First‐year Acacia seedlings are anisohydric “water‐spenders” but differ in their rates of water use

Abstract

Premise

First‐year seedlings (FYS) of tree species may be a critical demographic bottleneck in semi‐arid, seasonally dry ecosystems such as savannas. Given the highly variable water availability and potentially strong FYS–grass competition for water, FYS water‐use strategies may play a crucial role in FYS establishment in savannas and, ultimately, in tree–grass competition and coexistence.

Methods

We examined drought responses in FYS of two tree species that are dominant on opposite ends of an aridity gradient in Serengeti, Acacia (=Vachellia) tortilis and A. robusta. In a glasshouse experiment, gas exchange and whole‐plant hydraulic conductance (K _plant) were measured as soil water potential (Ψ _soil) declined. Trajectory of the Ψ _leaf/Ψ _soil relationship during drought elucidated the degree of iso/anisohydry.

Results

Both species were strongly anisohydric “water‐spenders,” allowing rapid wet‐season C gain after pulses of moisture availability. Despite being equally vulnerable to declines in K _plant under severe drought, they differed in their rates of water use. Acacia tortilis, which occurs in the more arid regions, initially had greater K _max, transpiration (E), and photosynthesis (A _net) than A. robusta.

Conclusions

This work demonstrates an important mechanism of FYS establishment in savannas: Rather than investing in drought tolerance, savanna FYS maximize gas exchange during wet periods at the expense of desiccation during dry seasons. FYS establishment appears dependent on high C uptake during the pulses of water availability that characterize habitats dominated by these species. This study increases our understanding of species‐scale plant ecophysiology and ecosystem‐scale patterns of tree–grass coexistence.

Savannas are characterized by a continuous grass canopy with a spatially heterogeneous tree cover. Understanding the factors that generate and maintain this tree–grass coexistence has long been of interest (e.g., reviews by Scholes and Archer, ¹⁹⁹⁷; Sankaran et al., ²⁰⁰⁵). Precipitation and its effects on soil moisture and grass productivity are considered primary drivers enabling this coexistence, although studies have also found support for the role of herbivory (Midgley et al., 2010), fire (Hoffmann et al., 2012; Holdo et al., ²⁰¹⁴), and soil characteristics (Rietkerk et al., 1997; Sankaran et al., ²⁰⁰⁸; Holdo et al., ²⁰²⁰). Numerous studies have also suggested that first‐year seedlings (hereafter, FYS) are a demographic bottleneck for savanna tree species (Chidumayo, 2013; Anderson et al., ²⁰¹⁵; Morrison et al., ²⁰¹⁹). Thus, survival of FYS may play a critical role in maintaining the structure of savannas, yet the ecophysiological mechanisms that shape the mortality of FYS, and ultimately, tree–grass coexistence, are poorly understood.

A newly emerging savanna tree seedling may experience strong limitations to growth and survival beneath the grass overstory, possibly due to both sunlight and soil water limitations. In seasonally dry savannas, highly variable water availability across seasons may strongly influence FYS–grass competition for soil water and accompanying nutrients (Chesson et al., 2004; Cramer et al., ²⁰⁰⁷). For example, in eastern Africa, productivity of perennial C₄ grasses is highest during the wet season (circa November–May; Anderson et al., ²⁰⁰⁸), creating potentially strong sunlight limitations for FYS beneath the canopy during these periods. FYS not only must escape the grass canopy to avoid light limitations during the wet season, they must simultaneously compete with grasses for soil water from the same soil horizon (February et al., 2013; Ketter and Holdo, ²⁰¹⁸). Savanna grasses are substantial water “spenders” that rapidly deplete soil moisture after a rainfall event, particularly within the top layers of soil due to their extensive, shallow root system (Williams et al., 1998). Experimentally removing grasses has been shown to increase Acacia seedling establishment (Cramer et al., 2007; Ward and Esler, ²⁰¹¹; Morrison et al., ²⁰¹⁹). During the dry season, these drought‐deciduous FYS may drop their leaves, and soil moisture remains low until the subsequent wet season (Xu et al., 2015). The result is a relatively short window of adequate sunlight and soil water after pulses of precipitation during which FYS must maximize carbon gain to ensure survival.

In addition to seasonality and considerable stochasticity of water availability across seasons, precipitation also varies across space in many savanna ecosystems. For example, in the Serengeti ecosystem, mean annual precipitation (MAP) ranges from ~500 mm in the southeastern Serengeti to >1200 mm in the northwestern region (McNaughton, 1985; Anderson et al., ²⁰⁰⁸). This substantially greater precipitation in the northwest supports higher primary productivity of grasses, creating strong tree–grass competition, particularly during the wet season, and therefore greater risk of FYS mortality (Morrison et al., 2019). In contrast, rainfall is more uniformly distributed across seasons in the northwest, with a greater proportion of the MAP falling during the dry season (McNaughton, 1985; Anderson et al., ²⁰⁰⁸). This interaction between the annual amount, seasonality, and stochasticity of precipitation creates pulses of resource availability and grass productivity, generating strong FYS–grass competition for soil water and sunlight (Chesson et al., 2004; Dwyer et al., ²⁰¹⁰).

Given the highly variable water availability and strong competition for soil water in the Serengeti, the role of water‐use strategies and the vulnerability of whole‐plant hydraulic pathways during drought could be a key mechanism affecting FYS survival (e.g., Johnson et al., ²⁰¹¹). This whole‐plant hydraulic pathway (K _plant) reflects xylem function (transport efficiency and vulnerability to embolism), stomatal sensitivity to declining water status (iso/anisohydry), resilience of non‐xylem pathways in leaves (i.e., “outside‐xylem” pathways of Scoffoni et al., ²⁰¹⁷), and even the hydraulic conductance of soil to supply water to roots. For example, their limited rooting volume subjects FYS to lower and more variable soil water potentials (Ψ _soil) compared to larger, deeply rooted plants (McDowell et al., 2008). This exposure to extreme Ψ _soil, combined with their limited capacity for carbohydrate and water storage, leaves FYS highly vulnerable to tissue desiccation and xylem embolism during drought (Grossnickle, 2012). Thus, maintaining functional hydraulic pathways is critical for maximizing gas exchange and for avoiding lethal tissue desiccation.

Generally, stomatal closure is a key mechanism by which plants control their water status and avoid the negative effects of drought. A common approach to describing these water‐use strategies is the spectrum of stomatal sensitivity to declining water potentials—the so‐called iso/anisohydry continuum (Tardieu and Simonneau, 1998; McDowell et al., ²⁰⁰⁸; Martínez‐Vilalta et al., ²⁰¹⁴; Hochberg et al., ²⁰¹⁸). On one end of this spectrum, isohydric species maintain water status (i.e., Ψ _leaf) by tightly regulating stomatal conductance (g _s) as xylem tension rises. Stomata close before water status drops below a critical threshold of Ψ _leaf that would lead to cavitation, yielding a higher hydraulic safety margin (McDowell et al., 2008; Martínez‐Vilalta and Garcia‐Forner, ²⁰¹⁷; Fu and Meinzer, ²⁰¹⁸). But, under strong competition for water with grasses during pulses of water availability, FYS might benefit from sustaining gas exchange despite high xylem tensions. This anisohydric strategy of a “water‐spender” maximizes C gain during brief periods of water availability, but at the expense of potential hydraulic failure as soil moisture declines. Even though anisohydric species often have lower hydraulic vulnerability, they also have greater risk of desiccation because they operate with lower hydraulic safety margins (Martínez‐Vilalta et al., 2014; Fu and Meinzer, ²⁰¹⁸).

Theory states that, in the absence of rooting zone partitioning (Kambatuku et al., 2013; Ward et al., ²⁰¹³), as reported between grasses and FYS in some African savannas (Ketter and Holdo, 2018), plants should maximize their utilization of precipitation pulses by maintaining high hydraulic conductance during droughts (Schwinning and Ehleringer, 2001). The ability to continue water transport and gas exchange would be particularly advantageous during the dry season due to a substantially greater incidence of sunlight after grass die‐back, but critically low plant water potentials in this scenario would lead to catastrophic desiccation of aboveground tissues and/or re‐sprouting when precipitation returns. This effect is likely to be compounded in high disturbance environments, such as those with frequent fire, in which dry season tissues are likely to be lost to fire until the following wet season.

Understanding these possible relationships for FYS would contribute to understanding the mechanisms governing FYS–grass coexistence in savannas and possibly FYS establishment and mortality elsewhere (Brodersen et al., 2019). Here, we investigated stomatal behavior and hydraulic traits in FYS of two coexisting tree species that are dominant on opposite ends of an aridity gradient in the Serengeti: Acacia (=Vachellia) tortilis, which is abundant in more arid regions, and A. robusta, which is abundant in higher rainfall regions. Stomatal sensitivity to declining soil moisture and hydraulic vulnerability were examined by measuring gas exchange and whole‐plant hydraulic conductance (K _plant) throughout a drought experiment. We hypothesized that both species would demonstrate relatively anisohydric behavior, but that A. tortilis would sustain K _plant into more severe drought (=lower hydraulic vulnerability) than A. robusta would.

MATERIALS AND METHODS

We investigated stomatal responses to drought and hydraulic vulnerability by repeatedly measuring gas exchange and water potentials throughout a drought experiment in a greenhouse. Subsequently, we conducted an accompanying study to validate the accuracy of in situ soil water potential (Ψ _soil) measurements. We also measured Ψ _soil in the field and analyzed the distribution of each species at seven long‐term sites across the Serengeti.

Plant material and growth conditions

The experiment was conducted in a glasshouse facing south‐southwest that was maintained at ~30°C. Seeds of Acacia (=Vachellia, family Fabaceae) tortilis (Forssk.) and A. robusta (Burch.) were collected across Serengeti National Park in Tanzania and transported to Wake Forest University. Seeds were collected opportunistically from multiple individuals at eight sites that are part of a broader study on tree dynamics in the Serengeti (Holdo et al., 2020). The seeds were scarified using sandpaper to remove a small portion of the seed coat and were then sown in cylindrical pots (40 cm tall, 10 cm diameter). To replicate the water‐holding capacity of clay‐rich soils in Serengeti, we used a growing medium with 12/5/3 (v/v/v) of Metro‐Mix 360 Professional Growing Mix (Sun Gro Horticulture, Agawam, MA, USA), Turface Mound & Plate All‐Purpose Clay (PROFILE Products, Buffalo Grove, IL, USA), and Quikrete Premium Play Sand (Quikrete, Atlanta, GA, USA), respectively. Seedlings (10 individuals of each species) were kept well watered for the first 74 days after sowing, then water was withheld for the remainder of the experiment. Measurements began on day 71 and continued until all leaves on a particular plant had dropped or until leaf water potential (Ψ _leaf) was less than –7 MPa (ending on day 168).

Gas exchange and hydraulic conductance

We measured whole‐plant hydraulic conductance (K _plant) using an evaporative flux method. K _plant was calculated using the formula of Sack et al. (2002):

$$K_{\text{plant}}=\frac{E}{{\Psi }_{\text{soil}}-{\Psi }_{\text{leaf}}}$$

In this application of Ohm's Law analogy for water transport in plants (Nobel, 2009), hydraulic conductance (K _plant) is the flux of water on a leaf‐area basis, given the water potential gradient across the plant. Thus, each measurement of K _plant required simultaneous measurements of steady‐state transpiration (E), leaf water potential (Ψ _leaf), and soil water potential (Ψ _soil).

Soil water potential was measured with Teros 21 sensors (METER Group, Pullman, WA, USA) buried in every pot at a depth of 23 cm (Appendix S1). Transpiration and photosynthesis were measured with an LI‐6400 fitted with an LED light source (6400‐02B, Li‐Cor, Lincoln, NE, USA). Chamber conditions were maintained at CO₂ concentrations of 405 ± 5 ppm and 2000 μmol·m⁻²·s⁻¹ PPFD. Gas exchange was measured on one to four pinnae from a fully expanded bipinnately compound leaf. Leaf tissue did not entirely fill the 6‐cm² chamber, so the leaf area was determined by scanning each sample, then removing artifacts in Adobe (San Jose, CA, USA) Photoshop 19.1.7. Leaf area was then determined using the LeafArea package in R (Katabuchi, 2015).

Leaf temperature during gas exchange measurements was calculated using an energy balance approach. First, chamber temperature was measured using a Type E thermocouple that was wired underneath the chamber gasket. We used boundary layer conductance values from the BLC Lookup Table (Li‐Cor) after performing a sensitivity analysis, which showed that the table values were appropriate for our samples. Immediately after gas exchange measurements, Ψ _leaf of the same leaf as gas exchange measurements was measured (Appendix S1) using a Scholander‐type pressure chamber fitted with an Almond compression gasket for short petioles (PMS Instrument Co., Albany, OR, USA).

Measurements were taken on 17 days between day 71 and 168 of the experiment. After the initial measurements when plants were well watered, plants were not measured if their soil water potential had not decreased adequately. On average, each A. tortilis was measured eight times, and A. robusta was measured 11 times.

Water potential curves

The trajectory of the relationship between Ψ _leaf and Ψ _soil was used to assess stomatal sensitivity to declining water status (iso/anisohydry). As soil moisture declines, the relationship between Ψ _leaf and Ψ _soil is often nonlinear and can be described as three phases (Meinzer et al., 2016; Fu and Meinzer, ²⁰¹⁸). Following the procedures of Knipfer et al. (2020), we used a piecewise linear regression (PLR) approach to estimate the boundaries between Phases I and II and Phases II and III. First, we visually estimated the breakpoint between phases and used these as starting values. We then iteratively searched values of Ψ _soil that minimized residual mean squared error, which provided the bounds for ordinary linear regression (Crawley, 2007).

Hydraulic vulnerability curves

Hydraulic conductance data throughout the drought experiment were used to calculate the percentage loss of conductance (PLC) using the formula:

$$\text{PLC}=100\times \frac{(K_{\max }-K_{\text{plant}})}{K_{\max }},$$

where K _plant values are a series of hydraulic conductance measurements throughout the experiment and K _max is each individual plant's maximum K _plant under ideal conditions (which we defined as Ψ _leaf > –0.2 MPa).

Vulnerability curves were constructed by plotting PLC as a function of Ψ _leaf, then calculating the water potential required to induce a 50 and 88% loss in hydraulic conductance (P ₅₀ and P ₈₈, respectively). We performed this analysis with the R package FitPLC (Duursma and Choat, 2017). First, starting values for P _x (the Ψ _leaf at x percent loss of hydraulic conductance) and S _x (the slope of the curve) were estimated from a linearized sigmoidal‐exponential model using linear regression. These starting values were then used to fit a Weibull model with the formula $K/K_{\max }={\left(1-x/100\right)}^p$, where $p={(P/P_x)}^{P_x{S}_x/V}$, $V=(x-100)\log (1-\mathrm{x}/100),$ and P is a range of Ψ _leaf values.

The model was fit for each species, using individuals as the random effect. Conductance declined rapidly within a narrow range of Ψ _leaf near the inflection point at P ₅₀. Thus, the Weibull function was weighted more heavily near values of P ₅₀ (Nolf et al., 2015) using a power function, Weight = |50 – PLC| ^1.2, which improved the quality of the Weibull fit (and therefore, statistical power), but did not significantly alter the P _x estimates themselves. Confidence intervals of the estimated parameters were calculated using bootstrapping (Duursma and Choat, 2017).

Leaf area estimation

A common approach to assess whole‐plant hydraulics is to measure the flux of water through the entire plant (e.g., sap flow or mass balance), then standardize K by the sapwood area or the whole‐plant leaf area (Yang and Tyree, 1994; Tsuda and Tyree, ¹⁹⁹⁷; Venturas et al., ²⁰¹⁷). Because gas exchange and Ψ _leaf were measured on individual leaves rather than the entire plant, we expressed E (and therefore, K) in terms of leaf area inside the Li‐Cor chamber rather than whole‐plant leaf area. Regardless of the units of K, PLC was expressed as a percentage of each individual's K _max when water potential was still high.

However, comparison of K _plant can potentially be confounded by differences in size, total leaf area, the degree of drought deciduousness, or the ratio of total leaf area to sapwood area. Previous work from our group has shown that these species have equivalent leaf area. In a prior experiment with these same two species (Rugemalila et al., 2020), we quantified the total projected leaf area (PLA), which is the total one‐sided leaf area of the entire plant. Experimental conditions during the previous experiment (e.g., size of pots, glasshouse temperature, sun exposure) were nearly identical to the present study, and seedlings were similar in age (measured at day 51 in the previous experiment; onset of drought in the present experiment was on day 71). Briefly, we used a subsample of leaves to estimate the average projected area per leaf, which we then multiplied by the total number of leaves to give the total leaf area of the plant. In full sunlight, the mean PLA ± SE was 48.51 ± 5.50 for A. robusta and 44.87 ± 10.74 cm² for A. tortilis (n = 9 and 10, respectively). There was no significant difference in PLA between the species (t = 1.1373, df = 11.4463, P = 0.2787), indicating that differences in evaporative surface area would not confound our interpretation of K _plant in the present study.

Validation of Ψ _soil measurements

Our calculation of K _plant requires accurate measurement of Ψ _soil, which is difficult to measure in situ, particularly at extremely low Ψ _soil because K _soil can decline by orders of magnitude during soil dry‐down. Thus, after the initial experiment above, we performed an additional experiment to validate the accuracy of our Ψ _soil sensors by comparing Ψ _soil measurements with predawn Ψ _plant. Theory suggests that in the absence of capacitance and nighttime transpiration, plant water status should equilibrate with the highest Ψ _soil that is available to the plant (Donovan et al., 2001). Ψ _leaf and Ψ _stem of bagged, nontranspiring plants have been used in lieu of direct measurements of Ψ at the soil–root interface (Tsuda and Tyree, 1997). To verify that our Ψ _soil sensors were equivalent to predawn Ψ _leaf (i.e., that our Ψ _soil sensors reflect Ψ that is actually available to the plant), we set up 11 pots with the same Teros 21 sensors, using the same soil and growing conditions as before. All pots had a sensor at the same 23‐cm depth, and six of the 11 pots had additional sensors at 10‐ and 35‐cm depths (data not shown).

Seedlings in this validation experiment were kept well watered for 126 days before water was withheld. At ~5‐day intervals after water was withheld (ending on day 171), we measured predawn Ψ _leaf for comparison to Ψ _soil sensor readings. To avoid any potential effects of nighttime transpiration, we covered the entire seedling (plus a moist paper towel) in a plastic bag on the day before predawn Ψ _leaf measurements. Due to a spider mite outbreak in the glasshouse, data were included from only seven plants (107 measurements total).

Species distributions and field Ψ _soil

To understand the relationship between the spatial distributions of Acacia species and temporal patterns of soil moisture availability, we examined field measurements of Ψ _soil and abundance of each species at seven long‐term sites across the Serengeti rainfall gradient (e.g., Anderson et al., ²⁰¹⁵). We expected that Ψ _soil would decline more quickly in sites where A. tortilis is more abundant, which would suggest that rapid utilization of pulsed resource availability is particularly adaptive for A. tortilis.

At each study site, we selected ~10 representative rainfall events during 2015–2019 that saturated the soil and were not immediately followed by another rainfall event, allowing Ψ _soil to decline monotonically. Ψ _soil was measured with Teros 21 sensors buried at a depth of 10 cm in soils that were representative of the site. After a typical rainfall event, Ψ _soil was >–10 kPa then declined over several days or until the next rainfall event. We defined “dry‐down days” as the number of days after a rainfall event for Ψ _soil to decline to –204 kPa, which is the mean Ψ _soil where Ψ _leaf begins to decline (Θ₁ from our WP curves, Figure 3).

Figure 3.

Water potential curves showing the trajectory of leaf (Ψ _leaf) and soil water potentials (Ψ _soil) during the drought for (A) Acacia robusta and (B) A. tortilis. Dashed vertical lines represent the transition points (Θ₁ and Θ₂) between Phases I‐II and II‐III. Thick dashed line is the 1:1 relationship between Ψ _leaf and Ψ _soil. Thick solid lines are the linear regression in Phase II. Initially in Phase I, Ψ _leaf fluctuated largely independently of Ψ _soil. After Θ₁, lack of stomatal closure affected the Ψ _leaf/Ψ _soil relationship of Phase II, with Ψ _leaf declining faster than the declines in Ψ _soil indicating extreme anisohydry. Beyond Θ₂, Ψ _leaf and Ψ _soil were effectively uncoupled in Phase III.

At the same seven sites where Ψ _soil was measured, abundance of adult trees >5 cm DBH and >2 m in height of both species was recorded in 1000 m² plots annually since 2010 (e.g., Anderson et al., ²⁰¹⁵; Morrison et al., ²⁰¹⁶; Rugemalila et al., ²⁰¹⁶). We used species abundance data from our 2018 surveys and plotted the density of each species (number per 1000 m²) as a function of the dry‐down days (Appendix S2).

Statistical analyses

All analyses were performed in R (R Core Team, ²⁰¹⁸). E and K _max were log‐transformed for statistical analyses. A two‐sample t‐test was used to test for differences in E, Ψ _leaf, and K _max between species, and data are expressed as means ± SEs. For vulnerability curves, the standard errors of P _x cannot be used for traditional statistical inference because these confidence intervals are often asymmetric. Thus, 95% confidence intervals of P _x and S _x were compared between species and were considered not significantly different if the confidence intervals overlapped (Duursma and Choat, 2017).

RESULTS

Species distributions and field Ψ _soil

The site‐level means of dry‐down days ranged from 7 to 13 days across the seven study sites. As we predicted, the density of A. tortilis was positively and A. robusta negatively related to the number of dry‐down days across the Serengeti (Appendix S2).

Ψ _soil validation

Our validation study showed that the difference between Ψ _soil and predawn Ψ _leaf was negligible, particularly when Ψ _soil and Ψ _{leaf_predawn} were ≥–0.2 MPa (Ψ _soil –Ψ _{leaf_predawn} = 0.088 ± 0.008 MPa, mean ± SE), meaning that our Ψ _soil sensor readings were sufficiently accurate to use in calculations of K _plant. Across the entire range of Ψ _soil (0 to –2.11 MPa), the slope was 0.918, and R ² = 0.6322 (Appendix S3).

Pre‐drought gas exchange

Soil water potential remained >–0.015 MPa during the first 74 days after sowing (Figure 1A). At the beginning of the drought experiment, when Ψ _soil was still high (>–0.2 MPa), there were significant differences in water‐use strategies between the species. Stomatal conductance (g _s) was initially ~50% higher in A. tortilis (Figure 1C), and transpiration (E) was also greater in A. tortilis (6.922 ± 0.632) than A. robusta (4.364 ± 0.231 mmol m⁻² s⁻¹; F _1,103 = 16.4, P = 9.835 × 10⁻⁵; Figure 2A). Despite having ~59% greater water loss through transpiration, A. tortilis also had higher Ψ _leaf than A. robusta (–1.18 ± 0.060 and –1.47 ± 0.044 MPa, respectively; F _1,103 = 16.3, P = 1.053 × 10⁻⁴; Figure 2B). During this initial period of the drought experiment, K _max (the maximum whole‐plant hydraulic conductance under ideal conditions) was 80% higher for A. tortilis than A. robusta (9.716 ± 1.598 and 5.403 ± 0.489 mmol m⁻² s⁻¹ MPa⁻¹, respectively; F _1,18 = 10.1, P = 5.219 × 10⁻³; Figure 2C). Similar to E (Figure 2A), A _net was ~77% greater for A. tortilis than A. robusta (20.10 ± 1.26 and 11.39 ± 0.55 μmol m⁻² s⁻¹, respectively; F _1,103 = 43.9, P = 1.618 × 10⁻⁹; Figure 2D).

Figure 1.

Declines in water status and gas exchange through the course of the drought experiment. Plants were well watered for 74 days after sowing, then water was withheld (thick dashed vertical line at day 74) until all leaves had dropped or until Ψ _leaf was < –7 MPa. (A) Soil water potential (Ψ _soil) remained high until water was withheld, then declined more rapidly for Acacia tortilis than A. robusta. Dotted vertical lines represent the 17 different days that (B) leaf water potential (Ψ _leaf) and gas exchange were measured. (C) Stomatal conductance (g _s) and (D) photosynthesis (A _net) was initially ~50% higher in A. tortilis, but declined more rapidly. Points in A are means of 10 Ψ _soil sensors per species. Points and error bars in B–D are means and standard errors of 3 to 10 plants per species, as some individuals were not measured on a particular day if its Ψ _soil had not decreased adequately since the previous measurement.

Figure 2.

Differences between Acacia robusta and A. tortilis in gas exchange and hydraulic traits at the onset of the drought experiment when Ψ _soil > –0.2 MPa. (A) A. tortilis had higher transpiration (E) yet remained at higher leaf water potential (Ψ _leaf) than A. robusta (B). Under these well‐watered conditions, A. tortilis had greater (C) maximum hydraulic conductance (K _max) and (D) net photosynthesis (A _net). Points are the means and standard errors of 10 plants per species for C. Points in A, B, D are means and standard errors of measurements where Ψ _soil > –0.2 MPa (A. tortilis: n = 40 and A. robusta: n = 65).

Drought responses

Due to its greater transpiration, Ψ _soil declined more rapidly for A. tortilis than for A. robusta (Figure 1A) during the drought experiment. Similarly, Ψ _leaf also declined more rapidly for A. tortilis (Figure 1B), reaching –5 MPa ~20 days earlier than A. robusta. After water was withheld, g _s of A. tortilis declined 74% within 1 week, from 0.4061 ± 0.1006 on day 75 to 0.1028 ± 0.0163 mol m⁻² s⁻¹ on day 82 (Figure 1C), whereas g _s and A _net in A. robusta declined more gradually throughout the drought (Figure 1C,D).

Water potential curves

When Ψ _soil was still high at the beginning of the drought (Phase I, Figure 3), Ψ _leaf varied largely independently of Ψ _soil, likely due to day‐to‐day differences in vapor pressure deficit and/or irradiance (Hochberg et al., 2018; Kannenberg et al., ²⁰²¹). Our PLR model identified the transition from Phase I to Phase II (Θ₁) at –0.185 for A. robusta and –0.223 MPa for A. tortilis (Figure 3). While soil drying progressed during Phase II, Ψ _leaf declined more rapidly than Ψ _soil. The result is that the gradient of Ψ _soil – Ψ _leaf increased as Ψ _soil decreased, indicating a lack of stomatal regulation at these moderate levels of drought. During Phase II, the Ψ _leaf/Ψ _soil relationship was y = 1.834x– 1.432 for A. robusta and y = 1.492x – 1.144 for A. tortilis. Beyond Θ₂ (at –2.004 and –2.043 for A. robusta and A. tortilis, respectively), the relationship between Ψ _leaf and Ψ _soil was weak in Phase III, suggesting that desiccation and/or xylem embolism were the primary limitation to K _plant and gas exchange.

Vulnerability curves

Despite major differences between the species in gas exchange and K _max when soil moisture was relatively high, we found no differences in their hydraulic vulnerability during drought (Figure 4). Acacia tortilis had a slightly lower estimate of P ₈₈, although this difference was considered statistically nonsignificant due to the overlapping 95% confidence intervals (Table 1).

Figure 4.

Vulnerability curves and estimates of leaf water potential (Ψ _leaf) required to induce a 50% (dashed line) and 88% (dash‐dotted line) loss in whole‐plant hydraulic conductance (P ₅₀ and P ₈₈, respectively). Acacia robusta (A) and A. tortilis (B) did not differ in their hydraulic vulnerability. Solid black lines are the Weibull model for each species. Pairs of thin vertical lines on either side of P ₅₀ or P ₈₈ represent the 95% confidence intervals.

Table 1.

Vulnerability curve parameter estimates and confidence intervals of Acacia tortilis and A. robusta.

P x , S x	Species	P x estimate (MPa)	95% CI	S x estimate	95% CI
P 50	A. robusta	–1.78	–1.68 to –1.89	66.27	47.65–88.96
	A. tortilis	–1.77	–1.58 to –1.99	48.30	31.38–80.21
P 88	A. robusta	–2.47	–2.25 to –2.85	35.03	22.29–51.11
	A. tortilis	–2.79	–2.26 to –3.61	22.56	11.67–44.59

Notes: P _x , leaf water potential (Ψ _leaf, MPa) required to induce a 50 or 88% decline in hydraulic conductance (K _plant); S _x, slope of the vulnerability curve at P ₅₀ or P ₈₈.

We also conducted an additional analysis of PLC data as categorical data, rather than continuous values of Ψ _leaf, by discretizing the PLC data into bins of Ψ _leaf (Appendix S4). This approach provides a simple between‐species comparison of PLC within important ranges of Ψ _leaf that would elucidate potential overfitting of our vulnerability curve models. This additional analysis also showed that PLC did not differ between species within each bin, further supporting our conclusion that the species do not differ in their hydraulic vulnerability.

DISCUSSION

We initially hypothesized that Acacia FYS in seasonally dry environments such as the Serengeti would invest in strategies to compete with grasses for water during pulses of resource availability. We expected both species to demonstrate relatively anisohydric stomatal behavior and to sustain hydraulic conductance even after water is withheld to compete with grasses after pulses of rainfall. Given that A. tortilis is widely distributed and more abundant in arid regions compared to A. robusta, we also predicted that A. tortilis would have a lower (=more negative) P ₅₀ and P ₈₈, thus prolonging C gain during drought.

Our data show that both species are strongly anisohydric “water‐spenders,” yielding rapid carbon gain after pulses of moisture availability. Despite their comparable stomatal responses to declining soil moisture and similar whole‐plant hydraulic vulnerability, they differed in their rate and timing of water use, particularly at moderate levels of drought. Acacia tortilis transpires and depletes soil moisture faster than A. robusta, despite being equally vulnerable to declines in K _plant. In our experiment, this higher E of A. tortilis was the primary driver of the faster decline in its Ψ _soil and Ψ _leaf. During wet periods, its greater g _s and K _max allowed for this higher E, which ultimately led to greater A _net than for A. robusta when soil moisture was still high. However, the drawback of this greater E in A. tortilis was that soil moisture was “spent” much faster than for A. robusta, leading to an earlier cessation of gas exchange.

This difference between the species in their timing of water‐use parallels the temporal trends in soil moisture after pulses of rainfall in the field. When we investigated the patterns of Ψ _soil in sites where these species occur, we found that the periods (number of days) of water availability are shorter (and less frequent) in sites where A. tortilis is more abundant (Appendix S4). Previous work has shown that A _net is strongly related to savanna tree seedlings' competitive ability against grasses (Campbell and Holdo, 2017), suggesting that a strategy like this—which maximizes K _max and A _net after pulses of water availability—is a particularly adaptive trait for A. tortilis. To our knowledge, this is the first time that physiological measurements at the leaf level (i.e., E) have been linked to species abundance and the rate of decline in Ψ _soil in African savannas.

Despite differences in the timing of soil drying and stomatal closure, both species demonstrated similar extreme anisohydric behavior as Ψ _soil declined. In phase II of the WP Curves (Figure 3), both species had a slope greater than 1, indicating that tension continues to increase as Ψ _soil declines (Martínez‐Vilalta et al., 2014; Meinzer et al., ²⁰¹⁶; Wu et al., ²⁰²¹). This steep slope implies that stomata were relatively insensitive to declining plant water status and that other mechanisms such as xylem embolism may be driving the drop in K _plant.

This anisohydric behavior has been observed in other arid and semi‐arid ecosystems that experience pulses of water availability (Fu and Meinzer, 2018). In a similar drought experiment, Wujeska‐Klause et al. (2015) compared seedlings of two Acacia species from contrasting habitats: A. aneura (arid habitats) and A. melanoxylon (humid, less‐arid habitats). They found anisohydric behavior in A. aneura, which had higher g _s during well‐watered periods, but experienced sharper declines after the onset of drought, similarly to A. tortilis in our study. These differences between species mirror our inference that tree seedlings in arid habitats experience greater selection for maximizing gas exchange immediately after pulses of water availability, but at the expense of earlier cessation of C gain.

Vulnerability of K _plant to declining Ψ _leaf and Ψ _soil is an important metric of water‐use strategies because it integrates hydraulic conductance through multiple potential resistors: water uptake from soil into fine roots, through coarse roots, shoots, and leaves, and ultimately through stomata. During drought, declines in conductance of this hydraulic pathway are typically thought to be driven by excessive tension in xylem conduits, resulting in embolism (Tyree and Zimmermann, 2002). While failure of xylem function is a key mechanism limiting conductance in roots and stems, other mechanisms may simultaneously limit K _plant during drought. For example, drops in K _plant could be driven by declines in conductance from soil to the fine roots. Their minimal rooting volume exposes FYS to steep gradients of water potential in the rhizosphere, particularly in very dry soils (which Acacia FYS experience in the field), effectively uncoupling water content near the roots from the bulk soil water content (Carminati and Javaux, 2020).

Leaves themselves potentially limit K _plant due to embolism in leaf veins (Milburn, 1966) and stomatal closure, but also by declines in conductance of non‐xylem pathways (Scoffoni et al., 2017). Water is transported into leaves via xylem, but must then pass through multiple tissues including vascular parenchyma and mesophyll before evaporating through stomatal pores. During drought, leaf shrinkage physically alters the non‐xylem hydraulic pathway through leaves, which increases hydraulic resistance (Scoffoni et al., 2014). Leaf conductance may also be limited by membrane permeability, which declines during drought due to reduced aquaporin activity (Sack and Holbrook, 2006; Maurel et al., ²⁰¹⁵). Regardless of which resistor is the primary driver of declining hydraulic conductance, the outcome for FYS's is the same: reduced capacity for gas exchange and increased probability of mortality.

Both of these species are drought‐deciduous, which may be an additional mechanism by which savanna FYS's can avoid lethal desiccation. Our study plants dropped their leaves at extreme levels of drought (when Ψ _leaf reached ~ –5 MPa), but otherwise appeared to be alive. The potential advantage of this strategy is that it reduces transpiration and resource utilization during extended periods of drought, which may protect other tissues (stems and roots) from dehydration (Santiago et al., 2016). However, a strategy of drought deciduousness also limits the ability of FYS to take advantage of pulses of water, since some time is required for rehydration and production of new leaves. Although we did not directly measure their drought recovery after a pulse of water, anecdotally, we have observed leafless FYS (~6 months without water in our glasshouse) produce new leaves within days of re‐watering. This ability to recover after re‐watering suggests that the ability to tolerate long periods of dormancy during the dry season—but to rapidly produce new leaves and resume photosynthesis after a pulse of water—may play a critical role in the survival of these FYS's.

Our method of constructing whole‐plant vulnerability curves and describing stomatal responses to drought is not without drawbacks. We used measurements of Ψ _soil to construct WP curves, rather than its more common proxy, Ψ _predawn. Even though Ψ _soil more accurately reflects the maximum Ψ available to the plant (Hochberg et al., 2018), it is rarely measured directly. We attempted to verify that predawn Ψ _leaf is approximately equivalent with our measurements of Ψ _soil, but this approach still assumes that Ψ _soil is homogenous throughout the rooting volume and that soil hydraulic continuity is sufficient (Donovan et al., 2001). Any potential error in Ψ _soil measurements, as well as inherent uncertainty in gas exchange and Ψ _leaf measurements, would ultimately be reflected in our calculated K _plant values.

CONCLUSIONS

With two tree species distributed on opposite ends of a precipitation and grass productivity gradient in a seasonally dry ecosystem, this study system is a valuable model for studying the mechanisms affecting FYS–grass competition. Both species maximize C gain during pulses of resource availability, which is likely stored belowground as nonstructural carbohydrates. During the dry season, a seedling with stored nonstructural carbohydrates can afford to lose aboveground tissues, whether it be through desiccation, senescence and drought deciduousness, fire, or herbivory, and then resume growth and/or resprout when precipitation returns. This ability to regenerate aboveground tissues and the high frequency of topkill and resprouting in this biome suggest that FYS establishment in savannas is, at least in part, dependent upon growing as fast as possible to establish belowground biomass from which a seedling can resprout after disturbance. Thus, competition with “water‐spender” grasses (Williams et al., 1998) causes strong natural selection against FYS that cannot gain enough carbon and store it belowground (the so‐called “Gulliver syndrome” sensu Bond and van Wilgen, ¹⁹⁹⁶). Although we only compared two species here, hydraulic traits may play a similarly important role in seedling establishment of other savanna species that experience strong competition with grasses. Our study provides important insight into the role of water‐use strategies and hydraulic traits in FYS survival and, ultimately, about the mechanisms enabling the tree‐grass coexistence observed in savannas.

AUTHOR CONTRIBUTIONS

S.T.C. designed the study, collected and analyzed the data, and wrote the manuscript. T.M.A. collected field data in the Serengeti. T.M.A. and W.K.S. contributed data interpretation and manuscript revisions.

Supporting information

Appendix S1. Diagram of whole‐plant hydraulic conductance (K _plant) measurements.

Appendix S2. Species distributions and field Ψ _soil.

Appendix S3. Equilibrium of predawn Ψ _leaf and Ψ _soil from the Ψ _soil‐sensor validation experiment.

Appendix S4. Categorical analysis of vulnerability curves.

Cory, S. T. , Smith W. K., and Anderson T. M.. 2022. First‐year Acacia seedlings are anisohydric “water‐spenders” but differ in their rates of water use. American Journal of Botany 109(8): 1251–1261 10.1002/ajb2.16032

DATA AVAILABILITY STATEMENT

The data that support the findings of this study are available from the Dryad Digital Repository: https://doi.org/10.5061/dryad.7h44j0zwx (Cory, 2022).