https://orcid.org/0000-0002-6376-0803
Abstract
The severity of droughts is expected to increase with climate change, leading to more frequent tree mortality and a decline in forest ecosystem services. Consequently, there is an urgent need for monitoring networks to provide early warnings of drought impacts on forests. Dendrometers capturing stem diameter variations may offer a simple and relatively low-cost opportunity. However, the links between stem shrinkage, a direct expression of tree water deficit (TWD), and hydraulic stress are not well understood thus far. In this study, we exposed two widespread conifers Pinus sylvestris L. and Larix decidua Mill. to lethal dehydration by withholding water and closely monitored TWD, midday water potential (|$\psi $|) and midday stomatal conductance (|${\textit g}_{\textit s}$|) under controlled greenhouse conditions. We found strong relationships between the three variables throughout the dehydration process, particularly suggesting the potential for continuous |$\psi $| predictions and stomatal closure assessments. However, the relationships decoupled during recovery from severe drought. We also identified TWD thresholds that signal the onset of drought stress and tissue damage, providing insights into stress impacts and recovery potential. While these findings are promising, challenges remain in practically transferring them to field set-ups by suitable TWD normalization. Importantly, we observed that midday |${\textit g}_{\textit s}$| was drastically reduced when TWD persisted overnight, providing a directly applicable drought stress signal that does not require normalization. In conclusion, while challenges remain, our results highlight the potential of dendrometers for monitoring tree water dynamics. Implementing dendrometer networks could support the development of early warning metrics for drought impacts, enabling large-scale monitoring in diverse settings, such as urban areas and forest ecosystems.
Introduction
The frequency and severity of droughts are expected to increase with climate change (Spinoni et al. 2018, Suarez-Gutierrez et al. 2023), resulting in global tree damage and mortality (Allen et al. 2010, Senf et al. 2020, Yi et al. 2022), negatively affecting the carbon balance and ecosystem functioning (Choat et al. 2018). Water limitation immediately affects tree carbon sequestration through species-specific water control mechanisms. Regulation of stomatal conductance (|${\textit g}_{\textit s}$|) reduces water loss at the expense of net photosynthesis (|$A_{\text{net}}$|). Severe droughts affect the hydraulic system of trees, resulting in poorly understood long-term legacy effects (Müller and Bahn 2022) and having lasting impacts on plant functionality and forest structures (Pretzsch et al. 2022).
Understanding the progressive impacts during and after drought enables adaptive measures. Monitoring tree physiological responses offers valuable information on species sensitivity and dependence on site properties, supporting management strategies such as selecting less susceptible species and timely irrigation in urban settings. However, appropriate techniques for operational tree water status monitoring are still not widely applied (Novick et al. 2022).
Usually, water potential (|$\psi $|) is measured to assess levels of drought stress. Generally, |$\psi $| gradients, primarily driven by soil water availability, leaf-level transpiration and cohesion forces (cohesion-tension theory; Dixon 1915), govern the water transport in vascular plants. Parameters such as |$\psi $| at which |$50\%$| of hydraulic conductivity is lost (|$P_{50}$|) provide insights into tree species’ susceptibility to hydraulic damage (Choat et al. 2012). However, notable limitations exist: traditional pressure chamber measurements for |$\psi $| are complex (Rodriguez-Dominguez et al. 2022) and destructive sampling from tree canopies is necessary. Though valuable for stress indication, they are not continuous, and simultaneously monitoring multiple individuals requires substantial logistic effort. Thermocouple psychrometers provide continuous measurements but are complex, temperature-sensitive and not universally applicable, especially for resin-rich gymnosperms (Martinez et al. 2011, Novick et al. 2022).
Therefore, identifying mechanisms for continuous tree drought stress evaluation can vastly improve forest monitoring. A promising candidate is tree water deficit (TWD; Zweifel 2015, Steppe 2018, Salomón et al. 2022), derived from stem diameter variations monitored by dendrometers capturing µm-scales, which are already widely applied (e.g. TreeNet; Zweifel et al. 2021a), cost-efficient and non-destructive. TWD represents the decrease between actual and recorded maximum stem diameter (Zweifel et al. 2016, Knüsel et al. 2021). It is sensitive to environmental conditions like vapor pressure deficit (VPD) and soil water content (SWC) (Zweifel et al. 2005, Sánchez-Costa et al. 2015, Dunkleberger et al. 2023, Oberhuber et al. 2023), water storage capacity of cell tissues (De Swaef et al. 2015) and water-use traits such as stomatal regulation (Drew et al. 2011, Liu et al. 2023). During the day, trees typically transpire more water than they can take up due to limitations in whole plant conductance or soil water availability, resulting in the mobilization of internal water reservoirs. On top of apoplastic water release (Mantova et al. 2022), symplastic water release from living elastic tissues (Knipfer et al. 2019), such as bark and phloem, results in turgor loss, reduction of cell volume and subsequent stem shrinkage (Zweifel et al. 2001, De Swaef et al. 2015, Steppe et al. 2015). While xylem shrinkage is often neglected, it can also contribute to reversible water release, largely depending on living cells within the xylem matrix (Skelton 2020, Oberhuber et al. 2023). With ample soil water, internal storage pools replenish at night, allowing stem diameter recovery or even a new maximum (Zweifel et al. 2021b). However, during periods of extended water limitation, growth is assumed to stop due to cell turgor losses inhibiting cell division and expansion (Zweifel et al. 2016, Krejza et al. 2022), while TWD accumulates, making it an indirect indicator of drought stress.
TWD is closely related to |$\psi $| (Cochard et al. 2001, Steppe et al. 2006, Drew et al. 2011, Peters et al. 2021, Lauriks et al. 2022), potentially enabling continuous |$\psi $| predictions (Dietrich et al. 2018) and stress monitoring. Additionally, TWD correlates with gas exchange, and TWD thresholds for plant mortality can be determined (Lamacque et al. 2020, Andriantelomanana et al. 2024). Knowledge of tree sensitivity thresholds can leverage existing data series, offering insights into the severity of past and current droughts.
However, the relations between TWD and physiological parameters have been evaluated within a limited range of species and drought-stress conditions, constraining their practical applicability. Most studies investigate mild drought conditions, excluding the impact of irreversible drought damage (Cochard et al. 2001, Drew et al. 2011, Dietrich et al. 2018). Under severe drought, dynamics might change as embolism events irreversibly release additional water from inelastic xylem conduits (Knipfer et al. 2019, Skelton 2020) and trigger the collapse of elastic cells (Mantova et al. 2022, Andriantelomanana et al. 2024). Understanding the |$\psi $|–TWD relationship during this critical phase is essential to indicate long-term damage (Vergeynst et al. 2015). Furthermore, correlations of TWD with gas fluxes are of interest but have received even less investigation than those with |$\psi $|. While |$\psi $| indicates the degree of drought stress and the current risk of irreversible damage, |${\textit g}_{\textit s}$| represents plant functioning and possible carbon assimilation. Lastly, the impacts of stress release on the relationship between TWD and physiological parameters are currently unknown. Depending on drought severity, recovery pace varies among different physiological processes, with gas exchange parameters requiring more time compared with |$\psi $| (Ruehr et al. 2019, Rehschuh et al. 2020). If TWD recovery timing (Lamacque et al. 2020) does not match that of |$\psi $| and |${\textit g}_{\textit s}$|, relationships decouple. Moreover, understanding how post-stress recovery dynamics affect susceptibility to future droughts is crucial to assessing forest functioning in a changing climate (Ruehr et al. 2019, Kannenberg et al. 2020).
Here we conducted a greenhouse study and exposed seedlings of two coniferous species, Pinus sylvestris and Larix decidua, to complete dehydration and variable rewetting. We chose the species due to their ecological relevance (Obojes et al. 2024) and their different strategies regarding water use, namely anisohydric vs isohydric behavior (Oberhuber et al. 2015). The experiment allowed us to analyze the relationships of TWD to midday |$\psi $| and |${\textit g}_{\textit s}$| during drought progression and how stress release modifies these relationships. Specifically, we targeted the following questions: (i) how are TWD and |$\psi $| related during lethal drought progression; (ii) does the relationship between TWD and |$\psi $| change during recovery (after re-watering); (iii) how is TWD related to gas exchange during drought and recovery; and (iv) can we derive early-warning metrics on tree drought stress intensities based on TWD?
Materials and methods
Experimental setup
We conducted an experimental study in an environmentally controlled greenhouse facility at the KIT-Campus Alpin in Garmisch-Partenkirchen (|$708\ {}m$| a.s.l., 47|$^{\circ }$|28′32.9′ N, 11|$^{\circ }$|3′44.2′ E) in June to August 2023. Three-year-old Scots pine (P. sylvestrisL.) and European larch (L. decidua Mill.) seedlings (|$n=24$| per species) were obtained from a local tree nursery in Gunzenhausen, Germany, in March 2023 and planted in individual |$5.7 \ \text{L}$| pots containing an organic clay substrate and perlite at a ratio of |$\text{5:1}$| supplemented with |$6\ \text{g}$| of slow-release fertilizer (Osmocote 5 8-9M, Micromax Premium; ICL Specialty Fertilizers, Geldermalsen, The Netherlands). On average, P. sylvestris seedlings were |$79.3 \pm 3.2\ \text{cm}$| high and had a stem diameter of |$15.5 \pm 0.7\ \text{mm}$|. For L. decidua, corresponding values were |$79.0 \pm 2.9\ \text{cm}$| and |$12.2 \pm 0.7\ \text{mm}$|, respectively.
We placed the seedlings inside the greenhouse compartment to acclimate to the new conditions 3 weeks before the experiment started and watered them to field capacity (twice a day |$100\ \text{mL}$| + individual watering if necessary). At the experiment start, the trees were assigned randomly to a control (|$n = 6$| per species) or drought treatment (|$n = 18$| per species) as depicted in Figure 1. The drought treatment included more trees to allow frequent sampling. The control trees further received water, while irrigation in the drought treatment was stopped during two drought phases and resumed during recovery. All drought-treated trees were first exposed to a 9-day drought period (D1). After a 3-day recovery phase, a second and more extended drought cycle followed (D2) until |$\psi $| ranged around the determined |$P_{50}$| values (see Table 1a), which took |$28$| days for L. decidua and |$30$| days for P. sylvestris. Then the drought treatment trees were equally and randomly distributed either to a drought–lethal or a drought–recovery treatment. To observe the full range of dehydration, the drought–lethal trees were further dehydrated until tree death occurred. In contrast, seedlings in the drought–recovery treatment were re-watered to investigate how the recovery potential after hydraulic damage occurrence develops.
Water potentials of vulnerability curve and stomatal closure thresholds for seedlings of L. decidua and P. sylvestris. (a) |$\psi $| at |$12$|, |$50$| and |$88\%$| (|$P_{12}$|, |$P_{50}$|, |$P_{88}$|) percentage loss of conductance (PLC) and the according |$95\%$| confidence interval in brackets according to Figure S2 available as Supplementary data at Tree Physiology Online. (b) |$\psi $| at |${\textit g}_{\textit s}$| of |$5$| and |$50\%$| relative to control means (|$\psi _{g5}$|, |$\psi _{g50}$|) and the according |$95\%$| confidence interval in brackets according to Figure S6 available as Supplementary data at Tree Physiology Online.
| Species | L. decidua | P. sylvestris | |
|---|---|---|---|
| (a) | P|$_{12} \, [\text{MPa}]$| | −2.74 (−2.34/−3.06) | −2.52 (−2.32/−2.66) |
| P|$_{50} \, [\text{MPa}]$| | −3.16 (−3.00/−3.24) | −3.11 (−3.04/−3.17) | |
| P|$_{88} \, [\text{MPa}]$| | −3.47 (−3.35/−3.63) | −3.58 (−3.45/−3.70) | |
| (b) | |$\psi _{g5} \, [\text{MPa}]$| | −1.87 (−1.58/−2.23) | −1.33 (−0.91/−1.57) |
| |$\psi _{g50} \, [\text{MPa}]$| | −1.06 (−0.62/−1.24) | −0.82 (−0.72/−0.88) |
| Species | L. decidua | P. sylvestris | |
|---|---|---|---|
| (a) | P|$_{12} \, [\text{MPa}]$| | −2.74 (−2.34/−3.06) | −2.52 (−2.32/−2.66) |
| P|$_{50} \, [\text{MPa}]$| | −3.16 (−3.00/−3.24) | −3.11 (−3.04/−3.17) | |
| P|$_{88} \, [\text{MPa}]$| | −3.47 (−3.35/−3.63) | −3.58 (−3.45/−3.70) | |
| (b) | |$\psi _{g5} \, [\text{MPa}]$| | −1.87 (−1.58/−2.23) | −1.33 (−0.91/−1.57) |
| |$\psi _{g50} \, [\text{MPa}]$| | −1.06 (−0.62/−1.24) | −0.82 (−0.72/−0.88) |
Water potentials of vulnerability curve and stomatal closure thresholds for seedlings of L. decidua and P. sylvestris. (a) |$\psi $| at |$12$|, |$50$| and |$88\%$| (|$P_{12}$|, |$P_{50}$|, |$P_{88}$|) percentage loss of conductance (PLC) and the according |$95\%$| confidence interval in brackets according to Figure S2 available as Supplementary data at Tree Physiology Online. (b) |$\psi $| at |${\textit g}_{\textit s}$| of |$5$| and |$50\%$| relative to control means (|$\psi _{g5}$|, |$\psi _{g50}$|) and the according |$95\%$| confidence interval in brackets according to Figure S6 available as Supplementary data at Tree Physiology Online.
| Species | L. decidua | P. sylvestris | |
|---|---|---|---|
| (a) | P|$_{12} \, [\text{MPa}]$| | −2.74 (−2.34/−3.06) | −2.52 (−2.32/−2.66) |
| P|$_{50} \, [\text{MPa}]$| | −3.16 (−3.00/−3.24) | −3.11 (−3.04/−3.17) | |
| P|$_{88} \, [\text{MPa}]$| | −3.47 (−3.35/−3.63) | −3.58 (−3.45/−3.70) | |
| (b) | |$\psi _{g5} \, [\text{MPa}]$| | −1.87 (−1.58/−2.23) | −1.33 (−0.91/−1.57) |
| |$\psi _{g50} \, [\text{MPa}]$| | −1.06 (−0.62/−1.24) | −0.82 (−0.72/−0.88) |
| Species | L. decidua | P. sylvestris | |
|---|---|---|---|
| (a) | P|$_{12} \, [\text{MPa}]$| | −2.74 (−2.34/−3.06) | −2.52 (−2.32/−2.66) |
| P|$_{50} \, [\text{MPa}]$| | −3.16 (−3.00/−3.24) | −3.11 (−3.04/−3.17) | |
| P|$_{88} \, [\text{MPa}]$| | −3.47 (−3.35/−3.63) | −3.58 (−3.45/−3.70) | |
| (b) | |$\psi _{g5} \, [\text{MPa}]$| | −1.87 (−1.58/−2.23) | −1.33 (−0.91/−1.57) |
| |$\psi _{g50} \, [\text{MPa}]$| | −1.06 (−0.62/−1.24) | −0.82 (−0.72/−0.88) |
Overview of the experimental design. Shown are the drought and recovery cycles for L. decidua (L) and P. sylvestris (P) seedlings that were randomly assigned to a control (|$n=6$| per species) or drought treatment (|$n=18$| per species). We irrigated the trees to field capacity or withheld irrigation as depicted. In the first drought D1, we induced mild stress and re-watered all drought-trees once stomatal conductance approached |$g_{s} \approx 0$|. In the more extended drought D2, we aimed at inducing severe stress. Once water potentials (|$\psi $|) reached values representing hydraulic conductance losses of about |$50\%$| (|$P_{50}$|), we separated the drought treatment into a drought–recovery (|$n=9$| per species) that was re-watered and a drought–lethal treatment (|$n=9$| per species), which further dehydrated until tree death.
Temperature and humidity (CS215, Campbell Scientific Inc., Logan, UT, USA), as well as photosynthetic active radiation (PAR; PQS 1, Kipp & Zonen, Delft, The Netherlands), were continuously measured half-hourly to verify stable conditions. Daytime air temperature was |$23 \pm 2 \,^\circ C$| (mean |$\pm $| SD), daytime VPD |$1.0\pm 0.3\ \text{kPa}$| (Figure S1 available as Supplementary data at Tree Physiology Online) and ambient light conditions in the greenhouse were supplemented by LED grow lamps (LED-KE 400 VSP, DHLicht, Wuelfrath, Germany) to ensure a constant PAR of |$600\pm 100\,\frac{\mu \text{mol}}{\text{s} \cdot \text{m}^{2}}$| for |$15$| h a day. Trees were supplied with water using an automated irrigation drip system, and soil water content (SWC; UMS Water Content Sensor 10HS, 10-minute frequency; logger CR1000, Campbell Scientific, Inc.) was recorded for 10 trees per species (control |$n=2$|, drought–recovery |$n=4$|, drought–lethal |$n=4$|).
Non-continuous measurements
Water potential
Midday water potential (|$\psi $|) was measured frequently between once a day and once a week, depending on the experimental phase with a pressure chamber (Model 600D, PMS Instruments, OR, USA; Scholander et al. (1965)). In the case of L. decidua, we sampled small twigs, and for P. sylvestris, we utilized needle fascicles to save plant material and maximize the number of possible measurements. For P. sylvestris, twigs were sampled only in the drought–lethal treatment during the last phase of dehydration, wherein |$\psi $| reached low values and needles were too desiccated to provide a trustful base for measurements. We measured |$\psi $| until values of |$<\! -7\,\text{MPa}$| were reached. Based on our vulnerability curves (Figure S2 available as Supplementary data at Tree Physiology Online), we considered the trees dead at this stage. Daily means presented in the results section include a minimum sample size of three for the drought and two for the control treatments, with higher sample sizes for most days.
Leaf gas exchange
Midday leaf gas exchange, including stomatal conductance |${\textit g}_{\textit s}$| (|$\text{mmol}\, \text{m}^{-2}\, \text{s}^{-1}$|), transpiration E (|$\text{mmol}\, \text{m}^{-2}\, \text{s}^{-1}$|) and net photosynthesis |$A_{net}$| (|$\mu \text{mol}\, \text{m}^{-2}\, \text{s}^{-1}$|), was measured using the portable LI-COR 6800 system (LI-COR Biosciences, Lincoln, NE, USA) on 10 specific days on the same needle cohorts for each species, covering different stages in the drought–recovery cycles (control |$n\!=\!3$|, drought–recovery |$n\!=\!3-5$|, drought–lethal |$n\!=\!3-5$|). In the case of P. sylvestris, 10 needles were formed to a flat surface and enclosed in a leaf cuvette (|$2\ \text{cm}^{2}$|). In the case of L. decidua, we enclosed one needle fascicle. Ambient cuvette parameters were kept constant, including a photosynthetic photon flux density of |$2000\,\mu \text{mol}\, \text{m}^{-2}\, \text{s}^{-1}$|, a CO|$_{2}$| concentration of |$410\ \text{p.p.m.}$|, a |$10,000\ \text{r.p.m.}$| fan speed, air temperature of |$25^\circ C$| and relative humidity of |$50\%$|. Due to stable VPD during the experiment (Figure S1 available as Supplementary data at Tree Physiology Online), |${\textit g}_{\textit s}$| and E dynamics did not differ substantially (Figure S3 available as Supplementary data at Tree Physiology Online), so transpiration is not individually discussed in this work.
Vulnerability curves
We conducted Cavitron measurements (Cochard 2002) on the control seedlings (|$n\!=\!4$| per species) after the experiment. Stem samples were cut and directly re-cut under water, wrapped in cling film, enclosed in a cooling bag and transported to a laboratory at the University Innsbruck. Prior measurements, we completely removed the bark underwater, and samples were re-cut again on both ends to a length of |$28\ \text{cm}$| to fit into a custom-build rotor placed in a centrifuge (Sorvall RC-5; Thermo Fisher Scientific, Waltham, MA, USA) following the methods from Beikircher et al. (2010). We determined the percentage loss of conductance (PLC) by measuring the decrease in hydraulic conductivity while we reduced step-wise the rotational speed to increase the xylem pressure. PLC was calculated as
where |$k_{i}$| is the initial hydraulic conductance and |$k_{t}$| is the hydraulic conductance at the respective xylem pressure.
Dendrometer measurements
A total number of |$13$| trees per species (control |$n\!=\!3$|, drought–recovery |$n\!=\!5$|, drought–lethal |$n\!=\!5$|) were equipped with dendrometers (Ecomatik Diameter Dendrometers, Type DD-S and DD-S1, 10-min frequency, |$1.5\,\mu \text{m}$| resolution) to capture relative variations in stem diameter. Before the sensor installation (Figure S4 available as Supplementary data at Tree Physiology Online), we removed the most outer dead layers of the bark and measured the initial stem diameters per tree with a calliper to calculate absolute diameter changes from the dendrometer records.
One of the P. sylvestris drought–recovery trees equipped with a dendrometer showed an unexpectedly weak TWD–recovery after the final rewetting. As we cannot exclude a secondary pathogen infestation we excluded this seedling from further analysis.
Tree water deficit
We derived the TWD from the continuous high-resolution stem diameter variations as follows:
where |$d$| is the current absolute stem diameter and |$d_{max}$| is the maximum diameter measured for an individual tree in the past, which sets the theoretical maximum diameter under conditions of fully hydrated tissues. We apply the zero-growth concept, assuming that growth cannot occur during periods of stem contraction (Zweifel et al. 2016, Dietrich et al. 2018). In addition, we calculated a normalized TWD for treatment means:
where TWD|$_{max}$| is the highest TWD occurring for drought–lethal trees at the end of the experiment.
Statistical analyses
We performed data processing, visualization and statistical analysis with the programming language R, version 4.2.2 (R Core Team 2022). The significance of treatment and species differences was evaluated using a linear mixed model likelihood ratio test (‘lme’ package; Bates et al. (2015)). All results presented are treatment means with standard error (SE). To derive the matching TWD with midday |$\psi $| and |${\textit g}_{\textit s}$| in the shown relationships, we averaged the TWD of all dendrometer-equipped trees between 1 and 4 p.m. on the corresponding days (midday TWD). Besides, we derived pre-dawn TWD, averaged from 3 to 4 a.m. Note that the |${\textit g}_{\textit s}$| and |$\psi $| treatment means also include measurements on trees not equipped with dendrometers to reduce destructive harvest per tree. Data for individual trees is depicted in the supplementary material (Figure S5 available as Supplementary data at Tree Physiology Online). Regression fits were produced with the nls() function, and the |$95\%$| confidence intervals for all regression parameters and regression curves were estimated via bootstrapping. For the relation between TWD|$_{norm}$| and |$\psi $|, we used the following sigmoidal regression:
where |$e$| is the Euler function and the parameters |$A$|, |$b$| and |$c$| determine the asymptotic behavior, steepness and inflection point (indicating |$50\%$| of water depletion). While in the literature, this approach is usually applied (e.g. Dietrich et al. 2018), its inversion is mathematically complex, particularly at the assumed plateaus where multiple |$\psi $| values belong to the same TWD, which hinders |$\psi $| retrieval directly from TWD measurements. Therefore, in a second approach, we reversed axes, separated TWD|$_{norm}$||$- \psi$| data into three ‘slope groups’ and fitted linear curves (|$\psi (TWD_{norm}) \!=\! a \cdot TWD_{norm} + C$|). We defined the cross-sections of the regression lines as thresholds for phase changes. For the relationship between |${\textit g}_{\textit s}$| and TWD|$_{norm}$|, we chose an exponential approach:
where the parameter |$a$| defines the vertical scaling, |$b$| the bending and |$C$| and the vertical offset of the function. Equation (5) was further used to retrieve thresholds for stomatal closure, in expression the TWD|$_{norm}$| at |${\textit g}_{\textit s}$| equaling |$5$| and |$50\%$| (TWD|$_{g5}$|, TWD|$_{g50}$|) of the |${\textit g}_{\textit s}$| control mean. Similarly, |$\psi _{g5}$| and |$\psi _{g50}$| (equaling the concept of |$G_{c50}$| in Peters et al. 2023) thresholds were retrieved from sigmoidal |${\textit g}_{\textit s}$|-|$\psi $| relationships (|$g_{norm}(\psi ) = 1 / (1 + exp(-a * (\psi + b)))$|).
The software package ’fitplc’ (Duursma and Choat 2017) was applied to create vulnerability curves from the Cavitron measurements. We used sigmoidal regressions and determined |$\psi $| values introducing a PLC of |$12$|, |$50$| and |$88\%$| (|$P_{12}$|, |$P_{50}$|, |$P_{88}$|), which function as key hydraulic thresholds (Urli et al. 2013, Dorji et al. 2024).
Results
Vulnerability to drought
The hydraulic vulnerability curves of both species, L. decidua and P. sylvestris, indicate similar |$P_{50}$| values of |$-3.16$| and |$-3.11 \, \text{MPa}$| (Table 1a) and thus comparable drought susceptibility, which is in agreement with ranges determined for mature trees in Peters et al. (2023). However, the curve for L. decidua is steeper, corresponding to a faster rate of conductance loss per decrease of |$\psi $| (Figure S2 available as Supplementary data at Tree Physiology Online). Besides, in the case of P. sylvestris|$\psi $| values corresponding to halfway and fully closed stomata (|$\psi _{g50}$|, |$\psi _{g5}$|) are higher, and the slope of the response curve at |$\psi _{g50}$|—a measure for isohydricity (Peters et al. 2023)—is steeper, indicating a stricter stomatal control (Table 1b, Figure S6 available as Supplementary data at Tree Physiology Online).
Tree water status during drought recovery cycles
Changes in SWC affected both |$\psi $| and TWD (Figure 2, Table 2). During D1, |$\psi $| exhibited a slight yet significant decrease compared with control conditions (|$P\le 0.01$|). Concurrently, stem diameter variations displayed increased diurnal amplitudes, particularly pronounced in L. decidua. By the end of D1, TWD did not fully vanish during night-time and started to accumulate. After stress release, TWD and |$\psi $| both returned to control levels. In the more prolonged drought period D2, |$\psi $| levels reached |$-3.0 \pm 0.2 \, \text{MPa}$| and |$-2.7 \pm 0.2 \, \text{MPa}$| for L. decidua and P. sylvestris, respectively. Initially, TWD exhibited an increased diurnal cycle again, but after approximately 2 weeks, stem shrinkage continued without any diameter expansion during the night. By the end of D2, TWD levels reached |$1.3 \pm 0.1 \, \text{mm}$| for L. decidua and |$1.6 \pm 0.2 \, \text{mm}$| for P. sylvestris, translating to a relative shrinkage of |$12.5 \pm 0.6$| and |$10.5 \pm 1.0 \, \%$| and a TWD|$_{norm}$| of |$0.76$| and |$0.73$|. Upon rewetting, plants of the drought–recovery treatment showed an instantaneous recovery of |$\psi $|. In contrast, TWD decreased rapidly after stress release but did not fully return to control values (P|$<0.01$|). For L. decidua, approximately |$85 \pm 3 \%$| of the diameter shrinkage was reversed within 1 day, while for P. sylvestris, this value was |$80 \pm 2 \%$|. Subsequently, both species resumed daily stem diameter variations gradually decreasing TWD. Ultimately, TWD returned to the control values after 8 days for L. decidua and after 9 days for P. sylvestris.
Water status and gas exchange variables during a drought and recovery experiment conducted on potted seedlings of L. decidua (L) and P. sylvestris (P) under controlled conditions. Depicted are water potential (|$\psi $|), tree water deficit (TWD), soil water content (SWC), stomatal conductance (|${\textit g}_{\textit s}$|) and net photosynthesis (|$A_{\text{net}}$|). The values correspond to Figures 2 and 4 and indicate the mean |$\pm $| SE of the control treatment across the experiment, as well as treatment means |$\pm $| SE of the drought treatment at the end of the first and second drought phase (D1 and D2) and of the drought–lethal treatment at the end of the experiment. The last two columns correspond to the time re-watered trees required to fully recover to pre-stress levels after D1 and D2, respectively.
| Parameter | Species | Mean control | End D1 | End D2 | End drought–lethal | Recovery D1 (days) | Recovery D2 (days) |
|---|---|---|---|---|---|---|---|
| |$\psi $| [|$-\text{MPa}$|] | L | 1.1 |$\pm $| 0.02 | 1.2 |$\pm $| 0.1 | 3.0 |$\pm $| 0.2 | 7.0 |$\pm $| 0 | < 3 | < 1 |
| P | 0.6 |$\pm $| 0.02 | 1.0 |$\pm $| 0.1 | 2.7 |$\pm $| 0.2 | 6.5 |$\pm $| 0.5 | < 3 | < 1 | |
| TWD [|$\text{mm}$|] | L | 0.01 |$\pm $| 0.0002 | 0.3 |$\pm $| 0.03 | 1.3 |$\pm $| 0.1 | 1.8 |$\pm $| 0.1 | 0 | 8 |$\pm $| 1 |
| P | 0.04 |$\pm $| 0.0002 | 0.2 |$\pm $| 0.04 | 1.6 |$\pm $| 0.2 | 2.1 |$\pm $| 0.2 | 0 | 9 |$\pm $| 1 | |
| TWD [|$\%$|] | L | 0.1 |$\pm $| 0.001 | 2.7 |$\pm $| 0.2 | 12.5 |$\pm $| 0.6 | 16.4 |$\pm $| 0.4 | ||
| P | 0.2 |$\pm $| 0.001 | 1.3 |$\pm $| 0.2 | 10.5 |$\pm $| 1.0 | 14.3 |$\pm $| 0.8 | |||
| SWC [|$\%$|] | L | 33.6 |$\pm $| 0.3 | 18.9 |$\pm $| 1.8 | 2.8 |$\pm $| 0.6 | 0.1 |$\pm $| 0.1 | 0 | 0 |
| P | 33.5 |$\pm $| 0.2 | 17.8 |$\pm $| 1.6 | 1.9 |$\pm $| 0.7 | 0.2 |$\pm $| 0.2 | 0 | 0 | |
| |${\textit g}_{\textit s}$| [|$\frac{\text{mmol}}{\text{m}^{2}\, \text{s}}$|] | L | 104 |$\pm $| 7 | 12 |$\pm $| 1 | 3 |$\pm $| 0.4 | 3 |$\pm $| 2 | < 3 | 20 |$\pm $| 3 |
| P | 76 |$\pm $| 5 | 10 |$\pm $| 3 | 3 |$\pm $| 0.3 | 3 |$\pm $| 0.5 | < 3 | 10 |$\pm $| 3 | |
| |$A_{net}$| [|$\frac{\mu \text{mol}}{\text{m}^{2}\, \text{s}}$|] | L | 5.4 |$\pm $| 0.3 | 1.9 |$\pm $| 0.2 | 0.0 |$\pm $| 0.04 | -0.1 |$\pm $| 0.03 | < 3 | 12 |$\pm $| 3 |
| P | 6.0 |$\pm $| 0.3 | 1.3 |$\pm $| 0.3 | 0.0 |$\pm $| 0.04 | -0.1 |$\pm $| 0.05 | < 3 | 6 |$\pm $| 3 |
| Parameter | Species | Mean control | End D1 | End D2 | End drought–lethal | Recovery D1 (days) | Recovery D2 (days) |
|---|---|---|---|---|---|---|---|
| |$\psi $| [|$-\text{MPa}$|] | L | 1.1 |$\pm $| 0.02 | 1.2 |$\pm $| 0.1 | 3.0 |$\pm $| 0.2 | 7.0 |$\pm $| 0 | < 3 | < 1 |
| P | 0.6 |$\pm $| 0.02 | 1.0 |$\pm $| 0.1 | 2.7 |$\pm $| 0.2 | 6.5 |$\pm $| 0.5 | < 3 | < 1 | |
| TWD [|$\text{mm}$|] | L | 0.01 |$\pm $| 0.0002 | 0.3 |$\pm $| 0.03 | 1.3 |$\pm $| 0.1 | 1.8 |$\pm $| 0.1 | 0 | 8 |$\pm $| 1 |
| P | 0.04 |$\pm $| 0.0002 | 0.2 |$\pm $| 0.04 | 1.6 |$\pm $| 0.2 | 2.1 |$\pm $| 0.2 | 0 | 9 |$\pm $| 1 | |
| TWD [|$\%$|] | L | 0.1 |$\pm $| 0.001 | 2.7 |$\pm $| 0.2 | 12.5 |$\pm $| 0.6 | 16.4 |$\pm $| 0.4 | ||
| P | 0.2 |$\pm $| 0.001 | 1.3 |$\pm $| 0.2 | 10.5 |$\pm $| 1.0 | 14.3 |$\pm $| 0.8 | |||
| SWC [|$\%$|] | L | 33.6 |$\pm $| 0.3 | 18.9 |$\pm $| 1.8 | 2.8 |$\pm $| 0.6 | 0.1 |$\pm $| 0.1 | 0 | 0 |
| P | 33.5 |$\pm $| 0.2 | 17.8 |$\pm $| 1.6 | 1.9 |$\pm $| 0.7 | 0.2 |$\pm $| 0.2 | 0 | 0 | |
| |${\textit g}_{\textit s}$| [|$\frac{\text{mmol}}{\text{m}^{2}\, \text{s}}$|] | L | 104 |$\pm $| 7 | 12 |$\pm $| 1 | 3 |$\pm $| 0.4 | 3 |$\pm $| 2 | < 3 | 20 |$\pm $| 3 |
| P | 76 |$\pm $| 5 | 10 |$\pm $| 3 | 3 |$\pm $| 0.3 | 3 |$\pm $| 0.5 | < 3 | 10 |$\pm $| 3 | |
| |$A_{net}$| [|$\frac{\mu \text{mol}}{\text{m}^{2}\, \text{s}}$|] | L | 5.4 |$\pm $| 0.3 | 1.9 |$\pm $| 0.2 | 0.0 |$\pm $| 0.04 | -0.1 |$\pm $| 0.03 | < 3 | 12 |$\pm $| 3 |
| P | 6.0 |$\pm $| 0.3 | 1.3 |$\pm $| 0.3 | 0.0 |$\pm $| 0.04 | -0.1 |$\pm $| 0.05 | < 3 | 6 |$\pm $| 3 |
Water status and gas exchange variables during a drought and recovery experiment conducted on potted seedlings of L. decidua (L) and P. sylvestris (P) under controlled conditions. Depicted are water potential (|$\psi $|), tree water deficit (TWD), soil water content (SWC), stomatal conductance (|${\textit g}_{\textit s}$|) and net photosynthesis (|$A_{\text{net}}$|). The values correspond to Figures 2 and 4 and indicate the mean |$\pm $| SE of the control treatment across the experiment, as well as treatment means |$\pm $| SE of the drought treatment at the end of the first and second drought phase (D1 and D2) and of the drought–lethal treatment at the end of the experiment. The last two columns correspond to the time re-watered trees required to fully recover to pre-stress levels after D1 and D2, respectively.
| Parameter | Species | Mean control | End D1 | End D2 | End drought–lethal | Recovery D1 (days) | Recovery D2 (days) |
|---|---|---|---|---|---|---|---|
| |$\psi $| [|$-\text{MPa}$|] | L | 1.1 |$\pm $| 0.02 | 1.2 |$\pm $| 0.1 | 3.0 |$\pm $| 0.2 | 7.0 |$\pm $| 0 | < 3 | < 1 |
| P | 0.6 |$\pm $| 0.02 | 1.0 |$\pm $| 0.1 | 2.7 |$\pm $| 0.2 | 6.5 |$\pm $| 0.5 | < 3 | < 1 | |
| TWD [|$\text{mm}$|] | L | 0.01 |$\pm $| 0.0002 | 0.3 |$\pm $| 0.03 | 1.3 |$\pm $| 0.1 | 1.8 |$\pm $| 0.1 | 0 | 8 |$\pm $| 1 |
| P | 0.04 |$\pm $| 0.0002 | 0.2 |$\pm $| 0.04 | 1.6 |$\pm $| 0.2 | 2.1 |$\pm $| 0.2 | 0 | 9 |$\pm $| 1 | |
| TWD [|$\%$|] | L | 0.1 |$\pm $| 0.001 | 2.7 |$\pm $| 0.2 | 12.5 |$\pm $| 0.6 | 16.4 |$\pm $| 0.4 | ||
| P | 0.2 |$\pm $| 0.001 | 1.3 |$\pm $| 0.2 | 10.5 |$\pm $| 1.0 | 14.3 |$\pm $| 0.8 | |||
| SWC [|$\%$|] | L | 33.6 |$\pm $| 0.3 | 18.9 |$\pm $| 1.8 | 2.8 |$\pm $| 0.6 | 0.1 |$\pm $| 0.1 | 0 | 0 |
| P | 33.5 |$\pm $| 0.2 | 17.8 |$\pm $| 1.6 | 1.9 |$\pm $| 0.7 | 0.2 |$\pm $| 0.2 | 0 | 0 | |
| |${\textit g}_{\textit s}$| [|$\frac{\text{mmol}}{\text{m}^{2}\, \text{s}}$|] | L | 104 |$\pm $| 7 | 12 |$\pm $| 1 | 3 |$\pm $| 0.4 | 3 |$\pm $| 2 | < 3 | 20 |$\pm $| 3 |
| P | 76 |$\pm $| 5 | 10 |$\pm $| 3 | 3 |$\pm $| 0.3 | 3 |$\pm $| 0.5 | < 3 | 10 |$\pm $| 3 | |
| |$A_{net}$| [|$\frac{\mu \text{mol}}{\text{m}^{2}\, \text{s}}$|] | L | 5.4 |$\pm $| 0.3 | 1.9 |$\pm $| 0.2 | 0.0 |$\pm $| 0.04 | -0.1 |$\pm $| 0.03 | < 3 | 12 |$\pm $| 3 |
| P | 6.0 |$\pm $| 0.3 | 1.3 |$\pm $| 0.3 | 0.0 |$\pm $| 0.04 | -0.1 |$\pm $| 0.05 | < 3 | 6 |$\pm $| 3 |
| Parameter | Species | Mean control | End D1 | End D2 | End drought–lethal | Recovery D1 (days) | Recovery D2 (days) |
|---|---|---|---|---|---|---|---|
| |$\psi $| [|$-\text{MPa}$|] | L | 1.1 |$\pm $| 0.02 | 1.2 |$\pm $| 0.1 | 3.0 |$\pm $| 0.2 | 7.0 |$\pm $| 0 | < 3 | < 1 |
| P | 0.6 |$\pm $| 0.02 | 1.0 |$\pm $| 0.1 | 2.7 |$\pm $| 0.2 | 6.5 |$\pm $| 0.5 | < 3 | < 1 | |
| TWD [|$\text{mm}$|] | L | 0.01 |$\pm $| 0.0002 | 0.3 |$\pm $| 0.03 | 1.3 |$\pm $| 0.1 | 1.8 |$\pm $| 0.1 | 0 | 8 |$\pm $| 1 |
| P | 0.04 |$\pm $| 0.0002 | 0.2 |$\pm $| 0.04 | 1.6 |$\pm $| 0.2 | 2.1 |$\pm $| 0.2 | 0 | 9 |$\pm $| 1 | |
| TWD [|$\%$|] | L | 0.1 |$\pm $| 0.001 | 2.7 |$\pm $| 0.2 | 12.5 |$\pm $| 0.6 | 16.4 |$\pm $| 0.4 | ||
| P | 0.2 |$\pm $| 0.001 | 1.3 |$\pm $| 0.2 | 10.5 |$\pm $| 1.0 | 14.3 |$\pm $| 0.8 | |||
| SWC [|$\%$|] | L | 33.6 |$\pm $| 0.3 | 18.9 |$\pm $| 1.8 | 2.8 |$\pm $| 0.6 | 0.1 |$\pm $| 0.1 | 0 | 0 |
| P | 33.5 |$\pm $| 0.2 | 17.8 |$\pm $| 1.6 | 1.9 |$\pm $| 0.7 | 0.2 |$\pm $| 0.2 | 0 | 0 | |
| |${\textit g}_{\textit s}$| [|$\frac{\text{mmol}}{\text{m}^{2}\, \text{s}}$|] | L | 104 |$\pm $| 7 | 12 |$\pm $| 1 | 3 |$\pm $| 0.4 | 3 |$\pm $| 2 | < 3 | 20 |$\pm $| 3 |
| P | 76 |$\pm $| 5 | 10 |$\pm $| 3 | 3 |$\pm $| 0.3 | 3 |$\pm $| 0.5 | < 3 | 10 |$\pm $| 3 | |
| |$A_{net}$| [|$\frac{\mu \text{mol}}{\text{m}^{2}\, \text{s}}$|] | L | 5.4 |$\pm $| 0.3 | 1.9 |$\pm $| 0.2 | 0.0 |$\pm $| 0.04 | -0.1 |$\pm $| 0.03 | < 3 | 12 |$\pm $| 3 |
| P | 6.0 |$\pm $| 0.3 | 1.3 |$\pm $| 0.3 | 0.0 |$\pm $| 0.04 | -0.1 |$\pm $| 0.05 | < 3 | 6 |$\pm $| 3 |
Temporal dynamics of water relations during a drought and recovery experiment conducted on potted seedlings (L. decidua, P. sylvestris) under controlled conditions. Depicted are water potential (|$\psi $|), tree water deficit (TWD) and soil water content (SWC) as treatment means |$\pm $| SE. The two drought phases are marked as D1 and D2, and the recovery phases are illustrated with a blue background. In the final recovery phase, the drought treatment was split equally and only drought–recovery seedlings were re-watered, while drought–lethal trees were fully dehydrated.
Relationship between TWD and |$\psi $|
We found a strong relationship between TWD and |$\psi $| persisting across all drought conditions until lethal dehydration in both L. decidua and P. sylvestris. It can be effectively characterized by a single species-specific sigmoidal fit (Figure 3a and b; pseudo-|$R^{2}$| of |$0.97$| for both species). To better assess the possibility of using TWD as a predictor for |$\psi $| we replotted the data with reversed axes (Figure 3c and d) and found three distinct phases. Phase (I) shows an initial decrease in |$\psi $| with a low TWD response, (II) a steep linear relation indicating decreasing |$\psi $| with increasing TWD and (III) a less pronounced increase in TWD with further decreasing |$\psi $|. The first change of phases occurs at a normalized TWD of |$0.02$| for L. decidua and |$0.04$| for P. sylvestris, respectively, with corresponding |$\psi $| close to |$\psi _{g50}$| (|$\psi $| at |$g_{s}=50\%$| of control). The second shift appears at a |$\text{TWD}_{\text{norm}}$| of |$0.74$| and |$0.81$|, corresponding to |$\psi $| values of |$-3.0 \, \text{MPa}$| and |$-2.9 \, \text{MPa}$| (in between P|$_{12}$| and P|$_{50}$|). The |$R^{2}$| is lowest with |$0.64$| in phase (I) for L. decidua, while in all other cases |$R^{2}$| is |$\geq \, 0.88$| for both species.
Relationship between midday water potential (|$\psi $|) and normalized tree water deficit (|$\text{TWD}_{\text{norm}}$|) during a drought and recovery experiment conducted on potted seedlings (L. decidua, P. sylvestris) under controlled conditions. In the upper row (a, b), we applied a sigmoidal fit with |$\psi $| as the predictor variable. In the bottom row (c, d), we reversed axes and split the data into three distinct phases with linear regressions (excluding the drought–recovery treatment). Data are treatment means |$\pm $| SE. The TWD was derived from measurements between 1 and 4 p.m. corresponding to the |$\psi $| measurements. Normalized TWD was derived relative to the largest averaged TWD in the drought–lethal trees. Shaded areas depict the |$95\%$| confidence intervals of the fitted regressions.
However, the responses of the drought–recovery treatment (green triangles in Figure 3a and b) could not adequately be represented by the |$\psi $|–TWD relationship as observed during continuous drought development. Upon rewetting after D2, immediate recovered |$\psi $| values correspond to TWD persisting at levels higher than anticipated by the sigmoidal fit. In contrast, rewetting after D1 did not exhibit such prolonged effects on TWD.
Taking a closer look at species-specific TWD–|$\psi $| responses, we found slight differences in timing and depletion rates. Specifically, TWD accumulates at lower |$\psi $| levels for L. decidua compared with P. sylvestris. The point at which |$50\%$| of water is depleted (c parameter in the sigmoidal fit) occurred significantly later for L. decidua compared with P. sylvestris: |$-2.3 \pm 0.1 \, \text{MPa}$| versus |$-2.0 \pm 0.1 \, \text{MPa}$| (P|$<0.01$|). Additionally, L. decidua exhibited a higher shrinkage rate per |$\text{MPa}$||$\psi $| (b parameter in the sigmoidal fit, P|$<0.01$|).
Leaf gas exchange during drought–recovery cycles
Leaf gas exchange was highly sensitive to both drought and rewetting conditions (Figure 4, Table 2). At the end of D1, |$\psi $| reached |$1.2 \pm 0.1 \, \text{MPa}$| for L. decidua and |$1.0 \pm 0.1 \, \text{MPa}$| for P. sylvestris. Accordingly, |${\textit g}_{\textit s}$| was reduced significantly (P|$<0.01$|) to |$11.5$| and |$13.3\%$| of control values, respectively, while |$A_{net}$| dropped to |$35.2$| and |$21.7\%$|. The |$\psi _{g5}$| of full stomatal closure was reached in the first half of D2. During the remaining drought period, |${\textit g}_{\textit s}$| remained at its minimum and |$A_{net}$| approached zero. Leaf gas exchange fluxes quickly returned to control values after D1 but required time for full recovery following D2 (Table 2). In particular, |${\textit g}_{\textit s}$| recovery after D2 took approximately |$20 \pm 3$| days for L. decidua and |$10 \pm 3$| days for P. sylvestris. Similarly, the recovery period for |$A_{net}$| was approximately |$12 \pm 3$| in L. decidua and |$6 \pm 3$| days in P. sylvestris. Hence, the gas exchange of P. sylvestris tended to recover faster compared with L. decidua, with |${\textit g}_{\textit s}$| recovering generally slower than |$A_{net}$| in both species.
Temporal dynamics of leaf gas exchange in the course of a drought and recovery experiment conducted on potted seedlings (L. decidua, P. sylvestris) under controlled conditions. Depicted are stomatal conductance (|${\textit g}_{\textit s}$|) and net photosynthesis (|$A_{net}$|) as treatment means |$\pm $| SE. The two drought phases are marked as D1 and D2, and the recovery phases are illustrated with a blue background. In the final recovery phase, the drought treatment was split equally and only drought–recovery seedlings were re-watered, while drought–lethal trees were fully dehydrated.
Relationship between TWD and |${\textit g}_{\textit s}$|
We found a strong exponential relationship between |${\textit g}_{\textit s}$| and both the midday (Figure 5a and b) and pre-dawn TWD (Figure 5c and d) during dehydration. Corresponding pseudo-|$R^{2}$| values are |$0.87$| in both cases for L. decidua and |$0.77$| and |$0.81$| for midday and pre-dawn TWD, respectively, for P. sylvestris. Small increases in TWD resulted in drastic decreases in gas exchange, which appeared more pronounced for pre-dawn than for midday TWD, highlighting the sensitivity of |${\textit g}_{\textit s}$| to water deficit. While midday TWD allows estimations of partial stomatal closure, the pre-dawn TWD signal provided a clearer response at full stomatal closure, indicated by a phase shift at TWD >0. In more detail, at a midday TWD|$_{norm}$| of |$0.06$| for L. decidua and |$0.08$| for P. sylvestris|$50\%$| stomatal closure was reached (md-TWD|$_{g50}$|), shortly after the first shift in the TWD–|$\psi $| relation and in a range where pre-dawn TWD is still |$0$|. Stomatal closure became apparent at midday TWD values of |$0.23$| and |$0.25$| for L. decidua and P. sylvestris, respectively, corresponding to pre-dawn TWD values of only |$0.02$| and |$0.07$| (md- and pd-TWD|$_{g5}$|).
Relationship between midday leaf-level stomatal conductance (|${\textit{g}}_{\textit{s}}$|) and normalized tree water deficit (|$\text{TWD}_{ \text{norm}}$|) during a drought and recovery experiment conducted on potted seedlings (L. decidua, P. sylvestris) under controlled conditions. In (a, b), |${\textit{g}}_{\textit{s}}$| is related to midday (md) TWD, derived from measurements between 1 and 4 p.m. corresponding to the |${\textit{g}}_{\textit{s}}$| measurements. In (c, d), |${\textit{g}}_{\textit{s}}$| is related to pre-dawn (pd) TWD, derived from measurements between 3 and 4 a.m. We applied exponential fits (black) neglecting the drought–recovery data points with the shaded area indicating the |$95\%$| confidence interval of the curve. Additionally, the |$\text{TWD}_{ \text{norm}}$| at |${\textit{g}}_{\textit{s}}$| equaling |$5\%$| (a–d; TWD|$_{g5}$|) and |$50\%$| (a–b; TWD|$_{g50}$|) of the control mean (Mean(C)) was determined. Data are treatment means |$\pm $| SE. Normalized TWD was derived relative to the largest averaged TWD in the drought–lethal trees.
Due to different recovery behaviors of TWD and |${\textit{g}}_{\textit{s}}$|, the drought–recovery treatment following the final rewetting was slightly outside the expectation of the exponential curves for both species. However, we found species-specific differences. For L. decidua, |${\textit{g}}_{\textit{s}}$| during drought–recovery was partly lower than anticipated by the regression curve due to a delayed gas exchange recovery. Conversely, for P. sylvestris, |${\textit{g}}_{\textit{s}}$| deviated from the curve with values higher than expected, indicating a delay in TWD recovery. While the recovery of |${\textit{g}}_{\textit{s}}$| and TWD showed similar trends, it differed in timing. |${\textit{g}}_{\textit{s}}$| exhibited a gradual recovery, whereas TWD initially showed a rapid response followed by a gradual decrease.
Discussion
Our data confirm strong relationships between |$\psi $| and TWD, as well as |${\textit{g}}_{\textit{s}}$| and TWD. However, these relationships became unstable during recovery after severe drought. While dendrometer-based TWD measurements show promise for predicting |$\psi $| (Dietrich et al. 2018) and can serve as a drought stress indicator, caution is required when including periods of drought recovery in the analysis.
How does TWD develop with decreasing |$\psi $|?
The relationship between TWD and |$\psi $| was robust for both investigated tree species, including severe and lethal drought stress, but not during the recovery phase after D2 (see section 4.2). Both species showed similar TWD patterns, supporting previous studies indicating minor differences in drought susceptibility (Dulamsuren et al. 2019, Feng et al. 2021). Nevertheless, we found species-specific disparities. L. decidua exhibited a weaker stomatal control (Peters et al. 2023) and responded more anisohydric than P. sylvestris (Oberhuber et al. 2015). Additionally, L. decidua demonstrated a greater rate of shrinkage per |$\psi $| decrease during dehydration, aligning with expectations of high transpiration rates even under dry conditions (Leštianska et al. 2020).
We observed shifts in the TWD–|$\psi $| relationship with increasing water shortage, representable by either a single sigmoidal curve or distinct linear fits. Sigmoidal regressions were proposed by Dietrich et al. (2018) and the authors expect a saturation of TWD at the lowest |$\psi $| they have measured (about |$-2.5\,\text{MPa}$| for L. decidua and |$-1.8\,\text{MPa}$| for P. sylvestris). In contrast, our data suggest that TWD has not reached half its maximum at this |$\psi $|. Besides an overall high correlation, the sigmoidal fit fails to capture the development under the lowest |$\psi $| levels for our data. After expected tree death, as suggested by |$\psi $| measurements, it seems that the remaining water availability in elastic tissues did not lead to an asymptotic TWD saturation but a continued shrinkage, as demonstrated in phase (III) (Figure 3 c and d) and supported by similar findings of Andriantelomanana et al. (2024) for Populus tremula. This also induces difficulties selecting TWD|$_{max}$|, which is crucial for normalization of the TWD signal (see section 4.4). Linear TWD–|$\psi $| relationships were also found for the tropical species Callitris intratropica (Drew et al. 2011) and the temperate species Juglans regia (Cochard et al. 2001), Picea abies and Larix decidua (Oberhuber et al. 2015). However, none of these studies captured severe drought stress and slope shifts in the TWD–|$\psi $| relationship were consequently not present. Overall, the linear approach appears more realistic and applicable. While the sigmoidal method requires a complex inversion for |$\psi $| retrievals, the linear procedure allows direct predictions. Although defining additional inclination points may be a practical limitation, these inclination points could serve as valuable stress-severity thresholds. However, caution is required when extending regression curves beyond their calibrated range. For instance, Vergeynst et al. (2015) suggest extrapolating a single linear |$\psi $|–TWD regression fit for continuous |$\psi $| determination. With a slope change appearing at the onset of hydraulic damage, as our data suggest, such an extrapolation will lead to errors under severe drought stress (here in phase (III)). Also, predicting |$\psi $| under near-zero TWD is difficult despite a linear correlation, while in phases (II) and (III), TWD becomes a better predictor as dehydration accumulates.
We propose that the two slope changes in the TWD–|$\psi $| relationship (Figure 3c and d) can serve as thresholds for the onset of initial drought stress and the loss of instant recovery potential. While often a subsequent release of water from elastic and inelastic tissues during dehydration is proposed, Lauriks et al. (2022) suggest an overlapping, and Andriantelomanana et al. (2024) a fully simultaneous succession. Our results support partially overlapped water release from elastic and inelastic tissues and distinct phases indicating different drought severity and recovery potential. For phase (I), we hypothesize a parallel release of apoplastic and symplastic water with apparent but low volume loss. At this stage, the TWD vanishes at night, indicating no soil water limitation. The shift to phase (II) seems to be triggered by stomatal closure, as it occurs around |$\psi _{g50}$|. Further dehydration results in complete stomatal closure within the phase. Here, accumulation of TWD due to water release purely stems from elastic tissues, leading to the highest depletion rate. In contrast to the observation of Andriantelomanana et al. (2024), our data suggest no water release from inelastic tissues until the end of the phase according to the determined vulnerability curves and measured |$\psi $|. We assume that the second shift in slope corresponds to the onset of hydraulic damages as it occurs between P|$_{12}$| and P|$_{50}$|. In phase (III), P|$_{88}$| is accompanied by less rapid but nonetheless persistent shrinkage, indicating further release of symplastic water, joined by water release from embolized xylem conduits. Mantova et al. (2022) argue that cavitation events trigger the collapse of living cells, i.e. cell necrosis due to mechanical or oxidative stress. In particular, the loss of meristematic cells is critical since they play a crucial role in the survival and recovery of trees. While embolized xylem conduits are lost, secondary growth building new xylem can improve conductance. However, this requires surviving meristematic cells with a functional water supply (Mantova et al. 2022). Water released by xylem embolism may be used to sustain hydration of elastic tissues (Skelton 2020), and hence delaying cell collapse (Knipfer et al. 2019, Lamacque et al. 2020). Further shrinkage in phase (III) suggests increasing PLC and meristematic cell loss, leading to longer recovery after stress release. At a certain threshold, the recovery potential may be lost completely, resulting in tree death (Lamacque et al. 2020, Andriantelomanana et al. 2024).
Is the relationship between TWD and |$\psi $| impacted by stress release?
Since physiological processes recover at different rates after severe drought (Ruehr et al. 2019, Rehschuh et al. 2020), we expected a changed TWD–|$\psi $| relationship during recovery. While the severe drought phase D2 caused changes, the mild drought conditions in D1 had negligible impact. During D1, the highest occurring TWD was within the early part of phase (II), and stress release led to an instantaneous recovery of all physiological parameters, shifting the TWD–|$\psi $| relationship back to its initial stage. In contrast, at the end of D2, P. sylvestris trees were within the latter part of phase (II) and L. decidua trees in the early part of phase (III). Despite the rapid increase of |$\psi $| to levels similar to controls, TWD did not fully recover for several days in both species. Interestingly, the diameter increase showed similarities to a capacitor recharge curve, and further investigations might enable valuable insights into tree-water relations from TWD recovery, such as capacitance determination. In the TWD–|$\psi $| relationship, we observed a recovery hysteresis effect. The drought–recovery treatment deviated from the curve (Figure 3a and b) but gradually approached it again while TWD recovered back to control values. Hence, during this phase, the relationship does not hold. The distinct recovery behavior after D1 and D2 further supports our argument that the trees lose their potential for instantaneous recovery at the shift from phase (II) to phase (III).
However, whether the incompletely recovered stem diameter after rewetting still indicates a water deficit is arguable. By definition, TWD is greater zero if the current diameter is smaller than the maximum diameter in the past (see Eq. (2)). On the other hand, TWD is supposed to reflect the condition where elastic tissues are not fully hydrated and, consequently, not at their maximum volume. The sharp diameter increase after stress release stopped after approximately one day without full recovery to pre-stress conditions, eventually caused by an impairment of the xylem-phloem connection increasing radial resistance (Steppe et al. 2006, Salomón et al. 2017). Another explanation might be the collapse of elastic cells, which can not rehydrate afterwards (Lamacque et al. 2020, Fortunel et al. 2023). In this case, the disparity in diameter observed one day after stress release would indicate accumulated cell damage rather than a persistent water deficit.
Thus, the incomplete recovery of stem diameter after stress release may be a proxy for assessing cell damage and, consequently, recovery potential or even mortality risk. Previous studies have identified specific thresholds for TWD beyond which mortality occurs. Lavandula seedlings perished once they reached a stem shrinkage of approximately |$21\%$| (Lamacque et al. 2020), while Populus tremula seedlings succumbed at values ranging from |$7$| to |$12\%$|, depending on growth conditions (Andriantelomanana et al. 2024). In our study, the maximum stem shrinkage reached during D2 was |$12.5 \pm 0.6\%$| for L. decidua and |$10.5 \pm 1.0\%$| for P. sylvestris. Although none of the trees died, they all required a prolonged recovery time and can be supposed to have experienced damaged tissue. In L. decidua, |$85\%$| of the diameter shrinkage reversed after stress release, which may indicate that roughly |$15\%$| of living parenchyma or meristematic cells were irreversibly damaged. For P. sylvestris this value corresponds to |$20\%$|. Despite higher |$\psi _{min}$| values before rewetting compared with L. decidua and similar |$P_{50}$|, P. sylvestris exhibited higher tissue damage susceptibility. Notably, damage was not directly quantified in this study, and the exact cause of the incomplete diameter recovery is debatable. Further investigations of partial TWD recovery upon various stress levels and TWD thresholds for mortality may yield promising metrics enabling drought damage predictions, especially if they include in-situ embolism and cell collapse quantification.
How are TWD and gas exchange related?
The exponential decline of |${\textit{g}}_{\textit{s}}$| with increasing TWD is consistent with findings from two Lavandula species (Lamacque et al. 2020). While the initial decrease in |$\psi $| does not directly limit |${\textit{g}}_{\textit{s}}$| (Rehschuh et al. 2020), the first increase in TWD already signified drought stress, triggering partial stomatal closure and subsequent declines in gas exchange. Once TWD did not fully recover during the night, leading to a pre-dawn TWD greater than |$0$|, stomata were closed or nearly closed the following midday—a stress signal directly applicable to field conditions. However, a direct quantitative prediction of |${\textit{g}}_{\textit{s}}$| appears challenging due to the sensitive response between |${\textit{g}}_{\textit{s}}$| and TWD. Alternatively, TWD could be used to estimate |$\psi $|, facilitating a more straightforward prediction of the gas exchange rates with hydraulic models (Anderegg et al. 2017, Eller et al. 2018). Besides quantifying |${\textit{g}}_{\textit{s}}$| relations, correlations between TWD and transpiration might reveal interesting insights, especially if one considers canopy-level transpiration as a measure of whole tree water loss under variable VPD (Peters et al. 2023).
Upon stress release after D2, we observed hysteresis effects in the |${\textit{g}}_{\textit{s}}$|–TWD relationships due to asynchronous TWD and gas exchange recovery. The restoration of pre-stress levels seems to depend on tissue damage and recovery strategies. In general, the recovery of stomatal conductance is complex and relies on species-specific factors such as genotype, abscisic acid concentration, hydraulic resistance and transpiration rates (Buckley 2019). In the case of L. decidua, a higher percentage of shrinkage was reversed directly after stress release, resulting in a slightly faster TWD recovery. Pinus sylvestris showed less immediate TWD recovery but a faster recovery in |${\textit{g}}_{\textit{s}}$| and the carbon assimilation rate |$A_{\text{net}}$|. This differential recovery may indicate distinct priorities, with P. sylvestris emphasizing gas exchange recovery. In contrast to D2, the mild drought D1 did not exhibit changes in the |${\textit{g}}_{\textit{s}}$|–TWD relationship, aligning with the observation that gas exchange only requires prolonged recovery after hydraulic impairments occurred (Skelton et al. 2017).
The thresholds at which stomata close in response to increasing TWD are likely species specific (Lamacque et al. 2020, Andriantelomanana et al. 2024) and reflect individual water-use strategies. Trees rely on internal water reservoirs during drought, resulting in varying degrees of stomatal closure depending on the species’ tendency to limit water loss. However, in our study, both L. decidua and P. sylvestris exhibited nearly identical |${\textit{g}}_{\textit{s}}$|–TWD relationships, despite differences in water use strategies (Oberhuber et al. 2015, Martín-Gómez et al. 2023) and different |$\psi _{g50}$| and |$\psi _{g5}$| values. Consequently, stomatal closure developed at similar TWD thresholds. These thresholds likely depend on stomatal density, leaf traits and the possibility of reducing internal conductivity by non-stomatal effects. For example, Lavandula species fully close stomata at a stem shrinkage of |${\approx} {5}\%$| (Lamacque et al. 2020), while in our case md-TWD|$_{g5}$| translates to values of approximately |$2\%$| for L. decidua and |$1.7\%$| for P. sylvestris, indicating a more sensitive stomatal response. Populus tremula reacts even faster, with full stomatal closure reported at the onset of midday TWD accumulation (Andriantelomanana et al. 2024). Investigating such responses offers a promising approach to assessing water-use strategies. Besides, TWD thresholds for stomatal closure could serve as warning signals in field conditions. Practically, monitoring the occurrence of pre-dawn TWD might be a simple and directly applicable measure to predict stomatal closure. Applying midday TWD|$_{g50}$| and TWD|$_{g5}$| appears to provide more detailed insights into the degree of stomatal opening. However, transferring these thresholds (and |$\psi $|–TWD relations) requires normalization of TWD signals, a challenge still limiting practical use.
The challenge of normalizing TWD
A significant challenge in applying TWD relationships in field conditions is the normalization with TWD|${}_{max}$|, especially when lethal conditions are not captured by measurements. In our study, we defined TWD|${}_{max}$| as the mean maximum TWD of drought–lethal trees at the end of the experiment. However, post-mortem shrinkage complicates this choice, as defining TWD|${}_{max}$| at a different time would alter thresholds and relationship functions. Additionally, transferring TWD|${}_{max}$| to different tree sizes and environmental conditions requires careful consideration, given that tissue shrinkage depends on factors such as bark-to-xylem ratio, wood density, elastic modulus and growth conditions (Andriantelomanana et al. 2024). For instance, P. sylvestris exhibited greater absolute shrinkage than L. decidua, but the latter showed a larger relative shrinkage (Table 2), possibly indicating more water storage in elastic cells, supported by a thicker bark and a lower bark-to-xylem ratio (|$0.48$| vs |$0.25$|; Figure S7 available as Supplementary data at Tree Physiology Online). Future research should aim for a more stable and transferable normalization, e.g. by differentiating between bark and xylem shrinkage, or by experimenting with different normalization variables such as maximum daily shrinkage (Peters et al. 2023).
Conclusions
Dendrometer-derived TWD provides direct insights into the water dynamics of individual trees by enabling continuous |$\psi $| determination, the assessment of stomatal closure and the application of tree drought stress thresholds. However, the normalization of the TWD data remains a challenge and affects transferability. Additionally, recovery phases following severe droughts can alter the TWD relationships to |$\psi $| and |${\textit{g}}_{\textit{s}}$|. On the positive side, the incomplete diameter recovery after stress release provides an opportunity to quantify drought-induced damage and estimate recovery times. Dendrometer networks, capable of collecting data from numerous trees, allow the evaluation of tree functionality at larger scales. These networks could serve as early warning systems, detecting initial drought onset as well as severe drought conditions. A persisting TWD not refilled during the night is a clear drought stress signal, accompanied by drastically reduced midday leaf gas exchange. Further TWD accumulation can then hint towards critical damage to tree functionality and, ultimately, the risk of tree mortality. Identifying susceptible species and locations could guide mitigation efforts and appropriate management actions. Applications extend to urban environments for tree watering practices and forest ecosystems to support comprehensive monitoring efforts.
Acknowledgments
We thank Prof. Stefan Mayr from the University of Innsbruck for the possibility of using the Cavitron infrastructure in his lab.
Authors’ contributions
Y.Z., N.K.R. and R.G. designed the experiment. Y.Z., supported by F.A. and T.K., performed the measurements. Y.Z. analyzed the data and wrote the manuscript, with input and approval from all authors.
Funding
The study has been funded through the Helmholtz Initiative and Networking fund (W2/W3-156).
Conflict of interest
None declared.
Data availability
The data that underlie this study are available upon request.
