Introduction

Under steady-state conditions, i.e. assuming plants are unbranched and lacking internal capacitance, whole tree water use can be determined by the soil-to-atmosphere water potential gradient and hydraulic resistance along the water transport pathway (van den Honert 1948), in which stomatal control plays a significant role in regulating water loss (Jarvis and McNaughton 1986; Meinzer 1993). However, it is well-known that most plants are not unbranched and have internal capacitance or water storage capacity (Goldstein et al. 1998; Köcher et al. 2013; Stratton et al. 2000; Tyree and Yang 1990). This suggests water uptake by roots has at least two ways out: (1) vertical upward movement into the atmosphere via stomata; and (2) radial movement of water to recharge storage tissue in the stem (Donnellan Barraclough et al. 2018). The functional linkage between leaf stomatal control and stem water storage has been the subject of numerous studies in seasonally dry tropical forest trees (Brodribb et al. 2003; Meinzer 2002; Meinzer et al. 2008; Pfautsch et al. 2011), but there has been less focus on riparian trees in arid environments (O’Grady et al. 2009).

The presence of apical stem water storage may allow leaves to maintain a water potential amenable to carbon uptake throughout the day (Goldstein et al. 1998; Stratton et al. 2000). Moreover, discharge and recharge of internal capacitances appears to provide a signal for stomatal closure preventing runaway embolism (Nardini and Salleo 2000), and generating dynamic changes in apparent leaf-specifc conductance via the variation of leaf water potential (Meinzer 2002; Meinzer et al. 2008). At the whole-tree level, stem water storage recharge is able to help alleviate unfavourable water potential gradients (Tyree and Yang 1990). Conversely, water storage discharge may buffer the system by reducing the xylem pressure decline when transpiration increases, allowing sufficient time lags for stomatal closure with the onset of drought stress, and potentially preventing xylem embolism during water deficit (Domec and Gartner 2001; Ogasa et al. 2013; Zeppel et al. 2014).

Populus euphratica Oliv. is a typically phreatophyte, mainly distributed across riparian areas or in areas with access to water tables in arid regions of Central Asia and, therefore, it does not typically experience soil drought (Thomas et al. 2008). Populus euphratica is considered the only dominant species of the riparian forests in these regions and plays an important role in protecting the basin ecosystem, preventing desertification and conserving biodiversity. However, with the excessive use of water resources by human being, the P. euphratica forest has declined in extent in recent decades (Wang 1996). Therefore, knowledge of the water use of P. euphratica and its control mechanisms is important to protect and restore this endangered species. Previous studies have shown that it has a relatively ‘conservative’ water use even when it grows under favourable water conditions (Pan et al. 2016). Little is known about how this species regulates its water loss. Some researchers suggest that down-regulation of stomatal sensitivity to leaf-to-air water–vapor pressure has a significant contribution to its conservative water use (Gries et al. 2003; Overdieck et al. 2013; Thomas et al. 2008), but others found that vapor pressure deficit (VPD) has a limited effect on variation of sap flow (Si et al. 2007; Yu et al. 2016; Zhao et al. 2017), suggesting that tree water use is potentially regulated by other biological processes, such as stem water storage, instead of stomatal control solely.

Recently, we observed substantial nocturnal sap flow in P. euphratica (Yu et al. 2016), of which approximately 80% was attributed to stem storage recharge (Yu et al. 2018). Therefore, we expected that water use of P. euphratica could be determined by leaf stomatal control and stem water storage together. Specifically, we aimed to demonstrate (1) the occurrence of stem water storage based on the hysteresis in sap flow across different plant path lengths (trunk and roots) and external climate drivers (e.g. VPD); (2) plant water use was determined but not entirely by climate drivers via stomatal control; and (3) leaf stomatal control can be coordinated with stem water storage to determine plant water use.

Materials and methods

Site and plant material

The study area was situated in the lower reaches of Heihe River Basin, NW China, normally referred to as the Ejin Oasis. The climate is extremely arid exhibiting a ratio of mean annual precipitation (37.0 mm) to mean annual pan evaporation (2216.5 mm) of < 0.02 mm mm− 1 (Yu et al. 2019). Mean monthly air temperature at the Ejin meteorological station ranges from − 11.5 °C in January to 27.0 °C in July, 1957–2016. This study was conducted at the Qidaoqiao P. euphratica Forest National Natural Refuge, Ejin county, Inner Mongolia, lat 42°59′N, long 100°10′E, 920.5 m AMSL (Si et al. 2007).

The study site is characterized by young P. euphratica trees (average 38 year) with a stand density of 350 trees ha− 1 (Yu et al. 2019). Eighty of the total 350 trees within the 100 m × 100 m quadrat were measured for height (m), diameter at breast height (DBH, cm) and projected crown area (Ac, m2), and subsequently sapwood area (As, cm2) was estimated from a previously-published relationship between As and DBH for the site (Si et al. 2009). Four trees were selected to measure trunk diameter variation and sap flow, and As was calculated by increment cores at the end of the experiment. The traits of the four selected trees were not significantly different to the stand means (t-test, P > 0.1) (Table 1).

Table 1 Summary of biological parameters for four selected Populus euphratica trees and quadrat (Mean ± SE) with the size of 100 m × 100 m for this study area
Full size table

Meteorological and hydrologic parameters

Meteorological variables, including net radiation (Rn, W m− 2), air temperature (Ta, °C), relative humidity (RH, %), and wind speed (U, m s− 1) were recorded at a height of 20 m using a CR3000 datalogger at 0.5 h intervals in 2014 in the P. euphratica stand (Yu et al. 2019). Total precipitation was acquired from the nearest meteorological station located in Ejin city, 20 km from the study site. VPD (kPa) was calculated from Ta and RH, and mean daily VPD was not significantly different from that simultaneously measured at the average canopy height (8 m) in 2017 (P = 0.71). The depth to the groundwater table (GWT, m) was measured automatically by pressure transducers (HOBO-U20, Onset Computer Corporation, Bourne, MA, USA) at 0.5 h intervals. GWT decreased gradually until a minimum of 2.22 m on the 18th September, 2014, after which it increased suddenly to 158 mm above ground following a flooding event (Yu et al. 2019).

Stem diameter and sap flow

Four selected P. euphratica trees (Table 1) were used to measure stem diameter variation (Ds, um) at breast height (1.3 m) using point dendrometers (DD-L, Ecomatik, Germany) during the growing season from May through 16 October (168 days), 2014. Ds was separated into two phases: a contraction phase (− Ds) and an expansion phase (+ Ds) following Xiao et al. (2014). Owing to the failure of power to the datalogger, Ds data from 26 May through 16 June (12%) were lost.

Owing to limitations of the instruments, three of the four selected trees (No. 1 to 3, Table 1) were used to measure sap flow using the heat ratio method (SFM1, ICT Inc., Armidale, Australia). Sensors were radially inserted into the xylem tissue of the trunk (canopy break and DBH, 3.5 m and 1.3 m height, respectively) and root (0.3 m depth) on the north side of the selected trees. After insertion, the exposed cambium was covered with aluminium foil to reduce the effects of ambient temperature fluctuation and direct solar radiation. Heat pulse velocity (Vh, cm h− 1) was recorded and all corrections related to probe wounds and misalignment were made according to Burgess et al. (2001). Sap velocity (Vs, cm h− 1) was also calculated following Burgess et al. (2001). Because the sapwood depths (26 ± 1.0 mm) were less than the needle lengths (35 mm), Vs of the inner depth measurement point (7.5 mm from the tip of the needle) was always low and smooth, and even negative (data not shown), suggesting that the probe was inserted in heartwood. Thus, Vs of the outer measurement point (22.5 mm from the tip of the needle) was only selected for analysis. All raw data were continuously recorded at 0.5 h intervals using a microSD card.

Zero flow could not be determined by tree xylem cutting methods because the research site was located in a national nature reserve (Yu et al. 2013). Instead near-zero sap flow was assumed during a night (3 June 2014) with low Ta and extended rainfall (17.2 mm in 3 h), causing VPD to be the lowest (< 0.18 kPa for 3 h) throughout the entire measurement campaign. Recorded Vs ceased quickly with the onset of rainfall and then stabilized at the lowest recorded velocities (1.6 cm h− 1). The As for each selected tree (Table 1) was calculated using measurements of sapwood depth from four increment cores at the same height but different aspects, extracted from each tree after completion of sap flow measurements. Whole tree sap flow (Fs, L h− 1) was calculated as the product of As and Vs of the outer measurement point (calculated as the mean of radial variation given the slight difference along the radius of stem section) (Si et al. 2007), and integrated over 24 h to give daily Fs (L d− 1).

Leaf gas-exchange and water potential

Three scaffold towers 2 m high were set up to access tree crowns for leaf measurements on selected trees. Once a month on clear sunny days (from June 17th to 19th, July 18th and 24th to 25th, August 24th to 26th, and September 11th to 13th) during the growing season, 2014, field experiments were conducted to determine diurnal variation in leaf gas exchange of the three selected P. euphratica trees. Three healthy, fully expanded leaves on each selected tree were marked and measured continuously under ambient conditions using a portable gas-exchange system (LI-6400XT, LI-COR Biosciences Inc., USA) at 1 h intervals during days and at 2 to 3 h intervals during the nights of 18th and19th June. For each measurement period, the leaf level photosynthesis (An, mmol H2O m− 2 s− 1), transpiration (Tr, mmol H2O m− 2 s− 1), stomatal conductance (gs, mol H2O m− 2 s− 1) and environmental factors were calculated. Diurnal variation in leaf water potential (\({\psi _{\text{L}}}\), MPa) was measured on three leaves collected from each tree to the west, south, and east of the crown. Samples were wrapped in a sealed plastic bag with moist paper towel before cutting from the petiole with a razor blade, to prevent sudden dehydration under high temperature conditions. \({\psi _{\text{L}}}\) was measured immediately using a plant pressure chamber (1505D, PMS Instrument Co., Albany OR, USA).

Statistical analysis

The mean and standard error (SE) of all variables were calculated. The difference between biological parameters (height, DBH, As and Ac) of the selected trees and population means was tested using a one-sample t-test at a significance level of P = 0.05. The differences in means of leaf gas exchange variables and \({\psi _{\text{L}}}\) among the months was examined via repeated-measures ANOVA at significance level of P = 0.05 in conjunction with Tukey’s post-hoc test for continuously measured data. Pearson’s correlation was applied to examine the correlations between paired variables (trait-to-trait or trait-to-climate). Given that the different response of plant physiology to envrionmental factors, a linear modelling was performed to assess the relationship between − Ds and + Ds, − Ds and Fs, Fs and VPD in the night; and a sigmoidal model with 3 parameters (f = a/(1 + exp(− (x − x0)/b))) was fitted to assess the relationship between + Ds and Fs and VPD, Fs and gs, Tr, and \({\psi _{\text{L}}}\). All statistical analyses were performed with the software package SPSS Statistics (version 19.0; IBM, Armonk, NY, USA).

Results

Stem radial variation and sap flow

Diurnal increases in Vs preceded VPD increase in the mornings. Variation in Ds was the opposite to both of Vs and VPD (Fig. 1). The time of onset (6:30 am) and cessation (6:00 am, the next day) of Vs was roughly equivalent to Ds, but the pattern of diurnal variation was obviously different, in that Vs increased quickly in the morning hours when VPD was less than 2.0 kPa, and stabilized until VPD had increased to the maximum of 4.0 kPa. After this Vs decreased quickly until around 23:00, and subsequently more slowly, with a significant difference in in the slope of the relationship early and later in the day (P < 0.001). In contrast, Ds initially decreased until around 14:00 and then increased gradually during the afternoon and into the evening. There were differences in the time of onset of increasing Ds between two days (14:00 pm v.s. 19:00 pm), but not in when Ds started to decrease (6:00 am).

Fig. 1
figure 1

Diurnal variation of stem diameter (Ds, µm), sap velocity (Vs, cm h− 1) and vapor pressure deficit (VPD, kPa) during the typical sunny days (June 19th to 20th, 2014). The Ds was separated to two section: contraction (−, grey) and expansion (+, dark grey) phase. The night was represented by the grey horizontal bar at the top. These data were collected synchronously with the consecutive 2 days measurement of leaf gas exchange and water potential

Full size image

Seasonal variation of daytime Fs (Fig. 2d) was also consistent with Rn, Ta (Fig. 2a) and VPD (Fig. 2c). It increased rapidly to a maximum in June and decreased gradually with declining Rn, Ta, VPD and the groundwater table especially after 10 days flooding (Fig. 2f). At night Fs was always stable throughout the growing season (Fig. 2d). RH was interrupted by rainfall and was negatively related to U (Fig. 2b, c). Heavy rainfall events had a greater effect on RH and subsequent Fs than light events (abrupt drop in mid-June, Fig. 2d). Contraction of Ds was higher than the expansion of Ds in May to mid-July, after which contraction and expansion were almost consistent with each other (Fig. 2e).

Fig. 2
figure 2

Seasonal variation of the 5-days mean of a net radiation (Rn), air temperature (Ta), b relative humidity (RH), wind speed (U), c vapor pressure deficit (VPD) and rainfall, d stem sap flow (Fs) at day and night, e stem diameter variation (Ds) at contraction and expansion phase, and f groundwater table (GWT) for the P. euphratica forest during the growing seasons, 2014

Full size image

In the hottest period, mid-July, the Fs measured at the bottom of the trunk was generally lower than at the top of the trunk in both day and night, but Fs in the stem was lower than that in roots during the day, and higher at night. Whether in the stem or roots, maximum Fs was always stable during the day especially on days with high VPD, but during the night the trend in Fs was similar to VPD and the opposite of Ds (Fig. 3a-c). There was no hysteresis in Fs between the top and bottom of the trunk and slight hysteresis (0.5 h) between stem and roots with a counter-clockwise pattern observed (Fig. 3d), suggesting Fs at both the top and bottom of the trunk was activated earlier than in the roots. However, a clockwise pattern was observed between Fs and Ds at the bottom of the trunk, suggesting Fs was activated earlier than Ds (Fig. 3e). Apparently Ds remained constant with increasing Fs in the morning and then dropped with constant Fs in the afternoon, and increased linearly with decreasing Fs in the evening (Fig. 3e). Hysteresis relationships between stem Ds, Fs either in stem or roots and VPD were consistent, with a clockwise hysteresis observed between VPD and Ds and Fs, suggesting VPD was increased earlier than both of Fs and Ds (Fig. 3e, f).

Fig. 3
figure 3

Diurnal changes in a sap velocity (Vs) at the bottom and top of the trunk and the roots, b the normalized stem diameter (Ds) by minimum of days, c vapor pressure deficit (VPD), and the hysteresis d between the ten-days mean of normalized Vs by maximum at the bottom and top of the trunk and the roots, eDs and f VPD with during the summer (July 11 to 20), 2014

Full size image

Leaf gas exchange and leaf water potential

Diurnal variation in leaf gas exchange (gs,An and Tr), \({\psi _{\text{L}}},\) VPD and Fs was also different among the months (Fig. 4). Conductance, gs, increased quickly from the onset of daylight until 11:00 and decreased gradually into the night, with a slight increase heading into dawn in June (Fig. 4a). Photosynthesis, An, lagged behind gs with a lower value and midday depression observed in June and July than in the following two months (Fig. 4b). Transpiration, Tr, increased in the morning until the peak at around 14:00, after which it decreased symmetrically in the afternoon and into the evening (Fig. 4c). Mean \({\psi _{\text{L}}}\) showed the opposite pattern to Tr during the day, with a minimum of − 2.3 MPa in July, which increased to close to − 0.2 MPa in the evening in June (Fig. 4e). Hourly Fs did not change much during day, but quickly decreased into the evening (Fig. 4f). This was consistent with the decreased VPD at night (Fig. 4d). There was a hysteresis and time dependence ‘apparent feedforward’ (Fig. 5), i.e. the relationship between transpiration (Tr) and water potential (\({\psi _{\text{L}}}\)) and VPD depended on the time of day, such that the same VPD yielded higher Tr and \({\psi _{\text{L}}}\) in the morning than in the afternoon, and when VPD increased beyond a certain point, there was decreasing Tr and hence reducing \({\psi _{\text{L}}}.\) With the exception of gs, daily mean An, Tr and \({\psi _{\text{L}}}\) were significantly different among months during the growing season (Table 2). An steadily increased over time although the difference was not significant between August and September. In contrast, Tr increased initially until it reached a maximum in July, after which it decreased, in contrast to the trend observed for \({\psi _{\text{L}}}.\)

Fig. 4
figure 4

Hourly change of leaf a stomatal conductance (gs), b photosynthesis (An), c transpiration (Tr) and d vapor pressure deficit (VPD), e leaf water potential (\({\psi _{\text{L}}}\)) and f stem sap flow (Fs) at different months during the growing season, 2014

Full size image
Fig. 5
figure 5

A hysteretic ‘apparent feedforward’ relationship between (a) transpiration (Tr) and (b) water potential (\({\psi _{\text{L}}}\)) and evaporative demand (VPD). Arrows indicate the order in which data points were obtained

Full size image
Table 2 Seasonal variations of daily mean ± SE (n = 12) of leaf gas exchange parameters included photosynthetic (An, mmol CO2 m− 2 s− 1), stomatal conductance (gs, mol H2O m− 2 s− 1), transpiration (Tr, mmol H2O m− 2 s− 1) and water potential (ѱL, MPa) during the growing season, 2014
Full size table

The relationship between stem water storage, water use and environmental variables

Daily stem water storage discharge as represented by − Ds was linear associated with increasing Fs both during the day (P < 0.001) and night (P = 0.0013) (Fig. 6a). Stem water storage recharge, as represented by + Ds, was significantly sigmoidal (P < 0.001) in the day but not at night Fs (Fig. 6b). Fs showed an exponential increase with increasing VPD to a maximum during the day (P < 0.001) and a linear increase with increasing VPD at night (Fig. 7). Discharge (− Ds) was also significantly sigmoidal (P < 0.001) associated with increasing VPD (Fig. 8a), and + Ds was significantly linear (P < 0.001) associated with -Ds (Fig. 8b).

Fig. 6
figure 6

Daily stem diameter a contraction (− Ds) and b expansion (+ Ds) as a function of sap flow (Fs) at day and night. The fitting line, determination coefficient (R2) and significance level (P) were shown. The outlier in b was induced by the high U at night

Full size image
Fig. 7
figure 7

Daily sap flow (Fs) as a function of vapor pressure deficit (VPD) during day and night

Full size image
Fig. 8
figure 8

Relationship between contraction of stem diameter (− Ds) and a vapor pressure deficit (VPD), and b expansion of stem diameter (+ Ds)

Full size image

At the leaf level, An and gs were positively correlated with Tr, and all leaf-level variables were negatively correlated with ѱL (P < 0.05). Stem Fs was also positively correlated with Tr and negatively with ѱL (P = 0.001) (Table 3). Hourly ѰL always decreased with increasing gs or Tr, but the relationships were significantly different among months, of which ѱL decreased in June and July in a curvilinear relationship with increased gs, but with a linear relationship in August and September (Fig. 9a). ѱL decreased exponentially with increased Tr in July but the decrease was linear in June (Fig. 9b). There was no significant relationship between stem Fs and leaf gs (Fig. 9c), that potentially was confounded by stem water storage. Stem Fs was significantly related with Tr except in July (Fig. 9d).

Table 3 The Person’s coefficient between leaf An, gs, Tr, ѱL and stem sap flow (Fs) with all data pooled
Full size table
Fig. 9
figure 9

Relationship between leaf stomatal conductance (gs), transpiration (Tr) and leaf water potential (\({\psi _{\text{L}}}\)) and sap flow (Fs). a gs v.s. \({\psi _{\text{L}}}\); bTr v.s. \({\psi _{\text{L}}}\); c gs v.s. Fs; dTr v.s. Fs; Data were pooled from 2 to 3 days of measurement for each month

Full size image

Discussion

Numerous researchers have shown that plant water use is the result of a strong coordination between the systems regulating water delivery (xylem) and water loss (stomata) (Brodribb et al. 2003; Meinzer 2002; Meinzer et al. 2009, 2008; Pfautsch et al. 2011). This study supported our hypothesis that (1) stem water storage occurred based on the hysteresis among the sap flow along different water pathways, stem diameter variation and meteorological factors (e.g. VPD) (Figs. 1, 3); (2) stomatal control has great significance but is not the sole driver of leaf and whole-tree water use (Table 3; Figs. 4, 5, 6); and (3) trees can use tissue-stored water preferentially but not immoderately, and this is strongly controlled by stomatal behavior suggesting gs can be coordinated with stem water storage to determine plant water use (Figs. 7, 8, 9).

Stem water storage

There is some controversy around using branch and basal trunk sap flow measurements alone to estimate whole-plant water storage (Burgess and Dawson 2008; Phillips et al. 2009), because lags between crown transpiration and basal sap flow have been observed in some species (Čermák et al. 2007; Goldstein et al. 1998; Köcher et al. 2013), but not in others (Burgess and Dawson 2008; Pfautsch et al. 2011). Other than sap flux, measurements of stem diameter variation are among the most useful plant-based measurements to evaluate plant water consumption, and indicate the discharge and recharge of internal stores (Čermák et al. 2007; Scholz et al. 2008; Zweifel et al. 2001). Observations from other studies suggest that the patterns of water use within a plant can be better understood by combining sap flow and stem diameter variations measurement (Köcher et al. 2013; Sevanto et al. 2008).

For P. euphratica, there was no phase shift between the traces at the bottom and top of the trunk, at least within the 30 min resolution period of our data (Fig. 3d). This may be a result of the short distance between measurement sites on the main trunk (about 2 m). Such a height effect on diurnal water storage has been observed in other tree species (Goldstein et al. 1998; Köcher et al. 2013). Even so, we conclude that the discharge and recharge of stored water for the studied species occurred as illustrated by the following:

First, we observed a ‘plateau effect’ in sap flow (Figs. 1, 3a) just as in many other species (Goldstein et al. 1998; Scholz et al. 2008), especially wet–dry tropical savanna ecosystem (Eamus et al. 2000; O’Grady et al. 2009, 1999; Pfautsch et al. 2011). This suggests that the discharge of stem water, evidenced by quickly declining Ds at the same time in the study (Figs. 1, 3b), can maintain maximum or near maximum transpiration rates for a longer period of time (Eamus et al. 2000; Goldstein et al. 1998; Pfautsch et al. 2011). The time of stable sap flow was about 9–10 h (Figs. 1, 3a), similar to Eucalyptus victrix in semi-arid regions (Pfautsch et al. 2011).

Second, there was a slight hysteresis between Fs in the bottom of the trunk and the roots (Fig. 3b). It is well-known that stored water is withdraw firstly from branches, and subsequently stem and roots, although we did not measure this for branches in our study. The apparent lag of approximately 30 min between Fs in stem and roots, was consistent with savanna trees, in which root diameter declined about 40 min after changes were observed at the base of the stem (Scholz et al. 2008).

Third, the stem diameter variation (Ds) was large (Fig. 2b) and closely related to Fs (Fig. 5), suggesting that discharge of and recharge to stem stored water occurred (Zeppel et al. 2014; Zweifel et al. 2001). The large daily amplitude of variation in whole-stem diameter (Fig. 2b) in our study was similar to whole-stem diameter variation in Cerrado trees (Scholz et al. 2008) and previous measurements on radius variation for P. euphratica (Xiao et al. 2014). The strong relationship between stem diameter variation during the contraction phase (− Ds) and VPD (Fig. 7a) and + Ds (Fig. 7b) also suggest the daily discharge and recharge of stored water (Scholz et al. 2008).

Stomatal and stem hydraulic regulation on diurnal water use

It has been shown that components of the tree water transport pathway (roots, trunks, branches and leaves) can also serve as water storage compartments (Scholz et al. 2011). Thus, the amount of water that is withdrawn and used on a daily basis will not only depend on the environmental conditions experienced by the plant, but also on stomatal regulatory behaviour and water storage (Meinzer et al. 2009; O’Grady et al. 2008; Scholz et al. 2011).

Hysteresis in the relationship between diurnal sap flow (or tree water use) and environmental factors (e.g. VPD) has been observed in a number of species (Hogg and Hurdle 1997; Meinzer et al. 1997; O’Grady et al. 1999, 2008; Zeppel et al. 2004), including old-aged P. euphratica, in which Vs came in advance of VPD by 0.5 h (Li et al. 2016). The potential reasons for hysteresis can be related to either increased water use in the morning (e.g. water storage capacity, ‘stem hydraulic regulation’) or reductions in the afternoon (e.g. stomatal response to environmental factor, ‘stomatal regulation’). The reason for declining water use in the afternoon have been explained as associated with stomatal sensitivity to VPD (Hogg and Hurdle 1997; Meinzer et al. 1997; O’Grady et al. 1999; Zeppel et al. 2004). Previously, the stomatal sensitivity to VPD was attributed to increased resistance in the soil-plant-atmosphere continuum, and more precisely to increased leaf resistance, given that it accounts for 30%–90% of whole-tree resistance (Brodribb and Holbrook 2006) with declining \({\psi _{\text{L}}}\) (Brodribb and Holbrook 2006; Buckley 2005; O’Grady et al. 2008). However, this explanation was not consistent with the observation in the current study that gs and Tr did not decrease significantly with increased VPD and/or decreased \({\psi _{\text{L}}}\) (Fig. 4; Table 2). If, following a reduction in VPD, embolism repair lagged behind the recovery of leaf water status, then the trajectory of Tr vs VPD was altered (Buckley 2005).The high Tr during the hottest months was consistent with high Fs (Fig. 4), which perhaps is necessary to avoid overheating, particularly for plants from hot dry habitats (Lin et al. 2017). This suggests that trees must withdraw more water (from soil and/or tissues) to prevent the catastrophic decline in \({\psi _{\text{L}}}\) (Fig. 4e). Obviously, constant and steady flow (i.e. the ‘plateau effect’, Figs. 1, 3a) may not be met by root uptake (‘steep’ pattern of Vs) (Fig. 3a), but by the intermediate sources of water from stem water storage for transpiring leaves (Goldstein et al. 1998; Stratton et al. 2000).

Populus euphratica, has an ‘anisohydric’ behaviour (Fig. 4e), that seems to use an ‘apparent feedforward’ mechanism (Buckley 2005) to control its water use under high evaporative demand in arid environments (Fig. 5). In addition, VPD had not completely determined influence on sap flow or stem diameter variation in this study (Figs. 7, 8a), in line with the previous work on P. euphratica (Li et al. 2016; Xiao et al. 2014; Yu et al. 2018, 2016), suggesting that the water use of this tree was partially regulated by the transpirational demand (VPD), and hence stomatal control (Fig. 9). In addition, the positive relationship between Fs and stem diameter variation (Ds) (Fig. 6) suggests increased sap flow was partly used to recharge stem water storage, and subsequent compensates the sap flow under high transpirational demand. This supports the hypothesis that water storage capacity or hydraulic capacitance play an important role in refilling the stem following depletion by transpiration (Ogasa et al. 2013; Scholz et al. 2011).

The significance for tree growth

Midday minimum \({\psi _{\text{L}}}\) (− 2.3 MPa) in our study was similar to the woody phreatophytes in riparian areas of the arid zone (Gries et al. 2003; Zeng et al. 2006) included P. euphratica seedlings under drought (Li et al. 2012), but higher compared to typical desert plants (see review in Thomas et al. 2008). A lower \({\psi _{\text{L}}}\) that induced 50% loss of hydraulic conductance (P50, ~ − 2.7 MPa), was reported for a shallower water table than deeper, suggesting a positive hydraulic safety margin defined as the difference between midday \({\psi _{\text{L}}}\) and P50 (Pan et al. 2016). This suggests that in this study trees did not experience drought stress under the current water table. This may explain why seasonal variation of gs remained constant during the whole growing season (Table 2) as it is strongly related to soil drought (Garnier and Berger 1987; Giorio et al. 1999). The early peak of gs and subsequent gradual decline, while VPD continued to increase (Fig. 4), helps plants to minimize the risk of leaf cavitation (Brodribb et al. 2003; Goldstein et al. 1998; Pfautsch et al. 2011). Recently, a multi-layer porous media model predicted that the use of stored water remains constant as a result of water depleted in plant xylem tissues can be completely recovered to its previous state within a single diurnal cycle (Huang et al. 2017). This was supported by the stable nocturnal sap flow in the stem (Fig. 2d), which was substantially attributed to stem water storage (Yu et al. 2018). Because diameter expansion (+ Ds), which of including the irreversible radial growth except the increases in stored water, was first increased and then declined, Thus, nighttime Fs is not related to + Ds (Fig. 6b). This suggests coordination of stomatal control and stem water storage to buffer xylem water deficit, and subsequent prevention of embolism risk.

The An was greater in the late growing season compared with the early for P. euphratica (Fig. 4b; Table 2). It was consistent with that observed in the lower Tarim River (Chen et al. 2012), suggesting higher carbon uptake in the late growing season compared with the early (Thomas et al. 2008). However, Tr and gs did not increase as An increased in study (Fig. 4b; Table 2), suggesting enhanced water use efficiency in the late growing season (Chen et al. 2012). Mean gs (Table 2) was higher compared to the same species growing in the desert environment (Thomas et al. 2008), but similar to that determined in the riparian zones (Chen et al. 2012). High leaf An was inconsistent with low stem radial increment in the late growing season (See Fig. 2b for the indistinguishableness between + Ds and − Ds). One possible explanation is that the carbohydrates synthesised by photosynthesis were transported downwards into the roots, as evidenced by increased belowground biomass and rooting depth with increasing groundwater table (Li et al. 2012). In all, insignificant difference in stomatal conductance during the growing season suggests the stomatal control has a limited effect on plant water loss. Alternatively, daily stem water storage from groundwater provides sufficient water supply for leaves to maintain high transpiration, thereby high photosynthesis and water use efficiency. In turn, transpiration reduces plant water potential that provides the water potential gradient needed to store the next day’s water and limit losses to hydraulic redistribution (Yu et al. 2018).

Conclusions

Populus euphratica, a deciduous broad-leaved tree distributed in the riparian zone in arid regions of Central Asia, has a relatively ‘conservative’ water use strategy even when growing in the favourable water conditions, which was not solely attributed to stomatal regulation in contrast to the previous conclusion. The lag time in sap flow between the bottom of the trunk and roots and diameter variation suggests that stem water storage discharge and recharge occurred. The relationships between leaf gas exchanges, stem sap flow and/or diameter variation and meteorological factors demonstrate that the pattern of water use of P. euphratica was the coordination of leaf stomatal control and stem water storage. This has great significance for buffering xylem water deficit, preventing the risk of embolism, maintaining high leaf production, and water use efficiency.

Author contribution statement

TY carried out the experiment, analysed the data, and wrote the first draft. QF and JS help to design the experiment and discuss the results. EAP helped to review and revise the article.