Temperature governs the relative contributions of cuticle and stomata to leaf minimum conductance

Summary

• During periods of stomatal closure, such as drought, plant leaves continue to lose water at a rate determined by the minimum leaf conductance, g_min. Although g_min varies with temperature, less is known about what drives this variation, including how the pathways of water loss (cuticle or stomata) vary with temperature.
• We used gas exchange and bench drying methods to measure g_min and cuticular conductance, g_cw, across a wide temperature range (20–50°C) in 11 broadleaf species. Vapour pressure deficit (VPD) covaried with temperature from 0.83 to 10.7 kPa.
• The dominant pathway of water loss for g_min shifted from stomatal transpiration towards cuticular transpiration as temperature increased. Leaf traits had variable, temperature-dependent relationships with g_min and g_cw, with trait–conductance relationships being generally stronger at higher temperatures. Cuticular thickness varied inversely with high-temperature g_cw. Simulation results showed that g_cw may impact photosynthetic capacity estimates, particularly in species with low stomatal conductance.
• The pathways of water loss in leaves during times of stomatal closure depend strongly on temperature. This effect may have large implications for landscape-scale water balance modelling and improving gas exchange measurements. We propose variation in VPD as a potential contributing factor in g_min and g_cw variation among studies.

Introduction

Climate change is increasing the frequency and severity of hot drought events in many parts of the world, with further increases forecast for the coming century (Intergovernmental Panel on Climate Change (IPCC), 2021). During periods of water stress, plants typically reduce their stomatal aperture, restricting both water loss and carbon substrate availability for photosynthesis (Cowan & Farquhar, 1977). However, even with stomata maximally closed, leaves still lose water at a rate described by the leaf minimum conductance to water vapour, g_min (mol m⁻² s⁻¹; Table 1) (Duursma et al., 2019). While g_min is typically more than an order of magnitude less than stomatal conductance (g_sw; mol m⁻² s⁻¹) during more favourable conditions (Slot et al., 2021), plants may lose substantial amounts of water even under maximal stomatal closure due to high evaporative demand (Vicente-Serrano et al., 2020). Improved understanding of g_min is necessary, as transpiration during hot drought events can have substantial effects on plant mortality and landscape-scale water balance (Park Williams et al., 2013; Rogers et al., 2017; Hammond & Adams, 2019).

Table 1. List of symbols.

Symbol	Definition	Units
A	Net assimilation rate	μmol m^−2 s^−1
al	One-sided (projected) leaf area	m^2
ca	Ambient air CO2 concentration	μmol mol^−1
ci	Leaf intercellular CO2 concentration	μmol mol^−1
E	Transpiration rate	mol m^−2 s^−1
gbw	Leaf boundary layer conductance to water vapour	mol m^−2 s^−1
gcw	Leaf cuticular conductance to water vapour	mol m^−2 s^−1
gmin	Leaf minimum conductance to water vapour	mol m^−2 s^−1
gsw	Stomatal conductance to water vapour	mol m^−2 s^−1
gsw,min	Minimum stomatal conductance	mol m^−2 s^−1
$g_{\mathrm{sw}}^0$	Initial value of $g_{\mathrm{sw}}^{\prime }$	mol m^−2 s^−1
$g_{\mathrm{sw}}^{\prime }$	Stomatal conductance in excess of gsw,min	mol m^−2 s^−1
gtw	Leaf total conductance to water vapour	mol m^−2 s^−1
h	Elevation	m
Jmax	Maximum rate of RuBP regeneration	μmol m^−2 s^−1
LDMC	Leaf dry matter content	mg g^−1
LMA	Leaf mass per unit area	g m^−2
lsp	Stomatal pore length	μm
mdry	Leaf dry mass	g
mwet	Leaf wet (fresh) mass	g
M	Leaf mass	g
M0	Initial leaf mass	g
Patm	Atmospheric pressure	kPa
SVP	Saturation vapour pressure	kPa
t	Time	s
Vcmax	Maximum rate of Rubisco carboxylation	μmol m^−2 s^−1
VPD	Vapour pressure deficit	kPa
wa	Water vapour concentration in ambient air	mmol mol^−1
wi	Water vapour concentration in leaf intercellular air spaces	mmol mol^−1
wgs	Guard cell width	μm
xcut	Cuticle thickness	μm
xepi	Epidermis thickness	μm
xl	Leaf thickness	μm
xpal	Palisade mesophyll thickness	μm
xspg	Spongy mesophyll thickness	μm
β	Conductance conversion factor	g m mol^−1
κ	Ratio of adaxial to abaxial ci	unitless
μ	Molar mass of water	g mol^−1
ρs	Stomatal density	m^−2
τ	Stomatal closure time constant	s

Even under complete stomatal closure conditions, water may still escape the leaf via the cuticle (i.e. cuticular conductance, g_cw; mol m⁻² s⁻¹) or via small apertures in incompletely closed stomata (i.e. minimum stomatal conductance, g_sw,min; mol m⁻² s⁻¹; Fig. 1) (Duursma et al., 2019). Careful measurements partitioning g_min into stomatal and cuticular components have yielded widely conflicting conclusions about the relative contributions of these pathways, from all water lost via the cuticle (Slot et al., 2021), to highly variable species-specific ratios (6–44% water lost via the stomata; Machado et al., 2021), to a majority of water lost via the stomata in one species (Márquez et al., 2022).

Fig. 1

Leaf water fluxes during times of high and low water availability. During times of water availability and photosynthetic activity, stomata open to allow gas exchange with the atmosphere (upper image). In this case, leaf total conductance to water vapour g_tw is the sum of the stomatal conductance g_sw and the cuticular conductance g_cw (ignoring the boundary layer). During times of low water availability or low photosynthetic activity, stomata close to restrict water loss (lower image). With stomata fully closed, g_tw is at its minimum and is referred to as leaf minimum conductance g_min, which is a sum of g_cw and a minimum stomatal contribution g_sw,min. All conductances are expressed in units of mmol m⁻² s⁻¹.

As droughts often coincide with high temperatures, it is vital to understand the temperature dependence of g_min, as well as that of its underlying component pathways. The short-term temperature response of g_min frequently exhibits a monotonically increasing relationship with temperature over the range of c. 25–50°C, often with a ‘biphasic’ relationship (e.g. Schuster et al., 2016; Bueno et al., 2019; Billon et al., 2020; Wang et al., 2024; n.b., in nearly every study which has investigated the temperature response of g_min or g_cw, vapour pressure deficit (VPD) covaries with temperature; see the Discussion section). However, invariant relationships, or even declining g_min with temperature, have also been observed (e.g. Bueno et al., 2019; Slot et al., 2021). The short-term temperature dependence of g_cw also generally shows monotonically increasing relationships with temperature (e.g. Schreiber, 2001; Riederer, 2006), but this is typically measured on enzymatically isolated cuticles or leaf discs, rarely in vivo (cf., Márquez et al., 2021). Crucially, the short-term temperature dependencies of both g_min and g_cw have never been measured concurrently in vivo; thus, little is known about how these different water pathways respond to temperature.

Understanding how the different pathways of water loss contribute to the observed temperature sensitivity of g_min is necessary for informing theory on alternative plant water use strategies (Blonder et al., 2023). Recently, researchers have reported stomatal ‘decoupling’ at high leaf temperatures (Aparecido et al., 2020; Krich et al., 2022; Marchin et al., 2023). Stomatal decoupling occurs when stomatal conductance increases, or fails to decrease, despite declining assimilation rates at high leaf temperatures, contradicting theory based on optimality principles (Cowan & Farquhar, 1977; Medlyn et al., 2011). It is not known whether this decoupling is an adaptive response of the plant intended to cool the leaves (Michaletz et al., 2016; Garen et al., 2023) or a passive ‘failure’ mechanism (Slot et al., 2021; Blonder et al., 2023). Improved understanding of the temperature response of g_min and its component processes will help in dissecting the mechanism underlying this high-temperature water use behaviour.

Furthermore, understanding the temperature response of g_cw is necessary for improving gas exchange measurements. Recently, Márquez et al. (2021) proposed a new model for leaf gas exchange that accounts for g_cw (the Marquez-Stuart-Williams-Farquhar or MSF model). Previous gas exchange models calculated leaf intercellular CO₂ concentration c_i based on the ratio of diffusivities of CO₂ and H₂O in air (c. 1.6), under the assumption that all transpiration occurs via the stomata (von Caemmerer & Farquhar, 1981). However, while CO₂ and H₂O both move readily through the stomata, the cuticle is nearly impermeable to CO₂ (Boyer, 2015). Given that some water vapour escapes via the cuticle, a model that attributes all transpiration to the stomata will overestimate c_i (Tominaga et al., 2018). Photosynthetic capacity metrics such as V_cmax (maximum rate of Rubisco carboxylation) and J_max (maximum rate of RuBP regeneration) are estimated using the relationship between assimilation rate A and c_i (i.e. A–c_i curves) (Farquhar et al., 1980; Sharkey et al., 2007), hence errors in c_i may affect estimates of V_cmax and J_max, with potential consequences for process-based modelling frameworks that employ these metrics (Stinziano et al., 2019; Hussain et al., 2024). However, the effects of g_cw temperature dependence on metrics of photosynthetic capacity have not previously been investigated. To quantifying the magnitude of error and improve estimates of photosynthetic capacity, it is necessary to know whether and when to account for g_cw and its temperature dependence.

Here, we address these questions by measuring the short-term temperature responses of g_min and g_cw. Our study has three goals as follows:

1. to describe the temperature dependencies of g_min and g_cw, and to describe how the pathways of leaf water loss vary with temperature;
2. to test whether g_min and g_cw depend on anatomical, structural, and morphological leaf traits; and
3. to test whether g_cw and its temperature dependence cause errors in measurements of photosynthetic capacity.

We demonstrate that the contributions of stomata and cuticle to g_min vary with temperature, with the cuticular water loss pathway dominating at higher temperatures. We find temperature-dependent relationships between leaf conductance and leaf traits, and we further show that photosynthetic capacity metrics depend on g_cw, particularly at low stomatal conductance.

Materials and Methods

Plant material and sampling methods

We collected branch cuttings of sun-exposed leaves from The University of British Columbia Botanical Garden (49.26°N, 123.25°W; 85 m elevation) between 14 June and 15 July 2022 at 08:00–09:00 h. Cuttings were taken from 11 species of broadleaf trees representing a wide range of leaf traits (Table 2). All species were hypostomatous. Cuttings varied in length from c. 1–4 m, as some species, especially Acer spp., have exceptionally long vessels. To avoid embolism, the cut branch ends were submerged, recut underwater, and immediately transported to the laboratory for measurement.

Table 2. List of species.

Species	Abbreviation	Family	Deciduousness
Arbutus menziesii Pursh	arme	Ericaceae	Evergreen
Cladrastis kentukea (Dum. Cours.) Rudd	clke	Fabaceae	Deciduous
Acer macrophyllum Pursh	acma	Sapindaceae	Deciduous
Umbellularia californica (Hook. & Arn.) Nutt.	umca	Lauraceae	Evergreen
Magnolia acuminata (L.) L.	maac	Magnoliaceae	Deciduous
Alnus rubra Bong.	alru	Betulaceae	Deciduous
Acer circinatum Pursh	acci	Sapindaceae	Deciduous
Malus fusca (Raf.) C.K. Schneid.	mafu	Rosaceae	Deciduous
Quercus garryana Douglas ex Hook.	quga	Fagaceae	Deciduous
Betula papyrifera Marshall	bepa	Betulaceae	Deciduous
Populus tremuloides Michx.	potr	Salicaceae	Deciduous

Cuticular conductance

Plant material was placed in a growth chamber (E15; Conviron, Winnipeg, MB, Canada) along with two LI-6800 Portable Photosynthesis Systems (Li-Cor Biosciences Inc., Lincoln, NE, USA; firmware v.2.0.04).

We measured g_cw following the method of Márquez et al. (2021). The two LI-6800 s were configured such that one instrument recorded gas exchange quantities (e.g. A, g_sw, c_i) from the abaxial surface only, and the other instrument recorded gas exchange quantities from the adaxial surface only (Supporting Information Fig. S1). Leaves completely filled the LI-6800 aperture, preventing mixing of adaxial and abaxial fluxes. Gas exchange quantities were then used in the model described by Márquez et al. (2021) to estimate g_cw, assuming that adaxial and abaxial g_cw are equal (cf., Márquez et al., 2022). Additionally, because all species were hypostomatous, we estimated g_cw from the adaxial gas exchange data by interpreting reported g_sw as g_cw using the default calculation in the LI-6800 for boundary layer conductance. Both the Márquez method and the adaxial flux method measure g_cw on a single-sided leaf area basis and are therefore directly comparable. These two measures of g_cw compared favourably with each other (Fig. S2), but the Márquez method gave a higher rate of measurements with g_cw < 0. This may be due to error in our assumption that adaxial and abaxial g_cw are equal, error in our assumption that c_i at the adaxial and abaxial evaporative sites are equal (i.e. κ = 1) given that the leaves are hypostomatous, or error in the assumed ratio of diffusivities of H₂O and CO₂ in the cuticle (i.e. 20). Thus, we used the adaxial g_cw measurements for analysis.

During gas exchange measurements, environmental controls on both LI-6800 s were identical, except for light level which was 0 for the abaxial instrument. Airflow rate was 300 μmol s⁻¹, reference CO₂ concentration was 420 ppm, and fan speed was 10 000 revolutions per minute. Relative humidity was kept close to ambient humidity in the chamber (measured with an RH820; Omega Engineering Inc., Norwalk, CT, USA), up to a maximum of c. 50% to prevent condensation. Photosynthetic photon flux density (PPFD) was set to 1000 μmol m⁻² s⁻¹ in the adaxial instrument (Márquez et al., 2021). LI-6800 heat exchanger temperature and growth chamber temperature were controlled together to increase the range of achievable temperatures and reduce measurement error (Garen et al., 2022). The temperature of the growth chamber and LI-6800 s were set to one of 20, 30, 40, or 50°C (Table S1), measured in a random order for each species. Five leaves were measured for each species at each temperature. All leaves of a given species were taken from the same individual. We were unable to control humidity in the growth chambers during measurement, resulting in increasing VPD with temperature. Mean VPD at the above temperatures was 0.83, 2.9, 5.8, and 10.7 kPa, respectively (Table S1; see the Discussion section).

Leaf minimum conductance

After gas exchange measurements, we performed paired leaf-level measurements of g_min using the ‘bench drying’ method (Sack & Scoffoni, 2010; Slot et al., 2021). Leaves were removed from their stems, and petioles were coated in paraffin wax to prevent water loss. Leaf area was measured using a LI-3100 leaf area metre (Li-Cor Biosciences Inc., Lincoln, NE, USA). Leaves were weighed and placed into a growth chamber that was kept completely dark to prevent stomatal opening. Leaves were suspended from transparent tape, and an oscillating fan was used to disrupt leaf boundary layers. Leaves were left in the growth chamber for 6–12 min intervals; shorter intervals were used at higher temperature and VPD conditions due to more rapid water loss. After each interval, leaves were removed from the growth chamber and reweighed. Leaves were stored in plastic bags with human breath added to prevent water loss when not being dried or weighed (Sack & Scoffoni, 2010). During the drying process, temperature and relative humidity in the chamber were measured using an Omega RH820 (Table S1). This process was repeated up to a maximum of 13 measurements. Final leaf area was again measured with the LI-3100.

The mass loss curves generated by this method were used to estimate g_min. Once the leaves are removed from their stems, the stomata begin to close resulting in a period of exponentially decreasing conductance, followed by a constant g_min. g_min can be estimated by manually removing nonlinear portions of the mass loss curve and fitting a linear model to the remainder (Sack & Scoffoni, 2010). However, this requires manual processing and subjective judgement to distinguish linear portions of the curve. To alleviate these problems, we derived the following model (Notes S1),

$$M=\mathit{\beta \tau }{g}_{\mathrm{sw}}^0\left(e^{-\frac{t}{\tau }}-1\right)-\beta g_{\min }t+M_0$$ (Eqn 1)

where M is the leaf mass (g) at time t (s), τ is a time constant describing the rate of stomatal closure (s), $g_{\mathrm{sw}}^0$ is a stomatal contribution to conductance at t = 0 (mmol m⁻² s⁻¹), and M₀ is leaf mass at t = 0 (g). The first term in this expression describes the exponential decrease in stomatal conductance after leaf excision, the second term describes the linear decline in leaf mass due to minimum conductance, and the final term represents the initial leaf mass. The parameter β is defined as

$$\begin{array}{c}\beta =2a_{\mathrm{l}}\mu \left(\frac{\mathrm{VPD}}{P_{\mathrm{atm}}}\right)\end{array}$$ (Eqn 2)

where a_l is the projected leaf area (m²), μ is the molar mass of water (18.02 g mol⁻¹), and P_atm is atmospheric pressure (kPa). Eqn 1 may be fit to dry-down curves including both an initial exponential portion and a final linear portion, and g_min may be estimated directly from the model fit without the need to manually truncate data (Fig. S3). We fit Eqn 1 to our dry-down curves with the nls.multstart function (Padfield et al., 2021). When using Eqn 1, g_min was constrained such that $g_{\min }\ge -\frac{1}{\beta }\frac{\mathit{dM}}{\mathit{dt}}$ at the time of last measurement, as occasional overfitting gave unrealistic values of g_min.

Anatomical and structural traits

We selected two leaves per species per treatment analogous to the leaves used for conductance measurements. We performed stomatal peels using clear nail polish on the adaxial and abaxial surfaces (Hilu & Randall, 1984). The stomatal peels were imaged using an optical microscope at ×400 magnification. Adaxial peels were examined to ensure that all species were hypostomatous. A photograph was taken of the stomata on each abaxial peel slide, avoiding regions with large veins. The images were processed manually using ImageJ to measure stomatal density (ρ_s), stomatal pore length (l_sp), and guard cell width (w_gs), making five measurements of the latter two quantities per slide. Next, we cross-sectioned the leaves with a fresh razor blade. The sections were imaged using an optical microscope at ×400 magnification. These images were processed manually using ImageJ to measure cuticle thickness (x_cut), epidermal thickness (x_epi), palisade thickness (x_pal), spongy mesophyll thickness (x_spg), and total leaf thickness (x_l), making three measurements of each quantity per slide. x_cut and x_epi were taken as the average of adaxial and abaxial values (three measurements per side).

For structural traits, we used the same leaves that were used for conductance measurements. After g_min measurements, leaves were dried at 60°C for a minimum of 48 h (Pérez-Harguindeguy et al., 2016). Dry mass was then measured with a precision mass balance. Leaf mass per unit area (LMA) was then calculated as the dry mass divided by the fresh leaf area, measured previously during bench drying. Leaf dry matter content (LDMC) was calculated as the dry mass divided by the fresh mass.

Effects of cuticular conductance on metrics of photosynthetic capacity

We performed simulations to test the effect of g_cw and its temperature dependence on metrics of photosynthetic capacity. We obtained a published database of A–c_i curves measured across wide temperature ranges (Kumarathunge et al., 2019). We began by eliminating observations with unrealistically high or low values of c_i (c_i < 0 or c_i > 2000 μmol mol⁻¹). The MSF model requires inputs of transpiration rate (E), VPD, ambient CO₂ concentration (c_a), boundary layer conductance (g_bw), g_cw, P_atm, and A (eqn 14 in Márquez et al., 2021); thus, datasets which lacked one or more of E, VPD, c_a, or A were excluded. We assumed g_bw = 2.93 mol m⁻² s⁻¹, based on the default calculation used in the LI-6800 (Li-Cor Biosciences Inc., 2019). While this assumption may be imperfect, a sensitivity analysis showed that varying g_bw by up to 50% resulted in less than 1% change in c_i (Fig. S4). We then obtained location information from each of the publications cited in the A–c_i database. Elevation of measurement site, h (m), was taken from the publication if provided, otherwise h was estimated using Google Earth. We computed atmospheric pressure as (Campbell & Norman, 1998)

$$\begin{array}{c}{P}_{\mathrm{atm}}=101.3\exp \left(-\frac{h}{8200}\right)\end{array}$$ (Eqn 3)

As we only measured g_cw of broadleaf angiosperms, we retained only angiosperms in the database for greater consistency with our data. The final dataset contained 1683 individual A–c_i curves representing 69 species, with leaf temperatures spanning 5–50°C.

We first recalculated c_i using the MSF model with g_cw = 0 to ensure that any differences detected in subsequent tests were due to the effect of g_cw alone, and this had a negligible effect on c_i values (Fig. S5). We then applied two models of g_cw, (1) a constant g_cw = 3.2 mmol m⁻² s⁻¹ (mean for Magnolia acuminata at 20°C), and (2) a temperature-dependent function, g_cw = 0.0105T²–0.581T + 10.58 (best fit model for M. acuminata; Table S2). These g_cw functions were selected to illustrate the possible extent of effect on V_cmax and J_max because g_cw of M. acuminata at 20°C is close to the average at 20°C among species measured here (mean 3.3 mmol m⁻² s⁻¹), but it also exhibits notable temperature dependence. We recalculated all c_i values using the MSF model under these two g_cw scenarios. We fit the Farquhar-von Caemmerer-Berry model to each A–c_i curve under each g_cw scenario using the fitaci function from the plantecophys library (Duursma, 2015) to estimate V_cmax and J_max (Farquhar et al., 1980). We computed percent change in V_cmax and J_max and estimated quantiles in a rolling 5-degree window across the temperature range.

Statistical analysis

To model the temperature dependence of g_min and g_cw, we fit second-order polynomials to each conductance metric as a function of temperature (n = 20 leaves per species and conductance metric combination). We additionally tried linear and constant models, but the Akaike information criterion (Akaike, 1998) consistently preferred the quadratic model in all but one case.

To investigate the dependence of g_min and g_cw on leaf traits, we computed species-level means of trait values and conductances, and we fit bivariate linear models using ordinary least squares regression. We fit linear models to g_min and g_cw at each temperature using separate linear regressions for each trait. Finally, we performed a principal component analysis (PCA) using the prcomp function on trait and conductance data after rescaling and centering.

All analyses were carried out in R v. 4.2.1 (R Core Team, 2019).

Results

Temperature response of g_cw and g_min

The responses of g_cw and g_min to increasing temperature and VPD are shown in Fig. 2. g_cw showed substantial variability among species and with temperature. At 20°C, species mean g_cw ranged from 0.9 ± 0.1 to 7.5 ± 1.2 mmol m⁻² s⁻¹ (mean ± SE; Arbutus menziesii and Quercus garryana, respectively), whereas at 50°C, g_cw ranged from 1.4 ± 0.1 to 8.5 ± 1.0 mmol m⁻² s⁻¹ (Umbellularia californica and Cladrastis kentuckea, respectively). The shape of the temperature response relationship was also variable among species. Some species showed a generally increasing relationship to temperature (M. acuminata and Malus fusca), while others exhibited U-shaped relationships (Acer circinatum, Acer macrophyllum, C. kentuckea, and Q. garryana), and others were roughly invariant with temperature (Alnus rubra, A. menziesii, Betula papyrifera, Populus tremuloides, and U. californica). Leaf minimum conductance was also variable among species and with temperature. At 20°C, g_min ranged from 1.0 ± 0.2 to 17.8 ± 1.0 mmol m⁻² s⁻¹ (U. californica and P. tremuloides, respectively). Variability in g_min diminished substantially at 50°C, ranging from 2.6 ± 0.02 to 7.8 ± 0.4 mmol m⁻² s⁻¹ (U. californica and C. kentuckea, respectively). In most species, g_min generally decreased or had a U-shaped response to temperature. Umbellularia californica had a roughly invariant g_min.

Fig. 2

Temperature responses of cuticular conductance, g_cw, and leaf minimum conductance, g_min, by species. Panels show temperature dependence of g_cw (light green) and g_min (dark green) for each species. Points indicate mean conductance (n = 5), and error bars show SE of the mean. Solid coloured lines show best fit temperature response models for g_min and g_cw. Dashed grey line shows an ordinary least squares regression fit to the ratio of g_cw : g_min as a function of temperature. A value of 0 indicates all water is passing through the stomata, and a value of 1 indicates all water is passing through the cuticle; some values greater than 1 are observed due to measurement uncertainty. Horizontal dotted lines indicate g_cw : g_min = 0 and 1.

The ratio of g_cw : g_min quantifies the proportion of transpiration occurring via the stomata and the cuticle (Fig. 2). We found variability in g_cw : g_min among species; however, most species showed a significant positive trend with temperature (P < 0.05). This indicates that as leaf temperature and VPD increases, proportionally more water is lost via the cuticle, rather than the stomata. The exceptions were A. menziesii, in which a g_cw : g_min was invariant (P = 0.07), and U. californica in which g_cw : g_min decreased with temperature (P = 0.0008). Averaging across all species, we found that the proportion of water lost via the cuticle increased from c. 50% to 100% between 20°C and 50°C (Fig. S6). At high temperatures, some species exhibited g_cw : g_min > 1, likely due to measurement uncertainty from the differing methods used to measure g_min and g_cw, as bench drying may somewhat underestimate g_min due to undersaturation in the intercellular airspaces (Tominaga et al., 2020).

Relationship of g_cw and g_min to leaf traits

The relationship between leaf traits and conductances was investigated with PCA (Fig. 3), revealing a high degree of coordination (PC1 explains 51% of variance at 20°C and 56% at 50°C). Most anatomical and structural traits loaded strongly along PC1 at both temperatures. g_cw showed modest loading along PC1 at 20°C, suggesting some amount of coordination with traits, whereas g_min showed very little loading on PC1. g_cw and g_min both loaded strongly along PC3 at 20°C (Fig. S7). At 50°C, g_cw and g_min both loaded more strongly along PC1, suggesting increased coordination between high-temperature conductance values and leaf traits. However, g_min still loaded strongly along the orthogonal axis PC3 at this temperature (Fig. S7).

Fig. 3

Principal components analysis (PCA) relating cuticular conductance and leaf minimum conductance to traits. Panels show PCAs performed with conductances measured at (a) 20°C and (b) 50°C. Points show individual species, with circles indicating deciduous and triangles indicating evergreen species. Arrows show variable loadings, green arrows indicate conductances, yellow arrows indicate stomatal traits, blue arrows indicate leaf thickness metrics, and red arrows indicate structural traits. a_l, leaf area; g_cw, cuticular conductance; g_min, minimum conductance; LDMC, leaf dry matter content; LMA, leaf mass per unit area; l_sp, stomatal pore length; w_gs, guard cell width; x_cut, cuticle thickness; x_epi, epidermis thickness; x_l, leaf thickness; x_pal, palisade thickness; x_spg, spongy mesophyll thickness; ρ_s stomatal density.

We then investigated bivariate relationships between leaf traits and conductance across temperatures (Fig. 4; Tables S3, S4). Most leaf traits exhibited weak to modest relationships with g_cw which became stronger at higher temperatures (Fig. 4a; Table S3). For example, cuticular thickness x_cut exhibited a weak, nonsignificant negative trend with g_cw when measured at 20°C (r² = 0.11, P = 0.28; Fig. 5a), but exhibited a much stronger, significant negative correlation at 50°C (r² = 0.46, P = 0.01; Fig. 5b). LMA showed a similar relationship with g_cw, with a nonsignificant negative trend observed at 20°C (r² = 0.12, P = 0.28; Fig. 5c), and a stronger, significant negative correlation at 50°C (r² = 0.48, P = 0.01; Fig. 5d).

Fig. 4

Leaf traits explain differing amounts of variation in conductance as a function of temperature. Panels show coefficient of determination r² for regressions between leaf traits and (a) cuticular conductance g_cw or (b) leaf minimum conductance g_min performed with data collected at different temperatures. Each line shows r² for one trait at each different temperature. Lines are coloured according to types of variables: Yellow lines indicate stomatal traits, blue lines indicate leaf thickness metrics, and red lines indicate structural traits. Line types differentiate traits for visualization purposes only. a_l, leaf area; g_cw, cuticular conductance; g_min, minimum conductance; LDMC, leaf dry matter content; LMA, leaf mass per unit area; l_sp, stomatal pore length; w_gs, guard cell width; x_cut, cuticle thickness; x_epi, epidermis thickness; x_l, leaf thickness; x_pal, palisade thickness; x_spg, spongy mesophyll thickness; ρ_s stomatal density.

Fig. 5

Relationship of cuticular conductance, g_cw, to cuticular thickness, x_cut, and leaf mass per unit area (LMA). Panels show ordinary least squares linear regressions relating (a, b) g_cw to x_cut, and (c, d) g_cw to LMA at (a, c) 20°C and (b, d) 50°C. Points are species-level means. Error bars show SE. Solid lines are significant (P < 0.05). acci, Acer circinatum; acma, Acer macrophyllum; alru, Alnus rubra; arme, Arbutus menziesii; bepa, Betula papyrifera; clke, Cladrastis kentuckea; maac, Magnolia acuminata; mafu, Malus fusca, potr, Populus tremuloides; quga, Quercus garryana; umca, Umbellularia californica.

Effect of g_cw on photosynthetic capacity

Accounting for g_cw had variable effects on apparent photosynthetic capacity (Fig. 6). When assuming a constant g_cw (Fig. 6a,b), the median change in V_cmax and J_max remained low across all temperatures (maximum for V_cmax, 2.3%; maximum for J_max, 1.0%; both maxima occurred at 38°C). However, the range of percent change values varied strongly with temperature. For V_cmax, the interquartile range remained narrow throughout the temperature span (25^th percentile percent change < 1%, 75^th percentile < 10%). The 5^th and 95^th percentiles ranged from −0.1% to +0.7% at 10°C, to their widest of +0.1% to +40.1% at 38°C, before narrowing again at very high temperatures. For J_max, the interquartile range also remained narrow throughout the temperature range (25^th percentile < 1%, 75^th percentile < 5%); the 5^th and 95^th percentiles ranged from −0.3% to +0.2% at 10°C up to 0.0% to 22.4% at 38°C.

Fig. 6

Effect of cuticular conductance, g_cw, on estimates of maximum rates of Rubisco carboxylation, V_cmax, and RuBP regeneration, J_max. Points indicate individual estimates of percent change in (a, c, e) V_cmax and (b, d, f) J_max. Panels (a, b) illustrate percent change in V_cmax and J_max when a constant g_cw is considered, relative to no g_cw. Panels (c, d) illustrate percent change in V_cmax and J_max when temperature dependence of g_cw is considered, relative to a constant g_cw (i.e. additional to change in panels a, b). Panels (e, f) illustrate change in V_cmax and J_max as a function of stomatal conductance, g_sw. Light and dark shaded bands in (a–d) show regions bounding 95^th and 75^th percentiles of percent change values. Solid lines show median values. Dotted lines are shown for reference and correspond to no change.

When assuming a temperature-dependent g_cw (Fig. 6c,d), we found relatively small changes in V_cmax and J_max throughout much of the temperature range relative to V_cmax and J_max calculated with a constant g_cw, with larger changes apparent only at higher temperatures. As in the case of constant g_cw, interquartile ranges remained narrow throughout the temperature range (For V_cmax, 25^th percentile < 1%, 75^th percentile < 3%; for J_max, 25^th percentile < 1%, 75^th percentile < 2%). Extreme percent change values in V_cmax and J_max showed peaks at 40°C and 49°C; the 95^th percentile values at these temperatures were 23.9% and 34.7% for V_cmax, and 9.1% and 6.6% for J_max, respectively. We emphasize that these percent change values are relative to V_cmax and J_max computed with a constant g_cw, to illustrate the effect of g_cw temperature dependence alone. The combined effect of g_cw and its temperature dependence results in greater percent change in V_cmax and J_max relative to the case where g_cw = 0 (Fig. S8); the 95^th percentile of change in V_cmax and J_max reach 61.5% and 29.4% at 40°C and 38°C, respectively, relative to V_cmax and J_max computed without g_cw.

Finally, we related percent change in photosynthetic capacity to reported stomatal conductance for each A–c_i curve (Fig. 6e,f). For both V_cmax and J_max, percent change remains near zero for individuals with relatively high g_sw, but as g_sw decreases towards zero, the percent change values spread out, covering a wide range of possible values. Thus, the V_cmax and J_max estimates that exhibited large changes when g_cw was accounted for were almost exclusively plants with low g_sw.

Discussion

In this study, we investigated the temperature dependence of different water use pathways in 11 species of broadleaf trees. By separately measuring leaf minimum conductance and cuticular conductance, we showed that the cuticular water pathway tended to dominate as leaf temperature increased, although individual species varied widely in their response to temperature. The relationship of these conductance metrics to leaf traits was also found to be temperature-dependent, with the strongest such relationships emerging at high temperature and VPD. Finally, we showed that g_cw and its temperature dependence can have substantial impacts on estimates of photosynthetic capacity, particularly when reported g_sw is low.

Temperature response of g_cw and g_min

We found that the pathway of water loss varied with temperature, such that the cuticular pathway increasingly dominated at higher temperatures. When measured at 20°C, we found that g_min was substantially higher than g_cw in all species except U. californica (20–88% stomatal contribution to g_min). Thus, g_sw,min remains a substantial contributor to g_min for many species at low temperatures, consistent with the so-called ‘leaky stomata’ hypothesis (Duursma et al., 2019). This is also consistent with prior studies finding variable, but nonzero contributions of g_sw,min to g_min (Machado et al., 2021; Márquez et al., 2022). g_min is often taken as a proxy for g_cw because it is easier to measure (Duursma et al., 2019), but our results suggest that this should not be done without first ensuring g_sw,min is negligible for the species under consideration. We further found that as temperature increased, the proportion of water escaping the leaf via the cuticle typically increased (Fig. S6); g_min may therefore serve as an increasingly good proxy for g_cw as temperature increases. However, we caution that some species continued to exhibit substantial g_sw,min even at 50°C (in A. rubra, 61% of g_min was due to g_sw,min at 50°C; A. menziesii, 75%; U. californica, 48%).

The temperature dependence of g_min or g_cw often exhibits a monotonically increasing, biphasic relationship in which conductance rapidly increases above a transition temperature, typically in the range of 35–40°C (e.g. Schönherr et al., 1979; Schreiber & Schönherr, 1990; Schreiber, 2001; Schuster et al., 2016; Billon et al., 2020; Hartill et al., 2023; Wang et al., 2024). Many of the species studied here did exhibit increases with temperature, particularly in g_cw, at 40–50°C, although the monotonically increasing, biphasic relationship was in general not observed. Such high-temperature increases in g_cw may be due to phase transitions in the cuticular wax, or thermal expansion of the cuticular matrix resulting in the opening of additional pathways for diffusion in the cuticle surface, for example (Schuster et al., 2016). Despite the apparent prevalence of monotonically increasing relationships with temperature in the literature, Bueno et al. (2019) found that the date palm, Phoenix dactylifera, exhibited an invariant relationship between g_min and temperature. Furthermore, Slot et al. (2021) found that only 7 of 24 tropical tree species studied exhibited a biphasic temperature dependence, with most exhibiting weakly decreasing, U-shaped, or invariant responses. Our results similarly show a diversity of responses to temperature in g_min and g_cw.

The mechanism causing elevated g_min or g_cw at low temperatures (resulting in U-shaped and decreasing trends) is less clear, but variation in VPD may provide one possible explanation. Many studies that measure g_min or g_cw temperature response do not control the humidity of the air surrounding the leaf, or do so in ways that still yield covariation of VPD and temperature (e.g. Bueno et al., 2019; Billon et al., 2020; Machado et al., 2021; Slot et al., 2021; Hartill et al., 2023; this study). The need for humidity control during measurements of g_min or g_cw is not immediately obvious, since the models employed for calculating g_min or g_cw (based on Fick's law of diffusion) explicitly account for the evaporative demand of air due to VPD (Sack & Scoffoni, 2010; Notes S1). Conductance is conventionally defined as a parameter relating VPD to transpiration which, according to Fick's law, should be proportionally related (Pearcy et al., 2000; Notes S1). In photosynthesizing leaves, increasing VPD may trigger stomatal closure, which reduces leaf conductance by altering the surface properties of the leaf (Urban et al., 2017). However, in nonphotosynthesizing leaves with closed stomata, altering VPD is unlikely to affect stomatal aperture and therefore seems unlikely to alter the surface properties of the leaf in a manner that affects its total conductance. Based on this reasoning, g_min and g_cw may be expected to be invariant with VPD.

However, Schreiber et al. (2001) found a two- to threefold increase in g_cw as relative humidity was increased from 2% to 100%. Wang et al. (2024) reported modest increases in g_min, which were nonsignificant for one species and marginally significant for another species, as humidity was increased from 30% to 70%, though the sample size in this study was quite small. While the mechanism for this effect is not known with certainty, it has been proposed that water molecules may adsorb more readily into cuticular waxes at higher humidities, resulting in swelling of the cuticular wax and consequent increase in the area of polar pores through which water may permeate (Schreiber et al., 2001). An effect such as this may explain the observed U-shaped response of g_cw in many of our species, and those in other studies, as low VPD at low temperatures may result in an elevation of g_cw relative to that which would be expected based on temperature effects alone. In this study, we did not control humidity during our measurements, resulting in increasing VPD as temperatures increased (Table S1). This may improve the realism of our measurements relative to studies with strictly controlled humidity, since VPD and air temperature often strongly covary in the field (Sadok et al., 2021). Nevertheless, the lack of consistency in humidity control among studies impairs our ability to directly compare results from different studies, and potentially adds noise to data within studies. However, little work has been done on the humidity dependence of g_cw and g_min, and there remains a need to characterize this response across species, as well as its interaction with temperature. Furthermore, sensitivity of g_cw or g_min to humidity may suggest a need to revise models to incorporate non-Fickian water transport processes (Aparecido et al., 2020; Stinziano et al., 2020).

We further note that typical methods used for estimating g_cw and g_min, including those used in this work, assume that relative humidity in the intercellular airspaces of the leaf is 100% (von Caemmerer & Farquhar, 1981; Sack & Scoffoni, 2010; Márquez et al., 2021). This assumption is expedient for calculating conductance since the actual humidity inside the leaf is not typically known. However, recent work has shown that the intercellular airspaces may not be fully saturated with water vapour (Cernusak et al., 2018; Wong et al., 2022; Diao et al., 2024); in cases where VPD is high, this effect may be especially pronounced. Under these conditions, conductance may be underestimated. Thus, g_cw and g_min reported here may be conservative estimates, particularly at high temperature and VPD.

The temperature sensitivity of g_min and g_cw is of interest because, inter alia, it offers a possible explanation for observations of stomatal decoupling at high temperature, that is a continued increase (or lack of decrease) in reported g_sw as A declines at supraoptimal leaf temperatures (e.g. Aparecido et al., 2020; Krich et al., 2022; Marchin et al., 2023). Such decoupling behaviour departs from predictions of stomatal behaviour based on optimality criteria (Cowan & Farquhar, 1977; Ball et al., 1987; Medlyn et al., 2011). It is unknown, however, whether this increase in water use corresponds to an adaptive response of the plant, for example to prevent overheating via evaporative cooling (Michaletz et al., 2016; Garen et al., 2023), or is rather a passive mechanism, for example a mechanical failure of the cuticle or guard cells (Slot et al., 2021; Blonder et al., 2023). The high temperature g_min and g_cw values observed in this study were in the range of 1.3–8.5 mmol m⁻² s⁻¹, which are much lower than reported g_sw values at high temperature in studies that have reported stomatal decoupling (e.g. Slot et al., 2016; Urban et al., 2017; Aparecido et al., 2020; Marchin et al., 2023). At high temperature and VPD, canopy transpiration can remain high even with reduced stomatal conductance, resulting in substantial cooling effects on canopies (e.g. 5–7.5°C cooling effects; Drake et al., 2018; Kibler et al., 2023). While it remains unclear whether or not this is due to stomatal opening, we speculate that the discrepancy between g_min and g_cw values at high temperatures and g_sw reported during observations of decoupling likely means that stomatal opening is necessary to explain these observations.

Relationship of g_cw and g_min to leaf traits

We found numerous relationships between traits and conductance values that varied with temperature, including a significant relationship between cuticular thickness and g_cw at high temperature. Standard mass transfer models state that membrane thickness is inversely related to rate of gas diffusion across the membrane (Incropera et al., 2007). This is intuitive, and it implies that g_cw or g_min should exhibit a negative relationship with x_cut, as a thicker cuticle would constitute a greater barrier to transpiration (Schuster et al., 2016). However, contrary to expectation, many previous studies have found no relationship between x_cut and g_cw or g_min (Riederer & Schreiber, 2001; Anfodillo et al., 2002; Duursma et al., 2019; Machado et al., 2021; Grünhofer et al., 2022). Similarly, we found a weak, nonsignificant relationship between x_cut and g_cw when measured at 20°C. However, we also found a significant negative relationship between x_cut and g_cw when measured at 50°C. This latter relationship contrasts with the prior findings in the literature but conforms to the standard mass transfer models (Incropera et al., 2007) and intuitive expectations. It has been hypothesized that sorption of water vapour at the surface of the epicuticular wax is the dominant limitation to water transport (Riederer, 2006; Schuster et al., 2016; Duursma et al., 2019), which could explain the previously observed invariance of g_cw with x_cut. However, prior work has shown that cuticles exhibit substantial thermal expansion and undergo phase transitions at temperature in the range of 35–40°C (Schreiber & Schönherr, 1990; Schreiber, 2001; Riederer, 2006), which has been hypothesized to open additional pathways to water transport in the cuticle. If this reduces the effectiveness of the epicuticular wax as a transpiration barrier, then the thickness of the cuticular membrane may become relatively more important for restricting transpiration at high temperature and may explain the temperature dependence of the relationship observed here.

Since g_min and g_cw have implications for plant water use, it has been hypothesized that these values should be correlated with other leaf traits related to overall plant economic strategy, such as LMA or LDMC (Reich, 2014; Slot et al., 2021), but prior studies have produced mixed results (Saito & Futakuchi, 2010; Machado et al., 2021; Slot et al., 2021; Wang et al., 2024). Similar to our results for x_cut, we found weak relationships between structural traits and conductance at low temperatures which became stronger at higher temperatures. To our knowledge, trait–conductance relationships that become stronger at higher temperatures have not been reported previously in the literature. We speculate that the temperature dependence of these relationships may relate to the differential importance of high- and low-temperature behaviour of the cuticle for plant performance and survival. Even in cases where g_cw or g_min is invariant with temperature, leaves typically will still lose substantially more water at high temperatures due to high VPD (cf., Slot et al., 2021). Since water loss during hot drought is an important determinant of plant mortality (Hammond & Adams, 2019), the behaviour of the cuticle under these conditions may have a greater effect on plant fitness and therefore be under greater selection than the cuticular behaviour at low temperature (and, consequently, low VPD). If so, then we should expect greater correlation between the high-temperature behaviour of cuticular or minimum conductance and traits that reflect the overall economic strategy of the plant (e.g. LMA or LDMC). The high-temperature relationships we found align with general patterns of variation between leaf traits and economic strategy, as lower g_min or g_cw and high LMA and LDMC are both expected to coincide with more conservative resource use strategies (Reich, 2014; Slot et al., 2021), but said relationships weaken or disappear at lower temperatures, consistent with this hypothesis. Similarly, Schuster et al. (2016) reported that Rhazya stricta, a desert plant, counterintuitively exhibited higher g_cw than Juglans regia, a mesic-climate plant, at low temperature. However, this pattern reversed at high temperature, with g_cw of J. regia exceeding that of R. stricta above c. 40°C, consistent with the hypothesis above. These findings may be useful for understanding and predicting the response of vegetation to hot drought events, as LMA and LDMC are widely measured traits and may also be measured at landscape scales via remote sensing approaches (Moreno-Martínez et al., 2018; Kattge et al., 2020). However, additional research on species spanning wider ranges in trait values would be useful for thoroughly evaluating the strength and generality of these relationships.

Effect of g_cw on photosynthetic capacity

Metrics of photosynthetic capacity, V_cmax and J_max, had variable responses to the inclusion of g_cw in gas exchange calculations, with the largest changes observed at relatively high temperatures and low conductance. Many individual measurements showed negligible change when g_cw was considered; large changes were only found when reported g_sw was low. The observed dependence of V_cmax and J_max estimates on reported g_sw is intuitive because in species with low reported g_sw, the assumed values of g_cw make up larger proportions of the overall total leaf conductance. This is consistent with Hussain et al. (2024), who reported large errors in c_i when g_cw exceeded c. 10% of reported g_sw. We found that the temperature dependence of g_cw may induce further errors in c_i, V_cmax, and J_max. While not all species exhibit substantial temperature dependence of g_cw (cf., Bueno et al., 2019; Slot et al., 2021), assuming a temperature-invariant g_cw is likely not appropriate for all species. Leaf traits, such as LMA, may be a useful starting point for screening species which require further investigation due to their temperature-dependent relationships with g_cw.

In our simulations, we applied representative g_cw functions across many species which differ in their g_cw values and respective temperature dependences. The lack of direct measurements of g_cw on the same individuals as the A–c_i curves represents a source of uncertainty. It is possible that confounding factors may modify the relationship observed here. For example, if g_cw varies positively with g_sw across species, then the effect of g_cw on species with low g_sw may be less than observed here, while the effect on species with high g_sw would be greater. However, if g_cw varies inversely with g_sw across species, then the opposite would be true. Some studies have found positive relationships between g_sw and g_cw across species (Machado et al., 2021; Wang et al., 2024); however, in the species investigated here we found a weak, nonsignificant negative relationship between g_sw and g_cw (Fig. S9). We caution that these results are not intended to be definitive, but rather illustrative, and we hope that they motivate future work into the effects of g_cw on V_cmax and J_max, as well as the relationship between g_cw and g_sw.

Acknowledgements

The authors would like to thank the staff at The University of British Columbia Botanical Garden, particularly Ben Stormes and Daniel Mosquin, for their assistance with sample collection. We are also grateful to Haley Branch and Milos Simovic for their assistance with anatomical measurement and imaging protocols, and to Julien Grebert and Neo Wu for their assistance with data collection. This research was supported by an NSERC Discovery grant to STM and an NSERC PGS-D grant to JCG.

Competing interests

None declared.

Author contributions

JCG designed the research with guidance from STM. JCG collected the data. JCG analysed the data with guidance from STM. JCG wrote the first draft of the manuscript. JCG and STM revised the manuscript.

Open Research

Data availability

Data and R code for reproducing the figures and analyses in this paper are publicly available for download at doi: 10.17605/OSF.IO/CZUEF.

Disclaimer

The New Phytologist Foundation remains neutral with regard to jurisdictional claims in maps and in any institutional affiliations.