Effects of surface and sub-surface soil horizons on the seasonal performance of Larrea tridentata (creosotebush)


  • Erik P. Hamerlynck,

    1. Rutgers University, Department of Biological Sciences, Newark, NJ, USA,
    2. University of Nevada, Las Vegas, Department of Biological Sciences, Las Vegas, NV, USA, andThe Desert Botanical Garden, Phoenix, AZ, USA
    Search for more papers by this author
  • Joseph R. Mcauliffe,

  • Stanley D. Smith

    1. University of Nevada, Las Vegas, Department of Biological Sciences, Las Vegas, NV, USA, andThe Desert Botanical Garden, Phoenix, AZ, USA
    Search for more papers by this author


1. In warm arid and semiarid environments, the accumulation of clay minerals produces increasingly well developed soil horizons with the passage of time. Differences in the strength of development of two prominent soil horizons, silt- and clay-rich surface vesicular (Av), and clay-enriched subsurface argillic (Bt), may strongly influence the amount and seasonal continuity of plant-available water and the physiological activity of long-lived desert shrubs. Three sites were selected on an alluvial piedmont (bajada) in the Mojave Desert that varied in surface and subsurface horizon development. The first site, a deep deposit of stabilized dune sand, entirely lacked soil horizon development. The second site had a well developed surface stone pavement and underlying Av horizon, but lacked an argillic horizon in the sandy subsoil. The third had a well developed surface pavement and Av horizon, and a deeper, well developed clay-rich argillic horizon. Seasonal water potential and gas-exchange responses of the evergreen desert shrub Larrea tridentata[DC.] Cov., and volumetric soil water content (θ), were measured monthly in 1996 on these three soils in order to test the hypothesis that desert pavements, Av and subsurface Bt horizons differentially affect the effectiveness and utilization of seasonal precipitation.

2. Predawn and midday water potentials (ψpd and ψmid), net photosynthetic rates (Anet), and stomatal conductances (gs) in L. tridentata were highest in the deep, sandy dune soils lacking horizons that could restrict surface and subsurface infiltration. Plants growing in these soils also showed no physiological response to summer precipitation events. Following a single large precipitation event during the growing season (3·8 cm), the water potentials, Anet and gs in L. tridentata were similar in the first (sand dune) and second (pavement and Av horizon) sites. However, plant performance on these soil surfaces showed marked seasonal declines, and did not respond to a small pulse of summer rainfall. Plants growing at the third site (older soils with strongly developed pavement, Av and Bt horizons) had very low gas-exchange rates and water potentials. However, following convectional summer thunderstorms L. tridentata showed improved water relations and gas exchange in these soils.

3. Midday water potentials were frequently anomalously higher than predawn water potentials, up to +6 MPa late in the growing season, especially on soil surfaces with well developed soil horizons. This anomaly was due to seasonal decreases in ψpd accompanied by invariant midday ψ. In general, ψpd was correlated with θ across the depths measured.

4. Correlations between ψmid with θ at 35 cm increased dramatically with increasing Bt horizon development, suggesting that seasonal ψmid may have been due to the vertical translocation of water (‘hydraulic lift’).

5. Our results show that subsurface and surficial soil horizon development differentially affects the seasonal availability of water for desert plants. During wetter parts of the season, subsurface horizons limit the degree of water availability, while surface soil characteristics have the greatest influence on the effectiveness of summer precipitation. These findings suggest that the effectiveness of climatic precipitation and attendant plant utilization of water resources in warm desert systems may depend on the physical soil condition.


Water availability is the limiting factor in many ecological processes in warm deserts (Ehleringer 1985; Smith, Monson & Anderson 1997). Soil conditions modify the precipitation signal by affecting surface and subsurface infiltration, depth of moisture storage, and the temporal persistence of plant-available moisture (Noy-Meir 1973). Consequently, understanding desert ecological processes requires knowledge of the spatial distributions and hydrologic behaviours of soils and the subsequent impact on plant performance. With this view, the effectiveness of seasonal precipitation is not exclusively a function of climate, that is, the distribution, frequency, intensity, and duration of rainfall, but is fundamentally linked to the edaphic features that differentially transmit the precipitation signal to various rooting depths in the soil (Gile, Gibbens & Lenz 1998; McAuliffe 1994, 1999a; Parker 1995).

The Mojave Desert is North America's most arid desert (Smith et al. 1997), and any feature influencing soil hydrology is likely to have a major effect on long-lived desert shrubs such as Larrea tridentata[DC.] Cov. (creosotebush) which can maintain physiological activity throughout the year (Oechel, Strain & Odening 1972; Schlesinger & Jones 1984; Schlesinger, Fonteyn & Reiners 1989). It is well known that cemented calcic soil horizons (caliche) affect the soil water balance, water relations and productivity of this evergreen shrub (Cunningham & Burke 1973; Parker 1995; Shreve & Mallery 1933; Smith et al. 1995). However, cemented calcic horizons are not present in all desert soils, and represent only a single and rather extreme aspect of soil horizon development in arid land soils. Two more widespread soil horizons include surface vesicular A (Av) and subsurface clay-enriched argillic (Bt) horizons (McAuliffe 1999a, 1999b; McDonald, McFadden & Wells 1995). Both horizons become increasingly well developed with the passage of time. The deposition of fine eolian material (silt- and clay-rich dust) and incorporation of these materials beneath surface stones over time produces the Av horizon beneath a stone pavement (McFadden, Wells & Jercinovich 1987). Clay minerals translocated from the upper soil are incorporated within deeper parts of the soil, forming clay-enriched Bt horizons (Gile & Grossman 1968; McAuliffe 1999b; McDonald et al. 1995). McAuliffe (1994) conjectured that well developed Bt horizons in the Sonoran Desert also limit the depth of water percolation following low-intensity, long-duration storms, thereby limiting the growth and abundance of L. tridentata in older soils, much as cemented calcic horizons do. Indeed, soil hydrological modelling efforts suggest that well developed Av or Bt horizons are highly effective in the limiting annual soil water flux water past 50 cm (McDonald et al. 1996). Gile et al. (1998) found that strongly developed Bt horizons resulted in limited vertical root distributions and a proliferation of lateral rooting in L. tridentata. However, the effects of these widespread soil characteristics on the seasonal performance of desert plants are unknown.

Here we present a study investigating the influences of soil horizon on the effectiveness of seasonal precipitation, and the physiological performance of L. tridentata on contrasting soils. We expected the ecophysiology of this evergreen shrub to correlate with soil characteristics that influence recharge of deep soil water. Larrea tridentata generally has deep, nonoverlapping root systems (Brisson & Reynolds 1994; Gile et al. 1998), and can potentially photosynthesize throughout the year using soil water derived from either frontal storms or isolated summer rains (Franco et al. 1994). We formulated two simple hypotheses regarding edaphic controls on the seasonal performance of L. tridentata. First, we expected that subsurface characteristics, specifically clay-enriched argillic Bt horizons, would probably limit the effectiveness of slow-moving frontal storms in recharging soils and providing deep soil moisture to L. tridentata, as postulated by McAuliffe (1994) and McDonald et al. (1996). Secondly, we expected the effectiveness of summer precipitation would probably be determined by soil surficial characteristics such as well developed desert pavement and Av horizons, which could limit rapid infiltration needed to effectively transmit water from such high-intensity, short-duration storms into the soil. This approach differs from past studies that have examined plant performance in broadly different soils and habitats, or in soils from one parent material (Ehleringer & Cooper 1988; Halvorson & Patten 1974; Monson et al. 1992; Smith et al. 1995). We believe that a landscape-level context encompassing detailed understanding of subtle differences of soil conditions, and the consequent variation in soil hydrological responses associated with them, will help clarify mechanisms underlying many desert ecological processes hitherto explained largely on the basis of plant attributes.

Materials and methods

Study sites

Research was conducted from January 16 to October 27 1996 on an alluvial piedmont (bajada) flanking the western slope of the Providence Mountains, East Mojave National Preserve, California (34°35′ N, 115°37′40″ W). Monthly measurements of precipitation, plant predawn and midday xylem water potentials (ψpd and ψmid), and midday photosynthetic gas exchange were made at three sites containing soils that differed in development of desert pavement, Av and Bt horizons (Fig. 1).

Figure 1.

Representation of desert pavement, vesicular A horizon (Av) and clay-rich argillic B horizon (Bt) development of three soils from bajada surfaces differing in age and parent material. Qe, Quaternary eolian; Qf, Quaternary fluvial; numbers represent age periods for eolian (e) or fluvial (f) process-derived surfaces; VX, mixed volcanic, and PM, mixed plutonic parent materials. Ages of the surfaces are from McDonald et al. (1995); the vertical scale has been exaggerated.

The first site was on a late Holocene eolian dune deposit (Qe3, Table 1; McDonald 1994). Qe3 soils are deep (>1 m), sandy, and lack any soil horizon development (Fig. 1). The second site was a fluvial deposit of reworked eolian sand mixed with gravel deposited 8000–17 000 years ago (VX Qf5; hereafter referred to as the Qf5 site). A well developed stone pavement covers this surface; the underlying soil has a moderately well developed Av horizon, but lacks any subsurface Bt horizon (Fig. 1). The Av horizon is enriched with silt and clay, but the underlying soil is a loamy sand approximately 0·5 m deep, beneath which is an older, fluvial deposit of coarse gravel. The third site was on a late Pleistocene alluvial deposit (PM Qf4 surface; hereafter referred to as the Qf4 site), estimated to have been deposited 50 000–150 000 years ago. The deposit consists of coarse, gravelly to cobbly alluvium of mixed plutonic lithologies. Soils on this surface have well developed pavements and Av and Bt horizons (Fig. 1). The Av horizon and associated pavement on these latter surfaces reduce infiltration rates from 1 to 8 cm h−1 compared to rates of 40–58 cm h−1 on surfaces lacking these well developed surficial horizons (McDonald et al. 1996). All sampling was done near the sites where McDonald (1994) had described the soil profiles.

Table 1.  Elevation, slope, aspect and soil horizon characteristics of the three study sites: a late Holocene dune deposit (Qe3); an early Holocene fluvial deposit of reworked eolian sand overlain by a strong Av horizon and pavement (VX Qf5); and a late Pleistocene alluvial deposit (PM Qf4). Soil horizons are illustrated in Fig. 1
 Qe3VX Qf5PM Qf4
Elevation (m)660750915
Clast sizeNoneSmall; surficial, uniform, tightly interlockedVariable; distributed throughout soil profile
AvNoneStrong, 4 cmStrong, 9 cm
BtNoneVery weakStrong, 20 cm thick, 34–56 cm deep

Environmental and plant functional measurements

Rainfall was measured at each site with rain gauges set ≈1·5 m above the ground, with mineral oil added to eliminate evaporative losses. Rainfall measurements were made within 24–48 h after notification of a storm by staff at the UC Riverside's Sweeny Granite Mountain Desert Reserve, located 24 km south of the study area. Air temperature was measured after shading a fine-wire copper/constantan thermocouple in the cuvette of a LiCOR Li 6200 portable photosynthesis system for 15 min. The cuvette was held open, and held 1·0 m above the ground with the chamber fans running.

Volumetric soil moisture (θ) was quantified at 10 and 35 cm depths using time domain reflectometry (TDR). Three sets of paired stainless steel waveguides spaced ≈8 cm apart at each depth were placed in each surface. θ was estimated by relating the signal reflectance measured with a Techtronix 1502B cable tester (Techtronix Industries, Corvallis, OR) under PC control, with software using the equations of Topp & Davis (1985) (tdrwave v. 2·0, Desert Research Institute, Las Vegas, NV). Since the study sites were located on National Park Service land, we were unable to excavate any soil or root distribution samples for this study.

Predawn xylem water potentials (ψpd) of L. tridentata were measured between zero and 04·00 h Pacific Standard Time using a Scholander-type pressure chamber (Soil Moisture Equipment Corp., Santa Barbara, CA). Five plants from each site were sampled, and branches of sampled shoots were marked for subsequent midday water potential (ψmid) and photosynthetic gas-exchange measurements. Field estimations of midday (1100–1430 h) net photosynthesis (Anet) and stomatal conductance to water vapour (gs) were made with a closed-system portable photosynthesis system (LiCOR 6200, Lincoln, NE) using the equations of von Caemmerer & Farquhar (1981). Each shoot was enclosed in a 0·25 l cuvette for a 10 s sampling period, orienting the shoot to receive full sunlight (≈1400–1800 µmol photons m−2 s−1). After gas-exchange measurements, the shoots were harvested and immediately measured for ψmid and leaf area correction for gas-exchange measurements.

Statistical analysis

Two-way and three-way anova (statistix v. 4·0, Analytical Software, St Paul, MN) was used to detect site differences in soil water content, xylem water potential, and photosynthetic gas exchange. A repeated-measures, split-plot design was used, with study site as the whole-plot factor and day of year (DOY) and the DOY × site interaction effect as subplot factors. For θ, additional subplot factors of depth (and additional interaction effects) were included. F-tests were constructed using the site × replicate plant or TDR probe interaction as the whole-plot F-test denominator, and the DOY × site × replicate plant or DOY × site × depth × replicate TDR probe interactions as the appropriate subplot F-test denominators. We were specifically interested in the site × DOY and site × DOY × depth interactions, which would indicate soil-specific differences in plant performance and soil water content through the season. All q data were arcsine transformed to meet anova assumptions (Zar 1974). Linear regressions (statistix v. 4·0) were used to correlate seasonal patterns of predawn and midday water potentials to changes in volumetric soil moisture.


Only four rainfall events occurred during 1996, totalling <5·25 cm of rain (Fig. 2). This made 1996 the driest year in over 12 years in this area (Sweeny Granite Mountain Preserve weather station data). Most of the total rainfall came from a single large frontal storm in early spring (DOY 76) that delivered ≈3·0 cm rainfall to all sites. The three summer storms were smaller convectional thunderstorms (Fig. 2).

Figure 2.

Seasonal rainfall (top) and volumetric soil moisture (q) at 10 cm (middle) and 35 cm (bottom) depths. Each point is the mean of three measurements; bars indicate ±1 SE.

Volumetric soil moisture varied significantly between surfaces (F = 242·7, P < 0·0001); depths (F = 87·9, P < 0·0001); and dates (F = 49·9, P < 0·0001). There were significant two-way surface × depth (F = 56·7, P < 0·0001); surface × DOY (F = 4·1, P < 0·0001); and depth × DOY (F = 10·2, P < 0·0001) interactions. Changes in q with depth were most pronounced at the older Qf4 site, which gave rise to the first interaction (Fig. 2). Increases in Qf4 soil θ following two summer thunderstorms drove the surface × DOY and DOY × depth interactions (Fig. 2), but differences between depths were not strong enough to result in the expected three-way interaction.

Seasonal predawn water potentials (ψpd) of L. tridentata varied significantly in a site-specific manner over time (F = 18·9, P < 0·0001). After three of the four storms (DOY 76, 146 and 209), ψpd in L. tridentata increased markedly in plants growing in soils on the older PM Qf4 site (Fig. 3). ψpd in VX Qf5 plants increased only after the largest rainfall event (DOY 76), then declined through the summer, while ψpd did not vary significantly at the Qe3 dune site, where ψpd was consistently highest (−3·2 to −2·5 MPa; Fig. 3). Xylem ψpd at the VX Qf5 site declined sharply from a season high of −3·0 MPa (DOY 94) to extreme lows approaching −10 MPa. Predawn ψ was consistently lowest in plants in soils at the PM Qf4 site, ranging from seasonal springtime highs of −3·0 MPa to lows of −8·0 MPa in the early summer (Fig. 3). Midday ψ in L. tridentata also showed a significant site × time interaction (F = 2·05, P = 0·023). This interaction was due to: (1) marked differences in ψmid between the surfaces early in the season; (2) increases in ψmid in L. tridentata in Qf4 soils that did not occur in plants at the Qf5 and Qe3 sites following the DOY 146 storm; and (3) no significant differences between L. tridentataψmid at these latter sites during the summer and early autumn (Fig. 3). In contrast to ψpd, with the exception of a marked decrease on DOY 176, there were no clear seasonal trends in ψmid, which was relatively high (−3·0 to −3·9 MPa) and relatively invariant across the surfaces (Fig. 3).

Figure 3.

Seasonal rainfall (top) and predawn (middle) and midday (bottom) water potentials of Larrea tridentata growing on bajada surfaces with soils possessing distinct pavement, Av and Bt horizon development (Fig. 1). Each point is the mean of five measurements; error bars indicate ±1 SE.

Predawn ψ and θ were significantly correlated at 10 cm depths on all surfaces, but ψmid was not (Table 2; Fig. 4). At 35 cm soil depths, correlation between ψpd and θ was highly significant in Qe3 and Qf5 soils, but was not for L. tridentata growing in Qf4 soils (Table 2, Fig. 4). In contrast, the closeness of the relationship between ψmid and θ at 35 cm increased from insignificant in Qe3 and VX Qf5 soils to highly significant levels on the Qf4 surface (Table 2, Fig. 4).

Table 2.  Linear regression coefficients (r2) between Larrea tridentata mean predawn and midday xylem water potentials and volumetric soil water content (θ) at 10 and 35 cm depths over the 1996 potential growing season. Relationships are presented in Fig. 4
SurfaceTimeθ (10 cm)θ (35 cm)
  • *


  • **

    , Significance at P < 0·05 and 0·0001, respectively.

Figure 4.

Relationship between mean predawn and midday xylem water potential (ψ) of Larrea tridentata and volumetric soil moisture (θ) at 10 and 35 cm depths on three different bajada surfaces. Significances of regressions are presented in Table 2, each point is the mean of five ψ and three θ measurements; bars indicate ±1 SE.

For seven of the eight dates sampled for predawn and midday ψ, ψmid was markedly higher than ψpd in plants from most of the sites (Fig. 5). The exceptions were DOY 94, closely following the major DOY 76 rainfall, and after a minor rainfall event (≈0·2 cm on DOY 176) during unusually cool early summer temperatures. This ψ anomaly was significant and soil-specific over time (F = 7·58, P < 0·0001). Early in the growing season, plants from the PM Qf4 site had the strongest ψ anomaly, while plants from the VX Qf5 and Qe3 sites showed little or no difference between predawn and midday levels (Fig. 5). After DOY 180 (early summer), plants at the VX Qf5 site had by far the largest anomaly, increasing from a +1·0 to +6·0 MPa differential between midday and dawn ψ. Over this period, L. tridentata at the PM Qf4 site had nearly identical increasing water potential differences (from 0 to +3·0 MPa). Plants at the Qe3 site remained relatively constant throughout the season, with little difference between ψpd and ψmid.

Figure 5.

Seasonal rainfall and air temperature (top) and the annual course of midday/predawn differential xylem water potential of Larrea tridentata growing on bajada surfaces with soils possessing distinct pavement, Av and Bt horizon development (Fig. 1). Each point is the mean of five measurements; error bars indicate ±1 SE.

Seasonal patterns in net photosynthesis (Anet) and stomatal conductance to water vapour (gs) were soil-specific (F = 4·62 and F = 5·25, P < 0·0001, respectively), in parallel with patterns of predawn water potential (Fig. 6). The highest Anet values attained were in plants growing at the Qe3 and VX Qf5 sites (≈7·5–8·0 µmol m−2 s−1) on DOY 94. After this peak period, Anet and gs of Qe3 plants declined ≈60% and then remained fairly constant, while gas-exchange rates declined sharply in VX Qf5 plants, reaching seasonal lows of around −2 µmol m−2 s−1. Anet in plants at the PM Qf4 site remained close to photosynthetic compensation through most of the study period. After one 0·8 cm summer rain (DOY 213), Anet rose slightly, while gs increased sharply, in plants on the PM Qf4 sites. Anet and gs did not change, and remained low, on the Qe3 and Qf5 sites after this rain event (Fig. 6).

Figure 6.

Seasonal rainfall and air temperature (top) and the annual courses of midday net photosynthesis (Anet) and stomatal conductance to water vapour (gs) of Larrea tridentata growing on bajada surfaces with soils possessing distinct pavement, Av and Bt horizon development (Fig. 1). Each point is the mean of five measurements; error bars indicate ±1 SE.


Distinct soil characteristics among these sites resulted in markedly different seasonal patterns of water use and physiological activity in L. tridentata during a period of intense climatic drought (Figs 3 and 6), especially the seasonal patterns of water use by this desert evergreen shrub. Qf4 plants showed increased ψpd after later-season storms, whereas plants from the Qf5 and Qe3 sites did not (Fig. 3). The minor ψpd fluctuations seen in plants at the Qe3 site suggest that the deep, sandy soil allows substantial infiltration and storage of moisture during winter storms; such deeply stored moisture is more continuously available throughout the year. Such a soil could also renew water reserves with summer storms of sufficient intensity and/or duration (Alizai & Hulbert 1970; Ehleringer 1985; Pavlik 1980), but also exhibit high surface temperatures, resulting in rapid surface evaporation following smaller rains (Lancaster 1994). The seasonal decline in L. tridentata Anet at the Qe3 site could be attributed to increasing high temperature limitation (Mooney, Björkman & Collatz 1978). Indeed, on DOY 155, an unusually cool day, a marked rise in Anet occurred in L. tridentata at this site, suggesting a release from temperature limitation (Fig. 6). Also, seasonally increased vapour pressure deficit (VPD) would limit Anet, as the stomatal response of L. tridentata to VPD is greatest in well watered plants (Sharifi & Rundel 1993).

In contrast to the Qe3 site, the well formed surface pavement and Av horizon of the VX Qf5 site may restrict effective recharge of subsoil layers, even by large precipitation events (Abraham & Parsons 1991; Ehleringer 1994; McAuliffe & McDonald 1995). The single winter storm (DOY 76) was apparently sufficient for deep soil recharge, as indicated by seasonal highs in ψpd at this site shortly after the storm event. The higher ψpd and Anet (Figs 3 and 6) in L. tridentata in these soils, compared to plant performance at the Qf4 site, also suggest that L. tridentata may be free from competition by drought-deciduous shrubs such as Ambrosia dumosa, which are common on the Qf4 site (McAuliffe & McDonald 1995). In Mojave Desert bajada areas where overland flow was excluded, a lack of drought-deciduous shrubs resulted in improved water status in L. tridentata compared to those in areas receiving overland flow, but with high densities of competing shrub species (Schlesinger & Jones 1984; Schlesinger et al. 1989). The results of these and the present study differ from Fonteyn & Mahall's (1981) findings that experimental removal of A. dumosa did not alter L. tridentata water status, even long after removal of these competitors. Possibly, variation in soil condition and soil hydrology could conceivably affect and even alter experimental outcomes in these aridland systems.

Larrea tridentata from the Qf4 site showed increased ψpd after late spring and summer rains (Fig. 3). This agrees with the notion that only major summer storms affect plant processes in the Mojave Desert (Beatley 1974). Our findings also suggest that, depending on the soil medium, L. tridentata can utilize summer rainfall during extremely dry years, an ability not shared by many long-lived perennial desert shrubs (Ehleringer 1994; Ehleringer et al. 1991). Also, surface and subsurface clasts at the Qf4 site were more irregular in size than at the Qf5 site, and distributed throughout the soil profile (Table 1; see also McAuliffe & McDonald 1995). Larger rocks could possibly act as infiltration foci or as barriers to surface evaporation, much as nurse rocks and nurse plants do for desert seedlings (Abrahams & Parsons 1991; McAuliffe 1984; Nobel 1989; Nobel, Miller & Graham 1992), or as stems do in established plants (Lyford & Qashu 1969).

Improved ψpd following summer storms did increase stomatal conductances, but did not result in higher net assimilation rates in L. tridentata (Fig. 6). The low positive Anet indicates that L. tridentata may improve water status enough to maintain low levels of productivity during extended hot, dry periods, but that chronic water stress in these soils leads to reduced photosynthetic capacity (Mooney et al. 1978; Oechel et al. 1972). Thus any improved plant assimilation may be more transient than plant water status after summer rains on these surfaces, as has been found with L. tridentata from the Chihuahuan Desert (Franco et al. 1994), where summer rainfall is a much more important component of annual rainfall than in the Mojave Desert.

The striking anomalous patterns in xylem water potential observed in this study (Fig. 5) have been reported before for L. tridentata in the Chihuahuan and Sonoran Deserts (Halvorson & Patten 1974; Syvertsen, Cunningham & Feather 1975), and in halophytic cold-desert shrubs (Donovan et al. 1999). Syvertsen et al. (1975) suggested dynamic soil temperature gradients associated with soil water phase changes might be responsible for this pattern. Manipulation of soil temperatures by these investigators did not alter the plant water potential anomaly but, given the extensive rooting volumes of L. tridentata (Brisson & Reynolds 1994), it may be that insufficient soil volume was affected by their treatments. The data of Syvertsen, Cunningham & Feather (1975) were also restricted in that they used a single soil type, and did not account for soil horizon development. Donovan et al. (1999) showed that increased soil salinity induced an 0·9–1·8 MPa positive difference between predawn and midday ψ, and suggested that nighttime canopy water loss, with or without stomatal opening, is an important contribution to disequilibrium between plant and soil predawn water potentials (Donovan et al. 1999). We (E. P. Hamerlynck and S. D. Smith, unpublished results) and Syvertsen et al. (1975) did not observe nighttime stomatal opening in L. tridentata when ψ anomaly occurred. However, even with L. tridentata's very high cuticular resistance (Meinzer et al. 1990), it may be that the evergreen habit could result in sustained nighttime water loss that might contribute to a water potential anomaly under extreme drought conditions.

Hydraulic lift may also be a potential mechanism for this pattern. Hydraulic lift occurs during the night when deep plant roots transport soil water to upper soil layers, where numerous shallower roots subsequently utilize the translocated water (Caldwell & Richards 1989; Dawson 1993; Richards & Caldwell 1987). It has been reported that L. tridentata is capable of hydraulic lift (Caldwell, Dawson & Richards 1998; Yoder & Nowak 1999). Therefore the soil-specific water potential anomalies seen here may be a reflection of the soil water potential drop being bridged by L. tridentata roots. If this is indeed the case, then the increasing correlation apparent between the sites between ψmid and θ at 35 cm (Table 2) may reflect the depth to which transported water is lifted. Hydraulic lift may therefore become increasingly restricted to shallower depths by subsurface horizon development, as indicated by the strong relationship between ψmid in L. tridentata and θ at 35 cm at the Qf4 site (Fig. 4). Concurrently, the lack of correlation between ψmid and θ at the Qe3 and Qf5 sites suggests that water was not lifted to depths monitored in this study. The varying degrees of correlation with xylem ψpd and θ at 10 and 35 cm in Qf4 soils (Table 2) suggest that most of the rooting mass at this site are restricted to shallower depths by Bt horizon development (Fig. 1), which has been found to limit rooting depths in L. tridentata (Gile et al. 1998). The similar correlations between ψpd and θ at different depths at the Qe3 and Qf5 sites suggest these soils allow deeper rooting. In addition, the generally strong correlation between ψpd and θ across both depths suggests a near complete withdrawal of water from these depths at these sites, a factor needed to maintain the soil water-potential gradient driving lift (Williams, Caldwell & Richards 1993).

However, hydraulic lift does not explain why ψpd does not equilibrate with improved soil water status, but stays low throughout the night (Syvertsen et al. 1975), or why anomalies are not more common in L. tridentata at the Qe3 site where deep soil moisture may be more available. It is possible that some mechanism separates above-ground from below-ground components, as happens in the transition zone between roots and stems of some succulents, where a zone of reversible xylem embolism severs stems from extremely dry roots during protracted drought (Ewers, North & Nobel 1992). However, the time constants of reversing such widespread xylem cavitation are far slower than the rapid predawn-to-midday rise in water potential seen here and by Syvertsen et al. (1975).

Regardless of potential mechanism, water potential anomalies did not result in improved physiological performance. Anomalies occurred mostly in plants with very low ψpd– some of the lowest predawn water potentials recorded for L. tridentata (Monson & Smith 1982; Oechel et al. 1972; Syvertsen et al. 1975) – at a time when these plants were rarely above photosynthetic compensation. Given that these water potentials reflect minute amounts of water, it is unlikely that such a response is an adaptation; it is probably a reflection of soil hydrology. This is supported by the fact that following the DOY 76 storm, ψ anomaly was most apparent in the clay-rich soils at the Qf4 site (Fig. 5). During the summer, the general lack of soil recharge resulted in ψ anomaly being most pronounced at the VX Qf5 site. Our data partially support the findings of both Syvertsen et al. (1975) and Donovan et al. (1999), as the ψ anomaly increased with seasonal temperatures when both soil temperature gradients and nighttime evaporative losses are highest (Nobel 1989), then markedly reduced under unusually cool summer conditions and with seasonal cooling (Fig. 5). But given the strong soil-specific nature noted in our study, it is likely that surface and subsurface soil physical characteristics play the strongest role in determining the hydrology underlying this commonly observed water potential anomaly in desert shrubs.

In conclusion, we found that seasonal plant–water relations and gas exchange in L. tridentata were strongly mediated by soil horizons that affect surface infiltration and subsurface moisture flux. During wetter parts of the growing season, soils with subsurface layers such as clay-rich argillic horizons probably limit the depth of the wetting front due to the high moisture holding capacity of clays (Alizai & Hulbert 1970). This is in agreement with earlier findings relating seasonal growth and water relations in L. tridentata, and site water balance and community structure to effects of cemented calcic horizons (Burk & Dick-Peddie 1973; Cunningham & Burke 1973; Shreve & Mallery 1933; Smith et al. 1995). Our findings are novel in that they suggest that subtle, often unapparent combinations of surface and subsurface horizon development can play an important role in determining seasonal water relations and utilization. It is important to note that during the summer dry season, control of L. tridentata plant water availability switches to soil surface characteristics such as desert pavement and Av horizons (Figs 2 and 3). This might explain some discrepancies in the literature about seasonal water use patterns by long-lived evergreen species. Indeed, utilization of such ‘non-seasonal’ precipitation has been found to be highly important in long-term desert vegetation dynamics (Reynolds et al. 1999). In addition, we have shown that surface and subsurface characteristics may determine the seasonal occurrence and degree of water potential anomaly (Fig. 5), which may be linked to the redistribution of water resources, a potentially important process in water-limited systems (Caldwell et al. 1998; Donovan et al. 1999). Overall, our findings highlight the importance of detailed knowledge of soils in addition to climate for understanding desert plant ecological function (McAuliffe 1994, 1999a). It may be that complex soil mosaics of desert environments will meditate the overall impact of global change (Kemp et al. 1997; McAuliffe 1999a), either atmospheric trace gas or climatic, in the future.


The authors wish to thank Jim André and Claudia Luke, co-managers, and David Lee, steward, of the University of California Riverside's Sweeney Granite Mountain Desert Preserve for their unfailing assistance throughout the course of this study. Eric McDonald freely shared his time and knowledge of the geomorphology and soils in this study. We also wish to thank Travis Huxman, Lynn Fenstermaker and Dean Jordan of the University of Nevada, Las Vegas, for help in the field. This research was supported by the NSF EPSCoR program and matching funds from the State of Nevada to S.D.S., and the Desert Botanical Garden to J.R.M.