.3–40.5 ± 6.6 cm roots litre−1 soil for all three time periods, with no significant changes in RLD during the year.

Figure 5.

Root length densities (RLD, means and standard errors, n = 6) around mature Larrea tridentata shrubs exposed to present CO2 levels in the field in the upper 40 cm of the soil in control locations (no resource enrichment; open bars), water patches (shaded bars), and water-plus-nitrogen patches (closed bars) after 2 wk of patch exposure in March, May, and October 1999. No significant differences were observed between control and patch locations or between water patches and water-plus-nitrogen patches at any time of the year.


The strong early (0–105 d) root system growth stimulation of L. tridentata seedlings observed in the rhizotrons with increasing CO2 (Fig. 1a) indicates that the increase in CO2 concentration that has occurred since preindustrial times to the present may have resulted in more rapid root system establishment in desert seedlings, and that this may persist into the future. The decline in this stimulatory effect observed toward the end of the experimental period (105–178 d) indicates that differences in root growth rates between CO2 treatments become smaller as seedlings grow older, a response also observed for roots in response to other environmental variables (Gedroc et al., 1996; McConnaughay & Coleman, 1998; McConnaughay & Coleman, 1999). The more pronounced differences in root system size observed at the end of the experiment between preindustrial and present day CO2 levels – relative to those occurring between present day and future CO2 levels – indicate that CO2 increases that have occurred up to the present may have resulted in plants that have larger root systems during their seedling stage but that root system size of seedlings in the future will likely be similar to or only slightly larger than those of present day seedlings. This assertion assumes that any genetic adaptation of L. tridentata populations to rising CO2 that may have occurred from the past to the present, or that may occur from the present to the future, has been or will be small. Observed temporal differences in the degree of root stimulation (i.e. stronger during early seedling development and weaker toward the end of the experiment) could explain contradictory results in root system responses of L. tridentata to future CO2 levels found in other studies (BassiriRad et al., 1997; Huxman et al., 1999). The consistent patterns of root system responses to increasing CO2 observed with nondestructive and destructive (at harvest) methods further indicate that nondestructive root observations using rhizotrons are good indicators for responses of overall root systems, despite some quantitative differences between the two methods.

The surprising lack of any stimulation of root intensities in the upper zone of the rhizotrons with rising CO2 (Fig. 1b), together with strong stimulation of root intensities in the deeper zones, translated into a shift in root allocation downward in the soil profile. One ecosystem consequence of this shift may be an increase in carbon and nutrient inputs as well as cycling deeper in the soil profile. These patterns contrast with those reported for more mesic ecosystems in which root biomass and RLD production shifted to upper soil layers under elevated CO2, possibly caused by higher soil moisture and associated increases in nutrient availability in the topsoil (Prior et al., 1994; Arnone et al., 2000).

The lack of any significant weight loss from the rhizotrons (Fig. 2) at the beginning of the experiment before substantial root intensities were present indicate that weight loss observed later in the experiment (when roots were present) was likely due to plant water uptake, and that water depletion from rhizotrons thus can serve as an indication for plant water acquisition. Our results indicate that the large CO2-induced stimulation of root intensities led to an acceleration of soil water uptake from the rhizotrons during early stages of seedling development. This acceleration in water acquisition may be of high ecological importance for seedling establishment and survival in deserts where water is a highly limiting resource. However, the acceleration of soil water uptake during that time period was relatively small and not accompanied by an overall increase in water depletion from the rhizotrons at the end of the experiment. This was surprising given that root system sizes were nearly doubled under future CO2 compared with past CO2 levels during early seedling development and soil moisture in the rhizotrons was clearly growth limiting (cf. Fig. 4). Possible explanations for a lack of stronger CO2 effects on water uptake include (1) that similar rooting depths across all CO2 levels resulted in equal exploration and access to the same vertical soil profile, (2) that near-maximum soil water extraction was possible with the relatively low root intensities produced under preindustrial CO2 levels, possibly due to relatively high soil hydraulic conductivity in this experiment (i.e. sandy soil), (3) that reductions in stomatal conductance and transpiration with increasing CO2 (e.g. Farquhar & Wong, 1984; Morison, 1985; Tyree & Alexander, 1993) partially compensated for greater leaf areas and root intensities that occurred with rising CO2, and (4) that root hydraulic conductivity decreased under elevated CO2 (Huxman et al., 1999). Our results showing unchanged water use by the end of the experiment in the presence of increased leaf area and biomass (Table 2) confirm that stomatal conductance likely declined with rising CO2. These data demonstrate that even under very dry conditions rising atmospheric CO2 may lead to reductions in stomatal conductance, a response previously only detected (detectable) in L. tridentata under relatively wet conditions (Huxman et al., 1999; Pataki et al., 2000; Nowak et al., 2001). Finally, the presence of equal levels of soil water extraction observed across all CO2 treatments (Fig. 2: 1.9% v/v) by the end of the experiment indicates that atmospheric CO2 levels did not affect the ability of L. tridentata roots to extract water from very dry soils.

The lack of root proliferation into small, water-enriched patches in both seedlings (in rhizotrons) and mature shrubs (in the field) under any CO2 concentration indicates first that L. tridentata may not rely on root proliferation into ephemeral or spatially patchy water-rich microsites to enhance water acquisition, and second that future increases in atmospheric CO2 are unlikely to alter this. These results were surprising since many desert perennials have been reported to respond to resource-rich soil patches with increased root proliferation (e.g. Eissenstat & Caldwell, 1988; Jackson & Caldwell, 1989; Caldwell et al., 1991; Bilbrough & Caldwell, 1997). Our results, however, do not exclude the possibility of increased rates of water uptake by roots in wetter soil patches – similar to increased rates of nutrient uptake from enriched nutrient patches that have been observed (Robinson & Rorison, 1983; Jackson et al., 1990; van Vuuren et al., 1996). One possible mechanism of increased rates of water uptake includes increased hydraulic conductance between the soil and the roots resulting from rapid rehydration of roots and reversal of drought-induced shrinkage (Nobel, 1994).

Root biomass, RLD, and root intensities observed in our rhizotrons and in the field were relatively low but typical for desert ecosystems (Rundel & Gibson, 1996) with low primary productivity (Smith et al., 2000). Lower root biomass in our rhizotron experiments, compared to those reported in another phytotron experiment with L. tridentata (BassiriRad et al., 1997), may be due to the desert-like (i.e. dry) growing conditions we imposed in our study.

Taken together, the results of our study indicate that the continuing rise in atmospheric CO2 may accelerate seedling root growth in L. tridentata, may lead to proportionally greater investment of roots in deeper soil layers, and that these effects may enhance water acquisition during early stages of plant development. Rising CO2 levels, however, should not affect rooting depth or root proliferation in favorable soil microsites. Our data further indicate that root proliferation in water-rich microsites may not be an important strategy for L. tridentata to enhance capture of soil resources.


We are grateful to J. Boc, J. Rosta, S. Sigrist, L. Sotoodeh and H. Weatherly for their help with the rhizotron experiment and to D. Schorran for building the rhizotrons and maintaining the controlled environmental facility. We are especially thankful to J. Coleman, R. Kreidberg and P. Verburg for constructive critique of an earlier version of this manuscript. We also thank L. Fenstermaker, E. Knight and S. Zitzer for support during the field experiment. Funding for this study was provided by the Andrew W. Mellon Foundation and the National Science Foundation's Experimental Program to Stimulate Competitive Research (EPSCoR) program, and the Desert Research Institute's Center for Arid Lands Environmental Management.