1 Introduction

The moisture content of vegetation, defined as the ratio of water to dry mass, is a major factor of fire danger across ecosystems worldwide (Finney et al. 2021). Live fuel moisture content (LFMC) refers to the moisture content of leaves and fine twigs in the canopy (Yebra et al. 2013), while dead fuel moisture content (DFMC) is the moisture content of fine dead material, such as litter (Matthews 2014). Recent evidence suggests that LFMC could play a substantial role in wildfires as it has been linked to increased burnt area (Nolan et al. 2016; Pimont et al. 2019a; Rao et al. 2022), the occurrence of extreme wildfire events (Ruffault et al. 2018a), and accelerated rates of spread in shrubland fires (Pimont et al. 2019b).

Predicting LFMC dynamics is a fundamental challenge in the face of climate change with increasing drought events (Ruffault et al. 2020; Torres‐Ruiz et al. 2024), but our understanding of its spatial and temporal patterns remains limited. Several factors contribute to this uncertainty. First, LFMC is not only a function of weather conditions but also depends on species-specific traits, soil characteristic, and microclimate conditions that influence vegetation response to drought (Jolly and Johnson 2018; Nolan et al. 2020; Ruffault et al. 2023). Standard meteorological-based indices, including the drought code of the fire weather index (Van Wagner 1987), inadequately account for this complexity, resulting in limited predictive performance (Ruffault et al. 2018b). Furthermore, the definition of LFMC remains somewhat ambiguous. Some studies define LFMC at the leaf level, aligning with ground-based measurements (Martin-StPaul et al. 2018; Gabriel et al. 2021), while others define LFMC at the canopy level, integrating canopy heterogeneity and therefore matching with the scale of remote-sensing measurements (Yebra et al. 2013). To address this discrepancy, Ruffault et al. (2023) proposed differentiating LFMC as the fuel moisture content of individual leaves from the canopy fuel moisture content (CFMC) at the plant or canopy scale, a definition that is adopted in this study.

Plant desiccation under drought depends on a suite of plant and leaf anatomical and functional traits, ranging from characteristics of leaf cells to whole plant hydraulic architecture (Jolly and Johnson 2018; Nolan et al. 2020). Pressure–volume (p–v) curves (Tyree and Hammel 1972; Bartlett et al. 2012), which depict the relative water content of living cells as a function of leaf water potential (Ψleaf), have shown to offer a relevant framework to predict LFMC dynamics (Nolan et al. 2018; Pivovaroff et al. 2019; Scarff et al. 2021; Ruffault et al. 2023). According to this framework, the response of LFMC to Ψleaf is driven by several leaf traits, including the modulus of cell wall elasticity (ε, MPa), the osmotic water potential at full turgor (π0, MPa), the leaf dry matter content (LDMC), and the fraction of symplasmic compartment in the leaves (αf). By considering these leaf traits, it is possible to evaluate how LFMC dynamics may vary among species as Ψleaf decreases during drought (Nolan et al. 2020).

Accumulating evidence also shows that physiological processes related to plant hydraulic mechanisms should be taken into account for CFMC predictions (Ruthrof et al. 2016; Ruffault et al. 2023; Torres‐Ruiz et al. 2024). Indeed, decreased plant water potentials induced by drought can lead to the cavitation of plant vessels and diminish CFMC either by decreasing the amount of water in organs affected by cavitation or by initiating foliage and branch mortality (Hölttä et al. 2009; Martin-StPaul et al. 2017). Hydraulic segmentation, which posits that leaves are more susceptible to drought-induced embolism than stems to protect vital perennial organs from irreversible hydraulic dysfunction (Pivovaroff et al. 2014; Johnson et al. 2016), is also a critical factor to consider to predict the rates of leaf mortality with increasing drought. However, the relative importance of these physiological mechanisms in CFMC dynamics has yet to be quantified, as well as the extent to which they might alter CFMC both during and after drought.

Four different hypotheses (summarized in Fig. 1) can be proposed to explain the physiological mechanisms underlying tree desiccation during extreme drought. The first hypothesis (H1) suggests that changes in CFMC are solely due to reversible variations in LFMC linked to the symplasmic water content (as explained by p–v curves). Conversely, assuming that hydraulic-induced damages are also involved in CFMC changes during drought, three alternative hypotheses can be made, depending on the scale at which hydraulic damages occur. If hydraulic damages occur at leaf scale only (H2), then apoplasmic leaf compartments are affected, and both LFMC and CFMC would not fully recover after drought. Alternatively, hydraulic failure could trigger foliage or branch mortality (H3), which means that the ratio of dead-to-live fuel (αDead) increases with decreasing Ψleaf, lowering the CFMC, but that LFMC fully recovers after rewatering. Last, a fourth hypothesis (H4) is that both mechanisms at the leaf and plant scales occur simultaneously, in proportions that remain to be quantified.

Fig. 1
figure 1

Schematic representation of the drought experiment and principal hypotheses (H) investigated in this study to understand the physiological mechanisms underlying seedings desiccation during and after drought. The figure depicts the theoretical dynamics of leaf-scale live fuel moisture content (LFMC, orange lines), plant-level canopy fuel moisture content (CFMC, green lines), and the proportion of dead foliage within the plant canopy (αDead, brown dashed line) for each hypothesis. H1: LFMC and CFMC fully recover to their initial levels after plant rewatering. H2: LFMC and CFMC exhibit incomplete recovery due to drought-induced leaf-scale damages. H3: LFMC recovers to its initial value, while CFMC remains incomplete due to drought-induced foliage mortality (increase in αDead). H4: LFMC and CFMC both exhibit incomplete recovery due to both drought-induced leaf-scale damages and foliage mortality (increase in αDead)

Full size image

In this study, we investigated the impacts of a severe drought on the dynamics of CFMC and LFMC for two Mediterranean tree species (Pinus halepensis Mill. and Quercus ilex L.) both during and after (recovery) drought. We subjected seedlings of these two species to varying levels of severe drought-induced soil water deficits and assessed their recovery under each of these drought conditions. Then, LFMC, αDead, and CFMC for the seedlings were simulated based on Ψleaf using process-based models derived from p–v curves and vulnerability to xylem embolism. Specifically, we aimed to investigate (1) the predictability of LFMC using p–v curves; (2) the influence of hydraulic damage on CFMC, including effects on leaf tissues and/or foliage and branch mortality; and (3) potential differences in LFMC and CFMC dynamics between the two species.

2 Material and methods

2.1 Plant material and experimental design

The experiment involved seedlings of two Mediterranean species: Aleppo pine (Pinus halepensis Mill.) and Holm oak (Quercus ilex L.). In February 2019, 84 2-year-old seedlings (42 per species) were carefully selected to ensure homogeneity in terms of size (height: 30 to 50 cm), crown volume, and foliage density. In early March 2019, the plants were transplanted into 2.5-l cylindrical plastic containers filled with organic soil comprising 50% topsoil, 32% peat, and 18% sand. From March 2019 to June 2019, plants were grown at the French National Forestry Office experimental forest nursery in Cadarache (lat.: 43.91°, long.: 4.87°; France). After this growth period, all potted plants were brought to the campus of INRAE at Avignon (lat.: 43.91°, long.: 4.87°; France) to start the experiment in the greenhouse facility of the campus. The experiment started on the 7th of July 2019 and ended on the 14th of September 2019. It was conducted in a greenhouse of type S2 comprising four compartments, each measuring 65 m2. This greenhouse is equipped with two ventilated air temperature and relative humidity ventilated loggers (HD 9817T1) as well as two radiation loggers. The greenhouse temperature was maintained between 25 and 35 °C, the relative humidity (RH) was maintained between 40 and 75%, and the maximum diurnal photosynthetically active radiation (PAR) below 1000 μmol.m−2.s−1. All pots were saturated with water and covered with a plastic film to prevent soil evaporation.

Plants were subjected to various levels of water deficit and ecophysiological variables, and fuel moisture content was monitored during the drought period and after plant rewatering. Starting from full watering at the beginning of the experiment (day 0), and without further precipitation, six measurement campaigns were conducted at different time intervals (from day 0 to day 45) to cover a wide range of water potentials for each species (see measurement days in Appendix Table 3). On each sampling date (referred to as date Dd), a batch of six plants per species was selected to measure various ecophysiological variables (described below). At the end of each measurement day, the plants were randomly divided into two subsets of three plants each. One subset of three plants (referred to as droughted plants) was used for destructive estimation of moisture content (see below), while the other subset (referred to as recovery plants) was rehydrated to saturation to assess the recovery of physiological variables and moisture content. For the three recovery plants, the soil was maintained above field capacity, and additional measurements were conducted 2 days (Dd+2) and 6 days (Dd+6) after rehydration. Six days after rehydration, destructive estimations of moisture content were performed. A batch of six plants per species was used as a control treatment from the beginning of the experiment. For these plants, soil water was kept close to or above field capacity by watering the pots every 2 to 4 days. In addition to measuring live plants, we also examined the moisture content of litter to determine if the moisture levels of recently deceased leaves and litter exhibit similar patterns. Two batches of dead pine needles and oak leaves obtained from litter collected at a nearby mature forest with the same species (Moreno et al. 2021) were placed in the greenhouse at the start of the experiment to monitor the moisture content of the litter. The datasets acquired for this study are fully accessible in Ruffault and Martin-StPaul (2024).

2.2 Ecophysiological measurements

For each measurement campaign, water potentials and leaf gas exchanges were measured at Dd, Dd+2, and Dd+6. Plant transpiration was estimated at pot level through weightings. Water loss at the pot level was estimated through manual weight measurements, which were conducted twice daily using a FC6CCE-HX Sartorius balance: once in the morning at 8 am and once in the afternoon at 5 pm. For each campaign, leaf water potential was measured at predawn (Ψpredaw) and at midday (Ψmidday) for the targeted plants at Dd, Dd+2, and Dd+6. Samples were collected before sunrise (between 4 and 5 am) and between 1 and 3 pm for Ψpredaw and Ψmidday measurements, respectively. Water potential measurements were made using a Scholander pressure chamber (PMS model 1505 D) immediately after sample collection.

To examine the hypotheses linking foliage desiccation to cavitation levels, we compiled previously published data on vulnerability curves related to cavitation for the two species under study. For Quercus ilex, we gathered vulnerability curves for both leaves and stems. Stem measurements were obtained from Sergent et al. (2020) conducted on 1-m-long branches sampled from 10 individual trees using Cavitron techniques (Burlett et al. 2022). Vulnerability curves for Quercus ilex leaves were extracted from Moreno (2022) and were derived from samples collected at the Font-Blanche site using optical techniques. The Font-Blanche long-term experimental monitoring site is located in a mixed Mediterranean forest in South-Eastern France (43°14′27″N, 5°40′45 ″E; altitude 425 m above sea level). The site is dominated by P. halepensis in the upper tree stratum (average height, 13 m) and Q. ilex in the intermediate tree stratum (average height, 5 m). Vulnerability curves for Pinus halepensis were extracted from Moreno (2022), and these measurements were obtained using Cavitron techniques on 15 samples collected at the Font-Blanche site in 2018. Vulnerability curves were modelled according to the procedure described by Sergent et al. (2020).

2.3 Moisture content measurements

At the end of each measurement day (i.e., droughted plants at Dd and recovered plants at Dd+6), plants were designated for destructive measurements of moisture content. The aboveground parts of each plant were promptly divided into distinct compartments: young leaves, mature leaves, dead leaves, and wood. The identification of dead leaves or needles was conducted by comparing their colors with those of the control plants. For pine trees, needles exhibiting a brown, yellow, or desiccated appearance were categorized as dead. However, for oak trees, only leaves displaying a brown or desiccated state were deemed dead. It is important to highlight that the distinction between live and dead leaves was more straightforward for oak trees due to their distinct brown coloration upon death. In contrast, the identification process for pine trees was more nuanced, as yellowing was also considered an indication of death. Each of these classes was weighed fresh and then put in the oven for 48 h at 60 °C to obtain the dry mass. The moisture content (FMC) of the plant and its organs (leaf and wood) were calculated on a dry mass basis such as the following:

$$FMC =100 \frac{{F}_{W}-{D}_{W}}{{D}_{W}}$$
(1)

where \({F}_{W}\) is the fresh weight (g) and \({D}_{W}\) is the dry weight (g).

Furthermore, for each measurement campaign, three pots (50 cl) of dead pine needles and dead oak leaves were collected from litter batches to estimate the moisture content of dead fuels.

2.4 Soil water content

For each measurement campaign, the normalized soil water content (\({W}_{norm}\)) was computed at Dd, Dd+2, and Dd+6 for each pot using the following formula:

$${W}_{norm} = \frac{W-{W}_{r}}{{W}_{Sat}- {W}_{r}}$$
(2)

where \(W\) is the soil mass of the pot, \({W}_{r}\) is the soil mass at residual water content, and \({W}_{sat}\) is the water content at saturation. \({W}_{r}\) was measured for each pot at the end of the experiment after drying the soil at 70 °C. \({W}_{sat}\) was estimated at the beginning of the experiment, after the pots were irrigated at saturation. The mass of fresh trees was considered to be constant throughout the experiment.

2.5 Testing LFMC recovery after drought

To investigate whether LFMC fully recovered after drought (in accordance with H1 and H3; Fig. 1) or did not fully recover due to leaf-level hydraulic dysfunction (in accordance with H2 and H4; Fig. 1), we examined the influence of the minimum water potential reached by the plant (Ψmin) on LFMC recovery. For this purpose, among all recovered samples, we selected two groups for each species: rewatered leaves that were previously submitted to extreme drought (i.e.,Ψmin <  − 4 MPa for oaks and <  − 3.5 MPa for pines, respectively) and those that were not previously submitted to extreme drought (Ψmin >  − 3 MPa for oaks and Ψmin >  − 2.5 MPa for pines). These thresholds were adjusted empirically based on observed effects. We compared the LFMC of rewatered leaves in these two groups.

2.6 Model analysis

We developed a semiempirical modeling framework for estimating the canopy fuel moisture content (CFMC) based on the hypothesis developed in Fig. 1. Following Ruffault et al. (2023), CFMC can be computed as the weighted sum of live fuel moisture content (LFMC) and dead leaf fuel moisture content (DFMC) in foliage as follows:

$$CFMC= \left(1-{\alpha }_{Dead}\right)LFMC+{\alpha }_{Dead}DFMC$$
(3)

where \({\alpha }_{Dead}\) is the mass proportion of dead foliage on a dry basis in the plant (foliage mortality).

LFMC is derived from ΨLeaf by p–v curves where it is a function of the moisture content in the symplasmic and the apoplasmic compartments weighted by their respective volumetric fractions such as the following:

$$LFMC = L{FMC}_{sat} \left[{{a}_{f}RWC}_{Apo} + (1-{a}_{f}){RWC}_{Sym}\right]$$
(4)

where \({a}_{f}\) is the fraction of the apoplasmic compartment in the shoots, \({LFMC}_{sat}\) (% dry mass) is the fine fuel moisture content at water saturation, \({RWC}_{Apo}\) is the relative water content of the apoplasm, and \({RWC}_{Sym}\) is the relative water content of the symplasm.

\({LFMC}_{sat}\) is determined as a function of leaf dry matter content (LDMC, the ratio of leaf dry mass to saturated fresh mass, g.g−1) as in Ruffault et al. (2023):

$${LFMC}_{sat}=\frac{1}{LDMC}-1$$
(5)

\({RWC}_{Sym}\) is calculated using the following equation depending on whether \({\psi }_{Leaf}\) is below or above the turgor loss point (\({\pi }_{TLP}\), MPa) also derived from p–v curves (Bartlett et al. 2012):

$${RWC}_{Sym}=\left\{\begin{array}{c}min\left(\frac{-\left({\psi }_{Leaf}+{\pi }_{0}-\varepsilon \right)\pm \sqrt{{\left({\psi }_{Leaf}+{\pi }_{0}-\varepsilon \right)}^{2}-4{\varepsilon \psi }_{Leaf} }}{2\varepsilon }\right), if {\psi }_{Leaf} > {\pi }_{TLP}\\ 1 -\left(1-\frac{{\pi }_{0}}{{\pi }_{TLP}}\right), if {\psi }_{Leaf} \le {\pi }_{TLP}\end{array}\right.$$
(6)

where \({\pi }_{0}\) (MPa) is the osmotic potential at full turgor and \(\varepsilon\) (MPa) the modulus of elasticity of the symplasm.

According to hypotheses H2 and H4 (Fig. 1), leaf-scale hydraulic damages would cause a decline in apoplasmic leaf water content under drought such that \({RWC}_{apo}\) could be variable over time. Preliminary results show that LFMC did not significantly differ between the plants that were or were not previously submitted to extreme drought (see Fig. 3 in results); this mechanism was not included in this model.

\({\alpha }_{dead}\) was modelled from an empirical sigmoidal function adjusted to the data as follows:

$${\alpha }_{dead}=\frac{100}{1+{e}^{\left(\frac{slope}{25}\left({\Psi }_{min}-{\Psi f}_{50}\right)\right)}}$$
(7)

where \({\Psi }_{min}\) (MPa) is the minimum water potential reached by the plant (for a recovery plant, this is equal to \({\Psi }_{midday}\) before rewatering).\(slope\) is the linear slope at the inflection point, and \({\Psi f}_{50}\) (MPa) is the water potential causing 50% foliage mortality.

DFMC is modelled according to Resco de Dios et al. (2015) as follows:

$${DFMC=FM}_{0}+{FM}_{1}{e}^{-mVPD}$$
(8)

where \({FM}_{0}\) is the minimum DFMC, \({FM}_{0}+{FM}_{1}\) the maximum DFMC, and \(m\) (%.kPa−1) is an exponent parameter setting the rate of moisture decay with increasing VPD.

2.7 Fitting model parameters

The submodels (LFMC, \({\alpha }_{dead}\), and DFMC) were parametrized using a genetic algorithm implemented in the GA (genetic algorithm) package (Scrucca 2013) within the R programming language (R Core Team 2023). For LFMC, three types of procedures were tested. First, the equations were fitted by optimizing the target parameters, namely π0, ε, LDMC, and αf. In this initial procedure, αf was allowed to vary between 0.2 and 0.4 for both species, according to the range of values expected for this trait. In the second approach, αf was held constant at 0, completely disregarding the apoplasmic compartment. The third approach involved allowing αf to be determined freely by the fitting algorithm. The performance of the models was assessed by evaluating them against the data using several metrics, including the root-mean-squared error (RMSE), mean absolute error (MAE), and coefficient of determination (R2).

3 Results

3.1 Dynamics of plant dehydration

In both species, we observed a rapid and consistent drying of the plants submitted to soil water deficits throughout the experiment, with a decrease in normalized soil water content (Wnorm), canopy fuel moisture content (CFMC), live fuel moisture content (LFMC), midday leaf water potential (Ψmidday), and an increase in foliage mortality (\({\alpha }_{Dead}\)) (Fig. 2). Appendix Fig. 5 provides additional information on the dynamics of predawn water potential (Ψpd) and plant-level transpiration, showing consistent patterns with the observations of Ψmidday. On the other hand, the dead fuel moisture content (DFMC) remained relatively stable (with some variability) throughout the experiment for litter samples and dead leaves and needles (Fig. 2E, F).

Fig. 2
figure 2

Dynamics of plant dehydration during drought and rewatering in oak (Quercus ilex) and pine (Pinus halepensis). The time dynamics of canopy fuel moisture content (CFMC, % dry mass) (A, B), live fuel moisture content (LMFC, % dry mass) (C, D), dead fuel moisture content (DFMC, % dry mass) (E, F), proportion of dead foliage (\({\alpha }_{dead}\), %) (G, H), midday leaf water potentials (Ψmidday, MPa), and normalized soil water content (Wnorm) (K, L) are shown. The median of measured values, along with the 5th and 95th percentiles, are shown. Red and blue circles indicate the droughted and recovery pots, respectively, while black circles represent the litter

Full size image

Differences in the dynamics of fuel moisture content were observed between the two studied species with pine consistently maintaining higher levels of CFMC and LMFC compared to oak (Fig. 2 A, B, C, D). Specifically, LFMC decreased from 176 to 64% and CFMC from 152 to 53% in pine during the experiment (Fig. 2B, D). In contrast, oaks exhibited a decrease in LFMC from 90 to 43% and a decrease in CFMC from 90 to 27% (Fig. 2A, C).

The dehydration dynamics also exhibited differences between the two studied species. Unlike oaks, which showed a decrease in LFMC and CFMC starting after the second measurement date, pines exhibited a steady decline in water content (LFMC and CFMC) during soil drying. Species-specific variations were also observed in the dynamics of Ψmidday and \({\alpha }_{dead}\). Measurements of Ψmidday revealed a relatively consistent decline, with oaks reaching lower values (approximately − 8.5 MPa) compared to pines (approximately − 7 MPa). Throughout most of the experiment duration, αDead maintained its minimum value (14% for pines and 1% for oaks). However, a significant increase was observed when Ψmidday dropped below − 6 MPa in both species.

Recovery patterns varied across the variables studied and the level of drought reached by the droughted plants. Thus, Wnorm, LFMC, Ψmidday, and Ψpredawn exhibited full recovery regardless of the measurement period or drought level. Conversely, αDead and CFMC did not fully recover during the driest period of the experiment, with this pattern being more pronounced for oaks than for pines.

3.2 LFMC dynamics in droughted and recovery plants

We further investigated whether LFMC fully recovered after drought (in accordance with H1 and H3; Fig. 1) or did not fully recover due to leaf-level hydraulic dysfunction (in accordance with H2 and H4; Fig. 1). The results showed that while the average LFMC was lower for the leaves submitted to extreme drought for both species, our results did not show significant differences in LFMC values after rewatering between treatments (t-test; p > 0.1 for pine and oak) (Fig. 3).

Fig. 3
figure 3

Comparison of live fuel moisture content (LFMC, % dry mass) of rewatered leaves previously submitted to extreme drought (minimum water potential reached by the leaves (Ψmin <  − 3 MPa) and not submitted to extreme drought (Ψmin >  − 3 MPa). The results are presented separately for the oak (Quercus ilex) and pine (Pinus halepensis). Significant differences were determined using t-tests. No significant differences are denoted as “ns”

Full size image

3.3 CFMC predictions

A main goal of this study was to predict the canopy fuel moisture (CFMC) and identify the plant trait values that best align with empirical data. Based on previous findings (Sect. 3.2), we hypothesized that hydraulic damages at t leaf scale are insignificant for surviving leaves, and that extreme drought primarily impacts the rate of foliage mortality (αDead).

Our full model, based on the prediction of several fuel moisture subcomponents (LFMC, DFCM, and αDead), performed well at predicting CFMC for both species. For CFMC, the prediction metrics for pine (R2 = 0.87 and RMSE = 13.2%, Table 1) and oak (R2 = 0.87, RMSE = 9.4%, Table 2) were accurate, and no noticeable bias was observed in the predictions of our models across the range of observed values (Fig. 4A, B).

Table 1 Fitted parameters and goodness of fit for the different submodels—live fuel moisture content (LFMC), proportion of foliage mortality in the canopy (αdead), and dead fuel moisture content (DFMC)—as well as the full canopy fuel moisture content (CFMC) model for the pine species (Pinus halepensis). The parameters for the CFMC model include the osmotic potential at full turgor \({\pi }_{0}\) (MPa), the modulus of elasticity of the symplasm (\({\varepsilon }_{Sym}\), MPa), the leaf dry matter content (LDMC, mg.g−1), and the leaf apoplasmic fraction (\({\alpha }_{f}\)). The parameters for the \({\alpha }_{dead}\) model are the slope (%.MPa−1) and water potential for which 50% mortality is observed (ψf50, MPa). The parameter for the DFMC model is the rate of moisture decay with increasing VPD (m, %.kPa−1). The performance metrics include the root-mean-square error (RMSE, %), the mean absolute error (MAE, %), and the coefficient of determination (R2)
Full size table
Table 2 Fitted parameters and goodness of fit for the different submodels—live fuel moisture content (LFMC), proportion of foliage mortality in the canopy (αdead), and dead fuel moisture content (DFMC)—as well as the full canopy fuel moisture content (CFMC) model for the oak species (Quercus ilex). The parameters for the CFMC model include the osmotic potential at full turgor \({\pi }_{0}\) (MPa), the modulus of elasticity of the symplasm (\({\varepsilon }_{Sym}\), MPa), the leaf dry matter content (LDMC, mg.g−1), and the leaf apoplasmic fraction (\({\alpha }_{f}\)). The parameters for the \({\alpha }_{dead}\) model are the slope (%.MPa−1) and water potential for which 50% mortality is observed (ψf50, MPa). The parameter for the DFMC model is the rate of moisture decay with increasing VPD (m, %.kPa−1). The performance metrics include the root-mean-square error (RMSE, %), the mean absolute error (MAE, %), and the coefficient of determination (R2)
Full size table
Fig. 4
figure 4

Comparison of observed and simulated canopy fuel moisture content (CFMC, % dry mass), live fuel moisture content (LFMC, % dry mass), dead fuel moisture content (DFMC, % dry mass), and foliage mortality (\({\alpha }_{dead}\), %) as a function of leaf water potential (ΨLeaf, MPa) or vapor pressure deficit (VPD, kPa) for pine (Pinus halepensis) and oak (Quercus ilex)

Full size image

However, when looking at the predictions of the CFMC subcomponents, we observed disparities in model performance as well as differences in fitted parameters between the two species. For the LFMC model, the osmotic potential at full turgor (π0) was higher for the pine species (− 2.6 MPa) than for the oak species (− 3.7 MPa). Conversely, the modulus of elasticity of the symplasm (εSym) and leaf dry matter content (LDMC) were lower for pine (12 MPa and 369 mg.g−1, respectively) than for oak (15.9 MPa and 516 mg.g−1, respectively). Our results also showed that the performance of the models with αf forced to 0 was similar or even higher than that of the original LFMC models. Moreover, when αf was freely determined by the algorithm, it consistently yielded lower-than-expected values (Appendix Tables 4 and 5).

For αDead, our models also performed well for both species (R2 = 0.87 and R2 = 0.87 for pine and oak, respectively) (Fig. 4G, H). The fitted parameters indicated that leaf mortality occurs at a higher water potential for pine (ψf50 =  − 6.9 MPa) compared to oak (ψf50 =  − 7.5 MPa). For oak, ψf50 is lower than the P50 estimated from stem vulnerability curve on mature trees (− 4.6 MPa, Appendix Fig. 6 and Table 6), a species for which we also observed important differences in P50 for leaves and stem (Appendix Fig. 7). For pine, ψf50 is similar than the P50 measure on mature trees (− 7.6 MPa, Appendix Fig. 6 and Table 6).

Finally, the models for DFMC demonstrated limited predictive accuracy in both species, as indicated by low-performance metrics (R2 = 0.22 and R2 = 0.34 for pine and oak, respectively) (Fig. 4E, F).

4 Discussion

Our findings indicate that under severe soil water-deficit conditions, foliage mortality (increased αDead) is a significant physiological driver of canopy fuel moisture content (CFMC) in plant seedlings. Because foliage mortality is irreversible, the influence of leaf mortality on CFMC persists beyond the drought period, that is to say as long as dead leaves and twigs remain attached to the plant and new leaves are not produced (Fig. 2). This highlights the crucial role of foliage mortality in shaping the dynamics of CFMC during both drought and post-drought conditions and thereby supporting the hypotheses H3 and H4 of our study (Fig. 1). Consequently, our results indicated that relying solely on the current leaf water potential is insufficient for accurate CFMC prediction. It is crucial to consider the minimum water potential reached by the plant during the drought season (Ψmin) to account for extreme drought events that may have resulted in foliage mortality. On the other hand, we did not find significant evidence of plant hydraulic damages in the surviving leaves (Fig. 3), excluding hypotheses H2 and H4 (Fig. 1). These results suggest that some leaves suffer from irreversible hydraulic damages leading to death, while the other (surviving) leaves do not seem to be affected at all and are therefore capable of recovering their initial functions.

A range of leaf and plant traits drive both the values and dynamics of CFMC during drought, either by shaping (i) the plant response to drought (ΨLeaf), (ii) the relationship between LFMC andΨLeaf, or (iii) the response of foliage mortality toΨLeaf. Next, we discuss the effects of plant traits on each of these processes and how physiological processes related to low water potentials differ between the two studied species. We also discuss the main limitations of our study, particularly its generalization to mature, natural forests, and suggest research pathways that could enable to investigate more deeply how leaf-level embolism processes impact the dynamics of fuel moisture content.

The species-dependent plant response to soil water deficit (i.e., ΨLeaf dynamics) partly explained the different fuel moisture dynamics observed between the two studied species (Fig. 2), thereby highlighting the differences in plant functional strategies previously described by Moreno et al. (2021). Aleppo pine (Pinus halepensis) is generally considered a drought avoider characterized by early stomatal closure and a low resistance to drought, while Holm oak (Quercus ilex) is considered as a drought-tolerant tree species with a later stomatal closure but higher resistance to drought. Such differences can lead to lower transpiration rates, higher soil water content, and higher ΨLeaf for Aleppo pine than for Holm oak (Fig. 2). In this study, we specifically focused on assessing the impact of soil water deficit on leaf water potential (ΨLeaf). The experimental conditions were tightly controlled, maintaining greenhouse temperatures between 25 and 35 °C and relative humidity (RH) levels between 40 and 75%. It is crucial to note that variations in vapor pressure deficit (VPD) also significantly influence ΨLeaf dynamics by controlling transpiration rates (Grossiord et al. 2020). Additionally, temperature plays a pivotal role in ΨLeaf dynamics by directly influencing cuticular conductance, which represents the primary source of water loss from plants when stomata are closed (Duursma et al. 2019).

Consistent with prior studies (Nolan et al. 2018, 2022; Scarff et al. 2021), our results confirm that the differences in LFMC between plant species, at a given water potential, are governed by the response of LFMC to ΨLeaf (p–v curves traits) as well as the leaf dry matter content (LDMC), which determines the maximum LFMC value (LFMCsat). It should be noted here that the values reported for these traits in this study slightly differ from what has been measured at experimental sites in southern France. For instance, for pines, Moreno (2022) reported values of − 1.5 MPa and 445 mg.mg−1 for π0 and LDMC, respectively, which are higher than the values estimated in this study (− 2.6 MPa and 369 mg.mg−1). For oaks, values of − 2.8 MPa and 580 mg.mg−1 for π0 and LDMC, respectively, were reported by Moreno (2022), which are similar to the values estimated in this study (− 2.9 MPa and 521 mg.mg−1). These discrepancies could be due to the fact that our experiment was carried out on plant seedlings and not on mature trees. One should note, however, that the relative difference between species was coherent across studies.

Our study highlights that under extreme soil water deficits, the response of foliage mortality to declining water potential plays a major role in the dynamics of CFMC, aligning with the findings of Ruffault et al. (2023). In our experiment, we observed that 50% of leaf mortality in tree seedlings occurred at a water potential of − 6.9 MPa and − 7.5 MPa for pines and oaks, respectively. Interestingly, the difference in this threshold (0.6 MPa) between species is much smaller than their difference in stem P50, which is 2.6 MPa (considering a stem P50 of 4.8 MPa for pine and 7.4 MPa for oak, respectively). This might be explained by a greater vulnerability to embolism in leaves compared to stems in Quercus ilex. Indeed, leaf vulnerability curves, available only for oaks (see Appendix Fig. 7), indicate that leaves are more vulnerable to drought-induced embolism than stems. Although we lack estimates of needle P50 for Aleppo pines, previous studies have reported minimal segmentation for another pine species (Pinus pinaster; Bouche et al. 2016). Exploration of key areas, such as hydraulic segmentation and phenomena like leaf persistence on the tree after death, has been so far limited. Addressing these aspects could significantly enhance our understanding of wildfires and their ecological impacts, thereby highlighting the potential for exploration through the integration of plant hydraulics into fire behavior models (Dickman et al. 2023; Torres‐Ruiz et al. 2024).

Our modeling framework hypothesized that the water content of dead leaves on the tree is exclusively driven by VPD. However, our findings suggest that this assumption may not be entirely accurate (Fig. 4E, F). Indeed, the fuel moisture content of dead leaves of seedlings had higher values and a wider range of variations compared to the litter moisture content (Fig. 2E, F), which can be explained by several reasons. Firstly, recent studies have shown that lethality threshold for plant is below 100% loss of hydraulic conductivity (Hammond et al. 2019), suggesting that there might be a system of conductive tissues still supplying water to leaves until very late. Hence, leaves desiccation is more likely a progressive process, so that leaves that got partly disconnected from the vascular system within the last few days before sampling may differ in terms of water content, particularly when compared to those that got brown weeks or months ago. In addition, distinguishing between dead leaves and needles from living ones, especially for Pinus halepensis, posed a challenge in our experiment. It was based on a visual appreciation of leaf coloring, suggesting that some leaves or needles as could have been classified as deceased when they were not. As a result, the CFMC of canopy suffering leaf mortality could be slightly underestimated with our approach.

One notable aspect that could improve the understanding and predictions of CFMC dynamics is the consideration of the impact of leaf phenology. Leaf development and senescence influence leaf moisture dynamics by modifying the leaf dry matter content (LDMC, Eq. 4) (Jolly et al. 2014; Resco de Dios 2020; Nolan et al. 2022). In our experiment, we observed higher level of LFMC for young leaves during early stages of drought. However, young leaves LFMC values tended to reach similar levels as mature leaves during deep drought. Further study should evaluate more in depth the part of the CFMC decreases during the summer season that is related to changes in LDMC and not to the direct impact of drought water content (Jolly et al. 2014). Another mechanism that would have worth to investigate more deeply in our experiment is preprogramed needle senescence (Balaguer-Romano et al. 2020). Whereas this phenomenon is known to occur in mature pine during leaf renewal, it has not been observed on our young potted pines. However, a better understanding of this phenomenon is important as it can modify the percent of dead needles and change the estimation of the impact of drought on foliage mortality. Further experiments or monitoring on mature tree should allow to quantify the effect of drought compared to the effect only due to the proportion of dead needles and its impact on CFMC.

More generally, the conclusions of our study are based on observations of seedlings in pots, raising questions regarding how applicable our findings are to mature trees in natural conditions. One crucial difference lies in the definition of CFMC, where our study includes leaf mortality but does not consider tree mortality, a phenomenon that could significantly impact CFMC dynamics at the ecosystem or landscape level. It is also crucial to interpret our results by recognizing disparities between experimental setups and the complexities of mature tree environments, including factors such as competition, root depth, and profile. To date, only few studies investigated the impact of drought on CFMC that includes leaf or plant mortality (Balaguer-Romano et al. 2020; Ruffault et al. 2023); we therefore advocate for further research in this domain to enhance our understanding of CFMC dynamics in forested ecosystems.

5 Conclusion

With ongoing climate change, identifying the most vulnerable species and areas to wildfires has become a critical challenge. Our results shed light on the crucial role played by both leaf and whole-plant traits in shaping the plant response to drought in terms of fuel moisture content. Given the anticipated rise in the frequency and intensity of drought events, gaining a better understanding of the plant-hydraulic processes that contribute to drought-induced foliage mortality is crucial for assessing wildfire risks in forests and shrublands. These findings highlight the need to develop fire danger models that incorporate the mechanisms underlying plant responses to drought.