Linking sap flow and tree water deficit in an unmanaged, mixed beech forest during the summer drought 2022

Abstract

• Temperate mixed forests are currently experiencing severe drought conditions and face increased risk of degradation. However, it remains unclear how critical tree physiological functions such as sap flow density (SFD) and tree water deficit (TWD, defined as reversible stem shrinkage when water is depleted), respond to extreme environmental conditions and how they interact under dry conditions.

• We monitored SFD and TWD of three co‐occurring European tree species (Fagus sylvatica, Fraxinus excelsior and Acer pseudoplatanus) in dry conditions, using high temporal resolution sap flow, dendrometer, and environmental measurements.

• Species‐specific SFD responses to soil drying did not differ significantly, while TWD was significantly higher in F. excelsior. Inter‐specific differences in wood anatomy and water use strategies did not consistently explain these responses. TWD and SFD responded both to soil moisture content (SWC) during wet (SWC ≥ 0.2) and dry (SWC < 0.2) phases, with SFD responding more strongly. There was a significant correlation for TWD and vapour pressure deficit (VPD) only in the wet phase, and for SFD and VPD only in the dry phase. During the dry phase, the incoming PPFD significantly correlated with SFD in all species, and with TWD only in F. sylvatica and F. excelsior. TWD negatively responded to SFD, showing hysteresis effects from which a decreasing sigmoidal phase along the soil drying gradient was observed.

• The nonlinear correlations between TWD and SFD may result from a time lag between the two variables, and their different sensitivities to SWC and VPD under different drought intensities. We conclude that, under drought stress, TWD cannot be used as a proxy for SFD or vice versa.

Sap flow and reversible stem shrinkage upon water depletion are strongly correlated under extreme drought conditions, but they respond differently to drought and cannot be used interchangeably when assessing tree response to drought.

INTRODUCTION

Drought‐induced stress in Central European forests has increased in recent years due to more intense and frequent heatwaves and drought events (IPCC ²⁰¹⁴; Seidl et al. ²⁰¹⁴; Skogen et al. ²⁰¹⁸; Schuldt et al. ²⁰²⁰), resulting in severe vitality losses (Mathes et al. ²⁰²³). To better understand climate change impacts on temperate forest ecosystems, we need to identify the drought response mechanisms at both species and tree levels so as to develop more efficient management strategies to limit forest degradation risks (Zweifel et al. ²⁰¹⁶; Nievola et al. ²⁰¹⁷; Choat et al. ²⁰¹⁸; Hajek et al. ²⁰²²; Karimi et al. ²⁰²²). Important physiological processes, such as sap flow and tree water deficit (TWD), the latter of which is defined as reversible stem shrinkage when water is depleted, are water‐related indicators of a tree's response to extreme environmental conditions (Ježík et al. ²⁰¹⁵; Zweifel et al. ²⁰¹⁶). The greater sensitivity of sap flow to drought compared to that of stem water deficit has been described in various tree species (Bovard et al. ²⁰⁰⁵; Brinkmann et al. ²⁰¹⁶). Studies that concurrently monitored sap flow and stem circumference change have addressed the sensitivity to drought of each process, with a special focus on species‐specific responses to environmental change and revealed an interaction between these two physiological processes and the importance of species‐specific traits in water use strategies (Ježík et al. ²⁰¹⁵; Sánchez‐Costa et al. ²⁰¹⁵; Brinkmann et al. ²⁰¹⁶). However, studies that combine high temporal resolution sap flow and TWD data with atmospheric characteristics and data on soil moisture in mixed temperate forests are rare, despite the opportunity that they offer to gain deeper insight in the timing and complex interactions between these two physiological processes, especially under soil water‐limited conditions. This information is essential to quantitatively evaluate the vulnerability and regulation strategies of different tree species under conditions of severe drought (De Swaef et al. ²⁰¹⁵; Aldea et al. ²⁰¹⁸; Ammer ²⁰¹⁹).

The role of stomatal regulation is critical for species‐specific hydraulic efficiency, with species typically characterized by either isohydric or anisohydric stomatal responses (McDowell et al. ²⁰⁰⁸). Studies have confirmed decreasing stomatal conductance in isohydric species, which maintain consistent leaf water potential under soil drying conditions, and reduced leaf water potential but stable stomatal conductance in anisohydric species as soil moisture declines (Sperry & Tyree ¹⁹⁸⁸; Tardieu & Simonneau ¹⁹⁹⁸; McDowell et al. ²⁰⁰⁸; West et al. ²⁰⁰⁸). Thus, anisohydric trees usually have higher and more constant sap flow rate during drought compared to isohydric trees, which reduce their water consumption as a water‐saving strategy. For instance, Yi et al. (2017) and Chen et al. (2021) demonstrated that sap flow reduction and drought stress vulnerability were more pronounced in isohydric species, particularly in the outer xylem. Other findings, such as those reported by Hölscher et al. (2005), found that the anisohydric species Fraxinus excelsior had lower sap flow reduction than other anisohydric species such as Fagus sylvatica and even some isohydric species, such Acer pseudoplatanus, during a dry period. The complex role of isohydry and anisohydry in tree water use strategies can be influenced by other factors, such as root depth, root or stem anatomy (ring‐ or diffuse‐porous trees) that may significantly impact tree responses to drought (Köstler et al. ¹⁹⁶⁸; McCulloh et al. ²⁰¹⁰). Overall, the literature suggests that tree species show considerable intraspecific variation in their water regulation and may not be characterized as strictly anisohydric or isohydric (Cocozza et al. ²⁰¹⁶). Many tree species seem to be rather plastic when it comes to stomatal regulation, which results in their shifting along an isohydry–anisohydry continuum (Schumann et al. ²⁰²⁴).

Wood anatomy of trees is an important trait that affects tree sensitivity to drought (Seleiman et al. ²⁰²¹; Dorji et al. ²⁰²⁴). Ring‐porous tree species, such as F. excelsior, Quercus sp., or Robinia pseudoacacia, are characterized by wider pores in the earlywood compared to the latewood and often maintain a consistent water flow even during periods of drought stress (Hölscher et al. ²⁰⁰⁵; Bader et al. ²⁰²²). In contrast, diffuse‐porous species, such as F. sylvatica or A. pseudoplatanus, have small and homogeneously distributed pores in the early and late wood (von Allmen et al. ²⁰¹⁵). The narrow pores in diffuse‐porous species make them less susceptible to cavitation but, importantly, at the same time contribute to hydraulic conductivity loss under water‐limiting conditions. In ring‐porous species the risk of cavitation under drought stress associated with their large earlywood vessels may also downregulate their expected high hydraulic conductivity (Hacke et al. ²⁰⁰⁶; Li et al. ²⁰⁰⁸). However, wood anatomy was also found to be tightly linked to stomatal regulation. For example, Oren & Pataki (2001) observed that stomatal conductance in diffuse‐porous species was higher than in ring‐porous species, with increased stomatal sensitivity in response to changes in environmental conditions. Furthermore, the high‐water conductivity loss in ring‐porous trees under drought stress, ca. 90% according to Li et al. (2008), might increase TWDs to rates higher than those observed in diffuse‐porous trees.

Stem sap flow and water deficit are strongly affected by edaphic drought factors, such as soil moisture content (SWC) and rainfall (Sitková et al. ²⁰¹⁴; Brinkmann et al. ²⁰¹⁶; Grossiord et al. ²⁰¹⁸; Grillakis ²⁰¹⁹), and by atmospheric drought factors, such as air temperature, vapour pressure deficit, and solar radiation (Deslauriers & Morin ²⁰⁰⁵; Köcher et al. ²⁰¹²; Dietrich et al. ²⁰¹⁸; Grossiord et al. ²⁰²⁰; Zhao et al. ²⁰²³). These factors are directly or indirectly linked to both below‐ and above‐soil water variability, which drives the changes in daily sap flow and stem circumference (Meinzer et al. ²⁰⁰⁶; Clausnitzer et al. ²⁰¹¹; Battipaglia et al. ²⁰¹³; Aldea et al. ²⁰¹⁸). When soil dries out, the resulting sap flow interruption triggers stem shrinkage (Sitková et al. ²⁰¹⁴), leading to a strong link between sap flow and TWD following a decreasing soil moisture gradient. Studies of the correlation between sap flow and TWD of temperate species throughout the vegetation period have found a negative linear to nonlinear relationship between these two variables (Sánchez‐Costa et al. ²⁰¹⁵; Brinkmann et al. ²⁰¹⁶). The nonlinear reduction in sap flow with increasing TWD results from the combined regulation of transpiration and stem water storage (Sánchez‐Costa et al. ²⁰¹⁵). The type of TWD responses to sap flow variation might depend on how these two processes respond to changing environmental conditions.

Here, we assessed the effects of the extraordinarily dry summer conditions in 2022 on sap flow density (SFD) and TWD on three deciduous tree species that are often assumed to represent different ‘drought response or sensitivity types’ resulting from their wood anatomy and stomatal regulation: F. sylvatica (diffuse‐porous, anisohydric), A. pseudoplatanus (diffuse‐porous, isohydric), and F. excelsior (ring‐porous, anisohydric). Fagus sylvatica displays stomatal regulation under certain conditions earlier than F. excelsior and would therefore be classified as less anisohydric than F. excelsior (Lemoine et al. ²⁰⁰¹; Köcher et al. ²⁰⁰⁹). The 21 measured trees (12 F. sylvatica, 4 A. pseudoplatanus, 5 F. excelsior) grow in an unmanaged mixed broadleaf forest in Germany (Hainich National Park). To quantify the influencing factor ‘summer drought’, three different variables were considered: SWC, vapour pressure deficit (VPD) and incoming photosynthetically active photon flux density (PPFD, i.e., the downward quantity of photons received by the leaf surfaces). Our study proposed the following hypotheses: (i) SFD and TWD both respond to soil drying, with SFD being more negatively affected in F. sylvatica and A. pseudoplatanus than F. excelsior, and TWD more positively in F. excelsior than the two other species; (ii) SFD and TWD responses to SWC variation are more significant than their responses to VPD and PPFD in all species; and (iii) SFD and TWD are negatively and nonlinearly correlated, and daily SFD peaks earlier than daily TWD.

MATERIAL AND METHODS

Study site and investigated species

The study site (DE‐Hai) is a permanent research site (51°05′N, 10°27′ E, 350 m a.s.l., ca. 2,14 ha), established in 1999 in the Hainich National Park (Thuringia, Germany). The main goal of the research site is to measure the carbon balance of a natural, unmanaged deciduous forest via the ‘eddy‐covariance method’ (Knohl et al. ²⁰⁰³), and is part of the international networks FLUXNET and ICOS. The strict protection of the site ensured that the processes investigated in the present study were not influenced by recent forest management activities.

Forest management in the site ceased in 1997 and since then no silvicultural interventions have taken place (Mund et al. ²⁰²⁰). The stand is uneven‐aged (1–250 years) and composed of F. sylvatica (82% of total basal area), and two co‐dominant species, F. excelsior (9%), and A. pseudoplatanus (5%). Some other rare deciduous species are admixed (Mölder et al. ²⁰⁰⁶; Mund et al. ²⁰²⁰). The soil is a silty‐clay cambisol, developed on limestone under Pleistocene loess deposits. The climate is suboceanic–submontane, and annual mean temperature varies from 6.5 to 8.0°C, while annual mean precipitation varies from 500 to 800 mm (Fischer et al. ²⁰¹⁰; Mund et al. ²⁰²⁰). We selected 21 mature dominant and co‐dominant trees (diameter at breast height (DBH) 55.6–95.4 cm) of the three species (12 individuals of F. sylvatica, 5 F. excelsior and 4 A. pseudoplatanus) to measure SFD and stem circumference changes from 19 June to 30 September 2022. The higher abundance of F. sylvatica compared to F. excelsior and A. pseudoplatanus at the experimental plot, combined with the setup complexity, i.e., distribution of electric lines for power supply to each sensor, determined tree selection.

The selected trees were scanned using a mobile laser scanner (ZEB Horizon; GeoSLAM, Nottingham, UK) in July 2022, and individual trees were identified using their precise geographic coordinates which were integrated into the scan data. To capture details of the trees, eight independent scan sessions were conducted. Each scan session consisted of initiating the scanner at a random position in the forest area and, after picking it up from the start position on the ground, carrying it through the area of interest while it scanned all objects within 100 m distance. This scan‐walk continued for ca. 20 min before returning to the starting point to close the loop. Once finished, the scans were processed in Geoslam HUB (GeoSLAM) to create xyz files (point cloud) of the forest. From these point clouds, several dendrometric characteristics (tree height, crown radius, crown volume) and crown competition indices (crown competition index and box dimension) were derived using previously established algorithms (Metz et al. ²⁰¹³; Heidenreich & Seidel ²⁰²²) using Wolfram Mathematica (Wolfram Research, Champaign, Il., USA) (Table 1).

Table 1.

Measured trees per species (Fagus sylvatica L., Fraxinus excelsior, Acer pseudoplatanus) and their characteristics.

species	xylem anatomy	stomatal regulation	DDW, g cm−3	sample size	tree_id	DBH, cm	H, m	crown volume, m3	%CCI
Fagus sylvatica	Diffuse porous	Predominant anisohydric behaviour Leuschner et al. (2022)	0.51	12	475 483 481 2 208 130 7 357 18 51 405 389 560 542 544	70.8 81.0 57.9 78.4 80.7 93.8 82.2 67.0 64.9 73.9 74.3 79.5 79.7 70.8 95.4	34.6 34.5 34.7 36.1 34.2 35.8 35.4 36.2 31.7 35.1 34.5 33.4 33.1 36.1 33.4	1475.7 858.9 641.0 317.4 858.2 1561.0 604.4 1099.1 692.6 982.1 1291.4 1735.6 1086.1 610.8 2311.4	9.4 11.3 10.3 14.7 18.0 7.8 20.5 16.5 6.1 24.6 10.5 12.8 6.4 26.4 6.6
Acer pseudoplatanus	Diffuse porous	Predominant isohydric behaviour Leuschner et al. (2022)	0.49	4	116 39 540 618	65.8 56.8 55.6 66.5	34.5 30.3 30.7 28.4	724.6 670.3 421.9 473.1	10.6 42.4 31.2 35.8
Fraxinus excelsior	Ring porous	Predominant anisohydric behaviour Leuschner et al. (2022)	0.41	5	477 186 350 379 1	69.5 61.2 90.2 91.5 81.7	35.8 35.0 38.8 35.7 36.1	1261.6 1277.5 232.7 587.4 1113.8	17.4 15.3 44.2 32.6 30.6

DBH, diameter at breast height; DDW, sapwood basic dry density (estimated from prior field measurements); H, tree height; CCI, crown competition Index.

Microclimate data

In the vicinity of the eddy covariance tower, microclimate data were measured during the study period at 10‐min intervals. Ambient air temperature at 2 m above the soil (T _air, [°C]) was measured with infrared remote temperature sensors (IR120 Campbell Scientific). SWC (given in percentage or as a dimensionless proportion of volume of water per total volume of soil) was recorded at 10, 15 and 30 cm soil depths with soil moisture profile sensors (SoilVUE 10 soil profiler, Campbell Scientific). For all analyses, and the threshold that separates the time series of SWC into a dry and a wet phase (SWC = 0.2, see below), the arithmetic means of the three soil depths were used. The soil moisture profile sensors operated based on the manufacturer's calibration and with ±2% accuracy. VPD (kPa) was calculated from measurements of relative humidity and T_air; the incoming PPFD (μmol m⁻² s⁻¹) measured above the canopy was obtained using a photosynthetically active radiation sensor (PAR‐Lite; Kipp & Zonen). Rainfall data (mm) were also measured above the canopy, and gaps that occurred from 20 August 2022 due to sensor failure were filled using precipitation data from the closest DWD station to our study site (Hörselberg‐Hainich‐Behringen, Station ID ‘00336’). For gap‐filling, a linear regression (with intercept = 0) between our existing rainfall data and the DWD data, both aggregated to daily sums beforehand, was derived. Then the missing data were modelled based on the derived regression.

Measurements of sap flow density and stem circumference change

Methods of sap flow measurements using a thermal medium have evolved since the 1930s (Huber 1932; Huber & Schmidt ¹⁹³⁷; Granier ¹⁹⁸⁵; Cohen et al. ¹⁹⁹³). The Dual Method Approach (DMA) is one of several heat pulse methods that captures a large range of sap flow velocities, from moderate to high (Marshall 1958; Cohen et al. ¹⁹⁹³; Forster ²⁰²⁰). We used high‐resolution sap flow sensors (SFM‐4, Umwelt‐Geräte‐Technik, Müncheberg, Germany & Implexx Sense, Melbourne, Australia) using the DMA to monitor sap flow densities in mature F. sylvatica, F. excelsior and A. pseudoplatanus trees. The basic component of the SFM‐4 sap flow sensor, installed at a standard height of 1.3 m above the soil surface, is a head to which three needles are attached: one heating probe in the middle, and two thermal probes, one above and one below the heating probe (UGT ²⁰²²). The three needles were inserted 3 cm deep in the sapwood, vertically aligned and spaced 0.6 cm apart, parallel to each other. The sapwood depth covered by the sensor was within the range of the highly conductive xylem sections for the investigated species: F. excelsior and F. sylvatica (0–4 cm) and A. pseudoplatanus (2–4 cm) (Gebauer et al. ²⁰⁰⁸). The following raw data measured by each sap flow sensor were used to calculate the inner and outer sap flow densities: alpha (natural log of ratio of temperatures at inner and outer thermistors), beta (natural log of the ratio of maximum temperatures at inner and outer thermistors), and Tmax (the time to reach maximum temperature at inner and outer thermistors). Three main variables were derived from the raw data: SFD inner (hereafter SFD [g cm⁻² h⁻¹]), SFD outer [g cm⁻² h⁻¹] and total sap flow [l h⁻¹]. However, total sap flow values directly measured by the sensor could not be used as they were prone to inaccuracies, considering that scaling factor adjustment (using precise sapwood area) was not effective. In the following, we will therefore refer only to the inner density values (at outermost 1.5–2.5 cm sapwood depth) because, unlike the outer density values, their patterns contained little noise and could be efficiently rectified through probe misalignment corrections. Interestingly, SFD at inner sap wood depth was always higher than at outer depth, but the correlation between inner and outer density values was generally high (R ² > 0.9) (Table S1). To compare SFD across species despite expected interspecific/intraspecific variation of absolute SFD, we also normalized values (see below). We corrected the effects of probe misalignment on the SFD values by adjusting the space value between the probes (initially 0.6 cm) until we obtained nighttime values close to zero, especially during rainy days (when rainfall > 0 mm, VPD = 0 kPa and RH = 100%). For most trees, the sensors were installed in the northwestern exposure side of the trees in order to attenuate insolation risks, and this risk was further minimized by covering all sensors with camouflaged radiation shields. To increase accuracy in sap flow measurements, we replaced the default tree parameters required for the sensors' configuration by values that were directly measured (trunk diameter in cm and bark depth in cm) and derived from literature (basic dry density of wood or DDW in g cm⁻³; Table 1). Direct measurement of DDW would have required core extractions, thereby causing additional damage to trees in the National Park. We therefore used DDW values that were measured close to our study site by Gebauer et al. (2008) (Table S3).

Tree girth increments were recorded using high‐resolution dendrometers (DRS 26; EMS Brno, Czech Republic), positioned at breast height on the same trees as the sap flow sensors, slightly below them. Before installation, the bark surface at the specific height was cleaned using a hard brush to ensure secure contact of the fixing tape (stainless band: 12 × 0.2 mm) with the tree surface. The dendrometer operated at 0.1 mm accuracy, 1 μm resolution, and with a working temperature range between −40 and 60°C (EMS Brno ²⁰²²). Each dendrometer delivered two values: circumference change [mm] and operating temperature [°C], with the circumference measurements automatically corrected for potential temperature effects on the measurement system. Stem circumference change data requiring correction consisted mainly of jumps (sudden value drops or rises), removal, and gaps in the time series. Only minor gaps were found and could be filled with values from previous observations from the same time series dataset. After data correction, reversible TWD of the stem (TWD, [mm]) (Zweifel et al. ²⁰¹⁶) was derived from the stem circumference change measurements and converted from circumference to diameter metrics. More specifically, if TWD becomes zero, the tree is fully rehydrated and can grow. Thus, TWD = 0 serves as an indicator for the absence of drought stress.

Power supply and data transmission

Both dendrometers and sap flow sensors were digital output SDI‐12C compatible and recorded data at 10‐min intervals. All sensors were powered via cables connected to a 24‐volt power source located on the plot. Eight trees were each equipped with a LoRa‐node (TBS12B LoRaWAN; TekBox Digital Solutions, Vietnam) and a junction or transmission box (TBS04‐FB—SDI12, TekBox Digital Solutions), ensuring data transfer and connectivity between cables from the sensors, the power source and the LoRa node. The LoRa nodes were powered with three batteries of 1.5 volts each, enabling wireless data transfer to a distant server (up to 60 m).

Statistical processing and analyses

We used the statistical programming environment R (R Core Development Team, v 4.0.2) to structure, visualize, and analyse dendrometer, sap flow, and microclimate data. The packages Dendroanalyst, nls2, and PerformanceAnalytics were then used to extract the values of TWD from stem circumference change data, to fit nonlinear models, and to plot matrix charts of simple linear models, respectively.

Some of our analyses were performed using absolute SFD and TWD in order to show actual SFD and TWD rates at the species level, even though sap flow can vary from one species (or individual) to another because of differences in anatomy and physiology. In further analyses we instead used relative values, especially to assess the correlation between SFD and TWD, but comparable analyses using absolute values did not significantly change the results; these can be found in the Data S1. Normalization was done for each tree individually following the min‐max approach and values in the interval [0, 1] were obtained. We specifically used the formula:

$$\mathrm{Rel}=\frac{\mathrm{Abs}-{\mathrm{Abs}}_{\min }}{{\mathrm{Abs}}_{\max }-{\mathrm{Abs}}_{\min }}$$ (1)

where the normalized values of SFD or TWD (Rel) are functions of actual absolute values (Abs), minimum absolute values (Abs_min), and maximum absolute values (Abs_max).

We calculated the median values of TWD and SFD per species and plotted their hourly patterns across the study period. The SFD decrease was also calculated for each tree through the difference between the arithmetic mean of five daily maximum observations at the beginning of the study period (when SFD was at highest rate) and the average of five daily maximum observations in the driest soil conditions of the study period (when SFD was at lowest rate). The SFD decreases and maximum TWD per tree were then compared between the three species in boxplots and by applying a Kruskal‐Wallis test.

We also assessed the absolute SFD and TWD responses to three target environmental variables: SWC, VPD, and incoming PPFD. SWC values involved in correlations with SFD and TWD were derived from averages of SWC measurements at different soil depths (10, 15, and 30 cm). Correlations were assessed per species using simple generalized linear models (GLM) and nonlinear least squares (NLS) models, separately for two sub‐periods of interest: dry (when SWC < 0.2) and wet (when SWC ≥ 0.2). We excluded damp days (when VPD < 1 kPa and net solar radiation < 500 W m⁻²), so as to consider only the physiological responses induced by soil dryness and not by air humidity.

The normalized TWD response to SFD variation was assessed at the species level and for each individual (Fig. 4). At the species level, we calculated median values of daily maximum TWD and SFD and assessed their correlations in three cases: (i) considering dry and wet periods separately, (ii) following a changing SWC gradient across the total study period, and (iii) following a changing SWC gradient only in soil drying conditions (when SWC was decreasing). In case (iii) we fitted nls models following the equation:

$$\mathrm{TWD}=1-\mathrm{a}\times {\mathrm{e}}^{\left(\mathrm{b}\times \mathrm{SFD}\right)}$$ (2)

where TWD is daily maximum TWD, SFD is daily maximum SFD, a and b are the coefficients.

Fig. 4.

Normalized daily maximum tree water deficit (TWD) as a function of sap flow density (SFD) considering: (A) species median values (Fagus sylvatica, Fraxinus excelsior and Acer pseudoplatanus) for the dry (red circles, when SWC < 0.2) and wet (blue triangles, when SWC ≥ 0.2) phases. (B) Species median values following a soil moisture (SWC) gradient for the whole study period. (C) Species median values following a SWC gradient in strictly drying soil moisture conditions, and (D) for F. sylvatica (12 trees) also in strictly drying conditions. The soil moisture gradient was considered from wetter (blue) to drier (red) conditions. In (A) and (B), rising and falling arrows added to the plots of F. sylvatica indicate the directional course of the hysteresic cycle from wetter (late June) to drier (mid‐September) and back to wetter (late September) soil moisture conditions. In (D), the plot headings starting with ‘Fs−’ represent tree IDs.

At the individual level the correlation SFD‐TWD was modelled with nonlinear logistic functions:

$$\mathrm{TWD}=\frac{\text{Asym}}{1+{\mathrm{e}}^{\left(\text{xmid}-\left(\frac{\mathrm{SFD}}{\text{scal}}\right)\right)}}$$ (3)

where Asym is the asymptote, xmid is × value at inflection of the curve, and scal is the scale parameter.

We plotted a matrix chart of simple linear models (method: ‘Spearman’) between tree characteristics (diameter at breast height, tree height, crown volume, crown competition index and box dimension) and the three parameters Asym, xmid and scal for each F. sylvatica tree to assess the main drivers of such correlations. We assessed the time of daily SFD and TWD occurrence peaks for each tree and the resulting temporal values (in HH:MM:SS) were averaged per species across the total study period, the dry phase (SWC < 0.2), and the wet phase (SWC ≥ 0.2). Outliers were identified and removed when required.

RESULTS

Drought‐induced sap flow decrease and tree water deficit

The daily maximum air temperature, VPD and incoming PPFD, from 19 June to 30 September 2022, ranged from 7 to 34°C, 0.2 to 4.7 kPa, and 497 to 2173 μmol m⁻² s⁻¹, respectively. There was generally low rainfall from 19 June to mid‐August, and the predicted rainfall for the rest of the study period indicated more intense precipitation. Gap‐filled rainfall values were somewhat imprecise, but they reflected actual precipitation events that had not immediately resulted in soil water replenishment. In fact, SWC at 0–30 cm depth decreased consistently and reached a minimum in mid‐September (Fig. 1A).

Fig. 1.

(A) daily mean air temperature (°C) and soil water content over the top 30 cm (SWC, %), and daily maximum water vapour pressure deficit (VPD, kPa), incoming photosynthetically active photon flux density (PPFD, μmol m⁻² s⁻¹) and rainfall (mm). (B) Hourly median absolute sap flow density (SFD, g cm⁻² h⁻¹, solid line) and tree water deficit (TWD, mm, dashed line) for Fagus sylvatica (top), Fraxinus excelsior (middle) and Acer pseudoplatanus (bottom) from 19 June to 20 September 2022. Rainfall values from 20 August to 31 October are indicated with dashes because they were gap‐filled as described in the Methods. The dry and wet phases of the study period are indicated by light‐red and light‐blue fonts, respectively.

SFD declined in parallel to soil drying, TWD increased to its highest level, and differences were found between species. Initial SFD values were higher for A. pseudoplatanus (max: 30.3 g cm⁻² h⁻¹) and F. sylvatica (max: 20.4 g cm⁻² h⁻¹) than for F. excelsior (max: 17.8 g cm⁻² h⁻¹). TWD, on the other hand, varied between species in the sequence F. excelsior (max: 1.4 mm) > F. sylvatica (max: 0.5 mm) > A. pseudoplatanus (max: 0.2 mm) (Fig. 1B).

From highest soil moisture (0.27) to lowest (0.17), SFD reductions in F. sylvatica (mean ± SE: 11.0 ± 1.4 g cm⁻² h⁻¹), F. excelsior (12.5 ± 3.0 g cm⁻² h⁻¹) and A. pseudoplatanus (10.8 ± 3.1 g cm⁻² h⁻¹) did not differ significantly (P = 0.78; Fig. 2A). When SFD values were normalized, the median SFD decrease for F. excelsior was slightly lower than for the other two species, yet the three species did not differ significantly (Fig. S2). In contrast, daily maximum TWD in F. sylvatica (0.21 ± 0.02 mm) and F. excelsior (mean ± SE: 0.62 ± 0.17 mm) did not differ significantly from each other but were significantly higher (P = 0.0003) compared to A. pseudoplatanus (0.06 ± 0.02 mm) (Fig. 2B).

Fig. 2.

Comparison of sap flow density (SFD, g cm⁻² h⁻¹) decrease (A) and daily maximum tree water deficit (TWD, mm) (B) between the species Fagus sylvatica, Fraxinus excelsior and Acer pseudoplatanus observed from 19 June to 30 September 2022 (ns: non‐significant, *: 0.01 ≤ P ≤ 0.05 & ***: P ≤ 0.01 based on the Kruskal‐Wallis test). Error bars ± SD.

Effects of meteorological conditions on tree water status

Analysis of SFD and TWD responses to changes in SWC, VPD and PPFD during the wet (SWC ≥ 0.2) and dry (SWC < 0.2) phases of the study period found contrasting relationships to specific meteorological variables. In both wet and dry phases, SFD and TWD responded significantly to SWC change for all three species following a generalized linear model (GLM) (Fig. 3). The negative correlation of TWD with SWC for F. excelsior in the dry phase (R ² = 0.62) was stronger than that of F. sylvatica (R ² = 0.22) and A. pseudoplatanus (R ² = 0.11), while SFD and SWC positively correlated in an opposite order: F. excelsior (R ² = 0.60) < A. pseudoplatanus (R ² = 0.76) < F. sylvatica (R ² = 0.84). In the wet phase the TWD and SFD responses to SWC increase did not deviate much from each other. VPD could partly explain TWD variation in the wet phase but not in the dry phase while, conversely, VPD accounted for some SFD variation only in the dry phase. The strength of these relationships was similar across species. PPFD correlated positively with SFD in all the species only during the dry phase, and negatively with TWD only during the dry phase and only for F. excelsior (R ² = 0.35) and F. sylvatica (R ² = 0.10).

Fig. 3.

Responses of: (A) tree water deficit (TWD, mm) and (B) sap flow density (SFD, g cm⁻² h⁻¹) to soil moisture content (SWC), vapour pressure deficit (VPD, kPa), and incoming photosynthetically active photon flux density (PPFD, μmol m⁻² s⁻¹) observed from 19 June to 30 September 2022 (data considered when VPD > 1 kPa and net solar radiation > 500 W m⁻²). Values are averaged (median) for each species, i.e., Fagus sylvatica (left), Fraxinus excelsior (middle) and Acer pseudoplatanus (right). Two phases are distinguished: dry (red, when SWC < 0.2) and wet (blue, when SWC ≥ 0.2). Curves were fitted using generalized linear models and nonlinear least squares models for significant non‐linear patterns.

Correlation between SFD and TWD along a changing soil moisture gradient

When dry and wet phases were considered separately, daily maximum TWD did not correlate significantly with SFD (Fig. 4A). Without differentiating between the dry and wet phases and for the entire study period, TWD correlated weakly but significantly with SFD in the order F. sylvatica (R ² = 0.31) > A. pseudoplatanus (R ² = 0.26) > F. excelsior (R ² = 0.13) with signs of hysteresis effects from soil drying to rewetting conditions (Fig. 4B). In the soil drying phase (decreasing SWC) of the study period, the response of TWD to SFD variation was even more significant, following a nonlinear least squares model (SFD = 1− a × exp(b × TWD)) and in the same order for the three species F. sylvatica (R ² = 0.80) > A. pseudoplatanus (R ² = 0.63) > F. excelsior (R ² = 0.56) (Fig. 4C). Similar analyses at the individual tree level instead followed a logistic distribution with periods of high SFD variation in dry and wet limits of the soil drying phase, opposite that of TWD, which varied more between dry and wet limits (Fig. 4D).

The time of daily SFD peak did not differ significantly between the three species and was similar for the daily TWD peak. Within species, however, a significant offset between the daily peaks of TWD and SFD was observed, with the daily maximum TWD occurring later than daily maximum SFD. In fact, SFD peaked on average around noon (ca. 11:00:00–14:00:00 h CEST) while TWD peaked a bit later (ca. 13:20:00 h CEST to 16:30:00 h CEST) (Fig. 5).

Fig. 5.

Tree water deficit (TWD) of Fagus sylvatica (Fs), Fraxinus excelsior (Fe) and Acer pseudoplatanus (Ap). Daily peaks were derived taking into account the entire study period (Total), the dry phase (Dry, SWC < 0.2) and the wet phase (Wet, SWC ≥ 0.2). Error bars denote ± SD. The Wilcoxon test was used for comparisons of average peak times within species. Resulting significances *, *** and **** correspond to P < 0.05, P < 0.0001, and P < 0.00001 respectively.

For the three species the average hourly patterns of normalized SFD and TWD were plotted during the wet (when SWC ≥ 0.2) and dry (when SWC < 0.2) phases. We found that daily maximum SFD lagged from daily maximum VPD by ca. −5–−4 h during the wet phase and by ca. −4–−3 h during the dry phase. TWD lagged from VPD by ca. −2–0 h during the wet phase and by ca. 0–3 h during the dry phase (Fig. 6).

Fig. 6.

Daily changes in median sap flow density (SFD), tree water deficit (TWD) and vapour pressure deficit (VPD, kPa) averaged hourly for Fagus sylvatica, Fraxinus excelsior and Acer pseudoplatanus during the wet period (A) and dry period (B).

DISCUSSION

Species‐specific differences in sap flow density and stem circumference change under drought stress

In this study we simultaneously monitored SFD and TWD of 21 trees to better understand the physiological mechanisms behind the drought responses of F. sylvatica, F. excelsior and A. pseudoplatanus. Our first hypothesis, that there would be species‐specific SFD and TWD responses to drought, was only partially confirmed. In response to soil drying, all three species exhibited considerable SFD decreases, with species medians exceeding 10 g cm⁻² h⁻¹. However, these decreases did not significantly differ among species. The SFD decrease for F. excelsior at our site was comparable to that of F. sylvatica and A. pseudoplatanus. This unexpected finding cannot be explained consistently by the concept of stomata regulation and/or wood anatomy as key mechanisms of overall drought resistance or water management strategies. For example, Köcher et al. (2009) and Hölscher et al. (2005) observed strong resistance of F. excelsior to extreme water‐limiting conditions and consistent sap flow patterns during the summer in the Hainich National Park. Fraxinus excelsior is known for its deep roots and dense fine root system in mixtures, its ring porous wood anatomy and its lower leaf area index compared to F. sylvatica (Petriţan et al. ²⁰⁰⁹; Jacob et al. ²⁰¹³; Bader et al. ²⁰²²). All these factors could have contributed to their efficient access to water and reduced capillary water loss. However, at our site, sap flow reduction in F. excelsior decreased, except in August when sap flow patterns were more consistent than those of F. sylvatica and A. pseudoplatanus (Fig. 1). A possible explanation why its SFD decrease resembled that of the two other species may have been its ring‐porousness, counteracting the effect of other factors such as anisohydry. Another influencing factor may be an infection by Hymenoscyphus fraxineus. This pathogen is responsible for the “European ash dieback” and this disease is distributed throughout Germany where F. excelsior grows (Nguyen et al. ²⁰¹⁶; Langer et al. ²⁰²²). After an infection, mainly through stomata, fungal hyphae spread into twigs, branches and stems via vascular tissue, potentially causing tissue death (Carroll & Boa ²⁰²⁴). Our tree selection was based solely on visual assessments of vitality, potentially including infected trees. Moreover, the high sap flow sensitivity of F. sylvatica to drought was associated with its large leaf area and more limited access to soil water by the roots (Gessler et al. ²⁰²²), exacerbating sap flow reduction under soil drying conditions (Schäfer et al. ²⁰¹⁹). The predominant anisohydric stomatal response of F. sylvatica could also have contributed to a delayed stomatal closure despite drought stress. resulting in rapid water loss (Tardieu & Simonneau ¹⁹⁹⁸), even though some recent studies showed that stomatal control is not a prominent factor affecting water use and that other factors, such as stem rehydration, should also be considered (Peters et al. ²⁰²³). Acer pseudoplatanus typically thrives in relatively moist to wet sites, resulting in high water consumption and increased xylem sensitivity to embolism under soil water‐limiting conditions (Lemoine et al. ²⁰⁰¹). Additionally, its isohydric stomatal response to drought stress may account for the considerable SFD decrease observed at our study site. Isohydric species keep a more constant leaf water potential through earlier stomatal closure during drought, which reduces sap flow (Chen et al. ²⁰²¹). Our conclusion, consistent with recent studies (Hartmann et al. ²⁰²¹; Leuschner et al. ²⁰²²), indicates that the concepts of classifying species' drought resistance or sensitivity based on single plant traits (wood anatomy, stomatal regulation, stem/leaf water potential, fine root biomass, or rooting depth) are overly simplistic and may result in serious misinterpretations and flawed risk assessments of a tree species under climate change conditions.

We observed a significantly higher TWD in F. excelsior compared to F. sylvatica and A. pseudoplatanus. This finding may have been related to morphological differences between the three species. In fact, the larger vessels in the outer xylem of the ring‐porous F. Excelsior, combined with thicker bark, may have enhanced its shrinkage intensity during water‐limiting periods compared to the diffuse‐porous F. sylvatica and A. pseudoplatanus, both of which have much thinner bark. Bark exposure to humidity can affect bark water‐holding capacity and, consequently, TWD in air‐drying conditions. For example, Ilek et al. (2017) observed that approximately 10%–30% of maximum water stored in the bark is attributed to its hygroscopic capacity, highlighting bark sensitivity to air moisture fluctuation. The combined effects of air and soil drying may therefore more significantly influence TWD in species with thicker bark, such as F. excelsior, than in species with thinner bark, as observed in this study.

Sap flow density and tree water deficit responses to extreme environmental conditions

The simultaneous decrease in SFD and the gradual increase in TWD observed for the three tree species during the summer confirmed the direct effect of drying conditions on tree water status. Our second hypothesis, predicting stronger responses of SFD and TWD to SWC compared to VPD and PPFD variation, was mostly confirmed for the dry (SWC < 0.2) and wet (SWC ≥ 0.2) phases in all three species. The significant reduction of SFD and increase in TWD resulting from soil drying confirmed previous findings from studies carried out in different species and under various environmental conditions (Köcher et al. ²⁰⁰⁹; Sánchez‐Costa et al. ²⁰¹⁵; Nadal‐Sala et al. ²⁰¹⁷; Montoro et al. ²⁰²⁰; Szatniewska et al. ²⁰²²). Interestingly, the increase in TWD upon soil drying was higher in the wet period compared to the dry phase for F. sylvatica and A. pseudoplatanus, whereas F. excelsior had a stronger TWD response during the dry phase. Because of lower TWD rates in F. sylvatica and A. pseudoplatanus compared to F. excelsior, most of the shrinkage–soil water relationship was observed when soil moisture conditions were still above 0.2, whereas F. excelsior extended its stem water loss to below 0.2 SWC, resulting in a much stronger TWD response during the dry phase. Studies on TWD responses to soil drying under different soil moisture limits are rare, but findings such as those of Steppe & Lemeur (2007) revealed greater elasticity of stem water tissue in the ring‐porous Quercus robur compared to diffuse‐porous F. sylvatica. Our results suggest that the same mechanism applies to ring‐porous F. excelsior, reflecting greater elasticity compared to diffuse‐porous F. sylvatica and A. pseudoplatanus during the driest phase of the study period. The thicker bark of F. excelsior compared to the other species may have also contributed to its higher TWD response to drought. On the other hand, the SFD response to soil drying for F. sylvatica and A. pseudoplatanus was somewhat higher during the dry phase than the wet phase, while F. excelsior responded more significantly in the wet phase. The latter result could be related to differences in SFD ranges for each species and at respective phases. The SFD range in F. excelsior during the dry phase (4.6–11.9 g cm⁻² h⁻¹) was smaller in comparison with those of F. sylvatica and A. pseudoplatanus (2.7–14.9 g cm⁻² h⁻¹ and 4.1–19.3 g cm⁻² h⁻¹, respectively), which could explain the weaker SFD response to SWC depletion in extremely dry conditions. In fact, when values were normalized (Fig. S3), the SFD range of F. excelsior during the dry phase was not lower than those of F. sylvatica and A. pseudoplatanus and, hence, the correlation with SWC was stronger than that seen using absolute values. Moreover, as previously mentioned, the SFD of F. excelsior in August (corresponding to the dry phase) was consistent, moderating its response to SWC in that period. Our findings were generally in line with other studies, which have found that increasing soil drying affects sap flow reduction more strongly in the diffuse‐porous F. sylvatica and A. pseudoplatanus compared to the ring‐porous F. excelsior (Hölscher et al. ²⁰⁰⁵; Brinkmann et al. ²⁰¹⁶).

In all three species, TWD correlated positively and significantly to VPD during the wet phase but not the dry phase. Several studies have also found that TWD increased with increasing VPD (Deslauriers et al. ²⁰⁰⁷; Ehrenberger et al. ²⁰¹²; Ma et al. ²⁰²¹; Zhao et al. ²⁰²³), and Nehemy et al. (2021), following a similar pattern to ours, also found very weak correlations of TWD to VPD in water‐limiting conditions compared to wet phases. The effect of VPD on TWD can be translated as the direct effect of air drying on stem water content and, when drying conditions are extreme and stem shrinkage close to its maximum, VPD increment yields little further TWD reaction. On the other hand, SFD was positively correlated with VPD only during the dry phase, with species‐specific SFD thresholds where VPD increment did not stimulate any further sap flow increase. Despite the somewhat similar correlation strengths between the three species, the saturation levels of SFD differed in the following order: ca. 10 g cm⁻² h⁻¹ for F. excelsior, ca. 15 g cm⁻² h⁻¹ for F. sylvatica, ca. 20 g cm⁻² h⁻¹ for A. pseudoplatanus. This indicates that the two diffuse‐porous species, F. sylvatica and A. pseudoplatanus, had a higher tolerance to VPD increment compared to the ring‐porous F. excelsior in extremely dry soil conditions. The infection by Hymenoscyphus fraxineus pathogens might here again justify the different SFD response of F. excelsior compared to the two other species. In the study of Hölscher et al. (2005), F. excelsior was the only species that did not respond to VPD; consistent sap flow was observed in that species during the dry period. However, in contrast to our findings, they found that the sap flow correlation to VPD during the wet phase was stronger than during the dry phase. This discrepancy could be related to their chosen VPD range (0–2 kPa), which in our study was broader (1–5 kPa) in order to exclude wet days. The inclusion of lower VPD values would have enhanced the SFD response during the wet phase of our study period. We observed that most of the VPD effect was evident within the range 1–3 kPa. Beyond this threshold, trees would likely close their stomata to minimize water loss (Lendzion & Leuschner ²⁰⁰⁸).

Between July and September, Tian et al. (2023) observed a delayed response of TWD to VPD (c. 3–5.5 h) in Larix gmelinii Rupr. They found a longer TWD – VPD lag during the recovery period compared to the period of water loss. In our study, the time lag between TWD and VPD was also longer in the dry phase (ca. 0–3 h) compared to the wet phase (ca. −2–0 h). In addition, during the wet phase TWD peaked earlier than VPD whereas during the dry phase TWD peaked later, which might reflect decoupling of TWD from VPD under drier conditions. Salomón et al. (2022) indicated that increasing drought intensity may increase stem dehydration, which would explain the delayed lag between TWD and VPD in extremely dry conditions. Interestingly the time lags between SFD and VPD were equal during the wet and dry phases (ca. −5–−4 h and ca. −4–−3 h, respectively), which might indicate a comparable sensitivity of SFD to VPD at both phases. As stated above, the lack of correlation between SFD and VPD during the wet phase might partly be caused by the exclusion of wetter conditions (VPD < 1 kPa) from our analysis.

Both SFD and TWD correlated significantly with PPFD only in the dry phase. Several studies in various forest ecosystems have already shown that PPFD has control over sap flow or TWD (Bovard et al. ²⁰⁰⁵; Huang et al. ²⁰²⁴; Perron et al. ²⁰²⁴). Perron et al. (2024) specifically acknowledged the important role of photosynthetically active radiation on transpiration that induces water loss and consequently stem radial shrinking on a daily basis as well as over several days to weeks. However, high radiation might help to sustain transpiration and thus water uptake from the soil despite low SWC. Such a ‘driving force’ of radiation might explain the relatively strong correlation of PPFD to SFD only in the dry phase and for all species in our study. Furthermore, high PPFD in combination with high SFD might promote soil water uptake and the replenishing of stem water resources despite low SWC, which, in turn, reduces TWD and might explain the observed, species‐specific decrease of TWD with increasing PPFD in the dry phase (Fig. 3). The negative relationship between TWD and PPFD was relatively strong for the ring‐porous F. excelsior, but weak or even absent for the diffuse‐porous F. sylvatica or A. pseudoplatanus, respectively. This would correspond with the finding that at low SWC F. excelsior can sustain higher sap flow (at least in relative terms) for a longer time than the other two species (see the discussion above). Nevertheless, we cannot exclude that the decrease of TWD with increasing PPFD does not reflect a functional relationship, but rather the coincidence of a decreasing mean PPFD since August (Fig. 1) related mainly to the solar altitude and an increasing TWD driven mainly by a decreasing SWC.

Correlation between tree water deficit and sap flow density

When the wet and dry phases were considered separately, no significant responses of TWD to SFD variation were found, except for F. sylvatica in the dry phase (Fig. 4A). However, the fact that no relationship between TWD and SFD was found is not meaningful because a continuity between dry and wet phases expressed as hysteresis effect is visible for each species. Figure 4B illustrates a stronger TWD response to SFD across the study period than when dry and wet phases are considered separately, in the order F. sylvatica (R ² = 0.31) > A. pseudoplatanus (R ² = 0.26) > F. excelsior (R ² = 0.13). There were clear signs of hysteresis effects from wetter (late June) to drier (mid‐September) and back to wetter (late September) soil moisture conditions. In Fig. 4B, the hysteresis effect appears to play a role in attenuating the TWD response to SFD throughout the entire study period. Brinkmann et al. (2016) also observed a significant relationship between TWD and SFD from May to September, except for F. excelsior, which did not show a significant decrease in sap flow under drying soil conditions. Brinkmann et al. (2016) mainly emphasized the species‐specific sap flow responses to TWD, although noticeable signs of hysteresis effects, which were not specifically addressed by the authors, were also observed. One potential reason for the hysteresis effect reflected in our data may be the time lag observed between SFD and TWD across the study period. While the daily SFD reached its maximum around noon, TWD peaked ca. 2–4 h later (Fig. 5). This observation suggests a complex interaction between tree physiological responses, soil moisture, and atmospheric conditions. Under drying conditions, the effect of stem water stress capacitance, as described by Preisler et al. (2022), favours the use of stem water reserves to mitigate TWD, helping to maintain sap flow and delaying the impact of water deficit. Tian et al. (2023) and Sitková et al. (2014) observed that stem shrinkage occurred as a consequence of sap flow breakdown, thus supporting our finding on the delayed response of stem water deficit to sap flow depletion under soil drying conditions. Furthermore, our previous observation of a stronger correlation of SFD with SWC and VPD during the dry phase (SWC < 0.2) compared to the correlation of TWD to the same variables in all three species, supported a greater drought sensitivity of SFD compared to TWD and thus its earlier peak.

According to Farrell (1999), the hysteresis effect is associated with a specific curve shaped by an increasing and then decreasing strength of correlations. In our study, the decreasing correlation corresponded to a TWD decrease in response to SFD increase. This case is illustrated in Fig. 4C, where the negative nonlinear TWD response to SFD was found from drier to wetter conditions, with saturation of TWD in the driest condition. The patterns in Fig. 4C were averaged per species but obscure the actual course of correlations between TWD and SFD observed at the individual tree level. At the individual tree level, the function of TWD response to SFD was sigmoidal, revealing two levels of saturation for TWD corresponding to points of important SFD fluctuation (Fig. 4D). In fact, SFD varied more under extremely wet and dry conditions, while TWD varied more between these extremes. The nonlinear relationship between SFD and TWD was even more variable for F. excelsior with three of five trees slightly deviating from the ‘S’ shape. This may reflect varying sensitivities of SFD and TWD in F. excelsior compared to F. sylvatica and A. pseudoplatanus, but with the low sample sizes of F. excelsior and A. pseudoplatanus, it was not possible to look more deeply into species comparisons. The parameters of sigmoidal functions of TWD variation in response to SFD change, i.e., asymptote, xmid (x value at the inflection points of the curve), and scale, (Table S2) were not significantly different between the species, and within each species none of the parameters could be explained by the trees' characteristics, i.e., diameter at breast height, total tree height, crown volume, crown competition index, or box dimension (Fig. S7). When the three species were considered together a significant effect of the box dimension on the xmid parameter was observed. The latter effect may, however, have been a result of a larger sample size compared to the case when species were assessed individually. Several studies have stressed the relevance of tree height and diameter for canopy stomatal conductance and SFD in various diffuse‐ and ring‐porous species (Schäfer et al. ²⁰⁰⁰; Chiu et al. ²⁰¹⁶; Schoppach et al. ²⁰²¹; Pretzsch et al. ²⁰²²). However, in our study, the complex interactions between TWD and SFD were not adequately explained by these important tree characteristics, leaving open questions about the drivers of TWD responses to SFD under soil drying conditions. It may also be that other tree measurements, such as stem and leaf water potential, sapwood area etc., could help to elucidate the complex mechanisms of simultaneous SFD and TWD responses and adaptation to drought, but that should be tested in future studies with larger sample sizes.

The nonlinear negative response of TWD to SFD suggested different levels of saturation and variability for the two variables under varying drought stress degrees. The stronger correlation of SFD with SWC and VPD during the dry phase of our study period compared to the correlation of TWD to the same variables suggests that TWD cannot be used as a proxy for SFD and vice versa under conditions of drought stress.

AUTHOR CONTRIBUTIONS

LSD, MM and CA planned and designed the experiment. LSD and MGH conducted fieldwork. DS and MGH provided scanner‐derived variables. LSD, FK, PS, CA and MM conducted the statistical analysis. LSD wrote the manuscript. LSD, MM, PS, CA, FK, DS and MGH reviewed the manuscript.

CONFLICT OF INTEREST STATEMENT

The authors declare that they have no conflict of interest.

Supporting information

Table S1. Correlation between inner and outer sap flow densities (SFD) method: Pearson. Values of R ² (coefficient of determination) below 0.9 result from unclear patterns of outer sap flow densities due to important noises.

Table S2. Summary of trees’ dendrometric variables (diameter at breast height, tree height, crown volume), competition indices (crown competition index, box dimension) and parameters of the correlations TWD‐SFD (asymptote, xmid point and scale).

Table S3. Values of basic density of wood per species from Gebauer et al. (2008).

Figure S1. Patterns of relative sap flow density (SFD, solid line) and tree water deficit (TWD, dashed line) during the same period and averaged for F. sylvatica (top), F. excelsior (middle) and A. pseudoplatanus (bottom). The dry and wet phases of the study period are indicated by light‐red and light‐blue color fonts, respectively

Figure S2. Comparison of relative sap flow density decrease (A) and maximum tree water deficit (TWD, mm) (B) between the species F. sylvatica, F. excelsior and A. pseudoplatanus observed from June 19 to September 30, 2022 (ns: non‐significant, : 0.01 ≤ P ≤ 0.05 & : P ≤ 0.01 based on the Kruskal‐Wallis test). The error bars denote SD values.

Figure S3. (A) relative tree water deficit (TWD) and (B) sap flow density (SFD) to soil moisture content (SWC), vapor pressure deficit (VPD, kPa) and incoming photosynthetically active photon flux density (PPFD, µmol m^–2 s^–1) observed from June 19 to September 30, 2022 (data considered when VPD >1 kPa and net solar radiation > 500 W m^–2). Values are averaged (median) for each species, i.e., F. sylvatica (left), F. excelsior (middle) and A. pseudoplatanus (right). Two periods are distinguished: dry (red, when SWC < 0.2) and wet (blue, when SWC ≥ 0.2). Curves were fitted using generalized linear models and nonlinear least squares models for significant non‐linear patterns

Figure S4. (A) tree water deficit (TWD, mm) and (B) sap flow density (SFD, g cm^–2 hr^–1) to soil moisture content (SWC), vapor pressure deficit (VPD, kPa) and incoming photosynthetically active photon flux density (PPFD, µmol m^–2 s^–1) observed from June 19 to September 30, 2022. Values are averaged (median) for each species, i.e., F. sylvatica (left), F. excelsior (middle) and A. pseudoplatanus (right).

Figure S5. (A) relative tree water deficit (TWD) and (B) sap flow density (SFD) to soil moisture content (SWC), vapor pressure deficit (VPD, kPa) and incoming photosynthetically active photon flux density (PPFD, µmol m^–2 s^–1) observed from June 19 to September 30, 2022. Values are averaged (median) for each species, i.e., F. sylvatica (left), F. excelsior (middle) and A. pseudoplatanus (right).

Figure S6. Normalized tree water deficit (TWD) to sap flow density (SFD) for F. sylvatica (twelve trees), F. excelsior (five trees) and A. pseudoplatanus (four trees), across soil ‘drying’ gradient. The soil moisture gradient was considered from wetter (blue) to drier (red) conditions. When the correlation was insignificant, no fitting curve was plotted. The plot headings starting with ‘Fs‐’, ‘Fe‐’, and ‘Ap‐’ represent the tree IDs. For all the plots P was < 0.05 except for tree ‘Ap618’ where the correlation was insignificant (P = 0.27).

Figure S7. Statistics resulting from the correlation matrix between the parameters of the correlation TWD‐SFD (asymptote, xmid and scale) and the trees’ characteristics (diameter at breast height, height, crown volume, competition, box dimension) when the three species are considered (a), for Beech only (b) and for Ash only (c). No correlation matrix was plotted for maple because of a low sample size. Between variables, simple linear correlations (method pearson) were applied.

DATA AVAILABILITY STATEMENT

Data are available on request from the corresponding author upon reasonable request.