Importance of tree diameter and species for explaining the temporal and spatial variations of xylem water δ18O and δ2H in a multi‐species forest

Identifying the vegetation and topographic variables influencing the isotopic variability of xylem water of forest vegetation remains crucial to interpret and predict ecohydrological processes in landscapes. In this study, we used temporally and spatially distributed xylem stable water isotopes measurements from two growing seasons to examine the temporal and spatial variations of xylem stable water isotopes and their relationships with vegetation and topographic variables in a Luxembourgish temperate mixed forest. Species‐specific temporal variations of xylem stable water isotopes were observed during both growing seasons with a higher variability for beeches than oaks. Principal component regressions revealed that tree diameter at breast height explains up to 55% of the spatial variability of xylem stable water isotopes, while tree species explains up to 24% of the variability. Topographic variables had a marginal role in explaining the spatial variability of xylem stable water isotopes (up to 6% for elevation). During the drier growing season (2020), we detected a higher influence of vegetation variables on xylem stable water isotopes and a lower temporal variability of the xylem water isotopic signatures than during the wetter growing season (2019). Our results reveal the dominant influence of vegetation on xylem stable water isotopes across a forested area and suggest that their spatial patterns arise mainly from size‐ and species‐specific as well as water availability‐dependent water use strategies rather than from topographic heterogeneity. The identification of the key role of vegetation on xylem stable water isotopes has critical implications for the representativity of isotopes‐based ecohydrological and catchments studies.

between patterns of tree water uptake (informed by the isotopic composition of tree stems) and variations of vegetation and topography in the landscape. Therefore, we sampled tree trunk water of around 350 trees scattered across a forested area in Luxembourg to study how water uptake differed between trees and how it was related to the vegetation (tree species and diameter) and the topography (e.g. elevation and slope).
Global change poses a considerable challenge for forest and water resource management due to shifts in precipitation regimes and temperatures that may affect tree water uptake and plant water availability (e.g. Capell et al., 2013), species composition and distribution  and forest productivity (Boisvenue & Running, 2006). Therefore, an improved understanding of the influence of biotic and abiotic factors on tree water uptake is needed to better evaluate the variations of tree water use (Frank et al., 2015) and manage forest ecosystems to enhance their adaptive response to environmental stressors.
Stable isotopes of oxygen (ratio of 18 O to 16 O) and hydrogen (ratio of 2 H to 1 H) of water have been widely used to understand hydrological and ecohydrological processes in catchments (e.g. Bögelein et al., 2017;Brinkmann et al., 2018;Fabiani et al., 2021;Goldsmith et al., 2012Goldsmith et al., , 2018Penna et al., 2018;Sprenger et al., 2016Sprenger et al., , 2018Tetzlaff et al., 2020). They are an important tool for describing the movement of water through catchments and ecosystems (Kendall & McDonnell, 1998;. Stable water isotopes (SWI) have proved to be essential for investigating water fluxes in the soil-plant-atmosphere continuum (e.g. Fabiani et al., 2021;Goldsmith et al., 2012), flow paths and associated transit times of water in the subsurface (Asadollahi et al., 2022;Knighton et al., 2019;Sprenger et al., 2016), along hillslopes (Asano & Uchida, 2012) and in catchments (Kuppel et al., 2018;Rodriguez & Klaus, 2019;Sprenger et al., 2022;Wang et al., 2023) showing their potential to decipher water movements in the critical zone. Analysing SWI in plant stem water (usually assumed to be xylem water) is important for quantifying water sources and water use by plants (Penna et al., 2018).
Xylem water is a mixture of the different water sources used by trees (i.e. soil water from different depths and groundwater) (Penna et al., 2018). The variations of xylem SWI are therefore related to variations in these sources and how they are taken up by trees. The temporal variability of xylem SWI is related to soil SWI that are themselves modified via mixing with infiltrating rainwater with its own interstorm and intrastorm variation in isotopic composition (Bertrand et al., 2014;Sprenger et al., 2018). Additional temporal drivers of xylem SWI are meteorological conditions (e.g. air temperature, net radiation and humidity) that influence evaporation-fractionation of soil SWI (e.g. Bertrand et al., 2014). Vegetation characteristics such as species, forest type, tree height, diameter and above ground biomass (e.g. Fabiani et al., 2021;Goldsmith et al., 2012Goldsmith et al., , 2018Snelgrove et al., 2021) also influence xylem SWI. The effect of vegetation variables on xylem SWI is attributed to possible species-specific differences in the timing and intensity of water use (Snelgrove et al., 2021) or depth of water uptake (e.g. Brinkmann et al., 2019;Fabiani et al., 2021;Goldsmith et al., 2022;Kahmen et al., 2021).
Some studies addressed the spatial variability of xylem SWI at plot (Goldsmith et al., 2018), hillslope Goldsmith et al., 2012) and catchment (Gaines et al., 2016) scales relying on 12 to 60 trees sampled on the same day. Those studies showed that xylem SWI were influenced by soil depth at the tree location, vegetation variables (e.g. Gaines et al., 2016) and the depth and lateral distributions of soil SWI (Gaines et al., 2016;Goldsmith et al., 2018).
Spatial variations of soil SWI can themselves be related to topography (e.g. aspect, slope and elevation) that affect water movement in soil, energy inputs for evaporation and in consequence soil SWI. Topography (e.g. elevation) can also influence patterns in vegetation, the depth and accessibility of tree water sources and, in turn, tree water uptake depth and xylem SWI. Beyer and Penna (2021) emphasised that spatial data of xylem and soil SWI are scarce. For this reason, we are lacking an understanding of the relationship between xylem SWI and topographic and vegetation variables. Particularly, it remains unclear what is the relative importance of topographic and vegetation variables in explaining the spatial variability of xylem SWI.
In this study, we address this need by exploring the temporal and spatial variations of xylem SWI and their relationship with vegetation and topographic variables in a mixed beech-oak forest. We measured δ 18 O and δ 2 H of xylem water over two growing seasons (17 sampling campaigns) in $350 trees in the Weierbach catchment, Luxembourg.
We went beyond the sampling strategy generally used in ecohydrological studies (i.e. four tree individuals per species on average on each sampling campaign; Goldsmith et al., 2018) and sampled on average, for each sampling campaign, 10 tree individuals per species. To the best of our knowledge, this was the first analysis with more than 300 xylem samples determining the relative importance of vegetation and topographic variables in explaining the spatial variability of xylem SWI and highlighting its interannual change. The high total number of xylem samples taken within the 42-ha forested area provided a high sampling density (>3.3 trees/ha per growing season) and variations in vegetation and topography on which we based the spatial analysis.
Our research questions are guided by a perceptual model of the influence of vegetation and topography on xylem SWI. We know that species influence xylem SWI through its effect on tree water uptake depth (e.g. Fabiani et al., 2021). We also know that tree diameter at breast height (DBH) is associated with depth of tree water uptake (e.g. Schoppach et al., 2021), and we thus expect DBH to influence xylem SWI. Specifically, in the Weierbach catchment, trees rely on soil water with no significant uptake of groundwater .
Soil water availability is linked to vertical and lateral redistribution mechanisms of infiltrating water along the hillslope Rodriguez & Klaus, 2019), and we therefore believe that elevation, TPI and slope influence tree water uptake depth and, in turn, xylem SWI. We also expect these variables, along with aspect, to influence xylem SWI by affecting evaporation-fractionation of soil SWI. Multiple landscape variables can therefore affect xylem SWI, and their possible interactions are challenging to untangle their respective influence on xylem SWI. We hypothesize that xylem SWI vary systematically following the perceptual model; we address this hypothesis by investigating the following research questions: 2 | MATERIALS AND METHODS

| Study area
The Weierbach is a 42-ha forested headwater catchment located in the northwest of Luxembourg   (Figure 1). The region is characterised by gently sloping plateaus cut by deep V-shaped valleys. Two landscape units are distinguished depending on their subsolum type and their slope: plateaus (about 30 ha, slopes between 0 and 5 ) and hillslopes (about 12 ha, slopes between 5 and 44 ) (Martínez-Carreras et al., 2016). Furthermore, there is a small riparian zone of up to 3-m wide surrounding most of the stream network and representing about 0.4 ha (Glaser et al., 2020).
Detailed topographic variables were calculated from a highresolution (1 m) digital elevation model (DEM) (Luxembourgish air navigation administration, 2017) and included aspect ( ), slope ( ), curvature (À) and drainage area (m 2 ) for each 1 Â 1 m DEM pixel. The aspect represents the direction the downhill slope faces (measured clockwise from 0 (north) to 360 (north)), the slope represents the steepness, the curvature indicates if the surface is upwardly convex (positive value), concave (negative value) or flat (value of 0) and the drainage area indicates the area from where water flows downslope   (Table 1). Based on these variables, the topographic position index (TPI; À) and topographic wetness index (TWI; ln(m)) were calculated (Wilson & Gallant, 2000) as follows: where E is the elevation at a specific location (m) and E avg50 is the average elevation in a 50 m radius circle around this location (m). The TPI value decreases from the catchment ridges to valley.
The TWI characterises terrain-driven propensity for saturation and is calculated as follows: where A s is the specific area (i.e. drainage area per unit contour length) in m 2 m À1 and β is the slope in .
The bedrock in the Weierbach catchment consists mostly of Devonian slate containing schist, phyllite, and quartzite (Juilleret et al., 2011). Pleistocene periglacial slope deposits cover the bedrock, and the soil developed from these deposits is Leptic Cambisol (Juilleret et al., 2011) according to the World Reference Base classification. The weathered and fractured bedrock starts on average at about 140-cm depth, with fractures closing at approximately 5-m depth .
The average annual stream discharge is 478 mm Pfister et al., 2017) with lower base flow occurring from July to September due to higher losses through ET (potential ET annual average of 593 mm for the period 2006-2014; Pfister et al., 2017). Snow can accumulate for a few days in winter, but it generally melts within a few days.
The vegetation in the Weierbach catchment is dominated by uneven-age deciduous hardwood trees (70% of the catchment area; European Beech Fagus sylvatica and Oak Quercus petraea x robur) and pure plantations of conifers (30% of the catchment area; European Spruce Picea abies and Douglas fir Pseudotsuga menziesii) Hissler et al., 2021) located in some areas of the catchment ( Figure 1). The deciduous hardwood trees rely on soil water with no significant groundwater uptake . Tree DBH of selected trees was measured within a 360 m Â 20 m inventory plot located in the beech-oak stand (

| Hydrometeorological monitoring and isotopic measurements
Precipitation volumes (P) and air temperature (T) over the study period were measured every 15 min at the Roodt station ( Figure 1) following the World Meteorological Organization standards (Sevruk et al., 2009); we computed the daily total P and daily average T. Gaps in the daily P time series (6% over 2019-2020) were filled using a linear regression between daily P at the Roodt and Holtz (located about 2.6 km from the Roodt station; operated by the Water Agency of

| Xylem sampling
We focused our analysis on hardwood trees that dominated the catchment. We sampled sapwood xylem from beech and oak tree trunks with a Pressler corer across the catchment during the growing seasons 2019 and 2020 ( Figure 1). We transferred the sapwood xylem samples into 30-mL glass vials sealed with caps and Parafilm ® and kept them at À22 C until xylem water extraction.
In each zone, we took the coordinates (X, Y) of randomly selected points. A similar number of uneven-aged beech and oak trees were then randomly selected and sampled within a 15-m radius circle around the points (X, Y). We sampled distinct trees during each campaign. For the spatial analysis, we later generated a unique partialrandom set of coordinates (X r , Y r ) for each sampled tree based on a measured angle and horizontal distance from the point (X, Y). For each campaign, we sampled on average 10 trees for each species (the number ranged between 2 and 18 individuals) ( Table 2). In total, 102 and 101 samples were taken from beech and oak, respectively, in 2019, while 69 samples were taken from both tree species in 2020 (Table 2).
We measured the DBH of each tree sampled. The topographic variables at each of the sampled locations and the DBH of each of the sampled trees spanned the distributions observed in the Weierbach catchment (Table 1 and supporting information Figure S1).

| Xylem water extraction and isotopes analyses
We extracted water from xylem samples using the cryogenic vacuum distillation leak-tight line protocol Orlowski et al., 2016). We submerged the vials containing the xylem samples in a 100 C oil bath and collected evaporated water in U-shaped tubes submerged in liquid nitrogen (À197 C) for approximately 3 h. The lines were connected to a pump that applied a vacuum to reach the suction of 0.03 hPa below which there was no water left to extract.
Extraction was stopped 1 h after the suction reached the constant value of 0.03 hPa. Water was then collected using a Paster pipette, stored in 2-mL threaded vials with fixed 300-μL glass inserts and kept at 4 C before laser spectrometry analysis. lab standards to avoid drift over the course of the analysis. The quality control lab standard water was 0.02‰ for δ 18 O and 0.3‰ for δ 2 H . The isotopic composition is given as the relative difference in the ratio of heavy to light isotopes of water samples (delta notation, ‰) to the Vienna Standard Mean Ocean Water (VSMOW).

| Data processing and statistical analyses
Data and statistical analyses were performed using R Studio Version

Analysis of landscape drivers
We carried out principal component regressions (PCR) (Liu et al., 2003) to reveal the landscape variables influencing xylem SWI using a combination of vegetation and topographic variables. We used species, DBH, elevation, aspect, slope, flow accumulation, curvature, TPI and TWI as independent variables ( p predictors) and the detrended xylem SWI as dependent variable (outcome). First, we recoded the categorical variable species using dummy coding and tested on a test set (30% of the original set) to assess model prediction error. Using the selected model, the raw data matrix X with p predictors columns was replaced by a smaller matrix T with k PCs columns: T ¼ X Ã P Finally, we fitted a multiple linear regression model using the noncorrelated k PCs of T as predictors and the detrended xylem water isotopic composition ŷ as the outcome.
For each k PC, we calculated the percentage of variance in the outcome explained by each predictor as the product of the variance in the outcome explained by the PC and the loading of each predictor in the same PC. The predictors with a loading >j0.45j (Hair et al., 1998) were deemed to contribute largely to a PC. To determine the percentage of variance in the model outcome explained by each predictor, we summed the respective results of each k PCs of the model. We calculated the error in prediction for each sampled tree in space as the difference between the predicted and the measured value. We then tested for normally distributed errors in prediction (Shapiro & Wilk, 1965) and calculated the mean absolute error (MAE) of the model for further evaluation.

| Precipitation and xylem SWI
Over the 2019 sampling period, precipitation median δ 2 H and δ 18 O values were À37.6‰ and À5.9‰, respectively, while median values were À32.7‰ and À4.6‰ over the 2020 sampling period ( Figure 4a,b,e,f). The variability of precipitation SWI was higher over the 2019 than 2020 sampling period due to the 2019 sequential rainfall sampling that captured an isotopically depleted event in October.

| Spatial autocorrelation
In 2019, beech and oak xylem water δ 18 O and δ 2 H showed significant positive spatial autocorrelation, while only δ 18 O of beech xylem water was significantly and positively spatially autocorrelated in 2020 (Table 3). The low Moran's I indicated a weak spatial autocorrelation of these xylem SWI data. The empirical variograms showed a high variance in the data that was as much as the nugget size ( Figure S4); this prevented the fit of any function to the variograms and the estimation of the ranges.

| Landscape drivers
The

| Species-specific temporal variations of xylem SWI
The different temporal variations of xylem SWI (Figure 4 and supporting information Tables S1 and S2) observed between beech and oak (both having similar DBH ranges, Table 1) during both growing seasons suggest different water use strategies. The results suggest that beech exploit shallower and seasonally less stable (due to more exposition to the evaporation-fractionation process) water sources than oak, as observed by Fabiani et al. (2021)  The results further suggest that beech may use more various water sources than oak in response to the varying hydrometeorological conditions observed throughout the growing seasons ( Figure 2). This is in line with the recent observation that beech can change its rooting patterns and water use strategy more easily than oak (Goldsmith et al., 2022). This is also consistent with the finding that beech has the same probability to use deep and shallow soil water, while oak has a higher probability to use deep than shallow soil water (Kahmen et al., 2021).
These species-specific water use strategies support the existence of different water uptake niches between the two co-occurring tree species, as recently suggested by Fabiani et al. (2021) in the study area. These water use strategies also likely demonstrate the higher niche plasticity of beech, as shown in previous research (Goldsmith et al., 2022;Kahmen et al., 2021).

| Dominance of DBH and species as landscape drivers of xylem SWI
The overall higher importance of DBH, and to a lower extent species, in explaining the spatial variability of xylem SWI compared with topographic variables ( Figure 5) further supports that trees use a species-F I G U R E 5 Percentage of variance (%) in xylem water δ 18 O and δ 2 H explained by each measured variable of the optimal model for 2019 (a, b) and 2020 (c, d) ("Other" include the nonmeasured variables). The sign of the explained variance indicates if the variable is positively or negatively correlated with the isotopic value.
specific mixture of water sources from different depths. The notable influence of DBH on xylem SWI is consistent with previous studies that showed that tree diameter was associated with the depth of water uptake, with larger trees using deeper water (Dawson, 1996;Goldsmith et al., 2012;Phillips & Ehleringer, 1995). Different depths of water uptake between larger and smaller trees can lead to differences in xylem SWI values due to vertical variations in soil SWI. These variations can result from shallow soil water mixing with recent precipitation (Bertrand et al., 2014;Sprenger et al., 2018) and evaporation-fractionation of soil SWI (Bertrand et al., 2014).
The smaller influence of species in explaining the spatial variability of xylem SWI is nevertheless in line with earlier research. As discussed above, previous studies reported species-specific vertical root access (e.g. Kahmen et al., 2021) and lateral root elongation and proliferation that could lead to a greater access to soil water pools (Poot & Lambers, 2003) with depth-specific isotopic compositions (Goldsmith et al., 2012). Similar to our finding based on xylem from tree trunks, a species-specific spatial variability of xylem SWI in tree branches has previously been observed (Goldsmith et al., 2018).
The remarkable lower influence of topographic variables on xylem SWI, compared with vegetation variables, is consistent with previous findings of Gaines et al. (2016) who found a relationship between xylem SWI and tree DBH and height but no effect of the slope position on xylem SWI. Similarly, in the same study area, Fabiani et al.
(2021) did not observe significant differences in xylem SWI between hillslope positions. However, these studies were carried out in areas with a small elevation range (about 50 m), and we may expect a higher contribution of topographic variables in explaining the spatial variability of xylem SWI in areas with higher topographic variations. Indeed, topography influences plant water status (Looker et al., 2018) and may, in turn, affect tree water source partitioning and associated xylem SWI in steeper areas, in addition to elevation effects of precipitation SWI.
The variance in δ 2 H explained with the PCR models was much higher than the explained variance with models of δ 18 O ( Figure 5).
The low spatial variability of xylem water demonstrates the ability of the PCR models to reproduce the overall spatial patterns of xylem SWI across the Weierbach catchment well.
Despite our sampling density that was not high enough to reveal the extent to which xylem SWI were spatially autocorrelated, the low Moran's I suggest that there is little spatial structuration of xylem SWI. This appears to be independent of the water availability for trees as we observed low Moran's I for the wetter and the drier growing seasons. Future investigations of the spatial patterns of xylem SWI should however preferably follow a spatially nested sampling design.

| Influence of water availability on the temporal and spatial variations of xylem SWI
The clear lower temporal variability of xylem SWI (Figure 4 and supporting information Tables S1 and S2) observed during the drier (2020) than the wetter (2019) growing season suggests that trees adapted their water uptake depths in response to drier hydrological conditions. This change is supported by the respective presence, although limited, and quasi absence of spatial autocorrelation of xylem SWI in the wetter and drier growing season (Table 3). This adaptation is also in line with the respective increase and decrease of the influence of vegetation and topographic variables on xylem SWI across space observed between the wetter and the drier growing season ( Figure 5).
During the wetter growing season (2019), precipitation was approximately 100 mm higher than during the drier growing season (2020) and the water volume available for tree uptake was higher (average SWC was equal to 0.129 m 3 m À3 in 2019 and 0.114 m 3 m À3 in 2020) and more evenly distributed across the catchment. With these conditions, trees had more easily access to shallow soil waterthat is seasonally more variable- (Goldsmith et al., 2012) and were less water-limited, leading to a higher temporal variability of xylem SWI compared with drier conditions. The higher use of shallow soil water by trees during wetter than drier conditions also led to a higher influence of topography on xylem SWI and to the spatial autocorrelation of these isotopes, although limited. Topographic variables such as slope influence the amount of rainwater infiltrating in soils and, in turn, soil water mixing with rainwater  and in consequence soil SWI (Bertrand et al., 2014;Sprenger et al., 2018). Elevation, slope and aspect can influence air temperature and humidity that affect shallow soil SWI evaporation-fractionation (Bertrand et al., 2014). It has also been observed that the topographic index was spatially autocorrelated over a wide range of spatial extents (Cai & Wang, 2006); this may explain the spatial autocorrelation of xylem SWI observed during wetter conditions.
On the opposite, with drier conditions, trees had to adapt their water uptake strategies (e.g. depth of water uptake, related to DBH; Dawson, 1996;Goldsmith et al., 2012;Phillips & Ehleringer, 1995) and use a higher fraction of deeper and seasonally more stable water sources (Goldsmith et al., 2012). This shift in tree water source led to a lower temporal variability of xylem SWI and a higher influence of vegetation variables on xylem SWI in drier than wetter conditions.
The higher use of deeper soil water by trees during the drier growing season is also in line with the quasi absence of spatial autocorrelation of xylem SWI. These observations are consistent with previous studies demonstrating that trees could shift their water sources from shallow to deep soil water (Brinkmann et al., 2019;Lanning et al., 2020) depending on the water availability to meet water requirements and regulate water status. Brinkmann et al. (2019) showed that beech was particularly able to adapt its water uptake depth depending on the SWC.

| Isotopes-based ecohydrological studies
Our results suggest that, in the Weierbach catchment, tree species and size (DBH) explained more variations in xylem SWI than the topographic variables we evaluated. Similarly, a recent study in the same area showed that species and DBH were the main drivers of the spatial variability of sap velocity .
Robust descriptions of the spatial variability in xylem SWI in ecohydrological studies are critical to provide reliable isotope-based estimates of water sources for vegetation root water uptake (Beyer & Penna, 2021

| Isotopes-based catchment transit times studies
Currently, the isotopic signals of evaporation and transpiration as well as the age composition (i.e. transit time distributions) from state-ofthe-art model applications (e.g. Hrachowitz et al., 2015;Rinaldo et al., 2015;Rodriguez et al., 2018Rodriguez et al., , 2020Soulsby et al., 2016;van der Velde et al., 2015;Wang et al., 2023) are more often only indirectly constrained by calibrating models to observed isotopic signals of stream flow. This implies that the relationship between the isotopic composition of the water in storage and the water that is evaporated and/or transpired from this storage remains unclear . Recent development in the calibration of such models involved the use of soil and/or xylem SWI from a limited number of tree individuals, species or locations (e.g. Asadollahi et al., 2022;Knighton et al., 2019;Kuppel et al., 2018;Sprenger et al., 2022).
Our results suggest that, to better reflect the vegetation behaviour compared with these studies and to account for the importance of species in influencing xylem SWI across space, the spatially distributed xylem SWI data from our study can be used to determine species-specific isotopic signals associated with transpiration. These signals can be exploited in lumped model approaches (e.g. using a weighted mean isotopic signal of the species-specific signals) to further constrain lumped hydrological models developed for the Weierbach catchment and calibrated so far exclusively using discharge and stream SWI data (Rodriguez & Klaus, 2019). Species-specific isotopic signals associated with transpiration can also be exploited in semidistributed models (e.g. separating beech/oak, douglas and spruce zones) (e.g. Kuppel et al., 2018).
For these purposes, a good understanding of the forest composition and structure (species, DBH) in the Weierbach catchment is required to design an optimal sampling strategy that provide a representative isotopic signature. Further work is needed in other catchments to improve the quality of the xylem SWI data used in hydrological models.

| CONCLUSION
In this study, the measurement of beech and oak xylem SWI during several campaigns in the Weierbach catchment revealed a higher variability over the growing season of the beech xylem water isotopic signature compared with oak. Using spatially distributed xylem SWI measurements and PCR, we identified DBH and species as the dominant variables influencing the spatial variations of xylem SWI; topographic variables had a minor role in explaining these variations. We also noted a minor presence of spatial autocorrelation between xylem samples, but our sampling density was not high enough to reveal its extent. By way of sampling xylem over two growing seasons, we observed a respective increase and decrease of the influence of vegetation and topographic variables in explaining the spatial variations of xylem SWI between the wetter and the drier seasons.
Our results suggest that, in the study area, the spatial variations of xylem SWI arise mainly from size-and species-specific as well as water availability-dependent water use strategies rather than from topographic heterogeneity. Trees can also adapt their water use strategies in response to lower water availability.
Overall, our findings highlight the importance of vegetation variables in influencing xylem SWI. We demonstrate the importance of accounting for different tree diameters and species in field sampling protocols to accurately capture the isotopic variability existing within a study area. However, there is still a need for evaluating the role of additional vegetation and forest structure variables to refine our understanding of vegetation influences on xylem SWI. This information will improve the accuracy of the lumped isotopic signal associated with transpiration used in hydrological models and will help to better predict how catchments will respond to future changes in land cover, vegetation or stand properties associated with global change. for their input that helped to improve this work and manuscript. We thank the Editor and two anonymous reviewers for their comments that helped to improve this manuscript.

DATA AVAILABILITY STATEMENT
The data used in this study are the property of the Luxembourg Institute of Science and Technology (LIST) and can be obtained upon request to the corresponding author, after approval by LIST.