Author for correspondence: Robert S. Nowak Tel: +775 784 1656 Fax: +775 7844789 Email:email@example.com
• A common response of plants to elevated atmospheric CO2 concentration (CO2) is decreased leaf conductance. Consequently, leaf temperature is predicted to increase under elevated CO2.
• Diurnal patterns of leaf conductance and temperature were measured for three desert perennials, the C3 shrub Larrea tridentata, C3 tussock grass Achnatherum hymenoides and C4 tussock grass Pleuraphis rigida, at the Nevada Desert FACE facility. Measurements were made on ambient and c. 550 µmol mol−1 CO2 plots through both a wet and dry year.
• Reductions in conductance were 35%, 20% and 13% for Pleuraphis, Achnatherum and Larrea, respectively. Decreased conductance occurred throughout the day only for Pleuraphis. Both C3 species had smaller CO2 effects during dry periods than wet. Leaf temperature did not differ significantly between elevated and ambient CO2 for any species. Comparisons of blower-control and nonring plots indicated that the FACE apparatus did not confound our results.
• All three species exhibited decreased leaf conductance under elevated CO2, although reductions were not uniform during the day or among years. Nonetheless, leaf energy balance was only minimally changed for these microphyllous desert perennials.
Atmospheric CO2 concentration (CO2) has increased by > 30% over preindustrial levels in the last 150 yr (Keeling et al., 1976). With the present rate of increase, atmospheric CO2 will be approx. 550 µmol mol−1 in the year 2050 and probably will double current levels within the next century (Intergovernment Panel on Climate Change, 1995). Theoretically, plant and ecosystem responses to elevated CO2 are increased under conditions of reduced water availability because of the proportionally greater enhancement of water-use efficiency (Strain & Bazzaz, 1983). The Mojave Desert is the driest ecosystem in North America and thus most likely to respond to elevated CO2. Because deserts are one of the most important ecosystems on a land surface area basis and because they are increasing in area (Dregne, 1991), relatively small effects of atmospheric CO2 across deserts over the entire earth surface can result in major global impacts. Furthermore, changes in plant water use, coupled with the increased productivity of the desert under elevated CO2 (Smith et al., 2000), could have large feedbacks on the hydrological cycle. Ultimately, these feedbacks will affect the function of dryland ecosystems.
Early studies generally found decreased stomatal conductance when plants were grown under elevated CO2 (Bazzaz, 1990; Bowes, 1993). However, a recent meta-analysis of 48 studies with woody species (primarily trees) found a modest, but not significantly different from zero, decrease in conductance of 11% (Curtis & Wang, 1998). Because woody shrubs are a dominant growth form of many deserts, uncertainty in their response to elevated CO2 limits our ability to predict the impacts of rising atmospheric CO2 on desert ecosystems. At the leaf level, conductance affects transpiration, and transpiration in turn influences leaf energy balance. For the same leaf-to-air vapor gradient, decreased conductance decreases transpiration; decreased transpiration, in turn, acts to increase leaf temperature. However, many perennial species in the Mojave Desert are microphyllous, and convective heat exchange generally dominates the energy balance of small leaves (Gates, 1980). Thus, even if decreased conductance occurs in microphyllous Mojave Desert species with increased atmospheric CO2, changes in leaf temperature may be insignificant. To help resolve these uncertainties in the response of desert plants to elevated CO2, we measured leaf conductance and temperature of three perennials during portions of three growing seasons after initiation of the Nevada Desert free-air CO2 enrichment (FACE) experiment. Species were chosen to represent two of the major growth forms and two of the photosynthetic pathways that are common in the Mojave Desert: the C3 shrub Larrea tridentata (creosotebush), the C3 tussock grass Achnatherum hymenoides (indian ricegrass), and the C4 tussock grass Pleuraphis rigida (big galleta grass).
Materials and Methods
Nevada Desert FACE Facility
The Nevada Desert FACE Facility (NDFF) is a state-of-the-technology research facility that is designed to study responses of an undisturbed Mojave Desert ecosystem to increasing atmospheric CO2 (Jordan et al., 1999). The NDFF consists of nine study plots, each 23 m in diameter: three elevated CO2 FACE rings (c. 550 µmol mol−1), three blower-control rings at ambient CO2, and three nonring control plots. This experimental design yields two comparisons of interest: first is the comparison of FACE rings and blower-control rings to determine the effect of elevated CO2, and second is the comparison between the blower-control rings and nonring plots to determine the effect of the FACE apparatus. The array of study plots is located on a broad alluvial fan in vegetation that is dominated by Larrea tridentata (DC.) Cov., an evergreen shrub that grows over 1 m in height. Other important shrubs include the drought-deciduous Lycium andersonii A. Gray and Ambrosia dumosa (A. Gray) Payne. Abundant perennial grasses include the C3 tussock grass Achnatherum hymenoides (Roemer & Schultes) Barkworth (formerly Oryzopsis hymenoides) and the C4 tussock grass Pleuraphis rigida Thurber (formerly Hilaria rigida). Up to 75 annual species may occur, depending on amount and seasonality of rainfall (Jordan et al., 1999).
The Desert FACE experiment began on April 28, 1997 with a step increase from ambient CO2 (c. 370 µmol mol−1) to a set point of 550 µmol mol−1 for the three elevated CO2 plots. The Desert FACE experiment has operated continuously at 24 h d−1, 365 d yr−1 since the initiation of the experiment, except for system maintenance, when wind or cold-temperature thresholds were exceeded, or when the control system was damaged (primarily from lightening strikes). Mean CO2 concentration at mid-canopy level in the centre of the plot is very close to the target value of 550 µmol mol−1 (Jordan et al., 1999).
Leaf conductance measurements
Leaf conductance was measured during mid-morning and mid-afternoon with a Li-Cor 1600 Steady-State Porometer (Li-Cor Inc., Lincoln NE, USA). These two sampling periods were chosen because leaf conductance of desert plants often peaks before solar noon when moisture availability is low (Smith & Nowak, 1990). All measurements were made at prevailing atmospheric conditions (bulk air rh, treatment CO2, and prevailing air temperature), and the sample chamber was orientated perpendicular to the sun in order to maximize solar irradiance. Because of the time needed to move the centre-pivot aerial sampling platform from plant-to-plant within a plot (Jordan et al., 1999) and to walk the > 100 m from plot-to-plot, a single sampling period required 1.5–2 h to complete. To minimize the potential for systematic error over a sampling period, plot sampling was rotated among treatments. Because most days in a desert environment have cloudless skies, leaf conductance was measured only if skies were mostly clear. Mean photosynthetic quantum flux (± standard error) over all measurements was 1.71 mmol m−2 s−1 (±0.03).
For this study of leaf conductance, Larrea, Achnatherum, and Pleuraphis were selected to represent the two major perennial growth forms and photosynthetic pathways at the study site. Leaves of similar age were sampled within each date: actively growing shoot tips for Larrea and the youngest fully expanded leaf blade for the tussock grasses. Two plants of each species were randomly selected from the available population of plants in each plot. Only one species was sampled on any one date.
In 1997, measurements were made in each of the nine study plots, but only on Larrea. The other species were not sampled in 1997 because design problems with the aerial sampling platforms restricted our ability to sample readily, and once those problems were corrected, green foliage was too sparse for Achnatherum and Pleuraphis. Leaf conductance was measured over three 2-day periods: May 2–3 (4–5 d after initiation of the experimental treatments), May 22–23 (4 wk after treatment initiation), and June 24–25 (8 wk after treatment initiation). For each sampling period, mid-afternoon measurements were made the first day and mid-morning measurements on the second. This sampling schedule was used in order to accommodate the placement and monitoring of leaf thermocouples for leaf temperature measurements.
In 1998 and 1999, measurements on the evergreen shrub Larrea began in mid-February 1998 and continued at approx. 4-wk intervals until the end of June 1999. For the two tussock grass species, measurements were also made at approx. 4-wk intervals, but only during portions of each year when plants had sufficient green foliage for measurement. During these 2yr, we focused primarily on the CO2 effect and thus measured plants on the three FACE and three blower-control plots. On two occasions during 1998, we also measured Larrea plants located in the nonring control plots on the same date as those in blower-control plots. Mid-morning and mid-afternoon measurements for any one species were completed within the same day, and measurements on all species were usually completed on sequential days, weather permitting. Note that occasional equipment failure (e.g. mid-May 1998) or cold weather (e.g. mid-February 1998) prevented completion of both mid-morning and mid-afternoon measurements within a particular date.
Leaf temperature was measured during 1997 and 1998 with fine-wire (0.127 mm diameter) thermocouples inserted into green leaves. Thermocouples were monitored to ensure that they did not come out of the leaf or that the leaf they were in did not senesce. In 1998, air temperature at plant canopy height was also measured with a fine wire thermocouple placed in a shaded location near the centre of the plot. Dataloggers (Models CR-7, CR-10 and CR21X, Campbell Scientific, Logan, UT, USA) converted thermocouple voltages into temperature measurements and logged data. In 1997, only three dataloggers were available; thus only one plot of each treatment was measured on any one date. In 1998, six dataloggers were available, and all three plots of the FACE and of the blower-control treatments were measured on any one date. In 1997, thermocouples were measured each minute, but only hourly mean temperatures were stored. Measurements reported in this paper are for five 24-h periods between early May (< 1 wk after initiation of FACE treatments) and mid-July (approx. 10 wk after treatment initiation). In 1998, thermocouple measurements were made and stored every 5 min between February 10 and July 14.
Leaf temperature measurements in 1997 were made for the same three species used for leaf conductance measurements as well as for a second shrub species, Lycium andersonii. Only Larrea leaf temperatures were measured in 1998. For Larrea and Lycium, fully expanded leaves near the top of the plant canopy were measured on two plants in each plot. For the tussock grasses, fully expanded leaves on tillers near the southern periphery of the tussock were measured on two plants in each plot.
We also estimated the difference between leaf and air temperature (ΔT) for Larrea, Achnatherum, and Pleuraphis using an energy balance approach. Monthly average wind speed, air temperature, and solar radiation during mid-morning (10 am) and mid-afternoon (3 pm) (Table 1; Jordan et al., 1999) were determined from hourly mean, in situ measurements during March, April, May, June, and July 1998, that is the time period when the difference between CO2 treatments was largest for leaf conductance. Our porometer measurements of leaf conductance and of air vapor pressure were used to calculate latent heat loss from leaves. Because we used dimensional analyses to determine leaf area of samples in the porometer cuvette, average leaf width was measured for Larrea (5 mm), Achnatherum (1.7 mm), and Pleuraphis (3.1 mm). The absorptivity in the shortwave band for Larrea leaves (0.487) was from Ehleringer (1981); for Achnatherum and Pleuraphis leaves, we used a shortwave absorptivity of 0.45, which was a typical value for Mojave and Sonoran Desert herbs and annuals (Ehleringer, 1981). Shortwave reflectivity for a canopy (0.20) and longwave absorptivity for leaves (0.97) were typical values reported by Campbell & Norman (1998: Tables 11.2 and 11.3). We rearranged Equation 14.6 in Campbell & Norman (1998: p. 225) to estimate leaf temperature minus air temperature for in situ ambient CO2 (ΔTamb) and elevated CO2 (ΔTelev) leaf conductance as well as for the theoretical condition of no transpiration (ΔTzero).
Table 1. Average micrometeorological conditions (wind speed, air temperature, and solar radiation) during mid-morning and mid-afternoon measured at the Nevada Desert free-air CO2 enrichment (FACE) Facility for selected times in 1998
Wind speed (m s−1)
Air temp (C)
Solar radiation (W m−2)
Conductance data were analyzed with 2-way ANOVA that had plot treatment and sampling date as main effects. Because we expected differences among species and differences between morning and afternoon, separate ANOVA were used for each species by time-of-day combination. In all cases, the error term was derived from plot means; individual shrubs within a plot were considered as subsamples. Because of the unbalanced sampling design among all three plot treatments, separate ANOVA were used to compare the FACE vs blower control treatments and the blower control vs nonring control treatments.
For 1997 leaf temperature data, hourly mean values over five 24-h periods were compared with ANOVA. For all species except Lycium, 2-way ANOVA were used with plot treatment and hour-of-day as the main effects, and the error term was derived from different sampling dates. For Lycium, only one 24-h sampling period occurred before this drought-deciduous shrub shed its leaves for the year, and a Kruskal–Wallis nonparametric ANOVA was used. For 1998 temperature data, our analyses were structured to maximize the likelihood of detecting significant differences among treatments. The analysis focused on those time periods during the day when leaf conductance was and was not affected by plot treatment (i.e. morning vs afternoon), excluded cloudy time periods, and accounted for effects of wind speed. To accomplish this structured analysis, we first selected measurements between 08.00 and 17.00 hours that coincided with time periods when photosynthetic quantum flux also exceeded 0.5 mmol m−2 s−1. Next, we subtracted air temperature from leaf temperature. The difference between leaf and air temperature was then analyzed with a covariance ANOVA where time of day (morning vs afternoon) and plot treatment (ambient vs elevated CO2) were main effects and wind speed was a covariate.
Precipitation and leaf phenology
Although the majority of precipitation in the Mojave Desert typically occurs in the winter and spring, very little precipitation occurred from mid-January 1997 until a large rainstorm in mid-June (Fig. 1). Two large rainstorms in September 1997 brought total precipitation for the 1997 hydrologic water year (i.e. October 1996 to September 1997) to slightly above the long-term, mean annual precipitation for the area (140 mm). By contrast 1998 was a very wet year induced by a strong El Niño, with cumulative precipitation between October 1997 and the beginning of March 1998 already exceeding the long-term annual mean for the site. An extensive series of small rain showers during spring 1998 plus several large summer storms contributed to the 1998 hydrologic year total being 2.4-fold higher than the long-term mean. The 1999 hydrologic year was dry, especially during the winter and spring: < 7 mm of precipitation occurred between late-November 1998 and April 1999. Total precipitation for the 1999 years was 107 mm, or 76% of the long-term mean.
Although spring 1997 and 1999 were relatively dry and spring 1998 was relatively wet, the timing of new leaf growth for Larrea was similar in all years. New leaves began emerging in late-April or early May, with the majority of new growth occurring between mid-May and mid-June. The timing of leaf growth for the C4 tussock grass was also similar among years: new leaves were initiated in April, and peak green biomass occurred before the end of July. Initiation of leaf growth for the C3 tussock grass Achnatherum was more plastic than the other two species: major flushes of leaf growth occurred in late-winter 1997, winter-spring 1998, and fall 1998. Peak green biomass for Achnatherum occurred between March and May, depending on late-winter and early spring precipitation. For all three species, the amount of leaf production was dependent on precipitation (Smith et al., 2000).
Over the three years of measurement, Larrea plants on FACE plots had significantly lower leaf conductance than plants on blower-control plots in mid-morning, but not in mid-afternoon (Fig. 2). Averaged over all sampling dates, mid-morning conductance was decreased by 17% under elevated CO2. CO2 effects were greatest during the wet year: over the June-October 1998-time period, mid-morning leaf conductance was decreased by 27% under elevated CO2.
For both mid-morning and mid-afternoon measurements on Larrea plants, conductance on nonring plots was not significantly different than that on blower-control plots (Fig. 2). Because both sets of control plots experienced the same ambient CO2, any differences between these two treatments would be due to the presence of the FACE apparatus on the blower-control plots, that is the pipes and fans that mix and distribute air (Hendrey et al., 1998). Because these two treatments were not significant, we conclude that the CO2 effects on leaf conductance were not confounded by the presence of the FACE apparatus.
Mid-afternoon leaf conductance for Achnatherum was significantly decreased under elevated CO2 by 26%, averaged over all sampling dates (Fig. 3). Although the CO2 main effect was not significant for mid-morning leaf conductance of Achnatherum, the treatment–date interaction term was significant. Mean comparisons by the least significant difference method indicated significant differences between CO2 treatments on five of seven sample dates. Averaged over the two sampling dates at the beginning of spring 1998 and over the fall 1998 sampling date, conductance under elevated CO2 was decreased by 30%. For the two sampling dates at the end of the spring 1998 growing season, conductance under elevated CO2 was increased by 62%. Over all sampling dates, leaf conductance on FACE plots averaged 87% of that on blower-control plots. As with Larrea, differences between FACE and blower-control were greater when soil moisture was more readily available (i.e. early spring 1998 and fall 1998).
Mid-morning leaf conductance for the C4 grass Pleuraphis was significantly decreased by 36% under elevated CO2 (Fig. 4). Although the magnitude of the difference (34%) was similar for mid-afternoon leaf conductance, this difference was not significant. However, over both morning and afternoon measurements, the CO2 effect was significant (P = 0.038).
Leaf temperature over the 1997 sampling dates did not differ significantly among the FACE, blower-control, and nonring treatments for any species (Fig. 5). During 1998, we focused our analysis of Larrea leaf temperature data to detect differences between elevated and ambient CO2 treatments based upon our observations of leaf conductance, that is we expected to observe elevated CO2 effects in the morning, but not the afternoon, and on clear days when irradiance would not limit stomatal opening. To reduce variability within the temperature dataset, we used the difference between leaf and air temperature to account for within-day and between-day variation in leaf temperature. We also used wind speed as a covariate to account for variations in convective heat exchange. Neither the treatment main effect nor the interaction between treatment and time of day were significant, and thus the difference between leaf and air temperature for Larrea at elevated CO2 was not significantly different from that at ambient CO2 during either morning or afternoon time periods (Fig. 6).
Based on energy balance calculations, sunlit leaves of Larrea, Achnatherum, and Pleuraphis absorbed, on average, c. 1 kW m−2 of shortwave and longwave radiation during time periods when leaf conductance differed between CO2 treatments in 1998 (Table 2). Because of the ‘First Law of Thermodynamics’, the absorbed radiation also represents the amount of energy that would need to be lost from the leaf assuming a steady-state leaf temperature. Thus, latent heat loss from transpiration accounted for 6% of energy loss from Larrea leaves and 15–20% of energy loss from Achnatherum and Pleuraphis leaves. Furthermore, the difference in latent heat loss from leaves under ambient CO2 and that under elevated CO2 was 2% of total energy loss for Larrea and Achnatherum and 5% for Pleuraphis. Thus, the effects of elevated CO2 on ΔT (i.e. leaf minus air temperature) were calculated to be < 0.1 C for Larrea and Achnatherum and < 0.2 C for Pleuraphis.
Table 2. Total absorbed radiation, latent heat loss under both ambient and elevated CO2, the difference between ΔT (i.e. leaf–air temperature) of elevated and ambient CO2, and the difference between ΔT of zero transpiration rate and ambient CO2 for three desert perennials at the Nevada Desert free-air CO2 enrichment (FACE) Facility
Latent heat loss (W m−2)
Absorbed radiation(W m−2)
ΔTelev − ΔTamb
ΔTzero − ΔTamb
Values were estimated using the energy balance approach described in the Materials and Methods section and averaged over the mid-morning micrometeorological conditions in Table 1 for Larrea and over both mid-morning and mid-afternoon conditions for Achnatherum and Pleuraphis.
All three desert species exhibited decreased leaf conductance under elevated CO2, a common effect when plants are grown at elevated CO2 (Drake et al., 1997). Among the three desert species in this study, the C4 tussock grass Pleuraphis rigida had the greatest overall reduction in conductance when averaged over all observations (35%), followed by the C3 tussock grass Achnatherum hymenoides (20%), then the C3 shrub Larrea tridentata (13%). Wand et al. (1999) also reported greater reduction of conductance in C4 grasses compared with C3 grasses, although the difference between the two groups of grasses in their meta-analysis (29% and 24% for C4 and C3, respectively) is smaller than our observations. The relatively small response of the woody shrub Larrea to elevated CO2 is similar to the 11% (but not significant) reduction noted for woody species by Curtis & Wang (1998).
The decrease in leaf conductance by elevated CO2 was not uniform during the day among the desert species in this study. Significantly decreased conductance under elevated CO2 occurred throughout the day for Pleuraphis, but only during mid-afternoon and occasionally during mid-morning for Achnatherum. For Larrea, a CO2 effect on conductance only occurred in mid-morning. By contrast, the effects of elevated CO2 were relatively uniform throughout the day for loblolly pine and wheat (Ellsworth et al., 1995; Garcia et al., 1998).
For the C3 species Larrea and Achnatherum, the difference between leaf conductance at elevated and at ambient CO2 became smaller as water availability decreased. For example, the difference was largest during 1998, which was a period of above-average precipitation associated with an intensive El Niño event, than during either 1997, which was near average precipitation, or 1999, which was a La Niña event with below-average precipitation. A smaller CO2-induced reduction in leaf conductance during dry periods was also observed when averaged over 10 tallgrass prairie species (Knapp et al., 1996). However, this pattern of smaller reductions under drought does not occur across all species (Knapp et al., 1996; Heath, 1998) or is reversed in other ecosystems (Goodfellow et al., 1997). For the C4 desert grass Pleuraphis, our conductance dataset is not extensive, but this C4 species did not exhibit smaller CO2 effects under dryer conditions, which is consistent with meta–analyses of interactions between drought stress and CO2 for C4 grasses (Wand et al., 1999).
Decreased conductance and an accompanying decrease in transpiration have resulted in 0.6–1.0 C increases in canopy temperature of wheat and cotton, respectively (Idso et al., 1987; Kimball et al., 1995), but leaf temperature of desert plants under elevated CO2 did not differ significantly from that under ambient CO2 in our study. Two factors likely account for the lack of difference in leaf temperatures for desert plants. First, desert plants with small leaves and open canopies often have leaf temperatures near air temperature because of their high convective coefficients (Gates, 1980). Second, leaf conductance was typically < 200 mmol m−2 s−1 for these desert plants, which corresponds to rather low transpiration rates and hence a small proportion of the leaf energy balance. For example, the mid-morning decrease in Larrea conductance under elevated CO2 during 1998 resulted in a decrease in transpiration of < 0.1 mmol m−2 s−1, which corresponds to c. 16 W m−2 for a leaf that is absorbing c. 1 kW m−2, that is a 2% change in energy balance. Only under the worst possible scenario in which transpiration was assumed to be completely stopped (ΔTzero, Table 2) would increases in leaf temperatures of these 3 Mojave Desert perennial species approach those observed in crop plants.
For all comparisons of blower-control and nonring treatments, neither leaf conductance nor leaf temperature differed significantly. Thus, the presence and operation of the FACE apparatus did not confound these measurements. Furthermore, measurements of leaf photosynthesis (Hamerlynck et al., 2000) as well as several measures of plant density and growth (Smith et al., 2000) did not differ between the two types of control plots. The lack of a ‘blower’ effect at the NDFF greatly simplifies the extrapolation of our results to future atmospheric CO2 conditions.
This research was supported in part by funds from the US DOE (DE-FG03–96ER62292), the National Science Foundation (IBN-9524036), and the Nevada Agricultural Experiment Station. US DOE (DE-FC002–91ER5667) as well as the Nevada Test Site provided funding to construct and operate the Nevada Desert FACE Facility. We thank J. Allen, C. Biggart, M. Hargrove, A. Linnerooth, D. Martin, P. Mebine, J. Sloan, and S. Zitzer for field assistance, the employees of the US DOE Nevada Operations Office and of Bechtel Nevada for logistical and technical support, and 2 anonymous reviewers for their suggested improvements to the manuscript.