Elevated CO2 does not offset greater water stress predicted under climate change for native and exotic riparian plants


Author for correspondence:

Laura G. Perry

Tel: +1 970 226 9196

Email: lperry@lamar.colostate.edu


  • In semiarid western North American riparian ecosystems, increased drought and lower streamflows under climate change may reduce plant growth and recruitment, and favor drought-tolerant exotic species over mesic native species. We tested whether elevated atmospheric CO2 might ameliorate these effects by improving plant water-use efficiency.
  • We examined the effects of CO2 and water availability on seedlings of two native (Populus deltoides spp. monilifera, Salix exigua) and three exotic (Elaeagnus angustifolia, Tamarix spp., Ulmus pumila) western North American riparian species in a CO2-controlled glasshouse, using 1-m-deep pots with different water-table decline rates.
  • Low water availability reduced seedling biomass by 70–97%, and hindered the native species more than the exotics. Elevated CO2 increased biomass by 15%, with similar effects on natives and exotics. Elevated CO2 increased intrinsic water-use efficiency (Δ13Cleaf), but did not increase biomass more in drier treatments than wetter treatments.
  • The moderate positive effects of elevated CO2 on riparian seedlings are unlikely to counteract the large negative effects of increased aridity projected under climate change. Our results suggest that increased aridity will reduce riparian seedling growth despite elevated CO2, and will reduce growth more for native Salix and Populus than for drought-tolerant exotic species.


Riparian ecosystems in arid and semiarid western North America may be dramatically altered by increased water stress under climate change (Rood et al., 2008; Stromberg et al., 2010; Perry et al., 2012). Most climate models of the southwestern USA predict increased aridity over the next century as a result of greater evapotranspiration and less precipitation (Seager & Vecchi, 2010). In addition, warmer temperatures are reducing snowpacks and increasing winter runoff, and thus reducing late spring and summer streamflows in western North America (Christensen et al., 2004; Stewart et al., 2005; Maurer, 2007; Miller & Piechota, 2008). These changes are likely to increase riparian plant water stress, which can hinder tree seedling recruitment (Mahoney & Rood, 1998; Shafroth et al., 1998; Horton & Clark, 2001; Amlin & Rood, 2002) and alter plant community composition (Rood et al., 2003; Auble et al., 2005; Lite & Stromberg, 2005; Strom et al., 2011). In particular, greater water stress may reduce the abundance of dominant, mesic, native Populus (cottonwood) and Salix (willow) species (Rood et al., 2003), and favor more drought-tolerant, exotic species, including Tamarix ramosissima Ledeb., T. chinensis Lour., and Tramosissima × chinensis hybrids (tamarisk or saltcedar; Stromberg et al., 2010) and Elaeagnus angustifolia L. (Russian olive; Reynolds & Cooper, 2010).

Increased atmospheric CO2 concentrations (hereafter [CO2]), however, may offset the effects of increased aridity on riparian plants. Atmospheric [CO2] rose from 280 to 368 ppm during the 20th century and may rise to > 700 ppm by the end of the 21st century (IPCC, 2007). Elevated [CO2] reduces stomatal conductance and increases water-use efficiency (WUE) in many plants, particularly under dry conditions, thus reducing their overall demand for water (Wand et al., 1999; Ainsworth & Long, 2005). In semiarid grasslands, elevated [CO2] can reduce transpiration enough to completely counteract the negative effects of moderate warming on soil water availability, and increases plant growth most strongly in dry years (Morgan et al., 2004, 2011). However, potential effects of elevated [CO2] on plant growth and WUE have not been considered in most predictions of climate change effects on semiarid riparian ecosystems (e.g. Rood et al., 2008; Stromberg et al., 2010; Yarnell et al., 2010), and research is lacking on riparian plant responses to elevated [CO2].

Predicting how elevated [CO2] might affect riparian plants is difficult, because responses to elevated [CO2] vary considerably among species (Leakey et al., 2009). For Populus deltoides Bartram ex. Marsh spp. deltoides (eastern cottonwood), a dominant eastern North American riparian tree, elevated [CO2] (720–1000 ppm) decreased stomatal conductance by 25–50% and increased instantaneous WUE by 20–30%, net photosynthesis by 20–60%, and growth by 25–80% (Will & Teskey, 1997; McDonald et al., 2002; Barron-Gafford et al., 2005; Murthy et al., 2005; Lewis et al., 2010). If western North American riparian species respond similarly, then elevated [CO2] may mitigate the negative effects of lower water availability on riparian plants under climate change. Further, if riparian species vary in their responses to elevated [CO2], then elevated [CO2] could favor some species over others, potentially altering community composition.

Riparian species responses to elevated [CO2] could vary as a result of interspecific differences in direct effects on photosynthesis, effects on stomatal conductance, or indirect effects of stomatal conductance on growth. There are no previous data to suggest that particular riparian plant species are likely to be more responsive than others to direct effects of elevated [CO2]. Species with greater drought tolerance, however, may benefit less from indirect effects of reduced stomatal conductance on growth, because they are less water-stressed than other species when water is moderately limited. For example, Tamarix spp. are better able to access shallow soil water, maintain turgor through osmotic adjustments, and avoid xylem cavitation than Populus and Salix spp. (Smith et al., 1998; Rood et al., 2003), and may benefit less from reduced stomatal conductance.

Effects of increased aridity and elevated [CO2] on seedlings may be particularly important for riparian tree populations. Seedling mortality strongly limits native riparian tree abundance (Lytle & Merritt, 2004). Riparian Populus and Salix seedlings establish on bare, moist surfaces created by floods and exposed by flood recession, growing roots rapidly downward to maintain contact with soil moisture from the declining water table (Mahoney & Rood, 1998; Shafroth et al., 1998). Most seedlings die in their first 12 months from water stress or flood and ice scour (Stromberg et al., 1991; Cooper et al., 1999; Johnson, 2000), and successful establishment occurs only once every 5 to > 60 yr depending on streamflow and geomorphology (Scott et al., 1996). Increased aridity as a result of climate change may further reduce the probability of establishment (Perry et al., 2012). Understanding how elevated [CO2] interacts with increased aridity to influence seedling water stress and growth is critical for predicting how climate change will alter riparian plant communities.

We conducted a glasshouse experiment to test the effects of elevated [CO2] and water availability on seedling growth, physiology, and drought tolerance of five western North American woody riparian species, including the two most common native species (Salix exigua Nutt. (sandbar willow) and Populus deltoides Bartram ex. Marsh spp. monilifera (Aiton) Eckenwalder (plains cottonwood)), the two most common exotic species (Tamarix spp. and Elaeagnus angustifolia), and an increasingly common exotic species (Ulmus pumila L. (Siberian elm)) (Friedman et al., 2005; Zalapa et al., 2010). Our objectives were to evaluate the potential for elevated [CO2] to offset increases in water stress in riparian woody seedlings; and favor establishment by some species more than others. We predicted that elevated [CO2] would increase seedling WUE, and therefore increase growth more for the relatively drought-intolerant, native species than for the exotic species, and more in treatments with lower water availability.

Materials and Methods

We grew seedlings of the five study species (hereafter, Salix, Populus, Tamarix, Elaeagnus, and Ulmus) in a glasshouse under ambient (440 ± 48 (SD) ppmv) and elevated (752 ± 85 ppmv) [CO2] and with four water-table decline rates (0.5, 1.5, 3.0, and 4.0 cm d−1). Each of the 40 treatment combinations (five species × two [CO2] treatments × four water treatments) was replicated eight times.

The experiment was arranged in a blocked, split-plot design, with water treatments in ‘whole pots’, species in ‘subpots’, and [CO2] treatments in separate glasshouse bays. One bay received CO2 inputs sufficient to maintain the elevated [CO2] treatment during 16 h, lighted days. Because it was not possible to apply [CO2] treatments at the whole-pot or subpot level, the experiment lacked true replication of the [CO2] treatment. To minimize the likelihood of spurious effects of location, we used wheeled platforms to rotate the [CO2] treatments between the two bays weekly, except during an 11 d period of equipment failure starting in the 10th week. Within each bay, the pots were divided into eight blocks, each consisting of four whole pots (one per water treatment). When rotating the [CO2] treatments, we also rotated the block locations within each bay, and the whole-pot locations within each block.

Whole pots were 30-cm-diameter × 100-cm-deep PVC pipes, capped at the bottom, equipped with a flexible drainpipe near the base, and filled with water. Whole pots served as water reservoirs for five sand-filled subpots (one per species) arranged inside each whole pot (Supporting Information, Fig. S1). Subpots were 8-cm-diameter × 105-cm-deep (5275 cm3) PVC pipes, capped at the bottom, with four, 1-cm-diameter holes drilled 3 cm above the base, covered with fine screen. Similar pot dimensions have been used in many studies of riparian tree seedlings (Mahoney & Rood, 1991; Rood et al., 2000; Horton & Clark, 2001), because riparian tree seedling roots grow primarily downward to follow declining water tables (Mahoney & Rood, 1998). Subpots were filled with 5 cm of medium/fine gravel (predominantly 5–8 mm particles) at the bottom, then 92 cm of medium/coarse alluvial sand (predominantly 0.4–2.0 mm particles) mined from the Cache la Poudre floodplain (LaFarge North America, Fort Collins, CO, USA), then 5 cm of medium/coarse alluvial sand mixed with Scotts osmocote plus fertilizer (Scotts Professional, Geldermalsen, the Netherlands), and finally 1 cm of a finer medium/coarse sand (predominantly 0.2–1.0 mm particles). The fertilizer was a 3- to 4-month-duration, slow-release fertilizer, with 15% nitrogen (N), 9% phosphorus, 12% potassium, 1.9% calcium, 1.4% magnesium, 4% sulfur, and trace elements. It was applied at a rate of 4 mg N kg−1 to mimic a 10 mg N kg−1 yr−1 mineralization rate, based on N mineralization on sandbars with establishing riparian trees (6.5–32.9 mg kg−1 yr−1) (Adair et al., 2004).

Salix and Populus seeds were collected in Fort Collins, and Tamarix seeds were collected in Loveland, CO. Most North American Tamarix are Tamarix ramosissima × T. chinensis hybrids (Gaskin & Kazmer, 2009), which are morphologically indistinguishable from the parent species, so the species identity of the Tamarix is uncertain. Elaeagnus seeds, collected in Montana, were purchased from F. W. Schumacher Co., Inc. (Sandwich, MA, USA). Ulmus seeds, collected in North Dakota, were purchased from Lawyer Nursery (Plains, MT, USA). Because Elaeagnus seeds are dormant and germinate more slowly than the other species, Elaeagnus seeds were stored in moist sand at 4°C for 80 d, and then planted in flats filled with moist sand 28 d before the start of the experiment. One week before the start of the experiment, three Elaeagnus seedlings, just emerging from the seed coat, were transplanted into one subpot in each whole pot. All other species were sown into subpots at the start of the experiment. Ulmus seeds were covered with a 0.25 cm layer of sand after sowing. Seedlings were thinned to the largest three seedlings per subpot 1 wk after sowing, and to one per subpot (the mid-sized seedling) 3 wk after sowing.

For the first week after sowing, water levels in the whole pots were maintained daily at 5 cm below the subpot soil surface. After 1 wk, water levels in the whole pots were lowered by 0.5, 1.5, 3.0, or 4.0 cm daily by lowering the position of the drainpipe. We used declining water tables for the water treatments because Salix, Populus and Tamarix germination and establishment commonly occur during flood recession on riparian floodplains (Fenner et al., 1985; Mahoney & Rood, 1998; Shafroth et al., 1998). River stage decline rates of 0.5–4.0 cm d−1 are realistic during flood recession (Mahoney & Rood, 1991; Shafroth et al., 1998), and Salix, Populus, and Tamarix seedling survival is often poor with water-table decline rates > 4.0 cm d−1 (Mahoney & Rood, 1991; Segelquist et al., 1993; Horton & Clark, 2001; Amlin & Rood, 2002; Stella & Battles, 2010). Water-table declines continued throughout the experiment for the 0.5 cm d−1 treatment and continued down to a 98 cm depth for the 1.5, 3.0 and 4.0 cm d−1 treatments (reached at 69, 38 and 31 d, respectively).

Soil moisture was determined as a function of height above the water table in two unoccupied subpots with the water table maintained at 64 cm below the soil surface. After 1 wk, soil was collected from each subpot in 64, 1-cm-deep increments down to the water table. The samples were weighed, dried at 105°C, and reweighed to determine soil moisture. Soil moisture in each water treatment was estimated by calculating the daily mean soil moisture for the entire soil column above the water table, accounting for the changing position of the water table over time. Mean soil moisture over the course of the experiment was 12.2, 7.7, 6.2, and 5.8% in the 0.5, 1.5, 3.0, and 4.0 cm d−1 treatments, respectively. Soil moisture at field capacity (−0.033 MPa) and at the wilting point (−1.5 MPa) was determined for two samples by the Colorado State University Soil Testing Laboratory, Fort Collins, to assess approximate full and depleted soil water conditions.

The glasshouse bays were maintained at a temperature (± SD) of 27 ± 3°C during the day and 18 ± 2°C at night, with a 16 h photoperiod, to approximate the June to August climate in Fort Collins (Western Regional Climate Summaries; http://www.wrcc.dri.edu). Relative humidity was 28 ± 6% during the day and 52 ± 8% at night. Each bay was lit by 12 high-pressure sodium lamps. Midday photosynthetic photon flux density at the soil surface before sowing, averaged across subpots, was 517 ± 60 μmol m−2 s−1.

The experiment was maintained for 12 wk, similar to the duration of a typical first growing season for Populus, Salix and Tamarix in Colorado (Cooper et al., 1999). Seedling survival and height were measured weekly. In the 10th week, the percentage of each plant's leaf surface area composed of necrotic tissue was estimated by counting the number of leaves within different classes (0, 1–5, 5–25, 25–50, 50–75, and 75–100% necrotic tissue) and averaging the percent midpoints of all leaves. Above-ground and below-ground biomass was harvested in the 12th week, dried at 60°C, and weighed. During the harvest, predawn and midday stem xylem water potentials (hereafter Ψpd and Ψmd) were measured for whole plants (above ground) using a Scholander pressure chamber (Model 1505D, PMS Instruments, Albany, OR, USA). Ψpd was measured between 02:30 and 05:00 h for four randomly selected blocks, and Ψmd was measured between 11:00 and 14:00 h for the remaining four blocks. Roots were harvested by gently emptying the subpot into a sieve and rinsing away the sand. Harvested roots were straightened over a meter stick to estimate maximum root length, but rooting depths within the subpots were not measured. Small leaf sizes, especially in the faster water-table decline treatments, prevented measurements of photosynthesis, stomatal conductance, and leaf (rather than stem) water potentials.

Leaf tissue was analyzed for δ13C to evaluate intrinsic WUE (C assimilation per unit stomatal conductance, iWUE; Farquhar et al., 1988) and for C and N concentrations (hereafter [C] and [N]) to evaluate N limitation. Dried leaves were separated from stems and ground with a Wiley® mill (Thomas Scientific, Swedesboro, NJ, USA) equipped with a 1 mm screen, or cut with fine scissors for very small samples. Necrotic portions of Populus and Salix leaves were removed before grinding. For each plant, a 3–5 mg subsample was weighed to the nearest microgram, and analyzed for [C], [N], and δ13C using a Carlo Erba NA 1500 elemental analyzer (Milano, Italy) and a VG Isochrom continuous-flow isotope ratio mass spectrometer (Isoprime Inc., Manchester, UK) at the Colorado State University Isotope Laboratory (observed precision for δ13C = 0.2‰).

An estimate of atmospheric δ13C in the glasshouse was required to calculate Δ13Cleaf from δ13Cleaf. Because the CO2 source used to create the elevated [CO2] treatment was depleted of 13C (δ13C ≈ −39.9‰), we calculated atmospheric δ13C in the elevated [CO2] treatment as

display math

where [CO2]amb is the [CO2] in the ambient treatment (440 ppmv), [CO2]elev is the [CO2] in the elevated treatment (752 ppmv), δ13CCO2 is the δ13C of the CO2 source for the elevated [CO2] treatment (−39.9‰), and δ13Cgh(amb) is the ambient atmospheric δ13C in the glasshouse. We assumed that δ13Cgh(amb) was −10‰, based on recent atmospheric δ13C measurements in Salt Lake City, UT, USA (Pataki et al., 2007), which is similar in population size to Fort Collins, but set within a denser metropolitan area. Atmospheric δ13C in Salt Lake City ranged from −8 to −15‰ (mean c. −10.5‰) downtown and from −8 to −13‰ (mean c. −9.5‰) in a residential neighborhood. Given this uncertainty, we also conducted a sensitivity analysis in which we calculated Δ13Cleaf with δ13Cgh(amb) from −8 to −15‰. The estimate of δ13Cgh(elev) assumes even mixing of ambient and injected CO2 in the glasshouse. While imperfect mixing could lead to spatial variation in Δ13Cleaf within the elevated [CO2] treatment, this probably did not influence treatment effects, because the treatments were arranged in blocks and blocks were rotated frequently within bays.

Seedling measurements were compared among treatments with ANOVAs in Proc Mixed (SAS 9.2, Cary, NC, USA), with [CO2], water, species and all interactions as fixed effects and block and block × water as random effects. Heights were analyzed with a repeated-measures ANOVA using a first-order autoregressive variance structure. The Kenward–Roger method was used to estimate degrees of freedom. To correct nonnormality and heteroscedasticity, total biomass and heights were cube-root-transformed, Ψ, Δ13Cleaf, [N], and C : N were log-transformed, root : shoot ratios were inverse cube-root-transformed, and percent cover of necrotic leaf tissue was arcsine, square-root-transformed for analysis.


Growth and survival

Elevated [CO2] increased mean total biomass by 15% (Fig. 1a–e; Table 1). The [CO2] effect did not vary significantly among species or water treatments. Lower water availability in the faster water-table decline treatments (Fig. 2) reduced total biomass. Salix and Populus were the most strongly affected by lower water availability; they were the largest plants in the 0.5 cm d−1 treatment and the smallest in the 3.0 and 4.0 cm d−1 treatments (water × species interaction, Tables 1, S1).

Table 1. ANOVA results for effects of water, [CO2] and species on seedling growth
FactorndfTotal biomassRoot : shoot ratioMaximum root lengthHeight
ddf F P a ddf F P ddf F P ddf F P
  1. Post-hoc ANOVA results for the significant water × species interactions are provided in Supporting Information Table S1.

  2. a

    Significant effects are in bold (P < 0.05).

  3. b

    Salix exigua, Populus deltoides spp. monilifera, Tamarix spp., Elaeagnus angustifolia, Ulmus pumila.

  4. c

    Weeks 2–11.

  5. ndf, numerator degrees of freedom; ddf, denominator degrees of freedom.

Water328332.6  < 0.0001 2155.5  < 0.0001 2128.2  < 0.0001 320232.6  < 0.0001
[CO2]124915.6 0.0001 2494.7 0.03 24611.0 0.001 32050.0  < 0.0001
Water × [CO2]32490.70.62490.20.92460.40.73200.90.5
Speciesb42495.3 0.0004 24971.3  < 0.0001 24614.5  < 0.0001 320300.8  < 0.0001
Water × species122499.2  < 0.0001 2499.6  < 0.0001 2463.1 0.0003 3204.6  < 0.0001
[CO2]  × species42491.60.22491.90.12461.50.23200.70.6
Water × [CO2] × species122490.90.62491.00.42461.30.23201.20.3
Weekc9         2493973.4  < 0.0001
Week × water27         250976.0  < 0.0001
Week × [CO2]9         24935.5  < 0.0001
Week × species36         251044.8  < 0.0001
Week × water × [CO2]27         25091.8 0.006
Week × water × species108         25053.0  < 0.0001
Week × [CO2] × species36         25101.7 0.008
Week × water × [CO2] × species108         25051.20.06
Figure 1.

Mean biomass and root : shoot ratio with each water-table decline rate and [CO2] treatment combination for: (a, f) Salix, (b, g) Populus, (c, h) Tamarix, (d, i) Elaeagnus, and (e, j) Ulmus. In (a–e), the lower, darker portions of the stacked bars are below-ground biomass, the upper, lighter portions are above-ground biomass, and the error bars are for total biomass. ANOVA results are presented in Table 1. Different lower-case letters above the bars indicate significant differences in mean total biomass or root : shoot ratio among water treatments for each species (Tukey HSD, P < 0.05; Supporting Information Table S1). Error bars are one standard error of the mean.

Figure 2.

(a) Mean soil moisture in each cm of soil above a static water table 64 cm below the soil surface. Heights above the water table for soil moisture (± SD) at field capacity (−0.033 MPa; 3.0 ± 0.4% moisture) and the wilting point (−1.5 MPa; 1.6 ± 0.1% moisture) are indicated with arrows. Error bars are one standard deviation. (b) Estimated mean soil moisture above the water table in each water treatment over the course of the experiment (calculated from the data in (a)).

Elevated [CO2] increased mean root : shoot ratios by 4% (Fig. 1f–j; Table 1). The [CO2] effect did not vary significantly among species or water treatments. Lower water availability affected root : shoot ratios of some species and not others (water × species interaction, Table 1). Specifically, lower water availability increased Tamarix, Elaeagnus, and Ulmus root : shoot ratios, but not Salix and Populus root : shoot ratios (Table S1). Salix, Populus, and Ulmus root : shoot ratios tended to be greater than Tamarix and Elaeagnus root : shoot ratios, but this difference decreased with lower water availability (Table S1).

Elevated [CO2] increased shoot heights in most weeks of the experiment (Table 1), including a 19% increase in the final week (Fig. 3a–e). The effect of [CO2] on heights varied with time for some water treatments and species, becoming more consistent later in the experiment (week × water × [CO2] and week × [CO2] × species interactions, Table 1). Specifically, elevated [CO2] did not affect heights in the 0.5 cm d−1 treatment in the first 7 wk (post-hoc week × [CO2], F9,632 = 3.4, P = 0.0004) and did not affect Tamarix and Ulmus heights in the first 2 wk (post-hoc week × [CO2], F9,502 = 3.3, = 0.0006; F9,504 = 4.3, P < 0.0001, respectively). Lower water availability reduced mean heights of all species starting in the fifth week (week × water × species interaction, Table 1), and reduced Salix, Populus, and Tamarix heights more than Elaeagnus and Ulmus heights (post-hoc water × species, P < 0.0001 for weeks 6–11). At the start of the experiment (week 2), Elaeagnus and Ulmus were considerably taller than Populus, Tamarix, and Salix across all treatments (7.7 ± 0.2 and 5.5 ± 0.1 compared with 1.5 ± 0.1, 0.39 ± 0.03, and 0.29 ± 0.02 cm, respectively) (post-hoc species, F4,273 = 1684.0, P < 0.0001). At the end of the experiment (week 11), species differences in height varied among water treatments (post-hoc water × species, F12,249 = 7.6, P < 0.0001). In the 3.0 and 4.0 cm d−1 treatments, Elaeagnus, Ulmus and Tamarix were taller than Populus and Salix, whereas in the 0.5 cm d−1 treatment, Tamarix was taller than Elaeagnus and Elaeagnus was taller than Ulmus, Populus, and Salix (Fig. 3a–e).

Figure 3.

Mean heights in week 11 and maximum root lengths at harvest with each water-table decline rate and [CO2] treatment combination for: (a, f) Salix, (b, g) Populus, (c, h) Tamarix, (d, i) Elaeagnus, and (e, j) Ulmus. ANOVA results for heights across all weeks and maximum root lengths are presented in Table 1. Different lower-case letters above the bars indicate significant differences in mean height or maximum root length among water treatments for each species (Tukey HSD, P < 0.05; post-hoc ANOVA results for heights across all weeks and maximum root lengths are presented in Supporting Information Table S1). Error bars are one standard error of the mean.

Elevated [CO2] increased mean maximum root lengths by 12% (Fig. 3f–j; Table 1). The [CO2] effect did not vary significantly among species or water treatments. Effects of the water treatments on maximum root lengths varied among species (water × species interaction, Table 1). Mean maximum root lengths were longer in the 1.5 cm d−1 treatment than in all other water treatments for Populus and Ulmus; longer in the 1.5 cm d−1 treatment than in the 0.5 and 4.0 cm d−1 treatments for Elaeagnus; and longer in the 1.5 cm d−1 treatment than in the 4.0 cm d−1 treatment for Tamarix and Salix (Table S1). Elaeagnus and Ulmus maximum root lengths were shorter than the other species in the 0.5 cm d−1 treatment, and Elaeagnus maximum roots lengths remained relatively short in the 1.5 cm d−1 treatment (Table S1). Salix maximum root lengths were longer than all other species in the 0.5 cm d−1 treatment.

At harvest, the water table in the 1.5, 3.0, and 4.0 cm d−1 treatments was 98 cm below the soil surface. Estimating from Fig. 2a, soil with moisture at field capacity (3.0 ± 0.4%; −0.033 MPa) was c. 70 cm below the soil surface, and soil with moisture at the wilting point (1.6 ± 0.1%; −1.5 MPa) was c. 57 cm below the soil surface. The water table in the 0.5 cm d−1 treatment was 44 cm below the soil surface and soil with moisture at field capacity was c. 15 cm below the soil surface. Mean maximum root lengths at harvest were long enough to reach the water table for all species in the 0.5 cm d−1 treatment and for Populus, Salix, and Tamarix (under elevated [CO2]) in the 1.5 cm d−1 treatment (Fig. 3f–j). Mean root lengths in the 3.0 cm d−1 treatment were too short to reach the water table, but long enough to reach depths with moisture > 3.0%, or at least > 1.6%. Mean root lengths in the 4.0 cm d−1 treatment were long enough to reach depths with 3.0% moisture only for Salix under elevated [CO2], and were too short to reach depths with 1.6% moisture for several species × [CO2] treatment combinations. However, maximum root length may be an overestimate or underestimate of maximum root depth, because the root length may have been concentrated near the soil surface, or deep, fine roots may have been broken and not measured.

Most Populus (90% of plants) developed necrotic lesions on their leaves that appeared to result from pathogen infection (Fig. S2). Elevated [CO2] increased abundance of the necrotic lesions, tripling mean percent cover on Populus leaves (22 ± 3% compared with 7 ± 2%) ([CO2], F1,28 = 20.0, P = 0.0001). Many Salix (66% of plants) developed smaller necrotic lesions, and also exhibited greater percent cover of the lesions under elevated [CO2] than under ambient [CO2] (4 ± 1% compared with 2 ± 1%) ([CO2], F1,28 = 5.5, P = 0.03). A few Elaeagnus (8% of plants) had small necrotic regions on some leaves under elevated [CO2] (2 ± 1%), but not under ambient [CO2]. Minor leaf necrosis on most Ulmus (89% of plants) did not differ significantly between [CO2] treatments (4 ± 1%). Tamarix did not exhibit leaf necrosis.

Only three of the 320 study plants died during the experiment. All three received ambient [CO2], including two Salix (4.0 cm d−1 treatment) and one Populus (3.0 cm d−1 treatment).

Water potential and tissue chemistry

Elevated [CO2] affected Ψpd and Ψmd at harvest for some species in some water treatments (water × [CO2] × species interaction, Table 2; Fig. 4). Specifically, elevated [CO2] increased Ulmus Ψpd and Ψmd, particularly in the 0.5 and 1.5 cm d−1 treatments, although the water × [CO2] interaction was not quite significant, and decreased Tamarix Ψpd and Ψmd in the 4.0 cm d−1 treatment, but did not affect Ψpd and Ψmd of any other species (Table S2). Lower water availability also affected Ψpd and Ψmd for only some species (water × [CO2] × species interaction, Table 2). Lower water availability reduced Elaeagnus and Ulmus Ψpd and Ψmd and Populus Ψpd; increased Tamarix Ψpd and Ψmd under ambient [CO2]; and did not affect Salix Ψ (Table S2). Many of the Salix in the 3.0 and 4.0 cm d−1 treatments were too small to measure in the pressure chamber, which reduced the power to detect a water effect on Salix Ψ. Differences in Ψ among species depended on water treatment and time of day (time × water × species interaction, Table 2). In the 0.5 and 1.5 cm d−1 treatments, Populus Ψpd was greater than all other species and Tamarix Ψpd was lower than all other species (Table S2). Further, in the 0.5 cm d−1 treatment, Elaeagnus Ψmd was greater than Tamarix Ψmd. In the 3.0 cm d−1 treatment, Ψpd and Ψmd were greater for Populus than for all other species, and in the 4.0 cm d−1 treatment, Ψpd and Ψmd were greater for Populus than for Tamarix and Ulmus (Table S2). Ψpd was greater than Ψmd except for Populus in the 3.0 and 4.0 cm d−1 treatments (time × water × species interaction, Table 2).

Table 2. ANOVA results for effects of water, [CO2] and species on seedling water potentials and tissue chemistry
FactorndfStem water potentialΔ13CleafLeaf C : NLeaf [N]
ddf F P a ddf F P ddf F P ddf F P
  1. Post-hoc ANOVA results for the significant interactions are provided in Supporting Information Tables S2–S4.

  2. a

    Significant effects are in bold (P < 0.05).

  3. b

    Salix exigua, Populus deltoides spp. monilifera, Tamarix spp., Elaeagnus angustifolia, Ulmus pumila.

  4. c

    Predawn vs midday.

  5. ndf, numerator degrees of freedom; ddf, denominator degrees of freedom.

Water3262.40.0926918.0 < 0.0001 2136.1 < 0.0001 2130.3 < 0.0001
[CO2]12010.60.5269209.9 < 0.0001 2497.5 0.007 2496.9 0.009
Water × [CO2]32003.2 0.03 2695.5 0.001 2490.40.72490.60.6
Speciesb420150.4 < 0.0001 269159.7 < 0.0001 24924.7 < 0.0001 24917.7 < 0.0001
Water × species122003.5 0.0001 2699.0 < 0.0001 2491.9 0.04 2491.60.1
[CO2] × species42002.30.062696.8 < 0.0001 2490.20.92490.40.8
Water × [CO2] × species122001.8 0.04 2691.40.22490.90.62490.70.8
Timec928117.7 < 0.0001          
Time × water27272.60.07         
Time × [CO2]92000.10.7         
Time × species362015.8 0.0002          
Time × water × [CO2]271990.20.9         
Time × water × species1082002.1 0.02          
Time × [CO2]  × species362000.30.9         
Time × water × [CO2] × species1081990.80.6         
Figure 4.

Mean predawn and midday stem water potentials (Ψpd and Ψmd) with each water-table decline rate and [CO2] treatment combination in week 12 of the experiment for: (a) Salix, (b) Populus, (c) Tamarix, (d) Elaeagnus, and (e) Ulmus. Ambient and elevated [CO2] treatments are offset horizontally to facilitate viewing. ANOVA results are presented in Table 2. Different lower-case letters in (b) indicate significant differences in Populus Ψpd among water treatments, averaged across [CO2] treatments (Tukey HSD, P < 0.05; Supporting Information Table S2). Different lower-case letters in (c) indicate significant differences in Tamarix Ψ among water treatments under ambient [CO2], averaged across times of day. Different lower-case letters in (d) and (e) indicate significant differences in Elaeagnus and Ulmus Ψ among water treatments, averaged across [CO2] treatments and times of day. Error bars are one standard error of the mean.

Assuming that ambient atmospheric δ13C in the glasshouses (δ13Cgh(amb)) was between −8 and −10‰, elevated [CO2] reduced Δ13Cleaf across all water treatments and species (Fig. 5a–e; Table 2). The magnitude of this effect was smaller for Elaeagnus than for the other species ([CO2] × species interaction, Table 2), and was smaller in the drier treatments than in the wetter treatments (water × [CO2] interaction, Table 2). The sensitivity analysis indicated that the negative effect of elevated [CO2] on Δ13Cleaf was robust to assumptions of a more negative δ13Cgh(amb), except for Elaeagnus if δ13Cgh(amb) ≤ −11‰ and for the 3.0 and 4.0 cm d−1 treatments if δ13Cgh(amb) ≤ −12‰. The δ13Cgh(amb) would have to have been ≤ −15‰ for elevated [CO2] not to have significantly reduced Δ13Cleaf for the other species and water treatments. Effects of water-table decline rate on Δ13Cleaf differed among [CO2] treatments and species (water × [CO2] and water × species interactions, Table 2). Specifically, lower water availability reduced Δ13Cleaf under ambient [CO2] but not elevated [CO2] (across species), and for Elaeagnus but not other species (across [CO2] treatments) (Table S3).

Figure 5.

Mean Δ13Cleaf and leaf C : N ratio with each water-table decline rate and [CO2] treatment combination for: (a, f) Salix, (b, g) Populus, (c, h) Tamarix, (d, i) Elaeagnus, and (e, j) Ulmus. ANOVA results are presented in Table 2. Different lower-case letters above the bars indicate significant differences in mean Δ13Cleaf or C : N among water treatments (Tukey HSD, P < 0.05; Tables S3, S4). Error bars are one standard error of the mean.

Elevated [CO2] decreased mean leaf [N] by 7%, from 1.29 ± 0.03 to 1.20 ± 0.03% (Fig. S3; Table 2), and increased mean leaf C : N by 8% (Fig. 5f–j; Table 2). Lower water availability increased leaf [N] by 36%, from 1.06 ± 0.02% in the 0.5 and 1.5 cm d−1 treatments to 1.44 ± 0.04% in the 3.0 and 4.0 cm d−1 treatments, and decreased leaf C : N for most species, but did not significantly affect Elaeagnus leaf C : N (water × species interaction, Tables 2, S4). Elaeagnus leaf [N] was less than in the other species (0.97 ± 0.04% compared with 1.32 ± 0.03%), and Elaeagnus leaf C : N was greater than in the other species, particularly in the drier treatments (Table S4).


Water and [CO2] effects on seedling growth

Increased drought and lower summer streamflows predicted under climate change may reduce riparian plant growth and recruitment, and favor drought-tolerant species, including exotics, in semiarid western North America (Stromberg et al., 2010; Perry et al., 2012). The large negative effects of faster water-table decline rates on seedling biomass (70–97%) support the prediction that lower water availability under climate change will severely limit riparian seedling growth. Our results do not, however, support the hypothesis that elevated [CO2] will ameliorate these effects by increasing plant WUE. Instead, the small positive effect of elevated [CO2] on biomass (15%) suggests that increases in seedling growth and WUE under elevated [CO2] will be insufficient to offset the negative effects of increased aridity. Where aridity does not increase, however, seedling growth may increase moderately under elevated [CO2].

Our results also support the prediction that increased aridity will hinder seedling recruitment of native Salix and Populus spp. more than exotic woody riparian species. The larger negative effects of the drier treatments on Salix and Populus biomass than on exotic species biomass support previous studies on Tamarix and Elaeagnus (Shafroth et al., 1995; Glenn & Nagler, 2005; Lite & Stromberg, 2005; Reynolds & Cooper, 2010) and suggest that Ulmus is also more drought-tolerant than the native species. The similar responses of the native and exotic species to elevated [CO2] suggest that elevated [CO2] will not ameliorate the lower drought tolerance of the native species relative to the exotics.

Although quantitative predictions of riparian soil moisture under climate change are not available, the differences in soil moisture among our water treatments were similar in scale to projected changes in summer streamflows in western North America under climate change. Summer streamflows, which strongly influence riparian soil moisture (Naiman et al., 2005), declined by as much as 20% in the last century (Miller & Piechota, 2008; Rood et al., 2008) and are projected to decline by up to 50% in the next century (Christensen et al., 2004; Dettinger et al., 2004; Maurer, 2007), as a result of smaller snowpacks and earlier snowmelt at the headwaters. Similarly, estimated soil moisture in the 4.0 cm d−1 treatment was 25% less than in the 1.5 cm d−1 treatment and 52% less than in the 0.5 cm d−1 treatment.

[CO2] effects on WUE and water status

Contrary to expectations, lower water availability did not reduce seedling growth less under elevated [CO2] than under ambient [CO2], except temporarily for heights in the first 7 wk of the experiment. Although plant growth in grasslands often responds positively to elevated [CO2] through improved water relations (Volk et al., 2000; Morgan et al., 2004, 2011), exceptions to this pattern do occur (Morgan et al., 2004; Hovenden et al., 2008). Plants exposed to a single, prolonged water stress event, such as naturally occurs in Mediterranean grasslands (Morgan et al., 2004) or the present experiment, may only briefly experience the moderate water stress that seems optimal for plant response to elevated [CO2] (Nowak et al., 2004), and may therefore exhibit little increase in growth from water savings.

Lower Δ13Cleaf under elevated [CO2] suggests that elevated [CO2] increased seedling iWUE by reducing stomatal conductance and/or increasing photosynthetic capacity, a common plant response (Polley et al., 2002; Morgan et al., 2004; Nelson et al., 2004; Leakey, 2009; Onoda et al., 2009). The increase in iWUE, however, was smaller in the drier treatments than in the wetter treatments, which may explain in part why elevated [CO2] did not improve growth more in the drier treatments. Very low soil water content can override effects of [CO2] on stomatal conductance and growth (Hunt et al., 1996; Knapp et al., 1996; Smith et al., 2000; LeCain et al., 2003). The floodplain sand and gravel bars where pioneer riparian species often establish are well drained, and soil moisture can decline rapidly during flood recession, especially in the absence of precipitation (Shafroth et al., 1998; Cooper et al., 1999; Taylor et al., 1999; Hultine et al., 2010). Very low soil moisture in the drier treatments may have limited seedling responses to elevated [CO2].

Although elevated [CO2] appeared to increase iWUE, it did not increase Ψ for most species at the time of harvest. Measured Ψ values were similar to published values for moderately water-stressed Salix, Populus, and Ulmus (Foster & Smith, 1991; Cooper et al., 2003; Dulamsuren et al., 2009; Hultine et al., 2010) but greater than published values for Tamarix and Elaeagnus (Busch & Smith, 1995; Horton et al., 2001b; Gong et al., 2006; Zhao et al., 2007). Elevated [CO2] may have increased growth mainly by increasing photosynthesis directly (Drake et al., 1997) rather than by improving water status. Alternatively, the larger seedlings under elevated [CO2] may have had less stomatal conductance per unit leaf area but greater total transpiration, resulting in unchanged Ψ (Morison, 1993).

Increased seedling root allocation and maximum root lengths under elevated [CO2] also did not alleviate effects of lower water availability in this experiment. Such effects may increase growth and survival under some conditions, however, by increasing access to water and reducing flood mortality (Mahoney & Rood, 1998; Cooper et al., 1999). Conversely, such effects might decrease growth and survival when water is abundant, by reducing canopy production and competitive ability for light (Sher et al., 2000).

Factors limiting responses to [CO2]

Low nutrient availability in the floodplain sand in our experiment may have reduced effects of elevated [CO2] on growth (de Graaff et al., 2006; Reich et al., 2006; Lewis et al., 2010; McCarthy et al., 2010). High leaf C : N in our experiment (40 ± 13) suggests N limitation. Further, increased C : N and root allocation under elevated [CO2] may reflect increased nutrient limitation (Reich et al., 2006; Dewar et al., 2009; Lewis et al., 2010). In other experiments with greater nutrient availability, elevated [CO2] increased Populus deltoides ssp. deltoides biomass by 40–100% (McDonald et al., 2002; Barron-Gafford et al., 2005; Lewis et al., 2010), compared with the 21% increase in Populus biomass in our experiment. Nutrient availability is typically low in floodplain sandbars (Adair & Binkley, 2002; Adair et al., 2004), but riparian areas with greater nutrient availability because of nutrient pollution (Carpenter et al., 1998) or fine sediment accretion (Adair et al., 2004; Naiman et al., 2010) may have larger seedling responses to elevated [CO2]. Older riparian saplings or adult trees could respond to elevated [CO2] either more strongly because soil nutrient availability increases as riparian geomorphic surfaces age (Adair et al., 2004), or less strongly because nutrient demand increases, and hence availability decreases, as forests age (Körner, 2006).

Elaeagnus growth appears to have been particularly N-limited in our experiment. Elaeagnus had lower leaf [N] and greater C : N than the other species and than reported in field studies (Follstad Shah et al., 2010), the slowest growth in the wettest treatment, and the smallest changes in growth and Δ13Cleaf under elevated [CO2]. Elaeagnus is actinorhizal (i.e. forms associations with N-fixing Frankia bacteria), but we did not observe Frankia nodules on our Elaeagnus roots. Nodulated Elaeagnus with greater N availability, and plants growing near nodulated Elaeagnus (Follstad Shah et al., 2010), may respond more strongly to elevated [CO2].

Increased necrotic lesions on Populus leaves under elevated [CO2] may also have reduced effects of elevated [CO2] on Populus growth (Newcombe, 1996). We were unable to identify the pathogen responsible, and therefore cannot assess its ecological relevance. Effects of elevated [CO2] on pathogen infection in other Populus species vary among host and pathogen species (Percy et al., 2002; Scarascia-Mugnozza et al., 2005; Chakraborty et al., 2008; Eastburn et al., 2011). More generally, effects of elevated [CO2] on the study species might be different in more complex environments with competitors, herbivores, and pathogens (Körner, 1995; Navas, 1998).

Mechanisms of drought tolerance

Tamarix drought tolerance has been attributed to its ability to maintain turgor via osmotic adjustments, prevent xylem cavitation, and use both deep and shallow soil water (Busch & Smith, 1995; Smith et al., 1998; Pockman & Sperry, 2000). As in other studies, Tamarix Ψ tended to be lower than for other species, and did not decline with lower water availability, perhaps because of low osmotic potential (Smith et al., 1998; Horton et al., 2001b, 2003; Gries et al., 2003; Xu et al., 2007). By contrast, Populus Ψ tended to be greater than for other species, and was similar at midday and predawn in the drier treatments, suggesting stomatal closure. Populus is vulnerable to xylem cavitation and relies on tight stomatal control to maintain adequate Ψ when water-stressed (Horton et al., 2001a; Cooper et al., 2003). Populus Ψmd approached reported thresholds for 50% xylem cavitation (c. −1.0 MPa; Tyree et al., 1994), whereas Tamarix Ψ did not (c. −4.5 MPa; Pockman & Sperry, 2000).

Mechanisms of Elaeagnus and Ulmus drought tolerance are less well understood. Our results suggest that they may benefit from larger seed size and greater plasticity in root allocation and iWUE. Large Elaeagnus and Ulmus seeds (87.9 and 6.3 mg seed−1 compared with 0.1, 1.3, and 0.1 mg seed−1 for Salix, Populus, and Tamarix, respectively) allowed them to grow faster initially, when water tables were still high in the drier treatments. Further, increases in root allocation with lower water availability in Elaeagnus and Ulmus (and, to a lesser degree, in Tamarix) may have increased access to soil water (Poorter & Nagel, 2000). Plasticity in root allocation has also been observed in other studies of Elaeagnus and Ulmus (Shafroth et al., 1995; Park et al., 2012), but also of Salix and Populus (Shafroth et al., 1995; Kranjcec et al., 1998; Stella & Battles, 2010), so this trait is not always limited to the exotic species. In addition, increases in iWUE (i.e. reduced Δ13Cleaf) with lower water availability in Elaeagnus may have increased water conservation and/or C assimilation (Farquhar et al., 1989). Although plasticity in root allocation and iWUE can simply reflect variation in water stress, the fact that Elaeagnus and Ulmus biomass was greater than that of the other species in the drier treatments suggests that in this case it may reflect greater drought tolerance.

Rooting depth is an important predictor of riparian seedling survival, because seedling roots must maintain contact with moisture from declining water tables as streamflow decreases through the summer (Mahoney & Rood, 1998; Stella & Battles, 2010). Differences in rooting depth are unlikely to explain species differences in drought tolerance in our experiment, however, as maximum root lengths were similar among species in the drier treatments. Maximum root lengths were greatest with a moderate water-table decline rate (1.5 cm d−1), presumably because root growth was stifled by low water availability in the drier treatments and by inundation (particularly for Elaeagnus and Ulmus and less for Salix) in the wettest treatment (Segelquist et al., 1993; Kranjcec et al., 1998; Horton & Clark, 2001; Amlin & Rood, 2002).


Global change effects on western North American riparian plants may involve interactions between increased aridity (resulting from greater evapotranspiration, lower streamflows, and shifts in streamflow timing), elevated [CO2], increased heat stress (Grady et al., 2011), and changes in flood regime and geomorphology (Perry et al., 2012). Our results suggest that elevated [CO2] is unlikely to counteract the negative effects of increased aridity on riparian woody seedling recruitment. Further, they corroborate previous research demonstrating that common exotic riparian species are more drought-tolerant than dominant native species, adding Ulmus to the list of drought-tolerant exotic species. Finally, they indicate that elevated [CO2] is unlikely to benefit seedling recruitment of mesic native species over drought-tolerant exotics. Increased aridity is likely to favor recruitment of more drought-tolerant species, including the exotics Tamarix, Elaeagnus and Ulmus, over native Salix and Populus spp., despite elevated [CO2].


Caitlin Brooks, Laura Dev, Sarah Hylander, Miranda McCutchen, Benita Mondragon, Matthew Parsons, Cynthia Pritekel, Mary Carolyn Weaver, and Daniel Willis assisted with experimental setup, maintenance and data collection. Yolima Carrillo measured δ13C of the elevated CO2 source. David Merritt provided the Scholander pressure chamber. David Williams, Kevin Hultine, and three anonymous reviewers commented on earlier drafts of the manuscript. Funding was provided by the US Geological Survey, Invasive Species Program, and the US Department of Agriculture, Agricultural Research Service Climate Change, Soils & Emissions Program. Any use of trade, product, or firm names is for descriptive purposes only and does not imply endorsement by the US Government.