Advancing the pneumatic method to assess xylem vulnerability to embolism among distinct growth forms

Abstract

• Drought‐driven plant mortality is closely linked to xylem embolism. Building useful, reliable datasets of xylem vulnerability to embolism requires methods that are practical, fast, accurate, widely accessible and robust across growth forms. We tested and advanced the pneumatic method for constructing xylem vulnerability curves (VCs) across contrasting growth forms to improve inference of drought resilience.

• Using an automated pneumatron, VCs were constructed for three species representing a small woody shrub (Erica monsoniana), a large woody shrub (Protea repens), and a reed‐like graminoid (Cannomois congesta). For graminoid culms, we compared three approaches for estimating xylem water potential (Ψ) and developed a non‐invasive method that couples repeated relative water content (RWC) measurements with Ψ–RWC models to obtain high‐temporal Ψ estimates.

• Percent air discharged (PAD)–Ψ relationships were well captured by sigmoid functions. Cannomois congesta showed the steepest curves and the least negative thresholds overall (P₅₀ = −2.91 ± 0.09 MPa), indicating early, rapid embolism progression, whereas Erica monsoniana was most resistant (P₁₂ = −5.91 ± 0.74 MPa; P₅₀ = −6.78 ± 0.76 MPa) with higher variability; Protea repens was intermediate. P₅₀ estimates were the most comparable with prior optical, pneumatic and centrifuge estimates, whereas P₁₂ and P₈₈ showed greater divergence. Ψ_TLP was less variable between species, but ranked similarly (−1.49 ± 0.03, −1.53 ± 0.03, −1.59 ± 0.02 MPa for C. congesta, P. repens, and E. monsoniana, respectively). Such variation yielded systematically wider hydraulic safety margins for the three species.

• By demonstrating that the pneumatic method can generate reliable vulnerability curves across small and large woody shrubs and graminoids, this study broadens the comparative evaluation of xylem vulnerability across growth forms with contrasting anatomy. A practical advance is the use of repeated RWC measurements paired with Ψ–RWC relationships to improve Ψ resolution in graminoid culms while minimizing disturbance.

Our study shows that pneumatic vulnerability curves can be constructed across contrasting growth forms, with P₅₀ showing the closest agreement across methods. Repeated relative water content measurements improved water potential estimation for graminoid culms.

INTRODUCTION

Increasingly frequent and severe drought events are pushing plants beyond their hydraulic limits and driving widespread dieback and demographic loss at large spatial scales (Breshears et al. ²⁰⁰⁵; Allen et al. ²⁰¹⁰; Choat et al. ²⁰¹²). Mechanistically, these dieback events are closely associated with xylem embolism and subsequent hydraulic failure (McDowell ²⁰¹¹; Anderegg et al. ²⁰¹⁶). Given that climate projections indicate further increases in drought frequency and duration, drought‐linked dieback is likely to become more prevalent under ongoing climate change (Field et al. ²⁰¹⁴; Intergovernmental Panel on Climate Change (IPCC) ²⁰²³).

Quantifying xylem resistance to embolism and hydraulic failure is a major component to assessing drought resilience (Tyree & Sperry ¹⁹⁸⁹). For this purpose, xylem vulnerability curves (VCs) and their parameters are widely used. Confidence in vulnerability curve measurements is therefore fundamental for comparing drought resilience across species, growth forms and sites (Brodribb 2009; Choat ²⁰¹³). Xylem VCs provide a direct representation of the relationship between hydraulic function and water stress by plotting the proportion of xylem embolized as a function of water potential (Ψ, MPa), illustrating the onset and progression of embolism with water stress. In theory, curves can take on various shapes, reflecting underlying xylem physiology, and often characterize linear, exponential, or sigmoidal forms (Pammenter & van der Willigen ¹⁹⁹⁸; Venturas et al. ²⁰¹⁹). Relatedly, reliable embolism thresholds are expected to be physiologically consistent, with turgor loss preceding major embolism—yielding positive safety margins (Martin‐StPaul et al. ²⁰¹⁷). Vulnerability curve parameters, including the Ψ at 12%, 50%, and 88% embolism (P₁₂, P₅₀, and P₈₈, respectively), and derived metrics such as hydraulic safety margins are increasingly embedded in vegetation models that simulate drought responses (Cramer et al. ²⁰⁰¹; Cochard et al. ²⁰²¹). Accurate representation is, therefore, necessary to provide reliable inferences about hydraulic strategies and modelled drought outcomes. Yet, constructing reliable VCs remains technically challenging.

Several methods have been established to induce and quantify embolism, each with distinct strengths and constraints. Embolism is typically induced by bench dehydration, air pressurization, or centrifugation (Cochard et al. ²⁰¹³; Venturas et al. ²⁰¹⁷), and then quantified using hydraulic conductance (PLC), acoustic emissions, direct imaging or airflow (Sperry et al. ¹⁹⁸⁸; Holbrook et al. ²⁰⁰¹; Ennajeh et al. ²⁰¹¹; Brodribb et al. ²⁰¹⁷). Approaches, however, differ in practical feasibility and sensitivity to artefacts, including open‐vessel effects, residual pressurized air and non‐physiological flow or pressure conditions, as well as sample destructiveness, efficiency and costly equipment (Cochard et al. ²⁰¹³; Venturas et al. ²⁰¹⁷). Developing a useful, reliable and comparable embolism‐resistance dataset therefore requires techniques that are practical, widely accessible, and demonstrably robust across taxa and growth forms.

The pneumatic method, introduced as an alternative to quantify embolism via gas discharge (Pereira et al. ^{2016, 2020}), builds on earlier work linking gas conductivity and embolism (Cochard et al. ¹⁹⁹²; Franks et al. ¹⁹⁹⁵; Ennajeh et al. ²⁰¹¹). This method uses bench dehydration to induce embolism and measures the gas discharged under partial vacuum from the cut end of a branch, while Ψ is measured independently. Measurements can be made manually using simple tubing, a syringe, and a pressure sensor (Pereira et al. ²⁰¹⁶) or using an automated pneumatic device (or ‘pneumatron’) (Pereira et al. ²⁰²⁰; Trabi et al. ²⁰²¹). This method provides high‐temporal resolution, is rapid, non‐destructive, low‐cost, allows repeated measurements on the same branch, and multiple samples can be measured together. Additionally, the method makes use of bench dehydration, which mimics natural drought and does not require additional equipment.

To date, the pneumatic method has been applied in cross‐method comparisons and assessments of embolism resistance across a range of species—including trees, shrubs, and graminoids—in temperate, tropical, and Mediterranean ecosystems (Pereira et al. ^{2016, 2020}; Zhang et al. ²⁰¹⁸; Sergent et al. ²⁰²⁰; Wu et al. ²⁰²⁰; Chen et al. ²⁰²¹; Paligi et al. ²⁰²³; Miranda et al. ²⁰²⁴; West et al. ²⁰²⁴). Multiple studies have shown close agreement between the pneumatic method and several other techniques. This agreement is strongest for P₅₀ estimates, particularly in diffuse‐porous and short‐vessel angiosperms (Pereira et al. ^{2016, 2020}; Zhang et al. ²⁰¹⁸; Sergent et al. ²⁰²⁰; Paligi et al. ²⁰²³); as well as comparable agreement in graminoid monocots (West et al. ²⁰²⁴). Across studies, the strongest cross‐method agreement is typically observed for P₅₀, whereas early‐ and late‐stage vulnerability curve parameters (i.e., P₁₂ and P₈₈) are more variable and appear more sensitive to sample anatomy and measurement choices (Pereira et al. ²⁰¹⁶; Zhang et al. ²⁰¹⁸; Paligi et al. ²⁰²³; West et al. ²⁰²⁴). Such discrepancies have been attributed to methodological artefacts affecting gas discharge dynamics and identification of maximum discharge.

However, comparative evaluation across contrasting growth forms characteristic of species‐rich shrublands remains limited. This is particularly relevant in the Cape Floristic Region (CFR) – a globally recognized biodiversity hotspot for conservation priorities dominated by structurally and functionally distinct growth forms (Myers et al. ²⁰⁰⁰). Assessing whether the pneumatic method performs consistently across these contrasting forms is therefore relevant not only within fynbos vegetation, but also for extending method application to shrub‐dominated systems more broadly, including other Mediterranean‐type ecosystems. To date, xylem vulnerability to embolism has been assessed to a limited degree across representative species of dominant growth forms in the CFR using centrifuge, optical, and pneumatic methods (Jacobsen et al. ²⁰⁰⁹; Skelton et al. ²⁰²³; West et al. ²⁰²⁴). This knowledge base is increasingly important given these growth forms are globally relevant, and because the CFR is projected to experience hotter, drier conditions that will intensify plant water stress (Midgley et al. ²⁰⁰³; McDowell ²⁰¹¹; Altwegg et al. ²⁰¹⁴). Yet, predictive capacity remains limited by insufficient understanding of drought resilience across these dominant growth forms (Midgley et al. ²⁰⁰³; Hannah et al. ²⁰⁰⁵; Skelton et al. ²⁰¹⁵). Early work applied the standard centrifuge method on several woody shrub species (Jacobsen et al. ²⁰⁰⁹). Flushed stem segments were used to determine maximum hydraulic conductivity, then repeatedly spun to impose specific Ψ, with gravimetric conductivity measured subsequently to calculate PLC. Vulnerability curves (PLC–Ψ) were fatigue‐corrected, fitted with second‐order polynomials and used to estimate P₅₀. Curve shapes and magnitude varied, including concave, convex, and linear forms, while interspecific variation in embolism resistance was relatively low (Jacobsen et al. ²⁰⁰⁹). Subsequent work using the optical method at the same site produced comparatively steep, sigmoidal curves and stronger contrasts in embolism resistance across representative proteoid, ericoid, and restioid taxa (Skelton et al. ²⁰²³). More recently, optical and pneumatic methods were applied to restioid culms, yielding broadly comparable P₅₀ estimates but indicating potential differences in curve shape (West et al. ²⁰²⁴). For the pneumatic curves, Ψ was derived from separate culms bench‐dried alongside the measured samples (a limitation the current study aims to address). Nonetheless, West et al. (2024) demonstrated the utility of both techniques for generating robust embolism curves in restioid culms. To date, however, no study has applied the pneumatic method to proteoid or ericoid growth forms, and further evaluation is therefore required to assess its suitability across distinct growth forms and its comparability with previously applied techniques.

Here, we test and develop the pneumatic method for constructing xylem VCs across three contrasting growth forms represented by a small woody shrub, a large woody shrub, and a reed‐like graminoid, enabling comparison with existing datasets and assessment of its broader applicability. Vulnerability curves were constructed for three species, representative of the restioid, ericoid, and proteoid growth forms, namely, Cannomois congesta Mast. (Restionaceae), Erica monsoniana L.f. (Ericaceae), and Protea repens (L.) L. (Proteaceae). Embolism resistance has previously been assessed at least once for each species at this site using centrifuge or optical methods, and the pneumatic method has previously been applied here only to the restioid species, providing a practical basis for methodological comparison. Further, it is reasonably expected that individuals experience turgor loss prior to P₅₀ (Martin‐StPaul et al. ²⁰¹⁷) and incur only limited embolism prior to turgor loss (Miranda et al. ²⁰²⁴). We therefore generated pressure–volume curves to compare Ψ_TLP with P₅₀, predicting that species would maintain positive safety margins. Ψ_TLP also provided a physiological criterion for interpreting whether gas discharge was likely to reflect embolism progression or a potential early‐stage residual sap artefact. This study extends the pneumatic method to woody fynbos shrubs and a particular focus is improving Ψ estimates for graminoid culms by enabling a more direct assessment of the Ψ driving embolism, allowing us to evaluate its performance across contrasting growth forms within a biodiversity hotspot and consider its broader utility for shrub‐dominated vegetation.

MATERIALS AND METHODS

Site, species, and collection

For this study, plant material was sampled from Jonaskop in the Riviersonderend Mountain Range, Western Cape, South Africa (33° 56.3′S, 19° 31.2′ E). The site is in the south‐western corner of the Cape Floristic Region (CFR), within South Africa's Mediterranean‐type climate area. Sampling was conducted on a relatively flat, north‐facing sandy plain, 992 metres above sea level. The vegetation is representative of mountain fynbos, composed of an open canopy of tall proteoid shrubs (ca. 3.5 m), with an understory of short ericoid shrubs (ca. 0.25–1 m) and various restioids (Agenbag et al. ²⁰⁰⁸).

Three species were examined in this study: Cannomois congesta Mast. (Restionaceae), Erica monsoniana L.f. (Ericaceae), and Protea repens (L.) L. (Proteaceae) (Fig. S1). These species are representative of the dominant fynbos growth forms—restioid, ericoid, and proteoid, respectively. The following species descriptions are drawn from Manning & Goldblatt (2012). Cannomois congesta is a dioecious, reed‐like, rhizomatous perennial typical of the restioid growth form. Its photosynthetic culms, which form dense clumps, are slender, unbranched, hollow, and extend ca. 1 m in length. Cannomois congesta is predominantly found on well‐drained flats and slopes at elevations above ca. 1,000 m in the northwest, southwest, and Agulhas Plain areas. Erica monsoniana is a small to medium in size shrub, reaching up to ca. 1.5 m tall, with a sparse branching structure. It is distinguished by its fine leaves and large, white, calycine flowers, which are ovoid‐conical in shape. This species is typically found on middle to upper slopes in the northwest and southwest areas, favouring well‐drained, acidic soils. Protea repens is a broad‐leaved, woody proteoid shrub that can reach up to ca. 4.5 m in height. It features thick stems and leather‐like linear‐spathulate leaves – narrow and spoon‐shaped. The flower heads are large and narrowly obconic, measuring ca. 100–160 mm in length and 70–90 mm in diameter. This species typically inhabits sandstone and clay flats as well as slopes.

Samples were collected within an area ca. 625 m² (25 m × 25 m). The same populations were sampled using consistent methods over multiple campaigns between April 2023 and January 2024. Care was taken to ensure that all samples were healthy, well‐hydrated, and in a relaxed state. Samples were collected in the morning, before ca. 11:00 am, and handled carefully to remain intact and minimize water loss. The collection was conducted during a prolonged wet season and timed with natural precipitation events to ensure the plants were well‐hydrated. Furthermore, only healthy‐looking samples, with no observable signs of senescence, were collected.

Previous studies have shown that the maximum vessel length of P. repens is relatively short (ca. 21 cm), while E. monsoniana exhibits longer vessels (exceeding 50 cm) (Skelton et al. ²⁰²³). Branches of P. repens were cut ca. 100 cm from the distal tip, and E. monsoniana branches were cut at least 50 cm from the distal tip or at the root collar. Similarly, the maximum vessel length of C. congesta is ca. 15 cm (Skelton et al. ²⁰²³), and the vessels, as well as the central cavity, do not extend across the nodes (West et al. ²⁰²⁴). Therefore, C. congesta culms were cut below the most basal node, ensuring at least four nodes from the spikelet, with cuts made close to the rhizome. In line with previous studies (Skelton et al. ²⁰²³; West et al. ²⁰²⁴) and to avoid sex‐based differences, only female individuals were sampled. This method ensured the stems and culms were collected with sufficient length and structural integrity for further analysis. For all samples, the cut end was immediately submerged in water and re‐cut to relax the xylem and reduce the risk of embolism spread. Samples were then placed in airtight plastic bags with moist paper towels to inhibit water loss and transported to the laboratory where they were stored overnight in a dark room at 4 °C. All samples were measured within 24 h of collection.

Xylem vulnerability to embolism curves

Pneumatic measurements

Two automated pneumatic instruments – or ‘pneumatrons’—were constructed for this study to measure the air discharge (AD) from the plant xylem during dehydration. The instruments were designed following Pereira et al. (2020) and Trabi et al. (2021), with minor modifications to the control system to improve signal stability (further details are provided in the Supplementary Material).

Plant material for AD measurements was cut from the field‐collected plants, which were stored overnight in airtight plastic bags, with moist paper towels, in a dark room at 4 °C. This handling helped maintain hydration and minimize embolism prior to measurement, although residual sap in cut‐open conduits can influence the initial air discharge (Miranda et al. ²⁰²⁴). The potential effect of this artefact was therefore assessed during PAD processing, as described below and in the Supplementary Material. Before AD measurements began, the cut plant material was kept in smaller plastic bags with moist paper towels to maintain hydration during sample preparation. For P. repens and E. monsoniana, samples were cut ca. 40–80 cm from the distal tip. This length varied to ensure sufficient plant material was available for Ψ measurements throughout the dehydration period while avoiding excessive removal that could potentially affect the plant's dry‐down process. Whereas C. congesta samples were cut just below the third node from the spikelet, ca. 30 cm in length. The base of each sample was shaved with a clean razor blade to ensure unobstructed air flow from the xylem, and then immediately connected to a pneumatron port using tight‐fitting silicon tubing. Each pneumatron could accommodate a maximum of six stems or culms simultaneously and the stopcocks of the manifold allowed for adjustments to the number of stems or culms measured and reservoir volume. General purpose contact adhesive was applied to seal the fittings and prevent air leakage—the same adhesive used to seal cut leaves and manage any leakage that occurred during the measurement process. For C. congesta, the saturated weight (SW, g) was measured (accurate to 0.01 g) immediately before attaching the culm to the pneumatron. Saturated weight was used to assess the relative water content (RWC, %) during the dry‐down.

The air pressure in the manifold was recorded over a 15‐s interval for each AD measurement, which was automatically taken every 15 min. A 15‐s interval has been shown to provide accurate estimates for several species in previous studies (Zhang et al. ²⁰¹⁸; Paligi et al. ²⁰²³; Zhao et al. ²⁰²³). Moreover, >90% of the gas discharged within the first 15 s originates from embolized xylem, supporting the reliability of this measurement window (Yang et al. ²⁰²³). Air discharge (AD) was calculated as the difference between the initial (P _i, kPa) and final pressures (P _f, kPa) ($\mathrm{AD}=P_{\mathrm{i}}-P_{\mathrm{f}}$). Samples were left to dehydrate on the benchtop under atmospheric pressure and room temperature. Monitoring air pressure during each AD measurement showed an initial more rapid pressure drop, followed by a slower, more gradual decrease. Well‐hydrated plants sustained vacuum more effectively; however, as the plant material desiccated, more gas was discharged, leading to noticeable pressure changes, reflected in the relatively stable P _i and the progressively decreasing P _f. Subsequently, the AD measurements over the dehydration period showed a clear progression, with a similar shape observed between individuals.

The pneumatron provided high‐temporal resolution, capturing the progression of AD during dehydration with minimal pressure value spikes and drops. Initially, AD values were low reflecting the minimum air discharge (AD_min), gradually increasing until reaching an observable plateau, indicating maximum air discharge (AD_max). Measurements ceased when samples were visibly desiccated, and AD had plateaued. The minimum and maximum air discharge (AD_min and AD_max) were determined after excluding values affected by known leaks and erroneous pressure spikes. AD was normalized by calculating the percent air discharged (PAD, %) using the following equation:

$$\mathrm{PAD}=100\times \frac{\mathrm{AD}-{\mathrm{AD}}_{\min }}{{\mathrm{AD}}_{\max }-{\mathrm{AD}}_{\min }}$$ (1)

To evaluate the potential influence of early‐stage artefacts associated with residual sap in cut‐open conduits, we conducted a sensitivity analysis in which PAD was recalculated after excluding early AD measurements recorded at Ψ ≥ −0.5 MPa, Ψ ≥ −1.0 MPa and Ψ ≥ Ψ_TLP, based on the rationale discussed by Miranda et al. (2024). No detectable effect of correction was observed for E. monsoniana or P. repens. In C. congesta, the correction affected P₁₂ only, whereas P₅₀ and P₈₈ showed no detectable difference. Accordingly, PAD for C. congesta was calculated after excluding AD values recorded at Ψ ≥ Ψ_TLP for the primary analyses presented here. For comparison among methods, both corrected and uncorrected estimates are presented for C. congesta to illustrate the sensitivity of the parameters to this methodological choice. Further details of the sensitivity analysis are provided in the Supplementary Material.

Water potential estimation

Xylem water potential (Ψ, MPa) was measured concurrently with AD measurements using a Scholander‐type pressure chamber (PMS Instruments, Corvallis, Oregon, USA). Measurements of Ψ were taken at intervals between several minutes and several hours, with more frequent sampling during the initial rapid phase of dehydration and longer intervals thereafter, extending at least to ca. 50 PAD.

Ψ was measured similarly across species, with slight adjustments to account for leaf morphology or culms. First, to ensure Ψ equilibrium, the entire sample was covered with a black plastic bag for at least 15 min prior to measurement. For P. repens, which has broad leaves suitable for the pressure chamber, the selected leaf was placed in a small plastic bag containing a slightly moist paper towel to minimize water loss during handling and measurement. After enclosing the leaf, it was cut at the petiole and measured in the pressure chamber while remaining in the bag. In the case of E. monsoniana, the short, narrow leaves are unsuitable for the pressure chamber, so a small branchlet was bagged, cut, and measured in the same manner as P. repens leaves. For both P. repens and E. monsoniana, wherever any leaves or branchlets were cut, general purpose contact adhesive was applied to the area to avoid air entry in the branch and minimize effect on AD measurements, following the methods of Pereira et al. (2020).

Water potential estimation for graminoid species

Direct measurement of Ψ was not possible for C. congesta, which has culms rather than stems and leaves, due to potential structural damage and artefacts resulting from repeated use of the pressure chamber. Instead, we tested three Ψ estimation approaches for this growth form. Following West et al. (2024), Ψ–time relationships were derived from additional culms dried simultaneously on the benchtop (or ‘bench‐culms’). Ψ was measured using the pressure chamber; each culm was measured only once to avoid cumulative damage, and a mean segmented linear regression was fitted. Although the bench‐culms and air‐discharge culms (AD‐culms) appeared similarly desiccated throughout the measurement period, the potential influence of their differing treatments remains uncertain. To address this, relative water content (RWC, %) was measured to provide a more direct, non‐destructive estimate of Ψ for the AD‐culms.

RWC was determined by measuring saturated weight (SW, g) of AD‐ and bench‐culms while fully hydrated (accurate to 0.01 g). Thereafter, periodic fresh weight (FW, g) was measured throughout the AD process. For AD‐culms, luer fittings enabled quick, non‐destructive FW measurements between AD measurements. For the bench‐culms, Ψ was measured only once per culm as mentioned above, and FW was not measured subsequently to avoid any artefacts. Upon completion of AD measurements, all the culms were dried at 60 °C for 48 h to determine dry weight (DW, g). Thereafter, RWC (%) was derived using the following equation:

$$\mathrm{RWC}=100\times \frac{\mathrm{FW}-\mathrm{DW}}{\mathrm{SW}-\mathrm{DW}}$$ (2)

Using the bench‐culms, Ψ was modelled as a function of the inverse RWC ($\text{IRWC}=100-\mathrm{RWC}$, %) by applying individual segmented linear regressions. Additionally, Ψ was similarly modelled as a function of IRWC using an independently generated pressure–volume (PV) curve (see below). These Ψ–RWC models were subsequently applied to the measured RWC values of the AD‐culms to derive Ψ estimates. The measured RWC enabled a more direct estimation of Ψ throughout the desiccation process. Segmented linear regression was used to generate piecewise estimates for Ψ over time. A mean fit was used for bench‐culms, whereas separate fits were generated for each sample for the AD‐culms_RWC and AD‐culms_PVC approaches. We then compared these Ψ–time estimation methods by their effects on the derived vulnerability parameters P₁₂, P₅₀, and P₈₈. For each parameter, a linear mixed model was fit with estimation method as a fixed effect and individual as a random intercept. An analysis of variance (ANOVA) with Satterthwaite degrees of freedom tested the fixed effect. Estimated marginal means were computed, followed by Tukey's‐adjusted pairwise comparisons among methods.

Curve fitting and parameterization

Time‐resolved PAD and Ψ measurements were combined and xylem vulnerability to embolism was expressed as the relationship between PAD and Ψ. This relationship was fitted individually by non‐linear least squares using the following sigmoidal function:

$$\mathrm{PAD}=\frac{100}{\left(1+e^{a\left(\mathrm{\Psi }-b\right)}\right)}$$ (3)

where a reflects the sensitivity of PAD to decreasing Ψ, corresponding to the slope of the sigmoid function, while b is the Ψ at 50 PAD (P₅₀, MPa) (Pammenter & van der Willigen ¹⁹⁹⁸). Parameters P₁₂ and P₈₈ were obtained in closed form from the fitted curve using the following equation (Domec & Gartner ²⁰⁰¹):

$${\mathrm{P}}_x=b+\frac{1}{a}\ln \left(\frac{100}{x}-1\right),x\in \left\{12,88\right\}$$ (4)

Species‐level VCs and distribution‐based confidence intervals (CIs) for P₁₂, P₅₀, and P₈₈ were obtained via a parametric bootstrap (1,000 iterations). For every iteration, P₁₂, P₅₀, and P₈₈ were calculated; the bootstrap median was taken as the point estimate and the 2.5th–97.5th percentiles as the 95% CIs. Furthermore, one‐way ANOVA tested species effects separately on P₁₂, P₅₀, and P₈₈. When the species effect was significant, Tukey's honestly significant difference post hoc pairwise comparisons were generated.

Comparison between methods

Published vulnerability datasets were re‐analysed to compare parameter estimates across methods and studies. For Jacobsen et al. (2009) point estimates and standard errors (SE) were digitized from the PLC–Ψ plot. Points were modelled with a second‐order polynomial as per the described methods, with a weighted least squares using inverse‐variance weights ${\omega }_i=1/{\mathrm{SE}}_i^2$. To account for specified fatigue correction, PLC was linearly rescaled 0–100% and SE scaled by the same factor. Uncertainty was propagated via a parametric bootstrap (1,000 iterations) by drawing pseudo‐observations ${\mathrm{PLC}}_i~\mathcal{N}\left({\mathrm{PLC}}_i,{\mathrm{SE}}_i\right)$, and re‐fitting the quadratic each iteration. Point estimates are derived from the mean curve and 95% CIs are the 2.5th–97.5th percentiles. Data for Skelton et al. (2023) and West et al. (2024) was modelled with a sigmoid function (Equation 3), and species‐level uncertainty was quantified with parametric bootstrapping (1,000 iterations). For each species and method, draws were made from normal distributions centred on the species mean parameters. Point estimates are derived from the mean curve and 95% CIs are the 2.5th–97.5th percentiles.

Pressure–volume curves and hydraulic safety margins

To contextualize the VCs and calculate hydraulic safety margins, pressure–volume (PV) curves were independently generated for multiple stems/culms per species using the repeat measurement bench drying technique (Tyree & Hammel ¹⁹⁷²; Bartlett et al. ²⁰¹²). Upon collection, each stem and culm was re‐cut under water ca. 30 cm from the tip to relax the xylem. The stems and culms were immediately wrapped in airtight plastic, transported to the laboratory and stored in a 4 °C cold room for no longer than 24 h to maintain tissue integrity.

Before measurement, stems/culms were kept wrapped in plastic with only the cut end exposed and submerged in distilled water and allowed to rehydrate for 30 min in a cool, dark room. After rehydration, each stem and culm was cut from above the water mark of rehydration. Saturated weight (SW, g) was measured immediately (accurate to 0.01 g) and Ψ at SW was measured using a Scholander‐type pressure chamber (PMS Instruments, Corvallis, Oregon, USA) while wrapped in plastic to prevent water loss during measurement. Fresh weight (FW, g) and Ψ of each stem and culm were repeatedly measured as the samples dried from a fully hydrated state under ambient laboratory conditions. Measurements continued until each stem and culm had lost ca. 60% of its initial SW or until reliable Ψ measurements were no longer feasible.

Once measurements were completed, stems and culms were dried at 60 °C for 48 h to determine the dry weight (DW, g). Relative water content (RWC, %) was derived using Equation (2) above and individual PV curves were plotted as the inverse water potential (${\mathrm{\Psi }}_{\mathrm{i}}=-1/\mathrm{\Psi }$, MPa⁻¹) as a function of inverse RWC ($\text{IRWC}=100-\mathrm{RWC}$, %). Ψ at turgor loss point (Ψ_TLP), osmotic potential (Ψ_O), and RWC at turgor loss point (RWC_TLP) were calculated by fitting a linear function to the discernible linear portion using standard methodologies (Tyree & Hammel ¹⁹⁷²; Bartlett et al. ²⁰¹²). Finally, hydraulic safety margins (HSM_ΨTLP) were calculated as the difference between each parameter and Ψ_TLP.

Between‐species differences were evaluated with one‐way ANOVA run separately for Ψ_TLP, Ψ_O, and RWC_TLP. Species was treated as the fixed effect. Where the ANOVA indicated a significant species effect, Tukey's honestly significant difference test was applied to perform all pairwise species comparisons.

Statistical analyses

All analyses were conducted using R – version 4.4.2 (R Core Team ²⁰²⁴). Data wrangling used the tidyverse (Wickham et al. ²⁰¹⁹), and figures were produced with ggplot2 (Wickham 2016) and ggpubr (Kassambara 2023). Segmented linear regressions were fit using segmented (Muggeo 2008). Linear mixed‐effects models were fit with lme4 (Bates et al. ²⁰¹⁵), Satterthwaite‐approximated tests with lmerTest (Kuznetsova et al. ²⁰¹⁷), and estimated marginal means with Tukey's‐adjusted contrasts with emmeans (Lenth 2025). Marginal R² for mixed models were computed with MuMIn (Bartoń 2025). Non‐linear least squares fits used minpack.lm (Elzhov et al. ²⁰²³). Parametric bootstraps requiring truncated‐normal draws used truncnorm (Mersmann et al. ²⁰²³).

RESULTS

Resolving air discharge and water potential measurements for the three species

Across all three species, the pneumatic method detected adequate AD, which increased progressively over time, resulting in AD curves consistent with previous pneumatic studies (Fig. S4).

For P. repens and E. monsoniana, water potential (Ψ) was characterized by an initial rapidly decreasing phase followed by a more gradual second phase in which Ψ continued to decrease at a slower rate. Segmented linear regression effectively captured these trends (R² ≥ 0.894, Table S1). In all cases, some extrapolation was required to estimate Ψ to the end of the PAD measurement period (Figs. S5 and S6).

For C. congesta, the Ψ–RWC relationships from single‐measure bench‐culms and repeat‐measure PV curves were closely aligned (Fig. 1a, Table S2). In RWC–time relationships, AD‐culms were measured at high‐temporal resolution until AD measurements ceased, while bench‐culms were measured for around 30 h, when Ψ reached ca. –2.5 MPa (ca. 60% RWC); this resulted in divergent extrapolation thereafter (Fig. 1b, Table S2). The resulting Ψ–time relationships from the three approaches diverged, particularly at more negative water potentials, producing slight differences in the derived PAD–Ψ curves (Fig. S7). The derived vulnerability parameters therefore showed some differences among approaches, although the bench‐culm and RWC–AD‐culm approaches showed the closest agreement, especially for P₅₀ (Fig. 2, Table S3). RWC provided a direct, non‐destructive, individual‐level estimate of Ψ with less variation. Therefore, the RWC–AD‐culm approach was used to construct subsequent VCs in this study. The final paired PAD and Ψ curves for all three species are collated in Fig. 3.

Fig. 1.

Relationships between water potential (Ψ) and relative water content (RWC) during dehydration of Cannomois congesta. (a) Ψ as a function of 100 − RWC (%) for bench‐dehydrated culms and the pressure–volume (PV) curve. (b) RWC (%) through time (h) during bench dehydration for pneumatron‐treated culms (AD‐culms) and control bench‐culms.

Fig. 2.

Method effects on embolism thresholds for Cannomois congesta. Boxplots show P₁₂, P₅₀, and P₈₈ (MPa) estimated from bench‐culms, AD‐culms_RWC (Ψ inferred from the bench dehydration Ψ–RWC relationship) and AD‐culms_PVC (Ψ inferred from the PV curve). Boxes show medians and interquartile ranges. Different letters indicate Tukey's‐adjusted pairwise differences (α = 0.05) based on estimated marginal means from linear mixed models fitted separately for each threshold (Ψ ~ method + (1|individual); Satterthwaite d.f.).

Fig. 3.

Construction of pneumatic vulnerability curves across species from dehydration time series. Columns correspond to Cannomois congesta, Erica monsoniana, and Protea repens. Top row: percentage air discharged (PAD, %) through time (h) during bench dehydration. Middle row: Ψ (MPa) through time; lines show individual fits used to interpolate Ψ at the time of each air discharge measurement. Bottom row: resulting PAD–Ψ relationships for each individual used to estimate embolism thresholds (P₁₂, P₅₀, and P₈₈).

Assessing vulnerability to embolism as estimated by the pneumatron

Sigmoid functions effectively captured the PAD–Ψ relationships for all three species (R² ≥ 0.93). Overall, the curves differed significantly among species, with clear differences in derived vulnerability curve parameters (Fig. 4). Cannomois congesta exhibited the steepest curves (a = 3.17 ± 0.81) and the least negative thresholds, illustrating earlier embolism onset and rapid progression over a narrow Ψ range.

Fig. 4.

Pneumatic vulnerability curves and embolism thresholds across species. (a) Bootstrapped mean sigmoidal fits describing PAD–Ψ relationships for each species; grey lines show individual curves. Symbols indicate P₁₂, P₅₀, and P₈₈, with horizontal error bars showing 95% bootstrap confidence intervals. The vertical dotted line denotes species‐level mean Ψ_TLP estimated independently from PV curves. (b–d) Boxplots of P₁₂, P₅₀, and P₈₈ (MPa) by species; different letters indicate Tukey's pairwise differences following one‐way ANOVA (α = 0.05).

Erica monsoniana had the most negative thresholds with late embolism onset (P₁₂ = −5.91 ± 0.74 MPa) and a more gradual progression (a = 2.67 ± 1.18). Additionally, variation was larger across all thresholds for E. monsoniana, potentially reflecting greater spread in late‐stage embolism events.

Protea repens showed intermediate curves, with a moderately steep slope (a = 2.30 ± 1.27) and threshold estimates between those of C. congesta and E. monsoniana. Variation was moderate, especially at P₁₂ and P₈₈ estimates, indicating potential early gas discharge or greater deviation in maximum AD plateau among individuals.

Deriving pressure–volume curves and hydraulic safety margins

The relative ranking of embolism resistance inferred from the VCs – C. congesta < P. repens < E. monsoniana in order of increasing resistance—was analogous to turgor loss point (Fig. 5). Water potential at turgor loss point (Ψ_TLP) varied significantly among species (F = 26.35, P < 0.001). Where C. congesta exhibited the least negative Ψ_TLP (−1.49 ± 0.03 MPa). Erica monsoniana had the most negative Ψ_TLP (−1.59 ± 0.02 MPa), suggesting tolerance of lower water potentials before turgor loss, while P. repens was intermediate (−1.53 ± 0.03 MPa). Although the absolute difference was small (0.10 MPa).

Fig. 5.

Pressure–volume traits across species. (a) PV relationships plotted as −1/Ψ (MPa) against 100 − RWC (%); lines show mean segmented linear fits. (b–d) Boxplots of water potential at turgor loss point (Ψ_TLP, MPa), relative water content at turgor loss point (RWC_TLP, %) and osmotic potential at full turgor (Ψ_O, MPa). Different letters indicate Tukey's pairwise differences following one‐way ANOVA (α = 0.05).

A similar result was found for relative water content at turgor loss point (RWC_TLP) which showed significant species differences (F = 109.49, P < 0.001). Cannomois congesta had the highest RWC_TLP, followed by P. repens and then E. monsoniana. The low RWC_TLP of E. monsoniana suggests greater capacity to maintain turgor at lower tissue water contents. The osmotic potential at full turgor (Ψ_O) was also significantly less negative in C. congesta than in P. repens and E. monsoniana (F = 41.98, P < 0.001), further supporting its relatively less negative water potential range.

Consistent differences in P₁₂, P₅₀, P₈₈, and Ψ_TLP resulted in systematic variation in hydraulic safety margins (HSM_ΨTLP), with all three species having positive safety margins, even from P₁₂ (Table 1). C. congesta had the narrowest HSM_ΨTLP, followed by P. repens, and E. monsoniana. The P₁₂ estimates indicated embolism onset at 0.74 ± 0.17 MPa after Ψ_TLP in C. congesta, compared with 0.84 ± 0.24 MPa in P. repens, and 4.32 ± 0.74 MPa in E. monsoniana. Similar trends were observed for P₅₀ and P₈₈.

Table 1.

Hydraulic safety margins relative to the water potential at turgor loss point (Ψ_TLP) for Cannomois congesta (n = 5), Erica monsoniana (n = 6), and Protea repens (n = 6).

parameter	species	HSMΨTLP ± SD
P12	Cannomois congesta	0.74 ± 0.17
	Erica monsoniana	4.32 ± 0.74
	Protea repens	0.84 ± 0.24
P50	Cannomois congesta	1.42 ± 0.09
	Erica monsoniana	5.19 ± 0.76
	Protea repens	1.88 ± 0.33
P88	Cannomois congesta	2.09 ± 0.31
	Erica monsoniana	6.06 ± 0.93
	Protea repens	2.91 ± 0.68

Hydraulic safety margins (HSM_ΨTLP, MPa) were calculated as Ψ_TLP – P_x, where P_x is the xylem water potential at P₁₂, P₅₀, or P₈₈ (12%, 50%, and 88% air discharged, respectively) from the individual vulnerability curves. Values are mean ± standard deviation (SD).

Comparison of vulnerability curves with other methods

Across all three species, the VCs showed broad alignment with previous pneumatic, optical, and centrifuge studies, although the degree of agreement varied among parameters and methods (Fig. 6, Table 2). Across methods, P₅₀ often showed the closest agreement, whereas P₁₂ and P₈₈ appear more variable.

Fig. 6.

Cross‐method comparison of vulnerability curves across studies. Panels show embolism (percent embolized, %) as a function of Ψ (MPa) for each species, including curves from the current study and published estimates (methods and sources in legend). Symbols indicate bootstrapped P₁₂, P₅₀, and P₈₈ with 95% confidence intervals. The vertical dotted line denotes mean Ψ_TLP estimated independently from PV curves.

Table 2.

Vulnerability curve parameters for Cannomois congesta, Erica monsoniana, and Protea repens, estimated using different methods and reported across studies.

species	method	references	P12 [CI]	P50 [CI]	P88 [CI]
Cannomois congesta	Pneumatic	Current study	−2.25 [−2.57, −1.80]	−2.91 [−3.09, −2.73]	−3.56 [−4.03, −3.28]
	Pneumatic – uncorr.	Current study	−1.65 [−2.04, −1.24]	−2.66 [−2.98, −2.33]	−3.68 [−4.07, −3.28]
	Optical	West et al. (2024)	−1.89 [−2.02, −1.75]	−2.29 [−2.49, −2.09]	−2.69 [−2.97, −2.41]
	Pneumatic	West et al. (2024)	−1.39 [−1.43, −1.32]	−2.25 [−2.30, −2.19]	−3.10 [−3.23, −3.00]
Erica monsoniana	Pneumatic	Current study	−5.49 [−7.56, −4.00]	−6.77 [−8.33, −5.10]	−8.05 [−9.53, −5.92]
	Optical	Skelton et al. (2023)	−3.64 [−7.05, −1.66]	−5.74 [−8.36, −3.05]	−7.87 [−9.71, −4.27]
Protea repens	Pneumatic	Current study	−2.38 [−3.36, −0.72]	−3.39 [−4.04, −2.75]	−4.40 [−6.04, −3.43]
	Centrifuge	Jacobsen et al. (2009)	N/A [−0.16, N/A]	−3.56 [−3.90, −3.19]	−5.61 [−5.76, −5.46]
	Centrifuge – corr.	Jacobsen et al. (2009)	−1.72 [−2.22, −1.25]	−4.13 [−4.41, −3.82]	−5.69 [−5.84, −5.55]
	Optical	Skelton et al. (2023)	−3.04 [−3.29, −2.83]	−3.51 [−3.73, −3.28]	−3.98 [−4.18, −3.72]

P₁₂, P₅₀, and P₈₈ (MPa) represent the xylem water potentials associated with 12%, 50%, and 88% embolized, respectively. Values are bootstrapped mean (1,000 iterations) with 95% confidence intervals (CI).

For C. congesta, the current study produced a steeper pneumatic curve with more negative thresholds than those reported by West et al. (2024), particularly after the correction was applied (Fig. 6, Table 2). The correction shifted the fitted curve towards more negative water potentials and increased curve steepness, making P₁₂ and P₅₀ more negative while having little effect on P₈₈. Compared with West et al. (2024), both pneumatic curves yielded more negative P₁₂ and P₅₀ estimates than the previous pneumatic method, while the optical method remained intermediate for P₁₂ and less negative for P₅₀ and P₈₈ (Fig. 6, Table 2).

For E. monsoniana, there was no clear difference in P₅₀ between the current study and the optical method of Skelton et al. (2023), although the variation was notably larger for E. monsoniana compared with C. congesta or P. repens (Fig. 6, Table 2). Both optical and pneumatic methods have reported high variability in VCs for E. monsoniana, potentially reflecting either greater biological variation in embolism resistance or greater error under more negative water potentials.

For P. repens, the parameter estimates from the pneumatic method of the current study were comparable with those from the optical method of Skelton et al. (2023). This was particularly the case for the P₅₀, with more variation in P₁₂ and P₈₈ (Fig. 6, Table 2). The curves derived from the centrifuge method of Jacobsen et al. (2009) differed in shape from those of the optical and pneumatic methods. The resulting differences in derived parameter estimates were influenced by whether the centrifuge curves were fatigue‐corrected.

DISCUSSION

This study demonstrates the pneumatic method's potential to construct reasonable xylem vulnerability to embolism curves across three contrasting growth forms. Across all three species, the pneumatron detected clear, progressively increasing AD during desiccation, while water potential decreased predictably. As a result, paired PAD and Ψ measurements produced physiologically consistent VCs, supporting the method as a practical tool for assessing embolism resistance in small and large woody shrubs, as well as in reed‐like graminoids, where early gas discharge from hydrated parenchymatous tissue may obscure embolism‐related gas discharge. Further, safety margins between Ψ_TLP and the onset of embolism recorded using the pneumatic method were consistent with the hypothesis that species should display positive safety margins. Such findings support the applicability of the pneumatic method for generating VCs across diverse growth forms and indicate broader relevance for shrub‐dominated systems, including Mediterranean‐type ecosystems, where comparable functional diversity occurs.

Several outcomes emerged from this study. PAD–Ψ relationships varied by growth form and revealed clear differences in embolism resistance. Cannomois congesta exhibited early embolism onset and rapid progression across a narrow Ψ range, indicating a steep transition from functional to dysfunctional xylem. Erica monsoniana showed later onset and a more gradual progression with greater variation, implying a wider operational margin as Ψ declines. While Protea repens was intermediate. Pressure–volume traits reflected this ranking (C. congesta < P. repens < E. monsoniana), where turgor loss point was least negative in C. congesta and the most negative in E. monsoniana, with analogous differences in relative water content at turgor loss point. Consequently, hydraulic safety margins appeared narrowest in C. congesta, widest in E. monsoniana and intermediate in P. repens.

Together, these findings are coherent with the ericoid, proteoid, and restioid functional type responses characterized by previous studies (Miller et al. ¹⁹⁸³; Higgins et al. ¹⁹⁸⁷; Jacobsen et al. ²⁰⁰⁹; West et al. ²⁰¹²; Skelton et al. ^{2015, 2023}). Whereby, deep‐rooted proteoids tap into more stable soil moisture, regulate stomata, sustain higher water potentials, and operate within hydraulic safety margins. Shallow‐rooted ericoids are tightly coupled to near‐surface soil moisture, tolerate lower water potentials, and exploit rainfall pulses. Shallow‐rooted restioids employ conservative water use, utilize atmospheric moisture, partially decoupling from near‐surface soil moisture, and maintain higher water potentials. Under increasing climatic water stress, these differences in VCs imply predictable shifts in risk and competitive balance because growth forms operating closer to P₁₂ and P₅₀ should respond most strongly to incremental drought intensification, whereas wider safety margins may buffer hydraulic failure until prolonged drought imposes sustained stomatal closure and carbon limitation.

Between methods there appeared modest divergence in the established curves, particularly in overall shape. The pneumatic and optical methods both provide high‐temporal sampling and produce high‐density datasets. However, the optical curves often appear steeper and more strictly sigmoidal, whereas pneumatic curves often show early‐stage AD and a late‐stage taper at lower Ψ, which together influence the overall shape unless corrected for. By contrast, centrifuge curves arise from a completely different experimental framework – short, flushed segments under imposed xylem tension with potential fatigue correction – providing lower data density and an entirely different curve shape. As a result, discrepancies tended to concentrate at the early‐ and late‐stages, with greater divergence in parameters that depend on curve shape (i.e. P₁₂ and P₈₈).

Across methods, P₅₀ was the most comparable parameter, whereas larger differences were observed for P₁₂ and P₈₈, consistent with previous multi‐method comparisons (Pereira et al. ^{2016, 2020}; Zhang et al. ²⁰¹⁸; Sergent et al. ²⁰²⁰; Paligi et al. ²⁰²³; West et al. ²⁰²⁴). In the present study, applying a correction for potential early‐stage artefacts to improve P₁₂ estimates had no discernible effect in the woody shrub species, but shifted inferred embolism resistance in the herbaceous species, particularly at the onset of the vulnerability curve. For C. congesta, and presumably other morphologically similar species with an atactostele, the presence of hydrated parenchymatous tissue at the cut end may contribute to early gas discharge not associated with embolism, although this remains inferred. It therefore seems prudent to assess the impact of this early gas discharge on the vulnerability curve and correct appropriately (see Supplementary Material). Resolving uncertainties in the early and late stages of pneumatic curves – whether associated with residual sap, surrounding hydrated ground tissue, or identification of maximum air discharge plateau – remains a priority for robust estimation of P₁₂ and P₈₈ (Brum et al. ²⁰²³; Miranda et al. ²⁰²⁴). This matters because P₁₂ and P₈₈ determine inferred hydraulic safety margins, curve slope, and the rate of progression towards hydraulic failure, all of which are key components of model parameterization of drought responses and interpretation of mortality risk under extreme events. Reducing the quantification of drought resilience to P₅₀ alone therefore limits ecological inference.

Reliable, high‐frequency, non‐invasive water potential estimation remains a practical limitation. Pressure chamber measurements require destructive subsampling, necessitating interpolation between time points, and although wounds were sealed promptly with adhesive following best practice, local hydraulic effects cannot be excluded (Pereira et al. ²⁰¹⁶). For all species, water potential during the final phase required extrapolation, which increases uncertainty around P₈₈. A pragmatic approach, offered here, is to pair repeated RWC measurements on PAD culms with individual Ψ–RWC relationships, enabling high‐temporal and late‐stage Ψ estimates without repeated cutting. This reduces disturbance, extends monitoring into late dry‐down and may improve parameter inference.

Extending VCs to contrasting growth forms enables broader ecological and comparative questions. Growth forms analogous to proteoids, ericoids, and restioids are characteristic of Mediterranean‐type ecosystems, shrub‐dominated vegetation, and drought‐prone regions more broadly. Embolism resistance is rarely quantified, and the pneumatic method provides a practical technique. Expanding such datasets across hydrological gradients and species diversity would allow explicit tests of how rooting depth, stomatal strategy and water‐relations traits co‐determine safety margins and drought sensitivity, and whether these relationships hold across environments. In the CFR, where the landscape varies greatly over short distances, such datasets would also enable stronger prediction of where hydraulic thresholds are likely to be exceeded under future drought regimes and whether hydrological refugia could buffer populations or drive trait and lineage persistence.

In conclusion, this study (i) demonstrates the utility of the pneumatic method for generating xylem vulnerability to embolism curves across three contrasting growth forms; and (ii) introduces a practical approach for further estimating water potential by pairing repeat relative water content measurements from pneumatron samples with individual Ψ–RWC relationships. More broadly, it extends comparative evaluation of the pneumatic method across small and large woody shrubs as well as graminoids, highlighting its relevance for species‐rich, shrub‐dominated systems more generally.

AUTHOR CONTRIBUTIONS

MMA, RPS, and AGW designed the study. MMA performed the experiments, collected and analysed the data. MMA, RPS, and AGW prepared the manuscript.

Supporting information

Fig. S1. Study species, including: (a) Cannomois congesta (Restionaceae); (b) Erica monsoniana (Ericaceae); (c) Protea repens (Proteaceae).

Fig. S2. Effects of alternative early‐stage correction thresholds on vulnerability curve parameters (P₁₂, P₅₀, and P₈₈) for Cannomois congesta, Erica monsoniana and Protea repens. Percent air discharged (PAD) was recalculated using the full air discharge (AD) without correction (‘no correction’), or after excluding AD values recorded at Ψ ≥ −0.5 MPa, Ψ ≥ −1.0 MPa, or Ψ ≥ Ψ_TLP (water potential at turgor loss point). Letters indicate significant differences among correction treatments within a species and parameter.

Fig. S3. Individual vulnerability curves for Cannomois congesta culms, indicating percent air discharge (PAD) as a function of water potential (Ψ), calculated from the air discharge (AD) without correction and after excluding AD measurements recorded at Ψ ≥ Ψ_TLP (Ψ at turgor loss point). Vertical dashed lines indicate mean Ψ_TLP generated independently.

Fig. S4. Individual air discharge curves over the course of dehydration.

Table S1. Segmented linear regression for water potential (Ψ, MPa) over time (hours) during bench dehydration. Estimated parameters from fits for each species. Slope₁ and Slope₂ (MPa h⁻¹) are the rates of change in Ψ before and after the breakpoint, respectively. Values are means ± standard deviation (SD).

Fig. S5. Vulnerability curve construction for Protea repens. (a) Percentage air discharge (PAD) through time. (b) Xylem water potential (Ψ) measured periodically; lines are individual segmented linear fits used to estimate Ψ over time. (c) Resulting PAD–Ψ relationships for each individual.

Fig. S6. Vulnerability curve construction for Erica monsoniana. (a) Percentage air discharge (PAD) through time. (b) Xylem water potential (Ψ) measured periodically; lines are individual segmented linear fits used to estimate Ψ over time. (c) Resulting PAD–Ψ relationships for each individual.

Table S2. Segmented linear regression parameters describing (i) water potential (Ψ, MPa) as a function of relative water content (RWC, %) and (ii) relative water content (RWC, %) as a function of time (h) in Cannomois congesta. Two‐segment piecewise linear models were fitted for each relationship. For Ψ–RWC fits, the breakpoint is reported in 100 – RWC (%) and slopes are in MPa per %; for RWC–time fits, the breakpoint is in hours and slopes are in % h⁻¹. R² is the coefficient of determination.

Fig. S7. Vulnerability curve construction for Cannomois congesta (Top row) percentage air discharge (PAD) through time (Middle row) xylem water potential (Ψ) through time for paired bench‐culms (measured) and for pneumatron‐treated culms with Ψ inferred from relative water content (AD‐culms_RWC) or from the pressure–volume curve (AD‐culms_PVC); lines are segmented linear fits used to estimate Ψ over time (Bottom row) PAD–Ψ relationships derived from each method within individuals.

Table S3. Vulnerability curve metrics by Ψ–estimation method for Cannomois congesta. Mean ± standard deviation (SD) of P₁₂, P₅₀, and P₈₈ (MPa) derived from PAD–Ψ fits for bench‐culms, AD‐culms_RWC and AD‐culms_PVC (Ψ).

Table S4. Test summary of method effects on embolism thresholds (α = 0.05). Results from linear mixed‐effects models with method as a fixed effect, fitted separately for P₁₂, P₅₀, and P₈₈ (MPa).

DATA AVAILABILITY STATEMENT

The data of this study are available from the corresponding author upon reasonable request.