Drought responses and their effects on radial stem growth of two co-occurring conifer species in the Mediterranean mountain range

Patterns of stem radial variations showed thatCedrus libaniA. Rich. was less limited by summer drought than co-occurringJuniperus excelsaM. Bieb.Cedrus libanirecovered faster from tree water deficit and showed significantly higher radial growth rates and annual stem increments thanJ. excelsa. However, the ability ofJ. excelsato grow more hours per year may indicate a potential benefit in more extreme conditions. Knowledge about species-specific drought responses is needed to manage productive forests in drought prone areas. Under water shortage, trees commonly show stem shrinkage, which is assumed to inhibit growth. We investigated whether the two co-existing conifers Juniperus excelsa M. Bieb. and Cedrus libani A. Rich. (growing at the Taurus Mountains, SW-Turkey) show differences in water relations and stem growth in order to evaluate their respective drought tolerance. Stem radius changes were hourly monitored over 2 years using high-resolution point dendrometers. Radial stem growth, tree water deficit-induced stem shrinkage, and maximum daily shrinkage were extracted from stem radius change measurements, investigated for their patterns, and related to environmental conditions. Cedrus libani recovered from tree water deficit under higher temperature and vapor pressure deficit than J. excelsa. The number of hours during which stem growth occurred was higher for J. excelsa; however, growth rates and annual increments were significantly lower than in C. libani. Both species showed highest maximum daily shrinkage during the driest months indicating the ability to maintain gas exchange all year round. Juniperus excelsa showed a more conservative growth strategy while C. libani was less limited by summer drought and showed more annual stem increment under the conditions investigated.


Introduction
Future climate scenarios predict that the Mediterranean region will be among the more heavily affected parts of the world in terms of increased frequency and intensity of drought events (Ozturk et al. 2015). In the early past, the Mediterranean region has already experienced an increase in mean annual temperature and severity of summer droughts (Abbasnia and Toros 2019) which affect the vitality and productivity of forests (Tognetti et al. 2019). Studying the growth behavior and tree response to environmental conditions might help to predict how tree species and forests might respond to increased temperatures and water shortage (Anderegg et al. 2015;Rigling et al. 2013).
One way of investigating tree responses to changing environmental conditions is the use of high-precision point dendrometers which allow for monitoring diel stem radius changes (SRC), intra-annual stem growth patterns, and immediate plant water responses (De Swaef et al. 2015;Drew and Downes 2009). Daily SRC are a combination of (1) reversible shrinkage and swelling of elastic tissues (bark and cambium); (2) reversible contraction and expansion of conducting xylem; (3) irreversible stem growth (including wood and bark increments); and (4) thermal expansion and contraction (Daudet et al. 2005;.
Since the reversible processes are mainly determined by water transport dynamics within the tree, stem shrinkage and swelling dynamics can be used to quantify drought stress (Aldea et al. 2018;Vieira et al. 2013;, canopy water status (Dietrich et al. 2018;Van Emmerik et al. 2017), plant hydraulics (Daudet et al. 2005;Ehrenberger et al. 2012), and even osmoregulation (Barraclough et al. 2019;Lazzarin et al. 2018). There are two variables we focus on: the tree water deficit-induced stem shrinkage called tree water deficit (TWD) ) and the maximum daily shrinkage (MDS). TWD quantifies the loss of water from elastic tissues due to an imbalance between water loss (transpiration) and water uptake and reflects the accumulated water deficit also over several days lasting drought periods (Dietrich et al. 2018). It can be used as an indicator for drought stress (Zweifel et al. 2005). In contrast, the maximum daily shrinkage (MDS) quantifies the stem shrinkage over 1 day only ) and is more an indicator for the actual plasticity of a tree to respond to the current environmental conditions in air and soil (Daudet et al. 2005;Giovannelli et al. 2007). A large MDS means a tree is exposed to a large gradient between water demand and water supply and is able to respond with a large stem shrinkage due to a well-saturated stem tissue (Hinckley et al. 1978;. A large MDS therefore indicates a day with open stomata and thus a large transpiration and a high assimilation. A small MDS means either a small gradient between water demand and water supply or a large gradient but under conditions of a low tissue saturation, i.e., already existing stem shrinkage. Thus, a small MDS indicates either a day with low light or a day with closed stomata and thus little transpiration and assimilation. The reversible shrinkage patterns are contrasted by the irreversible stem increment, in this work called growth (GRO) (Deslauriers et al. 2003;Downes et al. 1999). GRO includes radial size increase only and does not include lignification processes or other processes attributable to growth that increase the wood density (Cuny et al. 2015). According to the zero-growth concept of Zweifel et al. (2016), irreversible stem growth (GRO) is limited to time periods with no stem shrinkage. Both TWD and GRO are derived from this zero-growth approach and together (GRO minus TWD) they equal the original dendrometer data SRC.
Here we investigated the two evergreen conifers Juniperus excelsa M. Bieb. (Greek juniper, Cupressaceae) and Cedrus libani A. Rich. (Lebanon cedar or Taurus cedar, Pinaceae), both widely distributed along the mountainous areas of the Eastern Mediterranean Basin, especially in the Taurus Mountains (Southern Turkey), where they can form mixed stands at higher altitudes (Nuri and Uysal 2009). Juniperus excelsa is slow growing (Özkan et al. 2010) and can be found on poor soils in arid to semi-arid regions (Yücedağ and Gailing 2013), between 500 and 2700 m altitude, with total annual precipitation occasionally as low as 250 mm (Gültekin and Gültekin 2007). Cedrus libani is mainly distributed as pure natural stands between 800 and 2200 m altitude with total annual precipitation generally ranging between 600 and 1200 mm but occasionally being as low as 430 mm (Boydak 2003;Durkaya et al. 2013). Both species are known to be frost and drought tolerant (Boydak 2003;Gültekin and Gültekin 2005). Further, several studies showed that they are able to remain photosynthetically/physiologically active at water potentials as low as − 3.0 MPa; therefore, both species can be characterized as anisohydric (Aussenac 2002;Epron 1997;Ladjal et al. 2005;Moran et al. 2017;Voelker et al. 2018).
Despite the knowledge available about the two species (Akkemik 2006;Güney et al. 2017;Touchan et al. 2005), we still lack of understanding how drought affects their respective growth performance. In this study, we monitored hourly SRC of five J. excelsa and four C. libani trees over 2 years (2013 and 2014) including dry periods during which both species experienced stem shrinkage lasting for single days up to several weeks. We investigated whether the two co-existing conifers (J. excelsa and C. libani) show differences in water relations (TWD and MDS) and stem growth (GRO) in order to evaluate their respective drought tolerance. We assume that the ability to grow during dry periods is an indication for drought tolerance quantified as the number of hours with growth and hourly growth rates. We hypothesize (1) that the species with more growth hours during warm and dry conditions has also a larger annual increment. Further, we assume that a short time for recovering from tree water deficit after a several days lasting drought period rather than the maximum tree water deficit value indicates drought tolerance. We hypothesize (2) that the species with a faster recovery from tree water deficit shows an overall better growth performance. Finally, we take maximum daily shrinkage as an indicator for a maintained transpiration and thus assimilation activity and assume that a high maximum daily shrinkage during a dry period indicates a high drought tolerance. We hypothesize (3) that the species able to maintain higher maximum daily shrinkage during dry periods is also able to grow more.

Study area and tree selection
The study was conducted in the Cedar Research Forest (CRF, 2617 ha) which is located in the southwestern part of the Taurus Mountains in the Elmali district of Antalya (SW-Turkey). The CRF is characterized by an oro-Mediterranean climate with dry summers and cold winters. Mean annual air temperature is 7.4°C, and mean total annual precipitation is 725 mm, with rainfall mainly occurring during the winter (annual means were calculated for the period 1968-2000 from climate records of a weather station within the CRF at 1665 m asl, Basaran et al. (2008)). The study site was located in the western part of the CRF at 1350 m asl within a mature natural Cedrus libani-Juniperus excelsa stand. Mean stand inclination was 25°with NW exposure. Soil characteristics varied between clay and clayey loam with a pH (H 2 O) around 7.4 (Basaran et al. 2008). To measure seasonal dynamics of stem radius changes (SRC), five individuals of J. excelsa and four individuals of C. libani, which were representative for the forest stand at our study site, were randomly selected within a circular plot (radius = 30 m). The selected trees were of suppressed and co-dominant status and had no visible damage. Age at breast height (1.3 m) was estimated by extracting tree cores with an increment borer (Suunto 300/400 mm, Finland). For characteristics of the studied trees, see Table 1.

Growth monitoring
Stem radius changes (SRC) were measured with highprecision point dendrometers (linear displacement potentiometers; MMR 10_11 R5K, MEGATRON Elektronik AG & Co., Munich, Germany) on five J. excelsa and four C. libani individuals. The dendrometers were installed at breast height on the slope-parallel sides of the stem, and shielded from direct sunlight and rainfall by aluminum foil and styrofoam covers. The contact head of the dendrometers was placed at the cortex surface after removing as much of the outermost dead bark as possible without damaging living tissue (Zweifel et al. 2006). The dendrometer signals (resolution: 2 μm, trustable accuracy < 10 μm) were recorded hourly and stored using a data logger (DL2e, Delta-T Devices Ltd., UK) from October 2012 until December 2014.

Meteorological data
Meteorological data were collected with a WatchDog 2700 Weather Station (Spectrum Technologies Inc., Plainfield, IL, USA) which was installed in an open area close to the study site (1340 m asl, Güney et al. 2020). Air (T air ,°C) and soil temperature (T soil ,°C), precipitation (Pp, mm), soil water content (SWC, %), relative air humidity (RH, %), and global radiation (GR, W m −2 ) were measured at 10-30 min time steps and were later converted to hourly and daily values. GR was converted to daily sums and expressed as MJ m −2 day −1 . Daily means of vapor pressure deficit (VPD, kPa) were calculated from site-specific T air , RH, and air pressure values. Stem temperature (T stem ) was measured using thermocouples installed at breast height under the bark and recorded hourly by the same data logger.

Data analysis
Three parameters were extracted from hourly recorded stem radius changes (SRC) (Fig. 1). The rate of (irreversible) growth-induced expansion (GRO rate , in μm h −1 , Fig. 1b) of the stem and the hourly tree water deficit-induced shrinkage of the stem (TWD, in μm, Fig. 1c) were calculated based on the concept of Zweifel et al. (2016) which assumes zero growth (ZG) during times of stem shrinkage. GRO rate equals zero as soon as stem shrinkage occurs and TWD increases. TWD is calculated as the difference between the last maximum stem radius and the current stem radius (Fig. 1a, c). Consequently, summed up growth-induced expansion of the stem (GRO) is limited to periods when TWD is zero (GRO period) and the stem radius exceeds the last maximum of stem size (in radius) (Fig. 1a). Maximum daily shrinkage (MDS, in μm, Fig. 1d) was calculated as the difference between the daily maximum and the daily minimum radius of the stem. Absolute TWD and MDS values depend on a number of factors, e.g., a tree stem circumference, and thickness and elasticity of its bark , and we therefore used normalized data to compare TWD and MDS patterns between species (Brinkmann et al. 2016). Normalizing was performed by dividing all TWD and MDS values for each individual tree by its respective maximum value (which was defined as 95% of the measured absolute maximum value during the whole measurement period).

Statistical analysis
Since the data did not fit a normal distribution (Kolmogorov-Smirnov test, p < 0.05), the Wilcoxon signed-rank test was used to check for significant differences between years (within the same species). Spearman non-parametric correlation coefficients were calculated between daily values of TWD, MDS, and environmental variables (T mean , T min , T max , T soil , T stem , Pp, SWC, RH, GR, and VPD) using SPSS v20.0 (SPSS Inc., Chicago, IL, USA). Further, TWD and MDS were fit with MARS models and GRO rate was fit with a two-part model (covering both study years). Beforehand, a factor analysis between environmental variables was performed which showed a high correlation between T mean , T min , T max , T soil , and T stem . Therefore, among the temperature variables, only T max , which was the most representative variable, was included in further analyses. Since the relationship between our dependent and independent variables included nonlinearities, multivariate adaptive regression splines (MARS, caret package in R, Kuhn 2012), which are less sensitive to multicollinearity, were used to investigate the effects of environmental variables (predictors) on TWD and MDS (response variables). MARS also allows to estimate the relative importance of the variables on TWD and MDS by calculating generalized cross  Corresponding values of GRO rate (rate of irreversible growth). c TWD (tree water deficit-induced stem shrinkage) is the difference between the previous maximum stem radius and the actual stem radius. d MDS (maximum daily shrinkage) is the amplitude between daily maximum and minimum stem radius validation (GCV) and residual sum of squares (RSS) of the model. Further, partial dependence plots are used to better understand the relationship between the environmental variables and TWD and MDS. Optimal MARS models were obtained using 10-fold CV in caret. Since our data set of GRO rate consisted of a large number of zero values, we used the twopm command in the Stata software 14 (StataCorp, College Station TX, USA). Twopm is a command for fitting two-part models on mixed discretecontinuous variables. A binary choice model is fit for the probability of observing either a positive or a zero outcome (cf. Belotti et al. 2015), in our case positive GRO rate versus zero GRO rate . The positive outcome is then fit with an appropriate regression model. We used a two-part model that modeled the positive GRO rate (non-zero values) with a gamma distribution using the generalized linear model (GLM) function and zero GRO rate with a binomial distribution using the logit function.

Environmental conditions during 2013 and 2014
At the study site, mean annual air temperatures for 2013 and 2014 were 11.1°C and 11.5°C, respectively. Air temperatures reached a maximum of + 37°C (August 2014) and a minimum of − 13.8°C (January 2013) (Fig. 2). Paired t tests (p <

Stem growth and shrinkage dynamics
The year 2013 was characterized by prolonged times of stem shrinkage lasting for several days which allowed for only short periods of GRO (Fig. 3). During spring and summer 2013, only four out of nine trees (J1, J5, C2, and C4) showed little GRO while five trees showed no GRO. Accordingly, GRO rate was low for both species during 2013 (Fig. 3b, Table 4 in Annex). During 2014, all C. libani individuals showed a significant increase in stem radius which led to an average increment of 1021 μm (284 μm in 2013, Table 4 in Annex) and mean GRO rate of up to 55 μm day −1 (Fig. 3b). In contrast, J. excelsa showed an average increment of 259 μm (245 μm in 2013, Table 4 in Annex) and only reached a mean GRO rate of up to 12.2 μm day −1 (Fig. 3b). However, two individuals (J2 and J3) showed no GRO and one individual (J5) grew much more than the others in 2014 (Table 4 in Annex). Juniperus excelsa grew during 329 h within a year while C. libani grew 278.5 h. The linear regression between tree size (DBH) and average tree ring width for the period 2005-2014 (based on extracted tree cores, Table 4 in Annex) was not significant (p > 0.05, Fig. 7 in Annex). Interestingly, stem growth (GRO) was detected in almost any month of the year and was not restricted to a specific growth period (Fig. 3b). Compared with total stem growth during 2013 and 2014, the percentage of GRO during the warm season (April to October in 2013 and 2014) was 42% and 69% for J. excelsa and C. libani, respectively. Normalized TWD levels were higher in C. libani than in J. excelsa from April to July 2013 but then, during August to mid-September, TWD started to decrease in C. libani and to increase in J. excelsa (Fig. 3c). During 2014, normalized TWD was significantly lower in C. libani than in J. excelsa both during the spring and summer (Fig. 3c, Fig. 4a-b). In both years, MDS levels were significantly lower during the spring than during the summer in both species and MDS during the spring was significantly lower in C. libani than in J. excelsa. MDS showed no significant differences between J. excelsa and C. libani during summer 2013 but was significantly higher in C. libani than in J. excelsa during summer 2014 (Fig. 3d, Fig. 4c-d). Within the same species, mean annual MDS showed no significant differences between years (p > 0.05), while TWD was significantly lower in 2014 (p < 0.05) in both species. Mean values of all (absolute) parameters are shown in Table 4 in Annex for each tree, species, and both study years.

Growth distribution during 24 h and along climatic gradients
Both species grew most at night and in the early morning hours between 2 and 7 AM (Fig. 5). Least growth was observed around midday for C. libani and around the evening hours for J. excelsa. In J. excelsa, almost 90% of the total Fig. 3 Annual course of a hourly stem radius change (SRC) and daily means of b GRO rate (rate of irreversible growth c normalized tree water deficit (TWD) and d normalized maximum daily shrinkage (MDS) of Juniperus excelsa (black symbols and lines) and Cedrus libani (blue symbols and lines)). Bold lines in a represent mean SRC. Letters indicate individual tree ID growth (derived from hourly GRO rate data for the whole measurement period) occurred in a VPD range between 0 and 0.24 kPa while the same range accounted for only around 75% of total growth in C. libani (Fig. 6), meaning that C. libani grew at higher VPD than J. excelsa. For both species, 1.9 kPa was the highest VPD at which GRO was measured. The temperature range for GRO was the same for both species (0 to + 20°C). Although for both species hourly growth rates peaked at 11°C, J. excelsa had a higher proportion of GRO rate at lower temperatures compared with C. libani. Minimum RH and minimum SWC during GRO were 34.18% and 3.2% for both species, respectively (Güney et al. 2020).

Impact of environmental factors on stem radius changes
Generally, air temperature (T max ) had the most significant impact on stem shrinkage as high temperatures were related closest to high values of TWD and MDS (Tables 2 and 3). However, while TWD of J. excelsa was closest correlated with T max , TWD of C. libani was significantly negatively correlated with RH (Table 2) which is also in accordance with the results obtained by MARS models (Table 3). The importance of VPD on TWD was higher in C. libani than in J. excelsa (Table 3). Further, results from correlation analyses showed a more significant relationship between SWC and TWD in J. excelsa than in C. libani (Table 2). TWD was inconsistently affected by SWC in both species. The SWC threshold after which the relationship between SWC and TWD stayed significantly negative was 37.6% in J. excelsa and 40.4% in C. libani as shown by MARS models (Güney et al. 2020).
GRO rate showed a significantly positive relationship to RH in both species. Further, RH was the coefficient that had the highest explanatory power for GRO (non-zero GRO rate ) (highest z value, Table 5 in Annex). The probability of observing higher values of GRO rate also increased with increasing RH. Additional to RH, Pp was also positive and significantly related to GRO in J. excelsa ( Table 5 in Annex). In C. libani, additional estimates that increase the probability to observe GRO were T max , VPD, and SWC (significance in decreasing order, Table 5 in Annex).

Species-specific stem growth patterns
Juniperus excelsa grew during 15% more hours than C. libani, but hourly growth rates and the resulting annual increments were about 60% lower in J. excelsa compared with C. libani. Values sharing the same letter were not significantly different (p < 0.05). In each box plot, data is plotted as the median (midline) and percentiles (box, 25 th and 75 th percentiles), the whiskers and individual points (outliers) represent all values within and above 1.5 times the interquartile range, respectively. Additionally, means were plotted as red lines This is in line with the findings of tree ring studies from Turkey which also show lower annual growth increments of J. excelsa than C. libani (Avsar and Tonguc 2003;Carus 2004;Saatcioglu 1969). Further, in our study, C. libani was able to grow more during the warm period (April to October) and during times of higher temperatures and VPD compared with J. excelsa (Fig. 6). Our results therefore suggest that, at similar sites, C. libani is generally able to grow more than J. excelsa under the warm and dry conditions investigated.
Physiologically this means that low water potentials in the elastic stem tissues of C. libani are faster replenished and thus the negative water potentials and the lowered cell turgor are faster Zero GRO rate values are not included. Values are shown for both study years (2013 and 2014). Total range of measured VPD and Tair during the measurement period were 0 to 5.48 kPa and -11.2 to 36.3°C, respectively returning to the level allowing for cell division and cell expansion (Dietrich et al. 2018;Lockhart 1965;Mencuccini et al. 2017;Zweifel et al. 2006) than in J. excelsa. This could also explain why, during the 24-h cycle, conditions that allowed for GRO were reached earlier in the night in C. libani (Fig. 5). The ability to recover from low water potentials is not solely driven by passive processes related to the availability of water but can be altered by active osmoregulation (Tyree and Jarvis 1982) which has been reported for anisohydric species like C. libani (Ladjal et al. 2007) and also Juniperus spp. (Voelker et al. 2018;Willson et al. 2008). Under water deficits, plants accumulate osmolytes (e.g., sugars, amino acids) in the cytoplasm which increases the osmotic concentration inside the cell (Bray 1993). Due to the increased osmotic concentration, water can move into the cambium and bark cells so that a positive cell turgor is maintained despite the low xylem water potentials which allows plants to remain physiologically active during drought and might thus be able to grow (Arndt et al. 2008;Barraclough et al. 2019;Lazzarin et al. 2018;Mencuccini et al. 2013;Morgan 1984).
Another possible reason for the generally better growth performance of C. libani may be attributed to its deeper taproots that access deeper water reservoirs (Baquedano and Castillo 2007;Boydak 2003;Sevim 1961) than the ones of J. excelsa Tümen and Hafızoğlu 2003). Juniperus excelsa not only was a less good grower during dry periods, but also showed large differences in GRO among the individuals (especially during 2014). We speculate that rather the heterogeneous soil conditions, which lead to different water availability for the individuals, than the different tree sizes or the different social status (Fig. 7 in Annex) could explain the different growth reactions (Fisher and Gardner 1995;Sass-Klaassen et al. 2008). Mesozoic limestone, which is predominant at our study site, has been reported to show large variations in soil depth, SWC, and stoniness and a rapidly changing soil water availability within a year and at short spatial scales (Basaran et al. 2008;Özkan et al. 2010). Since all J. excelsa individuals (J1-J4) that showed a decreased or inhibited growth in 2014 were grouped on the east side of the study site (Güney et al. 2020), we suspect that soil water availability might have been lower there. Unfortunately, no subplot-specific SWC measurements were available to confirm this speculation.

Seasonal stem growth patterns
The Mediterranean climate with a distinct heat and drought period inhibits growth in the summer (Fig. 2) usually leading to the well-described bimodal growth pattern with peaks before and  .78 ** .74 ** .79 ** .76 ** .78 ** − .22 ** − .40 ** − .54 ** .65 ** .74 ** N = 458 days (due to data gaps). Spearman's rho coefficient was calculated. Level of significance: *p < 0.05, **p < 0.01. For each dependent variable, the strongest correlation is in bold GRO rate rate of irreversible growth; TWD tree water deficit; MDS maximum daily shrinkage; T temperature; Pp precipitation; SWC soil water content; RH relative air humidity; GR global radiation; VPD vapor pressure deficit after this period (Aldea et al. 2018;Camarero et al. 2010;Vieira et al. 2014). In this study, the trees of both species grew before and after the dry period (Fig. 3) as expected; however, there were no clear growth peaks (except for C. libani in spring 2014) and stem radial increment was detected in almost any month also during winter time. Particularly J. excelsa utilized many hours in these winter months. From a physiological point of view, it is not unlikely to detect stem growth in these cooler and moister days, since these periods obviously offer better growth conditions than the summer months. It could be seen as a strategy to survive in an area being increasingly exposed to drought. As long as the temperature remains above zero, there is no obvious physiological reason why the evergreen trees should not be able to grow, except for an inner clock eventually avoiding that (Rossi et al. 2006. The fact is our results are unexpected and so far hardly supported by other investigations with alternative growth measurement methods. Güney et al. (2015) investigated xylogenesis of C. libani from the same area during 2013 and stated the growth cessation to take place at the end of September, but sampling was performed until the end of October only and the winter months were not explored. However, there are also some reservations to be considered with dendrometer measurements in general (Drew and Downes 2009;Ilek et al. 2016) and with the zero-growth concept  in particular. The zero-growth concept assumes no growth during periods of stem shrinkage which appears to be physiologically reasonable (Hinckley and Lassoie 1981;Hsiao et al. 1976;Lockhart 1965). However, there is also evidence that a certain fraction of growth might be assigned to the wrong time periods with this approach. Zweifel et al. (2016) quantified this temporally wrongly assigned fraction of growth to be < 5% of annual increment for different species but not covering the two species of this study. The smaller an increment in a μm resolution is, the higher is the risk for a misinterpretation in general. In this work, the irreversible growth increments during winter months, however, were > 200 μm and thus far larger than being discarded as measurement-or method-induced errors.
Another unknown factor which potentially affects the interpretation of the results is the ignored degradation of the bark cells (Gričar et al. 2015) particularly during wintertime with frost . Such a process could mask growth in the sense that a potential increment is underestimated when the bark degradation led to a loss of bark volume (radial stem size) before a growth period. Unfortunately, there is very little known about such degradation processes and the related tissue size changes in the bark (Gričar et al. 2015). However, we question that these processes play such an important role for the species in this study.

Stem shrinkage as an indicator for growth performance
Stem shrinkage results from an imbalance between the water loss by transpiration and the water uptake from the soil. The shrinkage can either be quantified as tree water deficit (TWD) including the absolute shrinkage from the past maximum stem size  or as maximum daily shrinkage (MDS), reflecting the dynamics of a single day only (Garnier and Berger 1986). TWD normalized with its maximum over the 2-year investigation is thus an indicator for drought stress, whereas MDS is a proxy for the ecophysiological activity of a tree, i.e., transpiration and assimilation. We assumed that a faster recovery from TWD indicates a better growth performance under drought. Cedrus libani showed a faster recovery from high TWD particularly during the hot and dry summer months 2013 when the TWD of C. libani started to decrease together with the decreasing VPD in August (Fig. 2, Fig. 3, and Fig. 8 in Annex), whereas the TWD of J. excelsa remained high until October when SWC sharply increased with the first rain events. Therefore, C. libani was able to recover from high TWD with humid air induced by minor rain events and cooler temperatures only (eventually with the help of deep taproots), whereas J. excelsa needed also more humidity in the soil to show the same response. This finding was statistically further supported by a closer relationship between TWD and VPD of C. libani than of J. excelsa (Fig. 8 in Annex) and a closer relationship between TWD and SWC of J. excelsa than of C. libani (Table 2, Güney et al. 2020). The different determining weights of VPD and SWC on TWD have been discussed for other species before (Ehrenberger et al. 2012;Zweifel et al. 2005) and the closer VPD dependency of TWD indicates a closer coupling to air conditions which was shown to disappear with increasing drought stress. We interpret these results as a better ability of C. libani to deal with the prevailing conditions compared with J. excelsa. However, this finding is in contrast to the higher number of annual growth hours of J. excelsa compared with C. libani which indicates that both species have special abilities indicating a good drought tolerance.
The annual course of MDS showed no species-specific differences. MDS was highest during July and August for both species and years, indicating that gas exchange was maintained during the hottest and driest months. Heavily drought stressed trees were found to minimize their daily shrinkage and swelling dynamics (Zweifel et al. 2007) and thus show very low MDS only. However, normalized MDS was significantly higher in C. libani than in J. excelsa during summer 2014. Since C. libani showed a significantly higher stem growth than J. excelsa during the year 2014, this confirms our hypotheses that species that are able to maintain higher MDS during dry periods are able to grow more. Nevertheless, we interpret that both species, despite the different growth responses, were not yet at the edge of their capability to remain physiologically active.

Do we cover the full story?
Despite the fact that almost all investigated growth indicators point towards a better performance of C. libani, we critically ask whether this might hold under more severe drought conditions predicted for the near future when the species are pushed towards their physiological capability to survive. Studies using plant water potentials and the related wilting point by Genc et al. (2005) showed a higher survival rate of J. excelsa under drought stress than C. libani. It appears that J. excelsa is able to maintain physiological processes under extreme conditions but at a very low growth rate. This ability to downregulate physiology may have led to the poorer growth performance in this study but may also indicate a more conservative (growth) strategy (Moran et al. 2017) bearing the potential to survive even more extreme conditions. This could also explain why J. excelsa is found in arid regions with total annual precipitation occasionally as low as 250 mm (Fontaine et al. 2007;Gültekin and Gültekin 2007), whereas C. libani is reported to grow in areas with > 430 mm annual precipitation only (Boydak 2003;Durkaya et al. 2013). The ability to grow more during the cold season could be a further advantage of J. excelsa over C. libani in drier and hotter environments.

Conclusions
In our study, almost all the physiological indicators measured point towards C. libani which, compared with J. excelsa, grew more under increased temperatures and VPD and recovered faster from tree water deficit-induced stem shrinkage under higher VPD already, whereas J. excelsa needed rain and higher SWC to resume growth. However, there remains the question whether J. excelsa with its conservative way of keeping physiological processes low during hottest and driest periods might increase its probability to survive at the cost of a generally lower growth performance. One clue to such an argument is based on the ability of J. excelsa to grow more hours per year than its neighboring tree species. More studies under more extreme conditions and including aspects of tree mortality are needed to follow up on this question.
Acknowledgments Open Access funding enabled and organized by Projekt DEAL. We acknowledge the inspiring exchange of dendrometer analysis know-how with the network TreeNet.info.
Funding Field studies were carried out within the project funded by the German Federal Ministry of Education and Research (01DL12041) and in cooperation with the Southwest Anatolia Forest Research Institute (SAFRI) in Antalya.
Data availability The datasets generated during and/or analyzed during the current study are available in the Zenodo repository, 10.5281/ zenodo.3978175.

Compliance with ethical standards
Conflict of interest The authors declare that they have no conflict of interest.    Open Access This article is licensed under a Creative Commons Attribution 4.0 International License, which permits use, sharing, adaptation, distribution and reproduction in any medium or format, as long as you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons licence, and indicate if changes were made. The images or other third party material in this article are included in the article's Creative Commons licence, unless indicated otherwise in a credit line to the material. If material is not included in the article's Creative Commons licence and your intended use is not permitted by statutory regulation or exceeds the permitted use, you will need to obtain permission directly from the copyright holder. To view a copy of this licence, visit http://creativecommons.org/licenses/by/4.0/.