Stem-mediated hydraulic redistribution in large roots on opposing sides of a Douglas-fir tree following localized irrigation


Author for correspondence:
Kathy Steppe
Tel: +32 9 264 61 12


  • Increasing evidence about hydraulic redistribution and its ecological consequences is emerging. Hydraulic redistribution results from an interplay between competing plant and soil water potential gradients. In this work, stem-mediated hydraulic redistribution was studied in a 53-year-old Douglas-fir tree during a period of drought.
  • Sap flux density measurements using the heat field deformation method were performed at four locations: in two large opposing roots and on two sides of the tree stem. Hydraulic redistribution was induced by localized irrigation on one of the measured roots, creating heterogeneous soil water conditions.
  • Stem-mediated hydraulic redistribution was detected during night-time conditions when water was redistributed from the wet side of the tree to the nonirrigated dry side. In addition to stem-mediated hydraulic redistribution, bidirectional flow in the dry root was observed, indicating radial sectoring in the xylem.
  • It was observed that, through stem-mediated hydraulic redistribution, Douglas-fir was unable to increase its transpiration despite the fact that sufficient water was available to one part of the root system. This resulted from the strong water potential gradient created by the dry soil in contact with the nonirrigated part of the root system. A mechanism of stem-mediated hydraulic redistribution is proposed and its possible implications are discussed.


Evidence of hydraulic redistribution in plants and its ecological consequences is becoming more and more widespread. ‘Hydraulic redistribution’ (HR) is a general term which describes the passive movement of water controlled by competing soil and plant water potential gradients and corresponding pathway resistances. This phenomenon is known as hydraulic lift when the water movement occurs in the soil compartment and water is transported through roots from deeper moist layers to upper dry layers (e.g. Caldwell & Richards, 1989; Caldwell et al., 1998). Roots have been found to transport water not only in a vertical direction, but also laterally when water is transported from wet to dry parts of the soil (e.g. Brooks et al., 2002; Smart et al., 2005; Nadezhdina et al., 2006). Although substantial information is available with regard to the process of root-mediated HR, much less information is available about HR within above-ground tree parts. This lack of information is especially surprising because of their ease of accessibility compared with roots. Moreover, the first demonstration of reverse flow and the possibility of HR in tree stems was provided as early as c. 300 years ago by Stephen Hales (1727) in his ingenious sap ‘circulation’ experiments. The sap flow in a branch transporting water from roots to foliage under normal natural conditions switched direction when the branch was severed and its cut upper end placed into water (Fig. 1). Under natural conditions, such reverse flow of water in tree stems can, for example, be caused by foliar uptake during fog or drizzly weather (Burgess & Dawson, 2004). In addition to the role played by tree stems in the redistribution of water coming from the leaves, they can be involved in the redistribution of soil water. Observations of this stem-mediated HR remain, as far as is known, limited to the recently published paper by Burgess & Bleby (2006). In their study, they showed that considerable portions of the stem of a large Eucalyptus salmonophloia tree became involved in the redistribution of water between lateral roots on opposite sides of the stem following a rain event. Redistribution between lateral roots was also observed by Brooks et al. (2006) in Douglas-fir, but the stem sap flux density, unfortunately, was not measured.

Figure 1.

 Historical illustration of an experiment to demonstrate the possibility of bidirectional passive water movement in a large branch of an apple tree (modified from Hales, 1727). The sap flow in the branch transporting water from roots (1) to foliage (2) under normal natural conditions (a) switched direction [from (2) to (1)] when the branch was severed and its cut upper end put into water (b).

One of the techniques used to characterize HR is stable isotope labelling (Dawson & Pate, 1996; Smart et al., 2005; Meinzer et al., 2006; Schoonmaker et al., 2007). These studies have provided important insights into the understanding of dynamic water transport, but are rather demanding in terms of research time and equipment requirements. An alternative approach is to perform direct and continuous measurements of the water movement using sap flow sensors. In order to characterize HR, only sap flow methods that allow bidirectional measurements can be used. Currently, the heat ratio method (Marshall, 1958; Burgess et al., 1998), the heat field deformation (HFD) method (Nadezhdina et al., 1998, 2008; Saveyn et al., 2008) and the modified heat dissipation technique (Brooks et al., 2002) allow for such measurements. As other available sap flow measurement techniques are not able to detect flow reversals, some studies may not have detected HR.

The observation of HR in tree stems under natural conditions is, so far, a rare occurrence because, for HR to occur, the root system needs to cover soil layers/patches of different water potentials. Naturally occurring rain events tend to level out these water potential gradients in most cases. Burgess & Bleby (2006), however, observed HR after rain because variability in the soil water content on different sides of the eucalyptus tree remained, which they attributed to possible rain shadows, patches of nonwetting soil or varying soil textures. Given the fact that HR is entirely driven by water potential gradients, unique experiments can be designed in which the competing soil/plant water potential driving forces are artificially altered. One such approach is to locally irrigate the root system and, as such, increase the soil water potential to near zero at that specific location.

The aim of the present study was to investigate HR in a 53-year-old Douglas-fir by performing a manipulation experiment in which one side of the root system was locally irrigated. For Douglas-fir, HR was found to be one of the main reasons for resisting drought (Domec et al., 2004; Meinzer et al., 2004; Unsworth et al., 2004; Warren et al., 2005, 2007). As none of these studies considered the possible contribution of the stem, it was our aim to investigate stem-mediated HR using sap flow sensors positioned at root and stem level, and to elucidate possible radial sectoring in the xylem, represented by radial differences in sap flux density magnitude and direction. In addition, we also showed the effect of a rain event on the observed HR in the tree.

Materials and Methods

Site description

The experiment was conducted in the summer of 2005 in the Sobesice forest district, a suburb of the city of Brno, Czech Republic (49°16′N, 16°39′E). The experimental stand covers 0.5 ha of mixed forest and is located at an altitude of 360 m. Douglas-fir [Pseudotsuga menziesii (Mirb.) Franco] is an important tree species in this forest district, from which one sample tree was selected for our measurements. The soil is illimerized on weathered granodiorite with a cover of loess to a depth of 30–40 cm. The mean annual temperature is 7.8°C and the mean annual precipitation is 580 mm, 360 mm of which falls during the growing season (Vasicek, 1984).

Environmental measurements

The air temperature and relative humidity of the air (EMS 33, EMS, Brno, Czech Republic) were monitored in an open area close to the sampled tree (c. 70 m) at a height of 5 m. Global solar radiation over the waveband 300–5000 nm (EMS 11, EMS) was measured c. 1 km from the experimental site. All meteorological sensors were logged at 15 min intervals with a data logger (MIDI-12, EMS). These variables allowed the calculation of the daily Turc potential evapotranspiration (PET) (mm d−1) (Turc, 1961):

image(Eqn 1)

where Rg is the global solar radiation (MJ m−2 d−1) and Ta is the average air temperature (°C). Turc’s equation is one of the simplest and most accurate empirical equations used to estimate PET (Federer et al., 1996).

On two occasions during the experiment (August 5, 2005 at 14.00 h and August 6, 2005 at 20.00 h), four soil samples, c. 3 m from the base of the tree, were collected from the upper (0–35 cm) and deeper (35–75 cm) soil layers to quantify the volumetric soil water content.

Sap flux density measurements

The sap flux density (cm3 cm−2 h−1) was measured using sensors based on the HFD method (Nadezhdina et al., 1998; Čermak et al., 2004; Nadezhdina et al., 2008; Saveyn et al., 2008). Each HFD sensor consisted of a heater and three needles. Each needle contained five thermocouple junctions, at 10 mm distance from each other. The first thermocouple junction was located 5 mm below the cambium. The total length of each needle was 80 mm. The heater was 20 mm longer to avoid the deepest thermocouples being influenced by nonheated sap. The first needle was installed 1.5 cm above the heater (axial direction), the second needle 0.5 cm next to the heater (tangential direction) and the third needle 1.5 cm below the heater (axial direction). The first needle contained the upper junctions of the axial thermocouples, the second needle the upper junctions of the tangential thermocouples and the third needle the lower (reference) junctions, which were the same for the axial and tangential thermocouples. To ensure the parallel installation of the needles, a template was used to keep the drill bit perpendicular to the bark surface. Before placement in the trees, the probes were coated with silicone grease. The sensors were thermally insulated with open-cell foam and aluminium covers. Changes in the heat field caused by moving sap were characterized by the ratio of the axial and tangential temperature differences around the heater.

Sap flux density estimates of each thermocouple junction i (qi) (cm3 cm−2 h−1) were based on this latter ratio, on the geometry of the measuring point and on the thermal diffusivity of the wood (Nadezhdina et al., 2006):

image(Eqn 2)

where D is the thermal diffusivity of fresh wood (cm2 s−1), K is the absolute value of dTsym – dTasym under conditions of zero sap flow (°C), dTsym is the temperature difference between the axial thermocouple junctions (°C), dTasym is the temperature difference between the tangential thermocouple junctions (°C), Zax is the distance between the upper junction of the axial thermocouple and the heater (1.5 cm), Ztg is the distance between the upper junction of the tangential thermocouple and the heater (0.5 cm) and Lsw is the sap wood depth (cm). A nominal value for D of 0.0025 cm2 s−1 was chosen, according to Marshall (1958). K was determined empirically from the relationship between dTsym – dTasym and dTsym/dTasym (Nadezhdina et al., 2006).

The same sensor configuration was used to estimate the reverse flow. However, in this case, the sap flux density was calculated differently from the measured temperature gradients in order to take into account the changed flow direction. When recorded values of dTsym became negative, Eqn 1 had to be reorganized (Nadezhdina et al., 2006):

image(Eqn 3)

Data were recorded every minute and stored as 10 min averages with data loggers (DL2e; Delta-T Ltd., Cambridge, UK and MIDI-12, EMS).

Experimental design

In order to detect stem-mediated HR, we selected a 53-year-old Douglas-fir tree (35.6 cm in diameter at breast height; southern crown projected area < northern crown projected area) in which four HFD sensors were installed. Two of these sensors were installed at the northern and southern sides of the stem at breast height. The other two sensors were installed in two large roots at a distance of 16 cm from the northern and southern sides of the stem base. The widths of these roots were 10 and 12 cm, respectively. An overview of the sensor installation is depicted in Fig. 2a.

Figure 2.

 Schematic representation of the experimental set-up, indicating the positioning of the heat field deformation (HFD) sensors (three parallel lines) and the application of localized irrigation (a), and the concept of stem-mediated hydraulic redistribution in the Douglas-fir tree (b).

After an initial period of monitoring the sap flux density during dry weather conditions (up to August 5, 2005), the southern side of the tree was locally irrigated with 60 mm of water covering 10 m2 (on August 5, 2005 between 15.00 and 16.00 h). The measurements made during the initial monitoring period showed that only two of the five thermocouple positions were detecting sap flow on the southern stem side, indicating a shallow sap wood depth of 15 mm. This shallow sap wood depth corresponded to fewer branches connected with that side of the stem, and hence a smaller projected crown area compared with the northern side. Given that one of the aims of the present study was to investigate radial sectoring in the xylem, it was necessary to have deep sap wood in order to optimally observe the expected reverse flow during stem-mediated HR. As this was expected to occur at the side opposing irrigation, the southern side of the tree stem was chosen to apply localized irrigation (Fig. 2a).

No rain occurred during the 9 d following the irrigation event and the weather conditions continued to be warm, gradually drying out the soil. Thereafter, 57 mm of precipitation fell during two rainy days (August 15 and 16, 2005). Therefore, the period from August 17, 2005 onwards was characterized as wet.

Radial sap flux density profiles (i.e. sap flux density as a function of xylem depth) measured in stem and roots were analysed for three key days during the experiment at 06.00 h (night) and 13.00 h (day). These days were selected just before irrigation (August 5, 2005), just after irrigation (August 6, 2005) and just before the rain event (end of the dry period, August 14, 2005).


Environmental site characterization

As the objective of this study was to induce and study stem-mediated HR under natural conditions, a period of dry weather was required. Figure 3 shows the global solar radiation and vapour pressure deficit (VPD) of the air during the experiment. Two distinct periods can be discerned: a dry period from July 30 until August 14, 2005, and a wet period from August 15 until August 21, 2005, initiated by a 2 d rain event. During the rain event, a significant decrease in global solar radiation and VPD was observed. Outside this period, dry conditions were characterized by high values of VPD (up to 4 kPa) and global radiation (daily sums of 19.4 ± 5.2 MJ m−2). The application of localized irrigation (on August 5, 2005) did not influence the atmospheric driving forces, but significantly altered the soil water content and, by inference, the soil water potential. The volumetric soil water content at the irrigated site increased from 10.6% to 24.8% (an increase of 133%) in the upper soil layer, and from 9.3% to 14.3% (an increase of 53%) in the deeper soil layer.

Figure 3.

 Global solar radiation (black line) and vapour pressure deficit (VPD) of the air (grey line) during the experiment. The three experimental phases are marked. Localized irrigation on the southern root (indicated by the first broken line) divides the dry period into homogeneous and heterogeneous parts. A rain event (indicated by the second and third broken lines) separates the dry from the wet period.

Direct response of sap flux density to localized irrigation

Immediately after localized irrigation near the southern root, a strong response in sap flux density was observed in the tree, at both the root and stem levels (Fig. 4). The sap flux density in the southern root increased, whereas it decreased in the northern root. This occurred at all depths, but was most evident in the outermost xylem. Similar observations were made for the stem, although they were less pronounced.

Figure 4.

 Direct response of the sap flux density in Douglas-fir to localized irrigation (indicated by the broken line) in the roots (a) and stem (b). Measurements were conducted on both the northern (N) and southern (S) parts of the tree and at several xylem depths (indicated by different line thicknesses and values expressed in millimetres).

Of particular interest were the negative sap flux densities, which were observed during the night in the northern root and stem. This corresponded to a reversal of flow, suggesting that HR took place. Reverse flow first appeared in the inner xylem of the northern root (45 mm below the cambium) c. 5 h after irrigation, and was followed by shallower xylem layers (i.e. 35 mm and 25 mm below the cambium) c. 7 and 9 h after irrigation, respectively. No reverse flow was observed in the outer xylem (15 mm and 5 mm below the cambium). The flow behaviour in the northern root differed from that observed on the northern side of the stem: the first sign of reverse flow only occurred at c. 9 h after the irrigation event and was established in all stem xylem layers (within a period of 2 h).

Long-term response of sap flux density to localized irrigation

The localized irrigation not only caused short-term effects, but also affected the sap flux density in the long term (Fig. 5). In order to illustrate these long-term responses, sap flux densities were averaged over all xylem depths.

Figure 5.

 Long-term response of the sap flux density in Douglas-fir to localized irrigation and a rain event in the roots (a) and stem (b) for both the northern (black lines) and southern (grey lines) sides of the tree. For each measurement location, the sap flux densities were averaged over all xylem depths. Localized irrigation on the southern root and the rain event are indicated by broken lines.

A gradual decrease in sap flux density was observed at all measuring locations during the dry period, caused by decreasing soil water availability. The drying of the soil was similar on both sides of the tree (homogeneously dry soil water conditions), and hence no HR or reverse flow was recorded. Although, in the opposing roots, similar sap flux densities were measured, the sap flux density on the southern stem side was twofold lower than that on the northern stem side. This difference could be attributed to the lower corresponding crown projected area of the southern side of the tree.

As a result of the localized irrigation, gradients in soil water potential were created (heterogeneously dry soil water conditions). During the first night following irrigation, the difference in soil water potential forced both roots to react in opposite directions – a positive flow in the southern root and a reverse flow in the northern root – causing water to be redistributed from the southern side to the northern side. Daytime sap flux densities in the southern root were, as expected, higher than those measured during the dry period. The daytime values in the northern root, however, decreased by more than 50%. As time progressed and the irrigated water on the southern side was depleted, the daytime sap flux density in the southern root decreased. A gradual increase in daytime sap flux density in the northern root was possible because the water which became available through HR resulted in a less negative soil water potential. As a result of these two opposing trends, sap flux densities converged to their earlier state. Similar sap flux density patterns were observed in the stem, although less pronounced and without the gradual increase in the northern daytime values.

After some rain on August 15, 2005, homogeneous soil wetting caused an increase in sap flux density in both roots and stem. As a result of this soil homogeneity, no reverse flow occurred and the proportionality between the signals, as observed during the period before irrigation, reinstated itself. The absolute values, however, were higher, given the fact that more soil water was evenly available.

Radial response of sap flux density to localized irrigation

In contrast with earlier studies, one of the aims of this study was to investigate the radial sectoring of HR in the xylem, i.e. differences in the sap flux density magnitude and direction as a function of xylem depth. Therefore, radial sap flux density profiles were analysed (Fig. 6).

Figure 6.

 Radial profiles of the sap flux density measured in the roots (a, c) and stem (b, d) of Douglas-fir. Three key days during the experiment are shown: August 5, 2005 (before localized irrigation on the southern root); August 6, 2005 (just after localized irrigation); and August 14, 2005 (just before the rain event, end of the dry period). The three days are represented by a profile measured at 06.00 h (night) and 13.00 h (day). The arrows indicate the transition between the studied days.

The radial profiles confirmed the patterns of increasing and decreasing sap flux densities (indicated by the arrows in Fig. 6), which were also observed using the averaged values (Fig. 5). However, from the radial profiles, it became clear that these changes showed different magnitudes for different xylem depths. This was most pronounced during the day when changes were larger in the outer xylem than the inner xylem.

With regard to HR during the night on the northern side of the tree, a pronounced difference between root and stem was observed. In the root, only the inner xylem layers (45–25 mm below the cambium) contributed to HR, whereas, in the stem, the middle xylem layers (35–15 mm below the cambium) contributed most to HR. As such, this is clear evidence of radial sectoring in the xylem.

Integrated view of temporal and radial influences of localized irrigation

A more complete picture, integrating both the temporal and radial variability of HR in both the roots and stem, is depicted in the surface plot (Fig. 7) for the period between the localized irrigation and the rain event (heterogeneous soil water conditions).

Figure 7.

 Integrated view of temporal (x-axis) and radial (y-axis) influences of localized irrigation on the sap flux density (z-axis) in the roots (a, c) and stem (b, d) of Douglas-fir. Localized irrigation on the southern root is indicated by the broken line. The reverse flow is shown by green colour gradients.

The fact that no reverse flow was observed on the southern side of the tree (Fig. 5, averaged values) was confirmed in Fig. 7, even when all xylem depths were considered. By contrast, the northern side of the tree showed significant reverse flow, the temporal, radial and circumferential variability of which is described below.

When looking at the temporal variability, significant night-time reverse flow was detected on the northern side of the tree (Fig. 7a,b) despite nonzero VPD (Fig. 3), suggesting the occurrence of some night-time transpiration which might reduce HR. The daily night-time reverse flow period in the root was significantly longer than that in the stem (Fig. 7a,b). Moreover, a delay of c. 3.5 h was noticed between the onset of the reverse flow period in the stem when compared with the root. Reverse flow in the root and stem ceased sharply around sunrise. This coincided with the time at which the transpiration-induced tension overruled the prevailing night-time water potential gradient. During the heterogeneously dry period, the night-time reverse flow activity continued for a longer time (several days) in the root when compared with the stem.

With regard to the radial variability, it was apparent that the reverse flow values were higher in the root than in the stem (Fig. 7a,b: indicated by the green colour intensity). Differences in magnitude were observed not only between the root and stem, but also between subsequent xylem depths. Indeed, although, in the stem, all measured xylem layers contributed to the reverse flow, only the inner layers in the root showed negative sap flux densities (Fig. 7a,b). As the reverse flow decreased with time, the HR remained more and more restricted to the inner xylem layers in both the stem and root. Moreover, in the root, an opposite flow was detected, where the inner layers transported water towards the soil and the outer layers conducted water towards the stem base. Although a typical radial sap flux density profile already demonstrates some radial sectoring, a bidirectional flow pattern can be regarded as the ultimate evidence of radial sectoring (Fig. 7a).

In addition to this radial sectoring, circumferential sectoring was also distinguished, whereby the northern and southern sides of the tree showed differences in sap flux density pattern, even before localized irrigation (Fig. 7). Although this has already been reported in several studies (e.g. Čermak et al., 1992, 2004; Saveyn et al., 2008), the simultaneous observation of opposing flows in the northern and southern parts of the stem after the application of localized irrigation was especially remarkable. This particular observation demonstrates the existence of stem-mediated HR in Douglas-fir.

Sap flux density correlation analysis

As a result of the experimental design, three distinct periods could be discerned: the homogeneously dry period (before localized irrigation), the heterogeneously dry period (after localized irrigation) and the homogeneously wet period (after the rain event). In order to investigate how water transport in the roots related to the mean stem sap flux density during these three periods, an analysis of possible correlations was performed (Fig. 8). Strong linear relationships were found between the mean stem sap flux density and the mean root sap flux density (Fig. 8a) for the three periods. A clear distinction could be made between the wet and dry periods (i.e. homogeneously vs heterogeneously dry).

Figure 8.

 Sap flux density correlation analysis between roots and stem in Douglas-fir. The relationship between the mean (i.e. average of northern and southern sides) stem and root sap flux density is shown for the homogeneously dry, heterogeneously dry and homogeneously wet periods (a). The daily relationship (indicated by different colours) between the mean stem sap flux density and the northern and southern root sap flux densities (b) illustrates the response to localized irrigation on the southern root (solid arrows) and the return to equilibrium (broken arrows).

The slope that characterized the wet period was below the 1 : 1 line, indicating a higher mean sap flux density in the stem compared with that in the measured roots. This could be attributed to the contributions of nonmeasured roots which were able to access the rain water, evenly distributed across the entire root system. In contrast with the homogeneously wet period, the slope of the homogeneously dry period was very close to the 1 : 1 line. Interestingly, an almost identical slope was observed for the heterogeneously dry period with the slight difference attributable to HR. Indeed, this difference can be explained when looking at the daily relationships of the individual root water transport responses with the mean stem sap flux density (Fig. 8b), which are averaged in Fig. 8a. The similar response observed in both roots on the day before irrigation (black symbols in Fig. 8b; part of homogeneous dry data in Fig. 8a) was suddenly disturbed by the localized irrigation event (coloured symbols in Fig. 8b; averaged as heterogeneous dry data in Fig. 8a). This event caused the daily regression lines for both roots to clearly deviate (Fig. 8b): for the southern root they were above and at an angle with the 1 : 1 line, whereas, for the northern root, they were below and more parallel to the 1 : 1 line. As time progressed, the regression lines returned to the original situation. A clear difference was also observed in the daily range of sap flux density values when comparing both roots: for the southern root, an initial increase was observed followed by a gradual decrease, whereas, for the northern root, the range remained more or less the same. As already mentioned, the different behaviour of the southern and northern roots during the heterogeneously dry period (Fig. 8b), when averaged, resulted in the same behaviour as the homogeneously dry period (Fig. 8a). This illustrated that Douglas-fir was unable to utilize the available water supplied during localized irrigation to increase its mean stem sap flux density (as illustrated by the slope of the homogeneous wet period in Fig. 8a).

Sap flux density versus PET

In addition to the relationships within the tree, the relationship of the daily total of mean stem sap flux density with the daily total of PET was investigated for the three distinct experimental periods described above (Fig. 9). A clear linear relationship existed between PET and the mean stem sap flux density during the homogeneously wet period. This was expected as, during this period, a sufficient amount of water was available to the entire root system, allowing the tree to fulfil the atmospheric demand for water. Similarly, a smaller slope of the relationship between PET and the mean sap flux density was expected during the homogeneously dry period because of stomatal regulation or high water potential gradients, limiting the transpirational water loss. However, during the heterogeneously dry period, no relationship was detected, again indicating a limitation in leaf water loss. This was surprising as an amount of water, equal to that received during the rain event, was available to the tree. Contrary to the homogeneously wet period, only one part of the root system had access to the irrigation water. For this amount of water, there was competition between the nonirrigated dry roots and the above-ground tree parts.

Figure 9.

 Correlation between the daily totals of the mean (i.e. average of northern and southern sides) stem sap flux density and potential evapotranspiration (PET) during the homogeneously dry (July 30–August 5, squares), heterogeneously dry (August 6–August 14, triangles) and homogeneously wet (August 15-August 21, circles) experimental periods.

In this particular situation, the roots proved to have the upper hand in the competition, resulting in HR of the water. As a result of this situation, the tree continued to dry out, explaining the lower daily totals of mean stem sap flux density compared with those measured during the homogeneously dry period for equal PET values.


How do the water relations in Douglas-fir cope with drought stress?

From the literature, it is clear that Douglas-fir is highly drought stress adapted. Large Douglas-fir trees may store c. 20% of the daily total water use in their stems (Čermak et al., 2007). In two 110–120-year-old Douglas-fir trees, the difference between minimum and maximum relative bole water content in autumn and midsummer, respectively, deviated by only 5% (Beedlow et al., 2007). This might be explained by the known fact that Douglas-fir is able to lift water from deep soil layers. This was confirmed by Warren et al. (2005), who reported that, as summer drought progressed, water extraction shifted to deeper soil layers, and recharge of the shallower layers through HR approached 0.15 mm d−1. Brooks et al. (2006) reported that, under maximum HR rates, redistributed water by Douglas-fir replenished c. 40% of the water depleted from the upper soil on a daily basis. These authors used root sap flow measurements on one side of the tree to demonstrate the existence of the hydraulic lift of water from deeper to shallower soil layers. However, stem-mediated HR was not considered.

The concept of horizontal HR is known in soil research (Brooks et al., 2002, 2006; Smart et al., 2005), and was validated using sap flow measurements in Norway spruce roots (Nadezhdina et al., 2006). The possible involvement of the stem in HR has remained unclear, however, and, indeed, inspired us to perform the present study.

Stem-mediated HR was first introduced by Burgess & Bleby (2006) in eucalyptus trees, but it was not investigated further whether other species would also exhibit this phenomenon. Our work directly confirms that Douglas-fir can redistribute water horizontally by its roots, but also that the stem is involved in this process (Figs 5 and 7). Given the positioning of the HFD sensors, we can confirm that the stem contributed to HR to at least a height of 1.5 m above ground level. It is, however, logical to assume that higher parts of the stem will also be involved to some extent. Based on their measurements, Burgess & Bleby (2006) suggested that tissues at a height of 4 m in eucalyptus would still be partly involved in HR. Presumably, the portion of the stem involved in HR will be tree and species specific, but will also depend on the amount of water supplied and the level of the prevailing drought stress.

The experiment with localized irrigation, performed in the present study, showed that stem-mediated HR occurs in Douglas-fir. This could have significant ecological consequences. HR observed in Douglas-fir allowed the tree to better deal with drought, even when only a limited part of the root system had access to water. The situation created by the artificial manipulation experiment might also occur in reality in cases in which only a (small) part of the root system has access to water (e.g. a creek or ground water). Through HR, not only will water be supplied to the above-ground tree parts, but also to the whole root system, allowing the tree to take advantage of nutrient uptake from a larger area around the tree at the same time. More studies with additional replicates are, however, necessary to better quantify the impact of stem-mediated HR on tree and stand water budgets.

What is the mechanism behind stem-mediated HR in Douglas-fir?

The stem-mediated HR for eucalyptus species detected by Burgess & Bleby (2006) is now also confirmed for Douglas-fir. Based on our results, we can conclude that stem-mediated HR is indeed the result of a complex interplay of competing water potential gradients (Fig. 2b). During the day, an upward flow of water is observed in the entire tree, caused by the water potential gradient between the above-ground tree parts and the soil. In the evening, the above-ground water potential gradient progressively diminishes as the transpiration ceases. Therefore, the tension in the dry root becomes an increasingly important driving force for the water flow coming from the wet root. At the moment when the tensions of the above-ground and below-ground tree parts become comparable, water tends to flow along the path with the lowest resistance. In our case, this seems to be the stem base in connection with the inner root xylem, where the first reverse flow is detected (Figs 4a and 7a). As the water potential gradient of the above-ground tree part further diminishes, the tension in the dry root prevails, gradually exerting its influence on the stem water of the dry side of the tree, resulting in a reversal of flow from the bottom up (c. 5 h to reach breast height, Figs 4b and 7b). However, in order for reverse flow to exist, a source providing sufficient amounts of water should be available. In our case, this water was supplied by an upward flow on the wet side of the tree. As was also concluded by Burgess & Bleby (2006), a substantial axial flow in the stem is a prerequisite for stem-mediated HR. The axially transported water at the wet tree side becomes available to the downstream flow through circumferential water movement around the nonconducting xylem (heartwood). This circumferential water movement becomes possible whenever the pit resistance is overcome, allowing the intervessel exchange of water (Domec et al., 2006; Kitin et al., 2009).

Our measurements demonstrate that the situation can be even more complex: within the northern root, two opposite water flow directions existed simultaneously, indicating radial sectoring (Fig. 7a). Although the inner xylem showed the expected reverse flow behaviour, the water in the outer xylem flowed upwards, suggesting a flow into one of the adjacent dry roots (Fig. 2b).

Although some evidence suggests that HR might be limited, if not cancelled, when night transpiration occurs as a result of nonzero VPD, our study demonstrated that significant HR existed despite a VPD different from zero during the night (Fig. 3). Apparently, the canopy acting as a sink for water as a result of night-time transpiration did not create a sufficiently strong water potential gradient that could compete with that in the soil, hence resulting in reverse flow during the night (Figs 5 and 7).

How does stem-mediated HR relate to partial root zone drying?

Our results show that the system under study has a high tendency to maintain a hydraulically homeostatic situation by transporting water passively from the wet to the dry part of the root system (Figs 8 and 9). Indeed, the sudden disturbance in the opposing root’s sap flux density caused by localized irrigation gradually returned to the initial state within c. 9 d (Fig. 8b). The aim of the system to return to a homeostatic situation also became apparent from the absence of a linear relationship between PET and the sap flux density (Fig. 9) despite the availability of sufficient water on one side of the tree.

These observations raise important questions about the partial root zone drying (PRD) technique commonly used in fruit tree irrigation practices. During PRD, water is distributed alternately to the root system: one part is being irrigated, whilst the other part is left to dry out. This alternating application of water is believed to invoke the so-called root-to-shoot abscisic acid signalling mechanism, which decreases plant water use by inducing partial stomatal closure and increasing, as such, water use efficiency (e.g. Dry & Loveys, 1999; Fernandez et al., 2006; Dodd, 2007; Dzikiti et al., 2008). The nature of the signalling mechanism, however, remains an open research question. An additional hypothesis to explain the reduced transpiration might also be stem-mediated HR, which redistributes water from the wet to the dry part of the root system. This causes less water to be available for transpiration. Moreover, as a result of this redistribution, the effect of PRD is also partially cancelled by re-wetting the dry part of the root system, and hence reducing the strength of the abscisic acid signalling mechanism. Of course, we cannot conclude from our case study with Douglas-fir that this is a common phenomenon for all tree species. Nevertheless, the presence of stem-mediated HR should be considered when using and studying PRD, especially in fruit trees.


This work was partially supported by the Czech national projects NAZV QG60063 and MSM 6215648902 and by the Bilateral Exchange Programme of the Flemish Community and the Czech Republic (UA-BOF BWS-2004 and BWS-2006). We also wish to thank the Research Foundation-Flanders (FWO) for the Postdoctoral Fellow funding granted to the corresponding author.