Hydraulically integrated or modular? Comparing whole-plant-level hydraulic systems between two desert shrub species with different growth forms

Authors


Author for correspondence:
Susana Espino
Tel:+1 657 278 5288
Email: susanaespino@csu.fullerton.edu

Summary

  • • Hydraulic systems of shrubs vary between hydraulically integrated and modular architectures; the latter divide the shrub into independent hydraulic units. Hydraulic systems of two common North American desert shrub species, the multi-branched Ambrosia dumosa and the single-stemmed Encelia farinosa (both Asteraceae), were compared to test for division into independent hydraulic units and the implications of such a division for water loss through leaves and roots.
  • • Hydraulic systems of mature shrubs in the field were characterized using dye tracers and by documenting the degree of stem segmentation. Young pot-grown shrubs were subjected to heterogeneous and homogeneous watering. Spatial within-canopy variation of leaf water potentials and stomatal conductances, as well as soil water contents, were measured in response to manipulated soil water heterogeneity.
  • • Results show that young Ambrosia shrubs are divided into independent hydraulic units long before they physically split into separate ramets as mature shrubs, and that young and mature Encelia shrubs possess integrated hydraulic systems. No hydraulic redistribution was detected for eitherspecies.
  • • Our study shows that functional segmentation into independent hydraulic units precedes physical axis splitting, rather than being the consequence of split axes, and suggests that mature shrubs with round basal stems are likely to be hydraulically integrated.
Abbreviations: 
CV

coefficient of variation

F

hydraulic integration index defined in Eqn 2

gs

stomatal conductance (mmol m−2 s−1)

HM

homogeneous

HT

heterogeneous

IHUs

independent hydraulic units

S

stem segmentation index defined in Eqn 1

T

temperature (°C)

Tsoil

soil temperature (°C)

WT

watering trial

θv

volumetric soil water content (vol.%)

θv,pre

mean θv (vol.%) at 04.00 h in the main compartment averaged over the 3 d before watering

θv,post

mean θv (vol.%) at 04.00 h in the main compartment averaged over the 3 d after watering

θv,diff

θv,post − θv,pre

Ψleaf

predawn leaf water potential (MPa)

Ψsoil

soil water potential (MPa)

Introduction

For the growth and survival of plants, it is essential to maintain an uninterrupted water transport system throughout the xylem (Pockman & Sperry, 2000). Desert plants, in particular, have developed adaptations that prevent hydraulic failure caused by the introduction of air bubbles (emboli) into the xylem by resisting the formation of embolisms, repairing embolisms and maintaining hydraulic redundancy (Holbrook & Zwieniecki, 1999; Cruiziat et al., 2002; Sack et al., 2008). Some shrubs are structurally segmented and divided into independent hydraulic units (IHUs) (Schenk, 1999), leading to a modular type of hydraulic redundancy. Segmentation or physical splitting of stems and roots, also known as axis splitting (Fahn, 1964; Jones & Lord, 1982), is known to be the dominant growth form in arid ecosystems (Schenk et al., 2008). Although this phenomenon appears to be a common occurrence in desert ecosystems, there has been remarkably little study of the effects of the development of functional and structural axis splitting on canopy-level physiology. The purpose of this study was to test for division into IHUs in two desert shrub species, Ambrosia dumosa A. Gray Payne and Encelia farinosa A. Gray ex Torr, document the effects of such a division on canopy water relations in young plants, and test the hypotheses that division into IHUs reduces water loss through leaves and roots.

Woody plants from dry environments that divide into IHUs can be exceedingly long lived, ranging from many decades to hundreds, even thousands, of years (Vasek, 1980; Schenk, 1999; Larson, 2001). Division into IHUs could potentially increase the chance of genet survival during a drought by concentrating water from small pockets of soil moisture in a single IHU, and thereby allowing IHUs to survive, rather than spreading the water out over a whole water-stressed canopy (Jones, 1984). Other potential benefits of IHUs include a reduced loss of water to the soil as a result of lower hydraulic redistribution, because of the lack of hydraulic connections between roots with access to water and roots in dry soil, reduced loss of water through stomata by allowing independent stomatal regulation, and restricted spread of embolisms (Tyree & Sperry, 1998) and pathogens (Chatelet et al., 2006). In this article, we focus on the relationship between division into IHUs and water loss through leaves and roots.

Ambrosia dumosa and E. farinosa are very common and often coexisting plant species in the Sonoran and Mojave Deserts of North America, and are similarly sized, drought-deciduous semi-shrubs in the Asteraceae. Differences between them include a shorter lifespan, larger leaves and single basal stems in E. farinosa, compared with greater longevity, smaller leaves and branching from the ground in A. dumosa (Muller, 1953). Ambrosia dumosa tends to split along its axes into separate ramets as it matures (Jones & Lord, 1982), with plants of more than 0.5 m in diameter being normally split (Schenk, 1999), and mature shrubs have been found to respond to heterogeneous watering by developing a highly heterogeneous water status in the canopy (Jones, 1984). It remains an open question as to whether physical splitting in this species is necessary to achieve a functional division into IHUs, or whether it is merely a consequence of possessing a wood anatomy that functionally divides a young plant even before the axes split (Schenk, 1999). Encelia farinosa shrubs are very different from A. dumosa in that they possess single, round, basal stems (Schenk, 1999; Schenk et al., 2008), a trait that is more typical for shrubs from wet environments, in which it is normally associated with a high degree of hydraulic integration (Schenk et al., 2008). However, the round stems of E. farinosa could be anatomically divided into separate IHUs. The two species stood out in a previous study (Schenk et al., 2008) among all desert shrubs as possessing the most extreme opposites in their degree of stem segmentation.

This study was designed to compare the hydraulic systems of A. dumosa and E. farinosa plants, including young plants of both species with undivided stems, predicting that young A. dumosa shrubs would be divided into IHUs and that E. farinosa would have an integrated hydraulic system. We tested the hypotheses that, in the presence of heterogeneous watering, a division into IHUs would cause water not to be shared among all canopy branches, creating greater water status variability within the canopy, and causing lower overall conductance, as several branches would be at lower water status when compared with a homogeneous watering treatment. We also tested the hypothesis that division into IHUs would decrease the degree of hydraulic connectedness among roots and thereby reduce water loss from the roots as a result of hydraulic redistribution.

Materials and Methods

Field studies of mature Ambrosia and Encelia shrubs

The degree of physical segmentation of the stems and the hydraulic integration of five shrubs of each species, A. dumosa and E. farinosa, with canopy diameters of greater than 0.5 m, were investigated in January 2005 at a field site on the southern slope of the Cottonwood Spring Mountains in the transition area between the Mojave Desert and the Sonoran Desert in California, using the methods described in Schenk et al. (2008). Briefly, for each shrub, a single lateral woody root with a diameter of c. 4 mm (± 2 mm) was excavated at a soil depth between 0.1 and 0.3 m, and inserted into a vial containing acid fuchsin dye (0.5% in water) for 24 h. The shrub was then cut at the most compact section of the basal stem, usually right at the soil surface, and digital photographs of transverse sections of the basal stem (e.g. Fig. 1) at this point were analyzed using the software SigmaScan Pro (version 5.0, SPSS Inc., Chicago, IL, USA). Shrub growth forms were characterized by measurements of canopy height and widths, and canopy volume (calculated as an ellipsoid). The degree of physical segmentation of shrubs was characterized by the relationship between the cross-sectional area and perimeter of the basal stem's living sapwood and associated undecayed heartwood at the basal stem's most compact point, using the formula:

Figure 1.

Transverse sections of basal stems of mature shrubs of Ambrosia dumosa (top) and Encelia farinosa (bottom) cut at their most compact point. To better show the shape of xylem, bark and phloem were edited out of the images using Photoshop software (Adobe Systems, version 7.0). Acid fuchsin dye-tracer distribution is shown in purple.

image(Eqn 1)

(p, stem perimeter; A, basal stem cross-sectional area, such that S = 1 for a perfectly circular area). The degree of hydraulic integration within the basal stem was characterized by measuring the fraction of basal stem cross-sectional area (without bark) colored by dye and calculating a hydraulic integration index as:

F = d/A(Eqn 2)

(d, area of stem cross-section colored by dye; A, total area of the basal cross-section).

Controlled pot experiments of young Ambrosia and Encelia shrubs

To investigate the hydraulic architecture of young Ambrosia and Encelia plants, an experiment was designed to grow young plants under conditions of manipulated soil water heterogeneity at a soil depth of 0.2–0.5 m (Barbour et al., 1977; Nobel, 1997). Dormant young Encelia and young Ambrosia shrubs (canopy diameters, < 0.24 m) with round basal stems that showed no external signs of axis segmentation were collected in September 2005 from the same location as used for the field studies described above. They were transported to the outdoor area of the glasshouse complex at California State University Fullerton and planted into pots made of polyvinyl chloride pipes (1.2 m tall, 0.24 m diameter; see Fig. 2). Fifty pots were placed vertically within a 17-m2 area in 10 groups of five pots each (Fig. 2b), which were held together in a circle (Fig. 2c), with groups spaced c. 0.6 cm apart. Pots were filled to the top of the pot with a mixture of soil and washed sand, resulting in a texture of loamy sand similar in texture to that from the field site, with 10.3% clay, 3.7% silt and 86.0% sand, and a field capacity of c. 14% volumetric soil moisture content (θv, %). The field capacity, defined as the soil moisture content at a soil water potential of 0.033 MPa, was calculated from the soil texture data according to Saxton et al. (1986). Each group included one control pot without a plant [in the heterogeneous (HT) or homogeneous (HM) watering treatment], two pots with one Encelia each (one each for the HT and HM treatments), and two pots with one Ambrosia each (one each for the HT and HM treatments). Half of the 10 controls and half of the Encelia and Ambrosia shrubs (20 Ambrosia and 20 Encelia) were randomly assigned to the HT treatment; the other half were assigned to the HM treatment.

Figure 2.

(a) Design of pots for controlled heterogeneous and homogeneous watering treatments. (b) Set-up of the pots in 10 groups of five pots. (c) Close-up view of the top of a group of five pots with the no-plant control covered to prevent the infiltration of water.

Each pot included an internal cylindrical compartment (28 cm tall, 12.7 cm diameter) attached to the interior side of the pot, with its upper opening 19 cm from the top of the pot and top of the soil (Fig. 2a), creating a heterogeneous soil water environment in the HT watering treatment. For HT watering treatments, internal compartments were closed at the bottom, whereas, in HM treatments, they were sealed with Nitex nylon cloth (TETKO, Inc., Briarcliff Manor, New York, USA) with 25-µm openings to restrict root growth (as in the HT treatment), but to allow water flow out of the internal compartment. Three irrigation tubes (0.6 cm in diameter) were placed vertically from the top of the soil surface down to 25 cm, and their bottom halves were connected to porous irrigation tubes closed at the lower ends. One tube was placed so that it delivered water directly to the internal compartment, and the other two delivered water to the main compartment. A layer of small river rocks was placed between 15 and 19 cm from the top, just above the internal compartment, to decrease upwards and lateral capillary movement of water.

Measurements of soil water contents

Each pot was equipped with at least two soil moisture probes (model ECH2O-10, Decagon Devices Inc., Pullman, WA, USA), inserted parallel to the ground through slits cut into the pot, and θv was recorded every 30 min. Both probes were situated 30 cm from the top of the pot, one centered within the internal compartment and the second directly opposite the first, in the main compartment (Fig. 2a). In addition, 10 Encelia and 10 Ambrosia pots were randomly selected for the insertion of three additional sensors. One was inserted vertically into the soil surface and therefore integrated measurements over the top 10 cm; the other two were placed parallel to the ground at depths of 60 and 90 cm (Fig. 2a).

Soil moisture probes were calibrated for effects of soil type and soil temperature (T) by placing them into containers with the soil mixture with a range of gravimetrically determined θv, and by placing them into a growth chamber and gradually increasing T. The calibration equation [θv as a function of probe output (in mV) and T] was created by nonlinear regression using the software TableCurve 3D (version 3, SPSS Inc., Chicago, IL, USA). Soil temperature data for the pot experiment were only available after 2006, and therefore the missing Tsoil data for 2006 were estimated from an empirical relationship developed using TableCurve 3D between data for minimum daily T of the current day, maximum daily T of the previous day at Fullerton Airport, and existing Tsoil measurements at 04.00 h Pacific Standard Time.

Watering treatments

For 6 months after planting, all plants were watered frequently and exposed to natural rainfall to facilitate plant establishment and promote vigorous root growth throughout the pots and internal compartments. On 1 April 2006, watering was discontinued and shrubs were covered during subsequent rain events to allow the soil to dry. Root activity in the compartments was evaluated by monitoring changes in θv in the internal and main compartments. The presence of active roots in a compartment was detected as a decrease in θv during the drying-out period, whereas a lack of decrease in θv was interpreted to indicate the absence of active roots in the compartment. Consequently, 12 shrubs per species (half in the HM and half in the HT treatments) were chosen from the initial 20 because they appeared to have roots in both compartments. Watering trials (WTs) and measurements of plant water status began on 9 July 2006 (Table 1), after θv had declined to less than 2% in all replicates, because previous measurements had shown a θv value of 2% to be the minimum water content in these soils (Fig. 3). For each measurement trial, 24 shrubs were measured within a 3-d period; per day, eight randomly selected shrubs were measured.

Table 1.  Dates for measurements of stomatal conductance, leaf water potentials and volumetric soil water contents for hydraulic redistribution experiments (watering treatments were applied during the evenings before the dates listed)
Watering trialDates measured
Shrubs 1–8Shrubs 9–16Shrubs 17–24
Initial (no watering)10 July 200611 July 200612 July 2006
113 July 200614 July 200615 July 2006
210 August 200611 August 200612 August 2006
324 August 200625 August 200626 August 2006
411 September 200612 September 200613 September 2006
525 April 2007 (no shrubs measured, soil water contents only)
624 July 2008 (no shrubs measured, soil water contents only)
Figure 3.

Mean volumetric soil water contents (θv) measured daily at 04.00 h for pots with Ambrosia dumosa and Encelia farinosa shrubs in heterogeneous and homogeneous watering treatments from 2 July 2006 to 13 September 2006. For each watering trial, the six pots within each treatment were watered over a 3-d period, two pots per day. To synchronize the dates according to the watering schedule, day 1 on the x-axes corresponds to either July 2, 3 or 4, depending on the first date of initial measurements. Grey arrows indicate the day of initial plant water relations measurements and black arrows indicate the days of the watering trials. (a) θv for Ambrosia shrubs at 30 cm depth. (b) θv for Encelia shrubs at 30 cm depth. (c) θv for Ambrosia and Encelia shrubs at 60 cm. (d) θv for Ambrosia and Encelia at 90 cm.

Six watering treatments (WT 1–6, see Table 1) were applied via porous watering tubes (Fig. 2a) using a 60-ml syringe the evening before the day on which the measurements of plant water status were taken. Watering amounts for both treatments were determined as the volume of water that would bring θv in the internal compartment up to field capacity. In HT treatments, this volume was added only to the internal compartment, bringing it up to near field capacity. In HM treatments, this volume was evenly distributed over the internal and main compartments via the three watering tubes, thereby creating the same below-field capacity θv in the internal and main compartments (Fig. 2).

Measurements of predawn leaf water potentials (Ψleaf) and leaf stomatal conductance (gs)

The psychrometric dew-point technique was used to measure Ψleaf (in MPa) at 04.00 h Pacific Standard Time using a water potential system (model HR-33T, Wescor Inc., Logan, UT, USA) and psychrometric sample chambers. Five leaves of Ambrosia or portions of five leaves for Encelia (c. 1 cm2) were collected from four different compass directions on the plant (N, E, S, W) and from the top of the canopy for psychrometry measurements. Psychrometric chambers were then placed inside a styrofoam box, left to equilibrate for 12 h before measuring. Leaf gs (in mmol m−2 s−1) was measured at 09.00 h with a steady-state leaf porometer (Decagon Devices Inc., Pullman, WA, USA). For measurements of the small Ambrosia leaves, the porometer was equipped with a custom-built attachment that reduced the opening for gas exchange measurements to either 1 or 2 mm diameter, and which was calibrated by measuring moistened filter paper with openings ranging from 1 to 6.35 mm. For each shrub, eight leaves from different compass directions (N, NE, E, SE, S, SW, W, NW) were randomly chosen for porometry measurements. Rates of nocturnal gs were also measured occasionally.

Measurements of hydraulic redistribution

Hydraulic redistribution was tested during the WTs in 2006 described above, and again in April 2007 and June 2008. The pots used for measurements of hydraulic redistribution were those in the HT treatment (six per species), which were equipped with a soil moisture probe in the upper 10 cm of the soil, in addition to the two sensors at 30 cm depth placed into the internal and main compartments (Fig. 2a). A significant increase in θv in the top 10 cm of the soil surface, or at 30 cm depth in the main compartment, during the 3 d after watering into the internal compartment was interpreted to indicate hydraulic redistribution from the internal compartment. To determine whether water could move from the internal into the main compartment without the aid of plant roots, WTs 5 and 6 included a control treatment that consisted of watering into the internal compartment of pots without shrubs.

Statistical analyses

Indices of stem segmentation (S) and hydraulic integration (F) were compared between mature shrubs of the two species using t-tests. To test for effects of treatments on measures of leaf water status, two separate multivariate repeated measures analyses of variance (MANOVA) were performed using SYSTAT (version 12.02, SYSTAT Software Inc., San Jose, CA, USA), one with mean Ψleaf and gs (per individual), and the second with coefficients of variation (CVs) of Ψleaf and gs (per individual) of the test plants as dependent variables. Both analyses had species and treatments as fixed effects, with measurements repeated five times (days from initial measurements: 0, 3, 31, 45, 63), followed by planned comparisons to test for treatment effects at each measurement date within species and for the treatment × species interaction.

To test for hydraulic redistribution from the internal compartment to the main compartment after watering, the mean θv value in the main compartment measured at 04.00 h, averaged over the 3 d before watering (θv,pre), was compared with the mean θv at 04.00 h of the main compartment, averaged over the 3 d after watering (θv,post). The differences between these two measures (θv,diff = θv,post − θv,pre) was used as a measure of hydraulic redistribution if positive, or absence of hydraulic redistribution if equal to or less than zero. Paired, one-tailed t-tests for each species and each water trial were used to test for the presence of hydraulic redistribution within trial dates, and significance at the level of P < 0.05 was determined by adjusting for false discovery rate in multiple comparisons (Benjamini & Hochberg, 1995; Verhoeven et al., 2005). To test for differences in hydraulic redistribution between species, the θv,diff data measured between 0 and 10 cm depth and at 30 cm depth were analyzed in MANOVA using SYSTAT with species as the independent variable and measurements repeated five times (no measurements of θv were available for WT 2). θv,diff data for WT 5 were also analyzed in a separate analysis of variance using SYSTAT. This analysis included a no-plant control treatment in addition to the two species.

Results

Field measurements of adult shrubs

Adult shrubs of the two species that were sampled for stem segmentation measurements and dye tracer experiments were of similar sizes (means ± standard errors: Ambrosia: height, 43 ± 5 cm; width, 95 ± 12 cm; canopy volume, 0.23 ± 0.07 m3; Encelia: height, 64 ± 4 cm; width, 84 ± 12 cm; canopy volume, 0.26 ± 0.07 m3). Ambrosia shrubs had highly segmented stems and were physically split into several subunits (segmentation index S = 5.2 ± 0.4; Fig. 1), whereas basal stems of Encelia were nearly round (S = 1.3 ± 0.1; Fig. 1). Mean indices S differed between species (P < 0.001). Acid fuchsin dye tracer provided to a single root was detected in basal stems of three of five Ambrosia shrubs, with between 0.6 and 3.4% (mean, 1.7 ± 0.9%) of the wood cross-sectional area colored by dye. Dye tracer was detected in all five Encelia shrubs sampled, and between 0.3 and 12.7% of the wood area was colored by dye (mean, 4.1 ± 2.3%). Mean F indices did not differ significantly between the two species (P = 0.482), but the maximum radial and tangential spread of dye was much greater in Encelia than in Ambrosia, in which the dye was physically restricted to small lobes of light-colored sapwood (Fig. 1).

Soil water contents

Volumetric soil water contents (θv, %) in the controlled outdoor experiments declined to c. 2% by the beginning of the watering treatments in July 2006 (Fig. 3a–d) and, except for short-term increases by watering, remained at these levels throughout the experiment. Watering into the internal compartment of the HT watering treatment resulted in θv of 14–17% for Encelia and 10–15% for Ambrosia during the night following watering; by contrast, the main compartment for both species remained constant at c. 2%θv throughout the measurement period. HM watering treatments increased θv in the main and internal compartments to c. 6–9%θv (Fig. 3a,b).

Predawn leaf water potentials (Ψleaf)

The Ψleaf value in summer 2006 decreased over the time of the experiment for both shrub species (Fig. 4a). The watering treatment had no effect on the absolute values of Ψleaf of either species across all measurement times (P = 0.421; Table S1, see Supporting Information), but absolute values of Ψleaf were significantly lower for Ambrosia than for Encelia (P = 0.001; Table S1). Canopy variability in Ψleaf of Ambrosia and Encelia shrubs, as characterized by CV, was not different for Encelia shrubs in the HT treatment vs HM treatment at any time (Fig. 5a; Table S2B, see Supporting Information). No differences between treatments were observed for Ambrosia shrubs in the initial measurements or in WTs 1 and 2; however, significant treatment effects developed after WTs 3 and 4 (Fig. 5a; Table S2B), when Ambrosia shrubs in the HT treatments had higher canopy variability in Ψleaf than shrubs in the HM treatment.

Figure 4.

(a) Mean [± standard error (SE)] predawn leaf water potentials and (b) mean (± SE) stomatal conductance for Encelia farinosa and Ambrosia dumosa shrubs before watering treatments were applied and in response to watering treatments in four watering trials (WT 1–4) from 16 July 2006 to 13 September 2006. There was no significant treatment for either predawn leaf water potential or stomatal conductance, but there were significant differences between species (Table S1, see Supporting Information).

Figure 5.

Mean (± standard error) coefficients of variation in predawn leaf water potentials (a) and stomatal conductance (b) for Ambrosia dumosa and Encelia farinosa shrubs before watering treatments were applied and in response to watering treatments in four watering trials (WT 1–4) from 16 July 2006 to 13 September 2006. Statistically significant differences within a species between the homogeneous vs heterogeneous treatments are indicated by an asterisk (P < 0.01).

Leaf stomatal conductance (gs)

Absolute values of gs for Ambrosia peaked around mid-August and showed a tendency to decline over time, whereas gs values for Encelia stayed more or less constant (Fig. 4b). The watering treatments had no effect on the absolute gs values of either species (Table S1). However, Ambrosia leaves had higher gs values than did Encelia leaves (P = 0.001; Table S1). Canopy variability in gs of Encelia shrubs, as characterized by CV, did not differ in response to HT vs HM treatments (Fig. 5b; Table S2B). Canopy variability in gs of Ambrosia shrubs showed a significant treatment effect in WTs 2, 3 and 4, when Ambrosia shrubs in the HT treatments had higher canopy variability in gs than did shrubs in the HM treatment (Fig. 5; Table S2B). Overall, there were significant differences between the canopy variability in gs and Ψleaf of the two species in their reaction to the watering treatments (significant species × treatment interaction in Table S2A, see Supporting Information).

Hydraulic redistribution measurements

After watering into the internal compartments of pots in the HT treatment, significant decreases in θv in the main compartments of pots with Encelia and Ambrosia shrubs and without shrubs were observed (Fig. 6a,b), indicating normal drying of the soil. No considerable increases in θv were found for either species at any depth of the main compartment, indicating that these shrubs did not perform hydraulic redistribution from the side chamber to the main chamber (Figs 3a–d, 6a,b) under the experimental conditions (Table S3, see Supporting Information). WTs 5 and 6, which included a no-plant control treatment in addition to the two species, also showed no evidence of hydraulic redistribution (Table S4, see Supporting Information), except for one Encelia plant in WT 6, which showed a nonsignificant increase in θv,diff at 10 cm (Fig. 6a).

Figure 6.

Differences between mean volumetric soil water content (θv,diff) in the main compartment averaged over two 3-d periods before and after watering (n = 6) for all six watering trials (see Table 1). (a) Top 0–10 cm soil depth; (b) 30 cm depth. Statistically significant differences within a species between before and after watering are indicted by an asterisk (P < 0.05). Black bars, Ambrosia dumosa; grey bars, Encelia farinosa; white bars, control.

Discussion

The physical segmentation and splitting of mature Ambrosia shrubs divided them into complete IHUs (Fig. 1), as observed previously (Jones & Lord, 1982; Jones, 1984; Schenk, 1999), whereas the round stems of mature Encelia shrubs provided no morphological or anatomical indication for division into IHUs. Dye in at least some Encelia plants spread widely around the stem within several growth rings (Fig. 1), indicating abundant radial and tangential connections between vessels. Despite these structural differences in mature shrubs, the fraction of wood cross-sections colored by dye did not differ between the two species. This raises the question as to whether physical splitting is required for dividing plants into IHUs, or whether the physical splitting follows from an anatomical division into IHUs that is already present in young plants.

In answer to that question, this study showed that division into IHUs can be present long before physical splitting of stems is externally apparent. Young Ambrosia shrubs showed significantly higher canopy variability in leaf water potentials (Ψleaf) and stomatal conductance (gs) in the HT than HM watering treatments, especially in the latter (Fig. 5), indicating that young Ambrosia shrubs with undivided stems are functionally divided into IHUs. Axis splitting in this species apparently does not break any functioning connections within the vascular system, and it occurs because the decaying heartwood readily succumbs to mechanical stress (Jones & Lord, 1982). For Encelia, no significant differences in canopy variability of Ψleaf and gs values occurred between the HT and HM treatments, indicating that young Encelia shrubs possess integrated hydraulic systems. This type of hydraulic architecture allows unrestricted transfer of water throughout the shrub, creating a relatively homogeneous water status within the canopy, regardless of soil water heterogeneity.

Division into IHUs could potentially reduce transpiration of the whole plant if IHUs with access to soil water opened their stomata, whereas other IHUs kept their stomata closed. However, mean gs values between the HT and HM watering treatments in Ambrosia did not differ, because this species does not appear to close its stomata very much, even when under drought stress, as observed previously (Jacobsen et al., 2008). It remains to be seen whether division into IHUs in other species reduces water loss, but it may be that a modular hydraulic system is primarily a risk-spreading strategy that increases the probability of survival for the whole genet by spreading the risk of mortality over independent modules, which cannot negatively affect each other (Watkinson & White, 1985; Eriksson & Jerling, 1990; Schenk, 1999). In our experiment, plants were probably not stressed to the point that some IHUs would desiccate and die.

Few other studies have investigated within-plant heterogeneity in water status and its relationships to hydraulic traits. In temperate tree and shrub species, different rates of water supply to leaves and branches have been interpreted to be caused by restricted vascular pathways (Orians et al., 2004, 2005; Zanne et al., 2006). It has been hypothesized that these restrictions may be caused by a lack of intervessel connections (Ellmore et al., 2006; Zanne et al., 2006), by hydraulic isolation of vessels in a matrix of nonconducting cells, such as air-filled fibers, that surround the vessels (Sano et al., 2005; Schenk et al., 2008), or by the presence of heartwood between functional strands of sapwood (Keeley, 1975; Burgess & Bleby, 2006).

The significant effects of heterogeneous watering on spatial canopy water relations of Ambrosia occurred when the mean Ψleaf value decreased to less than − 2 MPa (Figs 4a, 5). A decreasing Ψsoil over time would be expected to cause a decrease in Ψleaf (Turner, 1974). However, throughout the measurement period, θv remained constant at c. 2% (Fig. 3a–d), except during days after watering, suggesting that some other factor, perhaps increasing vapor pressure deficit, must have caused the declining Ψleaf. Depletion of water stored in plant tissues over the course of the experiment, i.e. capacitance (Meinzer et al., 2001), could be an alternative explanation for the decreasing Ψleaf. Because young Ambrosia shrubs appear to be divided into IHUs, in the HT treatment, only the roots in the internal compartments had access to additional water, which may have kept tissues of IHUs that were supplied by those roots hydrated. The other IHUs with roots in the main compartment (Fig. 3a), and subjected to low θv (2%), may have used most of the available internally stored water to stay functional. In the HM watering treatments, the overall declining Ψleaf in Ambrosia suggests that transpiration demands were not met by watering, causing water status to decrease, and presumably depleting internally stored water.

Mean Ψleaf in Encelia also decreased over the course of the experiment, albeit to a lesser degree than in Ambrosia shrubs (Fig. 4a). Encelia shrubs may have also depleted internally stored water, but their stems possess abundant phloem parenchyma, which, in well-watered plants, forms a succulent tissue around the xylem. This probably causes the stem capacitance of Encelia to be higher than that of Ambrosia, and the integrated hydraulic system may have allowed Encelia shrubs to refill water storage cells throughout the plant in response to both watering treatments.

It may be expected that, in the HT treatment, branches and leaves located directly over the internal compartment would show higher gs and Ψleaf. To test this concept, we used circular statistics to test for a directional distribution of gs and Ψleaf with reference to the directional location of the internal compartment. No directional effects were found (results not shown), which came as no surprise, because Ambrosia stems often intertwine spirally and irregularly (Jones & Lord, 1982), and because our results show that Encelia appears to be hydraulically integrated.

The degree of spatial canopy variability may be expected to differ between Ψleaf at predawn and gs during the day. Predawn measurements are made when water status within a canopy has come as close to equilibrium as the hydraulic architecture of the plant and environmental conditions allow. By contrast, daytime gs may vary depending on the exposure of leaves to solar radiation and wind. Stomatal conductance may also vary in response to heterogeneous soil water when high transpirational demand causes axial xylem water flow to vastly exceed any radial or tangential water flow, thereby preventing the equilibration of water status among vessels. Axial hydraulic conductance tends to be orders of magnitude higher than radial or tangential conductance, even in hydraulically integrated plants (Zwieniecki et al., 2001). In contrast with these expectations, the observed CVs for Ψleaf and gs in this experiment were similar within and between species (Table S2), possibly because both species had substantial rates of gs at night (Encelia, 35–49 mmol m−2 s−1; Ambrosia, 39–64 mmol m−2 s−1), which may have reduced canopy equilibration at night.

The anatomical traits that may confer different degrees of hydraulic isolation of vessels in Encelia and Ambrosia shrubs are revealed in tangential sections of their basal stems (Fig. 7). Xylem in basal stems of Ambrosia has very straight, parallel vessels, with bands of thick-walled libriform fibers separating the vessels from each other and from ray cells. Fibers are considered to be nonconducting cells (Tyree & Zimmermann, 2002; Sano et al., 2005), acting as a barrier for water transfer, especially when they are air-filled (Utsumi et al., 1998; Umebayashi et al., 2007). Because vessel groups are arranged in radial bands in Ambrosia, the fiber bands probably restrict tangential flow in the xylem of this species, thereby creating radial vessel groups that function as IHUs. By contrast, the basal stem xylem of Encelia shrubs has wavy vessels with greater potential for vessel-to-vessel contact, which should allow for radial and tangential flow among vessels (Fig. 7). Moreover, the vessels are in direct contact with ray parenchyma cells, which are thought to aid in intervessel water transport (Tyree & Zimmermann, 2002). The wood anatomical traits found in Encelia are clearly indicative of a shrub with an integrated hydraulic system.

Figure 7.

Tangential wood sections from basal stems of mature Encelia farinosa (right) and Ambrosia dumosa (left) shrubs. Encelia has wavy vessels in direct contact with libriform fibers and ray parenchyma cells. Ambrosia has straight vessels isolated from one another and from rays by thick-walled libriform fiber cells (bar, 350 µm). Images reproduced by permission of Hugo I. Martínez Cabrera, University of Connecticut.

Reduced water loss caused by hydraulic redistribution may be expected to be one of the many potential benefits of division into IHUs for plants in a water-limited environment, as there would be less opportunity for water exchange within the root system. Our experiments yielded no evidence for hydraulic redistribution within root systems of either of the two species (Fig. 6a,b). Compared with average changes in θv of c. 7%, observed as a result of hydraulic redistribution in Artemisia tridentata (Ryel et al., 2003), the differences observed here (θv < 0.7%) are negligible. Hydraulic redistribution has been reported previously for Ambrosia shrubs in the Mojave Desert (Yoder & Nowak, 1999). However, the evidence from that study was not strong, because it lacked measurements of Ψsoil fluctuations in bare soil without plants. Two other studies found circumstantial evidence for hydraulic redistribution in Ambrosia based on the fact that annuals and other shrub species benefited from their close spatial association with Ambrosia (Holzapfel & Mahall, 1999; Schenk & Mahall, 2002). We know of no previous research on hydraulic redistribution in Encelia.

Why did hydraulic redistribution not occur in the shrubs of this study? Rectifier properties in roots that allow water flow into roots, whilst minimizing reverse flow (Caldwell et al., 1998), may be a reason. Night-time transpiration is another plant characteristic that potentially could restrain the magnitude or pattern of hydraulic redistribution (Ryel et al., 2002; Hultine et al., 2003; Bauerle et al., 2008). Both species in this study had substantial rates of nocturnal gs. Shrubs that transpire significant amounts of water at night are likely to prevent loss of water from roots to the soil because their xylem Ψ will not rise above Ψsoil (Hultine et al., 2003). This seems to be the most probable explanation for a lack of hydraulic redistribution in our study.

The degree of connectedness within a plant's hydraulic system, i.e. hydraulic integration, is becoming a major focus of plant hydraulics research (Pratt et al., 2008). Connectedness allows water sharing between plant modules that differ in water status (Marshall, 1990) but, at the same time, it allows embolisms to spread within a hydraulic system (Wheeler et al., 2005; Hacke et al., 2006; Loepfe et al., 2007), and does not allow for the hydraulic isolation of vessels that is required for embolism repair whilst the xylem is under tension (Holbrook & Zwieniecki, 1999). From this, one would predict that a low degree of hydraulic integration should be a common trait in desert shrubs, and this seems to be the case (Waisel et al., 1972; Schenk, 1999; Lambert, 2007; Schenk et al., 2008). Shrubs divided into IHUs tend to have high longevity (Schenk, 1999), suggesting that this trait is adaptive for perennial plants in a desert environment. Results from this study show that functional segmentation into IHUs precedes physical axis splitting, at least in Ambrosia, rather than being the consequence of split axes. Our study also supports the finding of Schenk et al. (2008) that shrubs with round basal stems are likely to be hydraulically integrated. Division into IHUs allows independent stomatal regulation of IHUs in Ambrosia, but this did not lead to reduced leaf water loss under our experimental conditions (Fig. 4b), and we found no effect of the degree of hydraulic integration on water loss from roots as a result of hydraulic redistribution. Determining whether division in IHUs translates into increased fitness or survivorship in the field will require long-term studies of branch-level and genet-level demography.

Acknowledgements

We thank David Ackerly and three anonymous reviewers for their helpful comments and suggestions. We owe thanks to Ed Read, Daisha Ortega, Christine Goedhart, Greg Pongetti, Jessica Hoang, Nichole Cervin, Maya Mezon, Calvin Threat, Victor Herrera, Ana Espino and Rosa Hernandez for their generous help in conducting the research. We also thank Hugo I. Martínez-Cabrera for providing the anatomical pictures used in this paper, Cynthia S. Jones for helpful comments on the manuscript, and Darren R. Sandquist and Sean E. Walker for their comments on earlier versions of the manuscript. Funding for this research was provided by grants from the Andrew W. Mellon Foundation and the National Science Foundation (IOS-0641765) to H.J.S.

Ancillary