Transpiration patterns and water use strategies of beech and oak trees along a hillslope

The role of landscape topography in mediating subsurface water availability and ultimately tree transpiration is still poorly understood. To assess how hillslope position affects tree water use, we coupled sap velocity with xylem isotope measurements in a temperate beech‐oak forest along a hillslope transect in Luxembourg. We generally observed greater sap velocities at the upslope locations in trees from average‐sized trees, suggesting the presence of more suited growing conditions. We found a lower difference in sap velocity among hillslope positions for larger trees, likely due to the exploitation of deeper and more persistent water sources and the larger canopy light interception. Beech trees exploited a shallower and seasonally less persistent water source than oak trees, due to the shallower root system than oak trees. The different water exploitation strategy could also explain the stronger stomatal sensitivity of beech to vapour pressure deficit compared to oak trees. Xylem isotopic composition was seasonally variable at all locations, mainly reflecting the contribution of variable soil water sources and suggesting that groundwater did not contribute, or only marginally contributed, to tree transpiration. Overall, our results suggest that trees along the hillslope mainly rely on water stored in the unsaturated zone and that seasonally shallow groundwater table may not necessarily subsidize water uptake for species that do not tolerate anoxic conditions. Contrary to previous studies, at our site, we did not find higher sap velocity downslope as the subsurface hillslope structure promotes vertical water flux over lateral redistribution in the vadose zone.


| INTRODUCTION
Water availability in space and time is one of the key elements shaping forest ecosystems and their adaptive response to environmental stress. Landscape position is a dominant factor that controls the spatio-temporal variability of water available for tree transpiration (Looker et al., 2018). Several studies observed that a variation in water availability along a topographic gradient can result in different tree species distribution, growth rate, and transpiration fluxes (Elliott et al., 2015;Fan, 2015;Hawthorne & Miniat, 2018;Tromp-van Meerveld & McDonnell, 2006). Topography critically controls plant transpiration via the influence of slope and aspect on the amount of incoming solar radiation (Renner et al., 2016) and on water availability through hillslope hydrological processes (Fan et al., 2017). The age distribution and composition of natural forests are typically adapted to the varying water availability in different parts of the ecosystem in order to meet species-specific growth requirements (Band et al., 1993;Lin et al., 2019). Trees covering the hilltop and ridgeline, which are often characterized by thin soils, are frequently less productive in biomass Kume et al., 2016) but can adapt in certain cases to utilize water from weathered bedrock and rock moisture Klos et al., 2018;Rempe & Dietrich, 2018). On the contrary, trees located in topographically convergent locations may take advantage of shallower groundwater tables (Barbeta & Peñuelas, 2017;Brooks et al., 2015;Eamus et al., 2006;Pettit & Froend, 2018). Shallow groundwater replenishes soil water via upward capillary fluxes or temporary groundwater rise, modulating the water content in the unsaturated zone (Brooks et al., 2015). Groundwater can also directly sustain plant transpiration when the water table is within the root zone (Brooks et al., 2015;David et al., 2013;Miller et al., 2010). Trees at footslope locations can also benefit from water that is laterally redistributed from upslope areas (Band et al., 1993;Hwang et al., 2020;Lin et al., 2019). Nevertheless, some studies have observed high drought sensitivity, a small basal increment, and a reduced leaf area index in trees growing at downslope locations (Elliott et al., 2015;Hwang et al., 2020). This behaviour has been associated with the combined effect of higher water consumption by upslope vegetation compared to downslope vegetation and consequent lower downslope water subsidy available for trees in convergent areas (Hawthorne & Miniat, 2018;Hwang et al., 2020). The lack of consistent findings on how hillslope position controls forest water use may originate from the interplay of the sitespecific structure of the Critical Zone, which is the near-ce environment extending from the tree canopy through the soil up to the weathered bedrock (Brooks et al., 2015;Rempe & Dietrich, 2018).
Trees can utilize different below-ground water sources to sustain transpiration (Barbeta & Peñuelas, 2017). The accessibility of different water sources by trees is not only determined by water redistribution, subsurface structure, and the degree of subsurface heterogeneity, but also by root distribution and activity (Fan et al., 2017). Along a topographic gradient, rooting depth and biomass may vary depending on moisture availability in the subsurface, in order to meet the nutrient and water needs of trees (Tsuruta et al., 2020). Plants have developed different below-ground strategies to access nutrients and water and to respond to changes in their spatio-temporal availability (Bardgett et al., 2014;Fry et al., 2018). Tree water consumption is related to several species-specific features including architectural root traits (i.e., rooting depth and root length density) (Fry et al., 2018), xylem architecture (diffuse-vs. ring-porous species) (Wang et al., 1992) and stomatal regulation (isohydric vs. anisohydric species) (McDowell et al., 2008a;Uhl et al., 2013). Since roots differ in their functional and physical roles (Fry et al., 2018), the mere presence of roots at a given depth is not a reliable indicator of their contribution to water use (Ehleringer & Dawson, 1992). Additionally, it has been shown that the correspondence between physically present and functionally active roots is species-and time-dependent (Volkmann et al., 2016).
Until today, a range of ecohydrological studies successfully employed stable isotopes (i.e., hydrogen ( 1 H/ 2 H) and oxygen ( 16 O/ 18 O)) in the water molecule in the soil-plant continuum to investigate tree water use (Penna et al., 2018). Xylem water isotopic composition is presumed to reflect the integrated isotopic composition of water sources accessed by the tree (Dawson et al., 2002). It was shown that trees can adapt their water source from shallow to deep soil water following water availability (Brinkmann et al., 2018;Lanning et al., 2020).
This species-specific plasticity might be key for their survival and competitiveness under increasing water scarcity (Volkmann et al., 2016).
The spatial and temporal water source partitioning between different species is still poorly tested and contrasting results have been found (Allen et al., 2019;Bello et al., 2019;Grossiord et al., 2014;Meißner et al., 2013;Volkmann et al., 2016). While Meinzer et al. (2001) highlighted the potential of niche complementarity as a competition avoidance strategy, Grossiord (2019) and Gillerot et al. (2020) suggested that tree diversity does not systematically increase the performance of forest communities. Indeed, it was only in drought-prone environments that forest resistance to drought was enhanced by higher diversity (Grossiord et al., 2014). Despite the body of previous work, we lack studies addressing plant water use along hillslope transects. Specifically, combined assessments of transpiration rates and water uptake depths along hillslope transects might help to better understand water use strategies and plasticity of different tree species. Such a combined assessment supports a more detailed comprehension of the seasonal interactions and feedbacks between vegetation and hydrological processes occurring at the hillslope scale.
To address this gap, we carried out an ecohydrological study along a hillslope transect populated by oak (Quercus petraea (Matt.) Lieb. x robur) and European beech trees (Fagus sylvatica L.) in Luxembourg. Beech and oak trees are coexisting species in Central Europe (Barbaroux & Bréda, 2002;Grossiord et al., 2014); however, they may become competitors during drought conditions, resulting in negative consequences on forest vitality and composition (Petritan et al., 2017). Although taxonomically related, the morphological and ecological differences between the two species can result in temporally and spatially different patterns of water utilization (Zapater et al., 2011). We monitored soil moisture, groundwater level, sap velocity and hydro-meteorological variables for one growing season (April-October 2019). Additionally, we determined the isotopic composition of precipitation, soil water, groundwater and xylem water in order to assess the influence of topography on water use. The combination of these measurements allows for discrimination between water sources used by tree species for assessing the physiological response of trees to water availability along a hillslope.
Our study was driven by the general conjecture that landscape position controls tree water use through subsurface water redistribution, including spatially-variable groundwater table depth along the hillslope. Furthermore, we conjectured that the characteristics of a species result in a different response between species at the same hillslope position. In particular, we addressed the following specific research questions: (i) How does landscape position affect the spatial and temporal patterns of water use? (ii) How do two co-occurring species (beech and oak) characterized by different physiological and hydraulic traits respond to a variable water supply over the growing season? The forest stand on the selected hillslope transect was characterized in a survey of 18 consecutive 20 Â 20-m plots along the hillslope transect ( Figure 1) recording the number of trees, species, and diameter at breast height (DBH). The basal area for each tree was calculated from the DBH. Based on the digital elevation model (DEM), we derived the topographic position index (TPI, Hoylman et al., 2018;Weiss, 2001) to classify landscape position with respect to a defined neighbourhood. The TPI was computed as follows: where z is the elevation for the ith pixel and z i_100 is the average elevation in a 100 m radius around the ith pixel. We defined the plateau area for TPI > 0.5, the midslope for 0.5 < TPI > À0.5, and the footslope for TPI < À0.5. The upper five plots fell into the category plateau, the subsequent nine plots into the midslope, and the lowest four plots were associated with the footslope. A fully stocked mixed forest with European beech trees (78% of the forest stand, 60% in basal area) and pedunculate and sessile oak hybrid trees (22% of the forest stand, 40% in basal area) populates the hillslope transect. The two oak species often form hybrids, which are phenologically difficult to differentiate. Hence, we did not differentiate between the species and refer F I G U R E 1 Site overview with the forest inventory plots (red squares). At each sampling area, a groundwater well (blue dot) was installed and two trees for each species (beech, oak) were equipped with sap flow sensors (green star for beech, yellow star for oak) with two different diameter classes to them as oak. Shrubs are absent and the understory mainly consists of blueberries (Vaccinium myrtillus). Trees were divided into three diameter classes (Table 1). Given the relatively small diameter differences between individual trees in the plateau area (the upper part of the hillslope, see Figure 1), we define the forest as even-aged. The forest becomes more heterogeneous downslope (Table 1). The plateau area is characterized by a lower basal area and lower forest density than the midslope and footslope ( Figure 2). The forest structure in the Weierbach catchment is the result of past and current management practices. Oak trees are evenly distributed across the whole area, while beech tree allometry strongly varies between the three areas with increasing density from the plateau to the footslope ( Figure S1). For each landscape position (plateau, midslope, footslope), we established one sampling area ( Figure 1).
Soils are shallow (<1 m) and the lithology consists mainly of a Pleistocene periglacial cover bed overlying Devonian slate from the Ardennes massif (Juilleret et al., 2011). The analysis of eight profiles showed that soil characteristics (e.g., structure, porosity, bulk density, particle density and texture) were similar across the catchment (Glaser et al., 2016). On the plateau, the subsolum is characterized by Regolithic Saprolite with gleyic properties, while the hillslope is characterized by a Regolithic Saprock substratum with dense vertical cracks . The solum is a stony loam soil with a mean thickness of 50 cm and an average porosity of 30%. In the subsolum, the size of schist/slate fragments strongly increases while the drainage porosity decreases. On average, the slate bedrock starts at a depth of 140 cm and is highly weathered  and permeable (Bonanno et al., 2021;Scaini et al., 2018). The subsurface structure leads to a dominance of vertical hydrological fluxes, while lateral flow occurs in the deeper hillslope (Glaser et al., 2019(Glaser et al., , 2016 and contributes to streamflow (Rodriguez et al., 2021;Rodriguez & Klaus, 2019).
Temperature (T), relative air humidity (RH), solar radiation and precipitation (15-min logging intervals) data were available from the weather station. Using T and RH, we calculated the daily mean vapour pressure deficit (VPD) using Equation 2: Volumetric soil moisture was measured at the plateau and at a footslope at a different hillslope close to the catchment outlet with water content reflectometers (CS650, Campbell Scientific, UK). Each profile consisted of four probes installed horizontally at 10, 20, 40, and 60 cm depth recording at 30-min intervals (Figure 1). At each location, one groundwater well was installed (10, 9, 3.5 m deep, and screened for 5, 4, 2.5 m from the bottom at the plateau, midslope and footslope, respectively) and was equipped with water pressure transducers (Orpheus Mini, OTT, Germany) recording data at 15 min logging intervals.

| Measurement of sap velocity
At each of the three sampling areas (Figure 1), we selected one tree from the most frequent diameter class (25-50 cm, referred to as average diameter class) and one tree from the diameter class 50-75 cm (referred to as large diameter class) of both species for sap velocity measurements ( Table 1). The selected trees were equipped with heatpulse sap flow sensors (SFM1, ICT International Pty Ltd., Australia) ( Table 2). We positioned the sap flow sensors at the north-east side of the trunk, 1.3 m above the ground and shielded them from direct sun exposure. The sensors consist of a central heating needle and two needles, each with two thermistors (located at 12.5 and 27.5 mm from the bark on a 35-mm-long needle) recording the temperature upstream and downstream of the heater. Needles were installed one above the other with a vertical distance of 0.5 cm. The heat pulse F I G U R E 2 (a) Daily total precipitation amount (mm/d) and daily mean air temperature ( C) and (b) daily mean net radiation (W/m2) and daily mean vapour pressure deficit (VPD) (kPa) observed at the Roodt weather station velocity (V h in cm h À1 ) was calculated with Equation 3 (Burgess et al., 2001) at the inner and the outer thermistor: where k is the thermal diffusivity (cm 2 s À1 ) set to 0.0025 (Marshall, 1958), x is the distance between the heater and either tem- where V h is the calculated heat pulse velocity (cm h À1 ), B is the wound correction factor set to 0.13 cm (Marshall, 1958), ρ b the basic density of wood set to 0.5 g cm À3 (Burgess & Downey, 2014), c w the specific heat capacity of the wood matrix (1200 J kh À1 C À1 ; Becker & Edwards, 1999), Cs the specific heat capacity of sap (water, 4182 J kh À1 C À1 ; Lide, 1992), ρ s the density of sap water (1 g cm À3 ), and m c the water content of sapwood (set to 0.5 g cm À3 ; Burgess & Downey, 2014).
Following Renner et al. (2016), sap velocities measured by the inner and outer thermistor were averaged to obtain the daily mean sap velocity, which is used as a proxy of tree transpiration (Smith & Allen, 1996). For our analyses, we were interested in the response of sap velocity to environmental conditions (i.e., soil moisture and VPD) as an indicator for stomatal control, therefore we scaled daily mean sap velocities between 0 and 1 for further analysis, where 0 and 1 were the minimum and maximum daily mean velocities recorded by each tree over the entire growing season.

| Wood core, soil and water sampling for isotopic analysis
We carried out 14 bi-weekly sampling campaigns over one entire growing season from 8 April 2019 (before leaf flush) until 21 October 2019 (when leaves were turning yellow). For each campaign, we randomly selected two beech and two oak trees from the dominant diameter class (25-to 50-cm DBH) at each of the sampling areas (Table 1).
The maximum distance between trees equipped with sap flow sensors and trees that were sampled for xylem water was 25 m. We collected trunk cores encompassing only the sapwood with a Pressler borer, removed the bark, and transferred the remaining xylem wood into 30-ml glass vials sealed with caps and Parafilm ® . Samples were stored in a freezer (À22 C) until water extraction (see Section 2.4). During five sampling campaigns (8 April, 23 April, 3 June, 17 June and 1 July), we also sampled soil cores to assess the soil water isotopic composition. At each sampling area, we extracted three soil cores from the top 60 cm divided into five depth classes (0-5, 5-10, 10-20, 20-40 and 40-60 cm) with a soil auger. A 60-cm depth was the lower limit of sampling due to the interface between the soil and rock clasts of the basal layer (Juilleret et al., 2011). Each soil sample was stored in zip bags with as little air as possible until analysis (see Section 2.4).
We sampled groundwater at the sampling areas and rainfall bi-weekly with a rainfall collector (Palmex Ltd.) placed in a clearing ( Figure 1).

| Water extraction from plant and soil material and isotopic analyses
We extracted xylem water from wood cores through a cryogenic vacuum distillation line (Orlowski et al., 2016) by submerging the sample in a 100 C oil bath for 3 h under a vacuum of 0.03 hPa. Evaporated water was collected in U-shaped tubes, which were submerged in liquid nitrogen (-197 C). The pressure was continuously recorded to assure that the lines remained leak-tight throughout the entire We analysed soil samples for their water stable isotopic composition with direct liquid-water-vapour equilibration (Wassenaar et al., 2008). We prepared the sampling bags with a blot of silicone on the outside to later serve as a septum. After each sampling campaign, sampling bags were heat sealed upon arrival at the lab, shown for the IAEA precipitation isotope station in Trier (Klaus et al., 2015;Stumpp et al., 2014), which is approximately 64 km from the study site. We calculated the lc-excess for xylem water following Landwehr and Coplen (2004): where a and b are the slope and intercept of the LMWL, respectively.

| Sap velocity: Species, site and size-specific differences
The daily mean sap velocity exhibited species, site, and size-specific variations. Due to instrument failure, not all sensors supplied a complete record over the growing season (Table 2). Beech trees showed earlier leaf emergence (approx. 18-26 April) than oak trees (approx. 1-7 May), consistent with increasing sap velocity during these periods ( Figure 4). We observed a strong seasonal pattern in sap velocity at the three sampling areas for both monitored diameter classes ( Figure 4). Sap velocity increased in spring, reaching the maximum at the end of June when atmospheric water demand and soil moisture were high. Concurrently with decreasing soil moisture and high evaporative demand, sap velocity decreased gradually and showed minimum values in mid-August. Short-term increases of sap velocity occurred during the summer following precipitation events (e.g., middle of August, cf. Figure 2a). From the end of August, we observed a progressive decline in sap velocity, as soil moisture remained low and VPD decreased (Figures 2 and 3).
The daily mean sap velocities of beech trees from the average diameter class (25-50 cm DBH) were consistently higher at the midslope than at the other locations (p < 0.05, Pairwise Wilcox test) ( Figure 4a). Beech trees at the footslope location experienced a delayed increase in sap velocity but reached the same velocities as plateau trees in summer. Sap velocity for beech trees from the large diameter class (50-to 75-cm DBH) (Figure 4c) (Figures 4b and 4d). For the average diameter class, sap velocity was statistically similar for oak trees at the footslope and midslope, while for oak trees of the large diameter class, the plateau tree displayed the highest, followed by the one midslope and footslope (Figure 4d). For oak trees, sap velocity was approximately 2.5 times higher for the large diameter class than for trees from the average diameter class. We did not find a statistical

| Relationship between sap velocity, vapour pressure deficit and soil moisture
The relationship between normalized sap velocities from trees from the average diameter class and daily average VPD and the relationship between normalized sap velocity and daily average soil moisture across the four measured depths (from the plateau cluster) showed anticlockwise hysteresis for both species (Figures 5 and 6). However, the shape of the loop was different for the beech and oak trees

| Isotopic composition of water, soil and xylem samples
The isotopic composition of the bi-weekly precipitation samples col-  (Figure 7), soil isotopic values plotted along the LMWL.
Xylem water from both species plotted below the LMWL in the dual-isotope space (Figure 7). This was a consistent occurrence across all sampling campaigns and sampling areas (Figure 7). In particular, the xylem samples from beech trees plotted farther away from the LMWL than those from oak trees (Figure 7). The isotopic composition of xylem water was highly variable throughout the growing season ( Figure 8). Xylem water from the first sampling campaign (8 April 2019) before the leaf flush was significantly higher in δ-values for both isotopes (Wilcoxon rank sum test, p < 0.05) compared to later dates ( Figure 8) and was situated in the upper right area of the dualisotope plot (Figures 7 and S3). Concurrently with leaf flush, the δ-values became markedly lighter (Figure 8a,b) and xylem water fell in the lower left part of the dual-isotope space (Figures 7 and S3). Xylem water from both species became progressively heavier in both isotopes over the growing season, although two sampling campaigns that were carried out one day after rainfall led to a deviation from this general trend (cf. late July and September sampling, Figure 8). For both isotopes, the interquartile range and variability over the sampling season were higher for beech trees than for oak trees (Figures 8 and 9).
Xylem water from beech trees displayed a more pronounced variability at the sampling date and seasonal scale than xylem water from oak in oak trees (p < 0.05) but lighter in δ 2 H (p > 0.05). We found a statistical difference in xylem lc-excess between the two species (Figure 9c) (Wilcoxon rank sum test, p > 0.05) over the whole study period. We did not observe significant differences (p > 0.05, Wilcoxon rank sum test) in beech tree xylem δ 18 O, δ 2 H, and lc-excess along the hillslope when xylem data were grouped together ( Figure 9). Additionally, no topographic effect was evident in lc-excess over time ( Figure 10).
However, from June to September, beech trees at the plateau location exhibited more negative lc-excess than the other locations. For oak trees, we did not observe significant differences in xylem δ 18 O and δ 2 H (p > 0.05, Wilcoxon rank sum test) between the three locations but lc-excess was significantly different between the plateau and the footslope (p < 0.05, Wilcoxon rank sum test) (Figure 9).

| DISCUSSION
4.1 | Spatial and temporal pattern in tree water use along the hillslope We generally observed higher sap velocities in trees from the average diameter class growing at plateau and midslope locations compared to those at the footslope suggesting that growing conditions at upslope areas were more suitable. Here, trees may benefit from a lower degree of competition for water and greater access to light due to the lower forest density compared to the footslope location (Pretzsch & Forrester, 2017) ( Figure S1). We observed minor differences in sap velocity between locations for large diameter trees, because they likely have access to a deeper and more stable water reservoir (cf. Gaines et al., 2016;Goldsmith et al., 2012) and occupy dominant positions which ensure greater light interception than the average trees in the stand. From the beginning of the growing season until late June when leaves were completely unfolded, sap velocities increased along the hillslope due to the increasing VPD and sufficient soil water supply (Figures 3 and 4). From June onwards, xylem water isotopes progressively became heavier ( Figure 8) and the lc-excess became more negative indicating that trees relied on water sources that had been increasingly affected by isotopic enrichment, such as shallow soil water. From July onwards, soil moisture decreased, the groundwater table at plateau and midslope locations receded, and sap velocities decreased. The response of sap velocity to environmental controls (VPD and soil moisture) was species-specific but similar in all locations ( Figures 5 and 6) indicating that the sampled trees mainly absorbed water from similar sources, irrespective of hillslope position. However, at the footslope, sap velocities were equal to or lower than at the midslope and plateau locations ( Figure 4) despite a seasonally high and stable groundwater table (Figure 3d). This may indicate that increased groundwater accessibility did not foster higher transpiration rates.
Xylem isotopic composition varied seasonally but differed between the sampling areas in only a few sampling campaigns ( Figure 7). This result suggests that at our study site trees might not directly rely on groundwater or used it to such a limited extent that no clear differentiation in isotopic composition between trees vegetating in higher positions in the landscape was detectable. In studies where trees were found to exploit groundwater, xylem water isotopic composition hardly varied over time because groundwater isotopic composition was rather stable compared to rain and soil water (Carrière et al., 2020;David et al., 2013). The observed high and seasonally stable groundwater table at the footslope may even restrict root development and confine roots to shallow soil and saprolite layers, since saturated environments can limit root expansion and plant productivity for species that do not tolerate permanent hypoxic or anoxic conditions (Fan et al., 2017;Hasenmueller et al., 2017;Rossatto et al., 2014;Roy et al., 2000) such as beech and oak trees F I G U R E 7 Dual-isotope (δ 2 H and δ 18 O) plot of xylem water from the two species studied and their potential sources (soil water at five depth, groundwater) for every sampling campaign conducted in 2019. Soil was sampled on 8 and 23 April, 3 and 17 June and 1 July. The black line indicates the LMWL (δ 2 H = 7.4 δ 18 O + 6.5) (Schmull & Thomas, 2000). Additionally, the lack of daily groundwater table fluctuations, even when sap velocity was high (Figure 3d), may indicate that trees did not rely, or only marginally relied, on groundwater (cf. Naumburg et al., 2005). Diurnal water table fluctuations have been mostly observed in riparian areas and used to directly assess groundwater consumption by plants (Martinetti et al., 2021;Moro et al., 2004;Soylu et al., 2012), but these studies involved wetland species that cope with anoxic conditions in the root zone. However, at our study site the potential tree water uptake from groundwater at the footslope might have been balanced by groundwater inflow from upslope areas (Rinderer et al., 2017). The progressively declining water table at upslope locations over the growing season may have left behind an aerated soil profile at field capacity, which became available for deeper root exploitation (Naumburg et al., 2005). Trees F I G U R E 8 (a) δ 18 O, (b) δ 2 H and (c) lc-excess of xylem water for each sampling campaign across the growing season. The data from each species were averaged over the three sampling areas. The centre line in the boxplot indicates the median, and the lower and upper extremes indicate the first and third quartile, respectively. The whiskers indicate points within 1.5 times the interquartile range above or below the median located in the upper portion of hillslopes are usually found to develop a dimorphic root system that allows them to access the weathered rock moisture during dry periods (Fan et al., 2017). Hahm et al. (2020) were able to demonstrate that the receding groundwater table progressively offers a large water reservoir held in the weathered bedrock, which is accessed by deeper roots, and that rock moisture can offer a significant contribution to transpiration in a variety of lithological settings (Rempe & Dietrich, 2018). However, the isotopic characterization of this potential water source is often neglected in isotope studies due to difficulties in accessing and measuring below-ground compartments (Rempe & Dietrich, 2018).
Xylem water in our study displayed a hydrogen isotope ratio more depleted than any of the water sources considered, as shown in other studies (Barbeta et al., 2019;Oerter et al., 2019;Oerter & Bowen, 2017). Some possible explanations for this offset have been proposed in the literature, including isotopic separation between bound and mobile soil water (Brooks et al., 2010), water compartmentalization between flowing and stored water in the stem (von Freyberg et al., 2020;Zhao et al., 2016), and isotopic fractionation at the soil-root interface (Barbeta et al., 2020;Poca et al., 2019). Other studies have provided some indications that the isotopic offset could arise from δ 18 O fractionation processes (Marshall et al., 2020;Vargas et al., 2017), challenging the argument of negligible 18 O fractionation during root water uptake (Rothfuss & Javaux, 2017). We do not have evidence that goes beyond the behaviours recently discussed (Beyer & Penna, 2021;von Freyberg et al., 2020), but we cannot exclude that unsampled saprolite and weathered bedrock waters could act as additional sources potentially explaining this offset.
Furthermore, the lack of higher sap velocity at the footslope location, which is contrary to observations in several other studies due to the lateral redistribution of soil moisture (Hawthorne & Miniat, 2018), deeper soil, and higher water holding capacity (Kumagai et al., 2008;Mitchell et al., 2012;Tromp-van Meerveld & McDonnell, 2006) provides some insight into the hydrological functioning of the Critical Zone at the Weierbach. Due to the high hydraulic conductivities in the Weierbach catchment (Glaser et al., 2016) and the lack of shallow impeding layers, the hillslope structure does not promote the lateral downslope redistribution of soil water via interflow (Klaus & Jackson, 2018). This results in reduced subsidies to soil moisture at footslope locations, contrary to what was observed in other studies (Hawthorne & Miniat, 2018;Lin et al., 2019). Due to the highly fractured bedrock (Juilleret et al., 2011;Martínez-Carreras et al., 2016), upslope patches are hydrologically connected only through the saturated zone with downslope areas allowing groundwater at the footslope location to maintain a constant and shallow water table over the investigated period (Rinderer et al., 2017). This behaviour shows the critical importance of landscape characteristics determining water redistribution, water availability for vegetation, and ultimately growing conditions.

| Species-specific response to variable water supply
Our study showed that beech and oak trees clearly have different water uses, representing a resource-driven niche partitioning. The F I G U R E 9 (a) δ 18 O, (b) δ 2 H and (c) lc-excess of xylem water for each sampling campaign at the three sampling plots for all sampling campaigns conducted in 2019. The centre line in the boxplot indicates the median, and the lower and upper extremes indicate the first and third quartile, respectively. The whiskers indicate points within 1.5 times the interquartile range above or below the median systematically more negative and seasonally more variable xylem lcexcess ( Figure 8c) of beech trees compared to oak trees suggests the use of a water source more exposed to isotopic enrichment, likely near-surface soil water. Such a difference can be associated with species-specific root architecture. This explanation is consistent with in situ root measurements at other sites (Coners & Leuschner, 2005;Leuschner et al., 2001). For instance, in a beechoak forest stand, Leuschner et al. (2001) found that beech was a superior below-ground competitor of the topsoil where the maximum fine root densities occurred to colonize the nutrient-rich organic layer. In contrast, oak was considered a deep-rooted species able to access deeper subsurface water (Lanning et al., 2020).
Through a metanalysis across several sites, Fan et al. (2017) found that the average rooting depth of the genus Quercus was 5.23 and only 0.8 m for Fagus.
The different timing of leaf emergence observed for the two species, as also shown by sap velocities (Figure 4), indicates that water use partitioning does not only occur spatially in the subsurface driven by root distribution, but also temporally (Meinzer et al., 2001). This phenological variation has been linked to wood hydraulic conductivity (Wang et al., 1992). Over the dormant season, ring-porous species like oak with large vessels may experience a higher loss in hydraulic conductivity than diffuse-porous species like beech (Cruiziat et al., 2002).
To overcome winter embolism and restore the water flow pathway, oak trees invest in the formation of early wood before leaf expansion (Bréda & Granier, 1996), while stem growth in beech trees starts after leaf flush (Barbaroux & Bréda, 2002). At the time when leaves flush out, the xylem water in our study displayed an extremely heavy isotopic composition. This is consistent with previous studies and results from evaporation through the bark during periods of limited sap flow Phillips & Ehleringer, 1995).
Accessibility to different water sources could also explain the different hysteretic response of sap velocities to the environmental drivers (VPD and soil moisture). The increasing VPD and high soil moisture content in the first part of the growing season supported an increase in sap velocity in both species (Figures 5 and 6). However, the high VPD values and reduced water supply from late July onwards induced stomatal closure in beech trees in order to minimize water loss. Conversely, oak trees showed a lower sensitivity to these environmental forces with sap velocity reaching its maximum when VPD approached 3 kPa, despite reduced soil moisture content. The different degree of stomatal sensitivity to the environmental forcing coupled with different xylem isotopic composition suggests the use of a deeper and more stable water source for oak trees and a shallower and ephemeral water source in beech trees, supporting the findings of Bakker et al. (2008) and Grossiord et al. (2017).
In order to meet water requirements and regulate water status, trees have developed different adaptations, like stomatal response (isohydric vs anisohydric species), xylem architecture (diffusive vs. ring-porous), and rooting depth (Matheny et al., 2017;McDowell et al., 2008b). The interplay of these characteristics defines tree transpiration response to a variable water supply. The deep rooting strategy of oak trees enables them to overcome the hydraulic failure risk given their anisohydric stomatal regulation (Matheny et al., 2017) and ring-porous xylem (Cruiziat et al., 2002) by exploiting deeper water sources. Anisohydric species have a weaker stomatal sensitivity than isohydric species such as beech trees, which reduce water consumption via stomata closure to avoid water stress damage (Magh et al., 2020).
F I G U R E 1 0 lc-excess of xylem water from (a) beech and (b) oak trees at the three sampling areas over the growing season

| CONCLUSIONS
In this work, we examined the role of landscape topography (hillslope position) on the spatial and temporal patterns of water use in a mixed forest of beech and oak trees through sap velocity and stable water isotope measurements. We showed different patterns of sap velocity in different hillslope positions, with trees generally displaying higher sap velocity in upper locations than in downslope areas where the groundwater Thus, we reject our first conjecture that hillslope position controls tree water use though subsurface water redistribution. In our case, high and stable groundwater table at the footslope location might even reduce root expansion in species that do not tolerate saturated environments.
Furthermore, our results confirm that beech and oak trees have different ecohydrological niches driven by their species-specific water exploitation strategies and hydraulic traits, which are crucial to determining a tree ability to recover from water shortage periods. Beech trees, although more drought-sensitive than oak trees, are a superior above-ground competitor. This characteristic, combined with a marked shade tolerance, allowed them to regenerate extensively at our study site, while oak seedlings and saplings were absent. This confirms our second conjecture that species-specific characteristics result in a different response in different species at the same hillslope position.
Overall, our findings highlight that the link between forest community and the Critical Zone structure is highly dynamic due to species-specific interaction with water availability and subsurface flow patterns. Future management practices should operate to create optimal conditions for forest resilience, accounting for subsurface structure to promote more drought-tolerant species in order to ensure ecosystem functioning in face of future climate change.