Concurrent measurements of change in the bark and xylem diameters of trees reveal a phloem-generated turgor signal


Author for correspondence:

Maurizio Mencuccini

Tel: +34 93 586 8474



  • Currently, phloem transport in plants under field conditions is not well understood. This is largely the result of the lack of techniques suitable for the measurement of the physiological properties of phloem.
  • We present a model that interprets the changes in xylem diameter and live bark thickness and separates the components responsible for such changes. We test the predictions from this model on data from three mature Scots pine trees in Finland. The model separates the live bark thickness variations caused by bark water capacitance from a residual signal interpreted to indicate the turgor changes in the bark.
  • The predictions from the model are consistent with processes related to phloem transport. At the diurnal scale, this signal is related to patterns of photosynthetic activity and phloem loading. At the seasonal scale, bark turgor showed rapid changes during two droughts and after two rainfall events, consistent with physiological predictions. Daily cumulative totals of this turgor term were related to daily cumulative totals of canopy photosynthesis. Finally, the model parameter representing radial hydraulic conductance between phloem and xylem showed a temperature dependence consistent with the temperature-driven changes in water viscosity.
  • We propose that this model has potential for the continuous field monitoring of tree phloem function.


It has been known for some time that small but easily detectable changes in tree stem and branch diameter reflect variations in plant water status (Klepper et al., 1971; Simonneau et al., 1993; Zweifel et al., 2001; Steppe et al., 2006), as the volumes of living cells (Nobel, 1991) and xylem conduits (Irvine & Grace, 1997) vary according to the pressures to which they are subjected. Cambial growth can also be clearly distinguished in the stem diameter measurements (Génard et al., 2001; Steppe et al., 2006). The changes in cell and xylem conduit pressure arise mainly from changes in the transpiration rate, which determine the water potential of the xylem (Perämäki et al., 2001). The water potentials in the extra-xylary tissues largely follow the water potential variation of the xylem (Sevanto et al., 2002), as water potentials tend to equilibrate within the tree with time scales of minutes to several hours. The elastic portion of the bark, the functional phloem and the cambium are the main extra-xylary tissues that contribute to the whole-tree stem diameter variations (De Schepper et al., 2012).

By assuming a linear relationship between tissue volume and pressure, that is, Hooke's law (Nobel, 1991), the changes in inner bark (cambium and phloem) and xylem pressure can be calculated from the measured xylem diameter changes (Perämäki et al., 2001). If thickness variations of the inner bark were driven only by changes in xylem water potential, the inner bark thickness, discarding cambial growth, would be expected to follow changes in xylem diameter in a predictable manner (Sevanto et al., 2003). We should then be able to predict its movements on the basis of the hydraulic coupling between the xylem and the inner bark and the elastic properties of the tissues (Sevanto et al., 2011). This prediction is largely confirmed by measurements (Zweifel et al., 2001; Sevanto et al., 2003; Steppe et al., 2006), but changes in xylem water potential alone cannot explain completely the diurnal changes in stem diameter (Sevanto et al., 2002, 2003, 2011). This has led to the hypothesis that phloem sugar dynamics might also influence inner bark thickness variations (Sevanto et al., 2003). However, no theoretical framework exists to infer sugar dynamics in the phloem region from thickness change measurements of the inner bark.

In this study, we develop a model that allows the determination of this putative signal generated in the phloem region, and use this model to extract the turgor/osmotic signal from simultaneous measurements of xylem and inner bark thickness variations. We link the inferred changes in this signal throughout one growing season with the measured gross primary production (GPP) determined by concurrent eddy covariance measurements. Finally, we test whether a link between GPP, tree radial growth, sugar loading and unloading, and phloem translocation can be established in our boreal pine forest.


A diagram illustrating the main geometrical features of our system is given in Fig. 1(b,c), where Dx and Db refer to the measured xylem diameters and bark thicknesses, respectively, at times t = 1 and t = 2, and dDx and dDb refer to the corresponding changes over time. To avoid confusion, we refer to ‘stem diameter changes to indicate measurements that incorporate both inner bark and xylem, whereas we use ‘inner bark thickness changes’ to indicate measurements derived from the difference between stem and xylem diameter changes. We also define the quantities inline image inline image. They are obtained by subtracting from dDx and dDb the corresponding values inline image and inline image, respectively, measured at a reference pressure.

Figure 1.

Diagram illustrating the main geometrical features of our measurement system. (a) Representation of a stylized tree, illustrating the two heights at which the two pairs of LVDTs (linear variable displacement transducers) were placed. (b) Cross-sectional view of the geometrical arrangement of the system. The metal frame (depicted in grey) is held around the trunk by means of a metal plate (not depicted), screwed in position above the section shown in the figure (see figure in Sevanto et al. (2005) for a diagram illustrating this arrangement). The system coordinates are defined by the black screw opposite to the position of the LVDT (left-hand side of figure), which provides a fixed starting point for the measurements by being fixed directly into the tree xylem. A similar screw is also employed below one of the LVDTs on the right-hand side to obtain direct contact with the xylem tissue without any further debarking damage. On its side, a second LVDT is placed in direct contact with the inner bark after carefully cutting away the dead bark. The bark diameter is obtained by difference between the two readings of xylem and xylem plus bark diameter made by the two LVDTs. (c) Readings are carried out for subsequent intervals. In this case, at time t = 2, both bark and xylem have shrunk to some degree. The LVDTs provide directly the reading of the diameter and thickness changes dDx and dDb. The final quantities employed in the analysis (cf. main text), however, are ΔDx and ΔDb. They are obtained by subtracting the diameter changes measured on the midnight of the first day of measurement (i.e. inline image and inline image) from all subsequent measurements in the time series. See text for further details.

Under conditions of slow, elastic changes of small magnitude, the changes in xylem diameter dDx over a time interval dt reflect changes in xylem water pressure dPx (MPa) according to Hooke's law (Irvine & Grace, 1997; Perämäki et al., 2001):

display math(Eqn 1)

(inline image, xylem diameter (m) at a reference pressure; Er,x (MPa), radial elastic modulus of the xylem tissue). The same can be written for the inner bark thickness Db:

display math(Eqn 2)

(dPb (MPa), change in the pressure (tissue-averaged turgor pressure) of the live inner bark; inline image (m), inner bark thickness at a reference pressure; Er,b (MPa), radial elastic modulus of the inner bark tissue). The inner bark tends towards water potential equilibrium with the xylem with a time scale dependent on the radial resistance between the two (Sevanto et al., 2011). Here, we treated inline image and inline image as two model parameters, measured at a reference pressure, although, strictly speaking, Hooke's law requires the use of the actual diameter/thickness at each time interval in the numerator of Eqns 1 and 2. In practice, this is not an issue, as the absolute changes caused by the shrinkage and swelling are extremely small relative to the xylem diameter and bark thickness (cf. Fig. 1), and the approximation of constant values holds.

Assuming a negligible osmotic content in the xylem, the water flux J (m3 s−1) between the xylem and phloem can be written as:

display math(Eqn 3)

(L (m MPa−1 s−1), hydraulic conductance of the cross-sectional area A (m2) of contact between the xylem and phloem through which radial water exchange takes place; Π (MPa), osmotic pressure of the inner bark).

We modelled the inner bark tissue as if it acted only as a dynamic water reservoir for the xylem with constant osmotic content, and then used the residuals from the predictions of this model against the observed bark thickness changes to obtain an estimate of the variations in the osmotic content. In other words, we calculated how the inner bark thickness would behave if no changes in bark osmotic concentration (which are primarily caused by phloem transport) took place over time. The difference between the inner bark thickness measured and that which was predicted on this basis should contain a turgor signal caused by the changes in osmotic concentrations taking place in the inner bark. This approach avoids the problem of having to formulate hypotheses on the unknown patterns of phloem osmotic transport and focuses on modelling the known processes.

An additional potential problem is that the bark will also change its thickness passively, simply as a result of the geometrical stretching caused by xylem diameter changes. This effect will occur independently and simultaneously to the changes in turgor and osmotic relationships of the bark, as a consequence of the changes in xylem diameter. For example, if the xylem swells by a certain amount, the bark thickness will be expected to shrink correspondingly to keep the bark volume constant (cf. diagram in Fig. S1). However, the magnitude of this geometrical effect is small (in our case, it is between 2% and 8% of the measured xylem diameter changes) and can therefore be discarded. The demonstration is given in Supporting Information Notes S1.

The expression of the bark thickness changes as a function of the pressure of the xylem and the turgor and osmotic pressures of bark leads to the following expression for inline image (see Notes S2 for derivation):

display math(Eqn 4)


display math

γ is a constant representing the bark thickness change that occurs at the reference time (which is taken to be the first midnight of measurements) as a consequence of water potential disequilibrium between bark and xylem, Π* and inline image are the reference osmotic pressure and the reference bark volume (also taken at the first midnight of measurement), respectively, and ΔDx and ΔDb are the changes in xylem diameter and bark thickness, respectively, relative to the reference measurement at the first midnight, as defined above. The reference state at the first midnight of the time series is used for practical convenience and has no bearing on model fitting or on the interpretation of the parameters from Eqn 4, and its choice has only a minute effect on the calculated values of the biophysical parameters (cf. Notes S2). In addition, τ is a time constant (i.e. α is measured in s−1), which reflects the velocity of the radial pressure propagation between xylem and inner bark. By contrast, β can be interpreted as the ratio of the changes in bark thickness to xylem diameter for a given change in xylem water potential. Parameters α and β are positive, because, in a realistic physiological range, inline image > 0. In addition, the value of parameter β should be much higher at the top of a tree, because it depends on the ratio of the inner bark thickness to the sapwood xylem diameter, which is much larger there.

In practice, the model fitting is performed using measurements collected over a finite time period (in our case, 30 min), which means that the fitting estimates parameters inline image and inline image, averages over 30-min periods and with corresponding different units. The equivalent parameters from Eqn 4 can be easily derived. The bark thickness at time (t + dt) is then estimated from the previous thickness at time (t) and the predicted thickness change (cf. Notes S3 for more details on the fitting procedure):

display math(Eqn 5)

where the caret is employed to emphasize the nature of the predicted variable for the bark thickness. Therefore, one predicts inline image, at the end of the first interval after midnight, from the measured ΔDb(t = 0) and ΔDx(t = 0) (both set to zero at t = 0) using the predicted values of inline image, β and inline image. Once inline image has been predicted, predictions for the subsequent 30-min intervals are obtained by iteration of the same procedure using the predicted bark thickness, inline image and β. The difference between the measured (ΔDb) and the predicted (inline image) change in inner bark thickness gives the change in the inner bark thickness caused by the change in turgor of the bark region that is not explained by the xylem water potential:

display math(Eqn 6)

By differentiating with respect to the measured values in Eqn 6, one confounds the estimate of the turgor term with the residual error in the time series. Assuming the time series contains only uncorrelated random errors (i.e. white noise), Eqn 6 can be used to obtain the changes in the osmotic pressure of the inner bark which are caused only by the change in the number of moles of solutes (i.e. ΔΠsol):

display math(Eqn 7)

(ns, number of moles of solute that are being diluted into volume V; R and T, molar gas constant (8.314 J K−1 mol−1) and temperature (K), respectively). If α is obtained by fitting the model to the data, the radial hydraulic conductance L can be calculated:

display math(Eqn 8)

Assuming that information is available on the xylem elastic modulus and diameters, one can invert Eqn 4 for β and substitute into Eqn 8, that is:

display math

The confidence intervals in the estimates of radial hydraulic conductance can be obtained by propagating the error terms contained in all the measured variables and coefficients employed in the calculations using standard derivatives calculus and appropriate variance estimates (Hughes & Hase, 2010).

Empirical field measurements

Simultaneous xylem and inner bark thickness measurements were conducted from 28 June to 4 October 2006 at the Helsinki University research station in Hyytiälä, southern Finland on three 47-yr old, 15-m-tall, Scots pine (Pinus sylvestris L.) trees. A general analysis of the behaviour of earlier time series from the same forest can be found in Sevanto et al. (2002, 2003). Here, we concentrate on testing the proposed theory.

The measurement system on each tree consisted of two pairs of pen-like linear displacement differential transducers (LVDT; Solartron AX/5.0/S, Solartron Inc., Bognor Regis, West Sussex, UK) attached next to each other onto two rectangular metal frames mounted around the stem on the opposite side from the transducers at two different heights (1.5 and 10 m; cf. Sevanto et al., 2002, 2005 for further details and additional drawings of the system). For each pair, one of the sensors was mounted onto the smooth inner live bark, whereas the other was mounted onto xylem tissue, each pair providing diameter data every 30 min. The xylem sensor yielded direct measurements of xylem shrinkage and swelling, whereas the net inner bark thickness change was calculated by subtracting the xylem diameter changes from the stem diameter changes for each time interval (cf. Fig. 1). Air, sapwood (at 1 cm depth into the wood) and metal frame temperatures were available from copper-constantan thermocouples, and the thermal expansion properties of wood and frame were considered (cf. Sevanto et al., 2005). The data were collected once per minute and averaged over 30 min. Stem diameters were c. 15 cm (at 1.5 m height) and c. 5 cm (at 10 m height), and the inner bark thickness was c. 2 mm at both heights for the three trees.

Before subsequent analyses, the time series of inner bark thickness was de-trended to subtract the seasonal signal of cambial growth (Zweifel et al., 2001) using piecewise linear regression between visually identified break points in growth. This procedure gave acceptable results for our time series, although more sophisticated approaches are possible (Zweifel et al., 2001).

Evapotranspiration (ET) was measured and ecosystem GPP (an estimator for canopy photosynthesis) and total ecosystem respiration (TER) were inferred from half-hourly measurements of net ecosystem exchange (NEE) measured with a three-dimensional sonic anemometer (R3IA; Gill Instruments Ltd, Lymington, UK) and a closed-path CO2/H2O infrared gas analyser (LI6262; Li-Cor Inc., Lincoln, NE, USA) installed above the stand at a height of 23 m. The instrumentation is documented in more detail in Vesala et al. (2005). Gap filling, quality control of ET, NEE values and calculation of GPP followed standardized protocols (Reichstein et al., 2005; Papale et al., 2006).

Statistical analyses

Equation 4 was modelled using nonlinear mixed effects modelling (Pinheiro & Bates, 2000; Gelman & Hill, 2007) employing the NLME library (version 3.1-96; Pinheiro & Bates, 2000) in R 2.9.0. This allowed the representation of parameter α of Eqn 4 as a random variable estimated separately for each day of the year, which allowed the examination of the possibility that the radial hydraulic conductance varied from day to day (Steppe et al., 2011) (see Notes S3 for further details on model structure and results of preliminary analyses).

The presence of a significant diurnal cycle in the predicted turgor term inline image was tested by computing estimates of the autocorrelation function for different lag intervals of inline image, using the function acf in R. Daily totals of the calculated inline image term were also computed and regressed against daily totals of canopy GPP. Finally, we investigated, by regression analysis, how the day-to-day variability in α related to mean daily temperature T (as expected based on the dependence of α on radial conductance L and water viscosity), and how cumulative daily totals of inline image related to cumulative daily totals of GPP. All analyses were conducted in R 2.9.0 (R Core Team, 2012).


As no alternative empirical method is available to determine the turgor properties in the field at the spatial and temporal scales of our measurements, we validated our approach using model simulations. We employed the finite-element model of Hölttä et al. (2006) to perform an initial validation of the predictions of the model presented here. The more complex finite-element model was initialized from assumed daily cycles of canopy photosynthesis and transpiration. It predicted the magnitude of the changes occurring in xylem and bark thickness and the likely timing of the transfer of sucrose and of pressure–concentration waves (sensu Thompson & Holbrook, 2004) from the canopy to the two points of measurement down the trunk. In the finite-element model, we also varied the numerical values of L, the radial hydraulic conductance between xylem and bark, and of Er,b, the elastic modulus of the bark. We then ran this model on the subsequent time series of bark and xylem thicknesses and verified whether we could retrieve the original parameter values.


Performance of the model

Plots of mean hourly values of measured xylem and the measured and predicted inner bark thicknesses for the upper and lower positions of tree 3 are given in Fig. 2(a,b). On average, at the upper position, the inner bark thickness (black triangles; Fig. 2a) followed the dynamics of the xylem diameter (blue circles) with significant lags, whereas, at the lower position, the two time series followed each other much more closely (Fig. 2b). The predicted changes in inner bark thickness based on the model of Eqn 4 (green triangles) showed a time course very close, but not identical, to the observed thickness changes at the bottom position, whereas it was intermediate between the observed bark and xylem values at the upper position.

Figure 2.

Average values (± SE) of the measured xylem diameter, measured inner bark thickness, predicted bark thickness and modelled phloem-induced turgor changes in the inner bark thickness for tree 3 (Scots pine, Pinus sylvestris) during the period June to September 2006 at Hyytiälä, southern Finland. Values are reported for (a) the upper position (height of 10 m) and (b) the lower position (height of 1.5 m). A small horizontal jitter was applied to some of the time series to reduce overlap. Note that the phloem-induced turgor changes occur in addition to the water potential changes causing the fluctuations in bark thickness. Actual phloem turgor can be obtained by integrating the two signals.

Overall, all components of the final model were highly significant (Table 1) and the model closely replicated the diurnal and seasonal time courses observed in the inner bark thickness data (not shown). More interestingly, without fitting the autoregressive (AR) terms, the residuals from the model showed significant temporal dependence for at least two time periods. The inclusion of AR terms of order 2 improved significantly the models at both the upper and lower stem positions for all trees (always < 0.0001) and eliminated the temporal autocorrelation among the measurements, suggesting unbiased parameter estimation.

Table 1. Statistics of the nonlinear mixed effects models for the regression of bark thickness at time (t + 1) against bark thickness and xylem diameter at time (t) for both the upper and lower position data
Lower position
  Fixed effectsAutocorrelation structureRandom effects
Tree number  inline image β ϕ1ϕ2Standard deviation αResidual error
1Coefficients0.482 ± 0.0470.900 ± 0.0400.770.170.2230.001
1P values< 0.0001< 0.0001< 0.0001< 0.00010.0001 
2Coefficients0.065 ± 0.0190.311 ± 0.0510.600.050.1920.02
2P values< 0.01< 0.0001< 0.0001< 0.00010.0001 
3Coefficients0.162 ± 0.0160.440 ± 0.0100.650.320.0650.0004
3P values< 0.0001< 0.0001< 0.0001< 0.00010.0001 
Upper position
  Fixed effectsAutocorrelation structureRandom effects
Tree number  inline image β ϕ1ϕ2Standard deviation αResidual error
  1. Data are given separately for the three Scots pine (Pinus sylvestris) study trees. All independent variables were centred before analysis. Significance levels for fixed factors refer to the significance of the χ2 test in an analysis of deviance of the model with the factor vs the model without the factor.

1Coefficients0.596 ± 0.0641.603 ± 0.0670.790.180.3000.001
1P values< 0.0001< 0.0001< 0.0001< 0.00010.0001 
2Coefficients0.391 ± 0.0252.178 ± 0.0410.760.210.1560.002
2P values< 0.0001< 0.0001< 0.0001< 0.00010.0001 
3Coefficients0.359 ± 0.0101.958 ± 0.0300.630.360.0650.003
3P values< 0.0001< 0.0001< 0.0001< 0.00010.0001 

The values of the two parameters α and β were highly significantly different from zero at both stem positions (Table 1, always at least < 0.01 for the three trees). As predicted by theory, β (the ratio of xylem to bark elastic moduli divided by the respective diameters) was substantially higher at the upper position (with values ranging from 1.6 to 2.2), whereas values at the lower position varied between 0.3 and 0.9, and the average difference between the two positions was a factor of c. 3.6 (Table 1). The values of α (the inverse of the time constant for pressure propagation from xylem to bark), however, were much more similar at the two positions, varying between 0.065 and 0.596 per 30 min−1, or 3.61 × 10−5 to 3.31 × 10−4 s−1. For both stem positions and for all trees, the inclusion of random variability from day to day in α increased substantially the model log-likelihood (Table 1). As explained in Notes S3, there was no clear indication that short-term variations in α and β were significant within individual days (mornings compared with afternoons and night-times).

The validation exercise showed that our model retrieved model parameters, on average, to within 10–15% of those imposed by the finite-element model, giving a very good fit to the modelled data. Closer agreement was possible, but it depended crucially on whether the changes in bark thickness were caused by changes in solute concentrations (which is the process represented in our regression model) or also by fast-travelling pressure waves (which are not accounted for). The finite-element simulations showed that local changes in trunk turgor (and therefore thickness) could occur in response to turgor changes at the tree top and before local changes in sucrose concentration.

Diurnal dynamics

Across the whole season, the calculated inline image term for the three trees showed significant temporal autocorrelation across the 24-h period at both positions (Fig. 3), revealing a clear diurnal pattern. This indicates that the residuals from the model were not pure white noise and probably contained biologically relevant information. As a consequence, a significant additional term (the calculated inline image based on Eqn 6), characterized by a diurnal cycle (labelled ‘Predicted turgor’ in Fig. 2, red squares), was present in the time series at both heights for all three trees.

Figure 3.

Autocorrelation functions of the modelled phloem-induced changes in the inner bark thickness for the upper (blue dashed line) and lower (continuous red line) positions for tree 3 (Scots pine, Pinus sylvestris). A significant autocorrelation among the residuals is apparent, suggesting that the residuals from the fit are not white noise. The 5% confidence interval bands are given as dotted lines around the zero line (i.e. values outside the confidence bands are all significant).

The mean diurnal course for GPP during the 15 wk of the study period generally followed a bell-shaped curve with a gradual morning increase and afternoon decline (Fig. S2), with the exception of weeks 6 and 7, characterized by two drought events (cf. subsequent section), when an early morning peak was seen with a prolonged midday and afternoon decline. During these two drought events, soil water potential decreased from close to zero to −2 MPa at the measurement site (Duursma et al., 2008).

At the upper position, inline image reached its peak during the early morning from weeks 1–6 (Fig. S3 for tree 3), with the maximum gradually moving towards midday in the second part of the growing season. At the lower position, the inline image term reached a daily minimum at around midday during the early part of the season or in the late afternoon during the latter part. At the upper position, instead, a minimum was observed at c. 22:00 h and a recovery began almost immediately after the minimum and continued throughout the night.

The amplitude of the diurnal cycle also varied significantly over time, with generally much larger daily amplitudes during the early part of the growing season and reduced fluctuations during the latter part (weeks 11–15). Concomitant with reductions in canopy GPP, the amplitude of the inline image term became smaller during weeks 6 and 7 (days of the year 31–60), and two diurnal peaks were observed during some days. At both positions, the elevation of the inline image term decreased systematically throughout the season, and corresponding to the droughts of weeks 6 and 7, although, at the lower position, the elevation changes of the inline image line were smaller, in keeping with the lower magnitude of the values there.

Seasonal dynamics

The parameter α varied significantly across days at both stem positions for the three trees (Table 1, Fig. 4a,b for tree 3). The variability in α followed the daily temperature patterns, with highly significant exponential regressions found at both stem positions (Fig. 5a,b: < 0.001 in both cases, albeit with low R2, between 0.10 and 0.15). The slopes of the exponential regressions yielded values of Q10 (i.e. the proportional increase in reaction rate for a 10°C warming) of 1.15 and 1.86, for the upper and lower positions, respectively, which are relatively close to the Q10 of water viscosity, that is, c. 1.20, given the scatter in the relationship (Haynes, 2010).

Figure 4.

Plots of the daily values (± confidence intervals) of the random parameter inline image for tree 3 (Scots pine, Pinus sylvestris). The parameter inline image (which is calculated over 30-min periods) was allowed to vary freely from day to day, that is, it was treated as a random variable (see text for further details). (a) Upper position, blue symbols; (b) lower position, red symbols. The values are given for 30-min intervals.

Figure 5.

Relationships between daily values of parameter inline image and mean daily air temperature at the site for tree 3 (Scots pine, Pinus sylvestris). The lines represent exponential regressions through the data (< 0.001 in both cases). The slopes of the two lines produced Q10 values of 1.20 and 1.08, respectively. (a) Upper position, blue symbols; (b) lower position, red symbols.

The measured morphological parameters were employed, together with the estimated parameters α and β, to calculate the values of the radial hydraulic conductance and of inline image for all three trees at the two heights (Table 2). The radial hydraulic conductance varied between 3.9 × 10−8 and 5.3 × 10−8 m MPa−1 s−1 and between 0.3 × 10−8 and 7.2 × 10−8 m MPa−1 s−1 at the upper and lower positions, respectively. Conversely, inline image varied between 9.2 and 12.5 MPa and between 7.4 and 21.4 MPa at the upper and lower positions, respectively.

Table 2. Calculated means (± SE) of the phloem physiological parameters for the three measured Scots pine (Pinus sylvestris) trees inferred from the modelling of bark thickness and xylem diameter changes
Tree numberPosition in the treeRadial hydraulic conductance L (m MPa−1 s−1)Bark elastic modulus corrected for osmotic pressure (Eb – Π*) (MPa)
1Upper(5.31 ± 0.41) × 10−812.48 ± 1.12
2Upper(4.73 ± 0.86) × 10−89.18 ± 0.85
3Upper(3.91 ± 0.40) × 10−810.04 ± 0.90
1Lower(7.23 ± 0.89) × 10−87.41 ± 0.67
2Lower(0.34 ± 0.11) × 10−821.44 ± 3.90
3Lower(3.36 ± 0.89) × 10−817.59 ± 1.45

The year 2006 represented an unusual meteorological year at Hyytiälä, with two periods of no precipitation lasting c. 2 and 1 wk, respectively, separated by two small rainfall events of c. 7 mm each. This resulted in the development of a significant soil drought, and reductions in canopy ET (Fig. 6, top panel, continuous red line), GPP (second panel, dashed blue line) and TER (third panel, continuous black line) were documented in the eddy covariance data (cf. Kolari et al., 2009). The diameter data also recorded significant shrinkage in both the inner bark and xylem tissues, concomitant with the development of the drought (weeks 6–7, days of the year 31–60), with sudden swelling recorded twice at both positions, corresponding to the rain events (data not shown). inline image showed distinct drought responses at the two positions (Fig. 6, bottom panel). At the upper position, the elevation and amplitude of the diurnal changes in inline image decreased dramatically during both drought periods (data for tree 3 were partially missing for the first drought cycle – Fig. 6 – but the same trend was observed for all three trees), whereas both features increased suddenly following the rain events, concomitant with parallel increases in GPP, ET and TER. Conversely, at 1.5 m height, changes were more muted and the elevation of the inline image line, if anything, tended to increase during the drought and then decline rapidly following the rain events.

Figure 6.

Enlargement of the time courses of the changes in ecosystem-level evapotranspiration (ET, a), gross primary production (GPP, b), total ecosystem respiration (TER, b) and modelled phloem-induced changes in the inner bark thickness (c) for both positions in tree 3 (Scots pine, Pinus sylvestris) during two drought episodes in weeks 6 and 7 of the study. The black vertical bars in (c) indicate the rainfall episodes which ended the droughts. The coloured red and blue areas in (c) mark the space limited by ± 1 SE for the phloem-induced change in bark thickness at the upper and lower positions, respectively.

To investigate the relationships between photosynthesis and the estimated inline image at the seasonal scale, daily integrals of GPP and inline image at both stem positions were calculated. Daily integrals of GPP (Fig. 7, green circles) reached a peak in the early part of the season, declined significantly around day 40 and day 60, corresponding to the two drought events (represented by the shaded areas in Fig. 7), and strongly declined as the season progressed. A similar trace was found for inline image at the upper stem position (Fig. 7, blue circles), whereas the inline image trace at the lower stem position (red circles) showed a more muted behaviour.

Figure 7.

Time courses of daily totals in canopy gross primary production (GPP) and daily totals of the phloem-induced changes in the inner bark thickness at two stem positions during the duration of the study for tree 3 (Scots pine, Pinus sylvestris). Upper position, blue circles; lower position, red circles; canopy GPP, green circles. The grey bars indicate the two periods of drought. The plot of canopy GPP has been moved upwards for clarity of presentation.

Highly significant correlations were found between the daily integrals of GPP and those of inline image at both stem positions, with significant lags present between 0 and > 30 d after GPP. At the top of the tree, the significance of these correlations reached a first peak at a lag of 1 d and a second peak at a lag of 10 d (Fig. 8, R2 = 0.68, < 0.0001, blue points), whereas, at the bottom of the tree, they peaked at a lag of 9 d and then again at a lag of 31 d (Fig. 8, R2 = 0.68, < 0.0001, red points). Interestingly, at the upper position, the days during the two drought periods tended to fall below the regression line (Fig. 8, marked with an ellipse).

Figure 8.

Relationship between daily totals of lagged canopy gross primary production (GPP) and daily totals of phloem-induced changes in inner bark thickness at the two stem positions for tree 3 (Scots pine, Pinus sylvestris). The phloem-induced changes at the upper position are plotted lagging behind GPP by 10 d, whereas the phloem-induced changes at the lower position are plotted lagging behind GPP by 31 d. Also shown are the best-fit regression equations for the two relationships. The ellipse identifies the days of the two drought periods at the upper position. Note that the cumulative changes in bark thickness are centred on their overall mean.

Stem diameter growth stopped at around week 5 for these three trees (data not shown) and growth at 10 m height was, on average, similar to that observed at 1.5 m (0.067 ± 0.021 mm and 0.078 ± 0.032 mm for the lower and upper positions, respectively).


It is generally accepted that diurnal variations in inner bark thickness can be understood on the basis of tissue elasticity and water exchange with the xylem, as outlined in our modelling section (Irvine & Grace, 1997; Perämäki et al., 2001; Zweifel et al., 2001; Steppe et al., 2006; De Schepper & Steppe, 2010). Our approach shows that the bark thickness changes are not entirely attributable to changes in xylem tension, and that an additional component must be invoked to explain the observations. We interpreted this additional signal as arising from the changes in inner bark osmotic concentration that could be related to phloem loading, transport and unloading of sugars.

In the absence of an alternative method to determine changes in phloem turgor and osmotic concentration at the time and spatial scale at which we worked, we employed a modelling validation approach. Our validation provided an important confirmation that the model presented here could retrieve the original parameter values. We also demonstrated that the discrepancy of c. 10–15% between the imposed and retrieved parameter values occurred because of fast-travelling turgor waves in the phloem. In other words, this model strictly yields the part of the bark thickness change that is not explained by changes in xylem water potential. However, phloem turgor pressure (and therefore partially also bark thickness) also responds to fast-travelling pressure waves caused by a change in sugar loading (Thompson & Holbrook, 2004; Mencuccini & Hölttä, 2010a).

Interpretation of model parameters

The parameter α (the inverse of the time constant for pressure propagation between the xylem and phloem) varied significantly over the course of the season, particularly in response to changes in air temperature (Fig. 5). It also appeared to decline significantly following the sudden rainfall events that broke the drought periods (Fig. 4). Its absolute value varied between 3.61 × 10−5 and 3.31 × 10−4 s−1, suggesting a time constant for the pressure transfer from xylem to phloem on the order of 1 to c. 5 h, much slower than the speed of sound, but as expected from previously reported values of xylem–phloem resistance (Génard et al., 2001; Sevanto et al., 2011). The radial conductance between the xylem and phloem, the main variable controlling α, is at least partly controlled by aquaporin activity (Martre et al., 2001), which might be expected to vary over space and time (Sevanto et al., 2011; Steppe et al., 2011), particularly in response to the need to regulate turgor in phloem cells following sudden perturbations of plant water potential. Our analysis, however, accounted for these dynamics, at least at the daily time scale, because α was allowed to vary freely from day to day. Conversely, we found no evidence of short-term (night-time vs mornings and afternoons) variability in α. We also failed to detect a significant temperature effect on α within a day (i.e. the coefficient for temperature was either nonsignificant or had a very small numerical value). However, when we calculated the temperature dependence of α based on its day-to-day variability in relation to the mean daily temperature, we obtained Q10 values varying between 1.15 and 1.86, with the value at the top of the tree being very close to the temperature dependence of the water viscosity.

Our measurements at the two heights were also consistent with other theoretical predictions, that is, values of β were much higher at the upper position (Table 1), by a factor of c. 3.6 on average, which corresponded well with the ratio of the xylem diameter to phloem thickness at the two heights, which averaged 3.0 (cf. Eqn 4). In addition, the values estimated for the radial hydraulic conductance L and for the bark elastic modulus for the three trees using Eqn 4 were of the same order of magnitude as found in previous studies, that is (4.65 ± 0.41) × 10−8 and (3.64 ± 1.99) × 10−8 m MPa−1 s−1 for the hydraulic conductance at the upper and lower positions, respectively (Génard et al., 2001; Sevanto et al., 2011).

Daily patterns in bark osmotic concentration

Inner bark thickness decreased during the day, primarily because of the transpirational pull from the needles, and increased during the afternoons and evenings, when water potential recovered. However, in addition to changes in xylem water potential, another signal with a diurnal rhythm could be detected (Figs 2, 3), which we interpreted as arising from the changes in the osmotic concentration of the phloem region. These phloem-induced turgor changes occur in addition to the water potential changes causing the fluctuations in bark thickness. Actual phloem turgor is given by the integration of the two signals. During the summer, the diurnal changes in the modelled inner bark osmotic concentration were found to be closely related to changes in GPP (Figs S2, S3). This is in accordance with the observed trends of diurnal variations in phloem sugar concentrations, which typically peak during the day and decrease during the night (Sharkey & Pate, 1976; Hocking 1980; Pate & Arthur, 2000; Cernusak et al., 2002). It is possible that the turgor increase detected in the mornings down the stem may have resulted from fast pressure propagation in the phloem caused by direct sucrose loading at the top. Similarly, the early decline in bark turgor at midday may have been a consequence of reduced phloem loading, perhaps caused by reductions in leaf water potential.

When the leaf water potential drops, phloem turgor also drops if phloem loading does not increase (Hölttä et al., 2006). Nikinmaa et al. (2012) showed that the observed stomatal regulation of leaf gas exchange on these very same trees acted to maximize the assimilate transport from leaves. Under such a scheme, the changes in water potential caused by transpiration and the changes in phloem loading caused by assimilation control stomatal aperture. If transpiration causes the water potential to drop, whereas assimilation does not increase sufficiently to compensate for the turgor loss via increased phloem loading, stomata should close until a new equilibrium is reached. This closure would prevent further leaf water potential declines, but would also result in decreased phloem loading and decreased turgor, as observed here.

The exact timing of this turgor peak varied during the season. In particular, inline image peaked before GPP during the period of active above-ground growth, which, in Scots pine at Hyytiälä, takes place between May and early July (see Sevanto et al., 2003). Instead, inline image typically followed GPP by c. 0.5–1.5 h after the cessation of above-ground growth and whilst root activity continued below ground during August and September (Iivonen et al., 2001; Niinistö et al., 2011).

At this daily time scale, the time lags observed in the late summer between the CO2 assimilation rate and changes in the osmotic concentration are in accordance with the time lags predicted theoretically (Thompson & Holbrook, 2004; Hölttä et al., 2006) and observed experimentally (Zimmermann, 1969; Mencuccini & Hölttä, 2010a; Heinemeyer et al., 2011; Vargas et al., 2011) for pressure concentration waves.

Phloem transport during drought

July and August 2006 were exceptionally dry at our measurement site, and clear reductions in GPP, TER and ET, resulting from stomatal closure, were observed for canopy fluxes (Kolari et al., 2009). Interesting dynamics of the inner bark sugar concentration were observed during and immediately after the drought period. Our analysis indicated that sugars tended to accumulate somewhat at the bottom of the stem during the drought, whereas the sugar concentration (and turgor) of the phloem region simultaneously decreased at the tree top (Fig. 6). We can hypothesize two reasons for this behaviour. First, osmoregulation may have played a role. As the drought progressed, sugar accumulation was probably required to prevent turgor loss, particularly in the living cells in the cambium and phloem of the roots. Closer to the top of the tree, such a strategy is required to a much lower degree, because Scots pine behaves isohydrically and stomatal closure allows the maintenance of stable water relations under most circumstances (Magnani et al., 2002). Second, the reduction in GPP and water potential during the drought must also have reduced the phloem loading in the leaf, leading to smaller osmotic driving gradients in the phloem and lower rates of translocation. This behaviour is also in accordance with the proposal that stomata act to maximize assimilate transport from leaves under drought conditions (Nikinmaa et al., 2012). If the drought-induced drop in leaf water potential cannot be matched osmotically by increased phloem loading, the stomata will close to avoid further declines in assimilate transport because of the hydrodynamic pressure drop caused by transpiration, especially so in isohydric plants (Nikinmaa et al., 2012). The increase in inline image at the base of the stem is also in accordance with the often noticed reduction in soil autotrophic respiration during drought (Tang et al., 2005; Carbone et al., 2011).

The response of the inline image term to rainfall was very rapid and may also have been caused by changes in the plant water potential and canopy GPP (in this case, mediated by the rapid propagation of the water potential wave caused by the sudden increase in soil water content). Interestingly, at the bottom of the tree, the rainfall events were immediately followed by two sudden episodes of rapid reduction in the bark inline image term, which could only have been caused by reductions in osmotic concentrations, suggesting a rapid release of solutes from the tree base following re-watering. In other words, when the rain ended the drought, sugars at the bottom of the phloem transport pathway were probably unloaded. A rapid increase in soil heterotrophic respiration after a precipitation event has been noticed in numerous studies (Law et al., 2001), with more recent studies showing that plant autotrophic respiration also responds rapidly (within 3 d) to rain pulses (Carbone et al., 2011). From the point of view of the optimization of phloem translocation, it also seems necessary for the sugars accumulated at the stem base during the drought to be disposed of externally to increase the phloem axial osmotic gradient to match the phloem sugar transport rate with the increased sugar production rate after the drought, and to avoid osmotic bottlenecks that would prevent a recovery in the sugar production rate (Nikinmaa et al., 2012). Conversely, the equally rapid increase in the inline image term at the top of the tree could have been caused by two, not mutually exclusive, processes. First, the rapid increase in leaf water potential may have elicited rapid changes in leaf phloem loading and turgor, which propagated down the phloem to the point of measurement. Second, stomatal re-opening and consequent increases in photosynthesis may also have increased phloem loading and turgor.

Seasonal patterns in phloem transport and relationships with GPP

At the seasonal scale, the osmotic signal was very well coupled with canopy photosynthesis (Fig. 7). The significant relationship between the daily totals of GPP and the daily totals of inline image (Fig. 8) is remarkable, because it suggests that the absolute value of inline image changed over time depending on the levels of GPP attained in the canopy. The fact that the first maximum correlation between these two variables was obtained with lags of 1 and 9 d at the top and bottom of the tree, respectively (as suggested by the significance level of t values for the correlation coefficient), suggests that it may take more than a week for the concentration profile of the tree to adjust to changes in canopy GPP, again supporting the distinction between fast information transmission and slow solution transport (Hölttä et al., 2006; Mencuccini & Hölttä, 2010b). The presence of a second peak of even higher significance of the correlation coefficient at longer lags (10 and 31 d, respectively) supports the existence of multiple temporal scales over which canopy photosynthesis, phloem transport and below-ground processes may be coupled (Kuzyakov & Gavrichkova, 2010; Heinemeyer et al., 2011; Vargas et al., 2011). Our model predicts that about a week is necessary for the concentration profile to achieve steady state after a sudden perturbation, but it cannot reproduce the time scale of 30 d for the peak in correlation coefficients at the bottom of the trees. It is therefore likely that additional processes not accounted for in our model, such as sucrose loading and unloading in the transport zone and dynamic changes in sucrose storage along the pathway, caused these longer delays in the equilibration of the phloem concentration profile.

The slope of the relationship between the cumulative daily canopy photosynthesis and daily totals of inline image was steeper in the upper part relative to the lower part of the tree. This may reflect the downward dampening of the sugar concentration signal from canopy GPP (Thompson & Holbrook, 2004; Mencuccini & Hölttä, 2010a). During the periods of cambial growth and drought, phloem loading and transport seemed to be less coupled to photosynthesis. This result also appears logical, as the location of the major sinks varies spatially during the season, and the sugar to starch conversion dynamics in the needles and along the transport pathway also varies (Rühr et al., 2009).


The model presented here appears to be capable of capturing one important dimension of bark thickness change which had been overlooked previously, that is, the role of phloem transport. Because the majority of inner bark thickness variations are caused by water exchange with the xylem, we were able to model the thickness variations by omitting the changes in osmotic content. Comparing our model results with the measured inner bark thickness variations revealed a diurnal signal that correlated strongly with photosynthetic production. We interpreted this additional signal in relation to phloem transport, and several lines of evidence supported this interpretation. Critically, the model assumes that all the relevant processes affecting changes in the xylem diameter and bark thickness are correctly represented in a mechanistic fashion, that is, that the discrepancies between the observations and model predictions can be attributed entirely to an osmotic-mediated process taking place in the phloem region. This seems to be justified on the basis of our analyses. The model presented here can be further improved to provide additional insights into the behaviour of the phloem transport system of trees. In particular, empirical validation can be obtained on small seedlings where continuous independent measurements are possible using nuclear magnetic imaging under laboratory conditions (Windt et al., 2006; De Schepper et al., 2012). Second, additional manipulations, such as canopy shading or phloem girdling (cf. Sevanto et al., 2011), either in the field or on small plants, should also be very informative, and should allow the determination of the extent to which bark thickness shrinkage reflects the processes of osmotic adjustment in the bark vs the purely passive adjustment to water potential changes in the xylem.


We thank Pasi Kolari for providing the eddy covariance data from the Hyytiälä field site, and the Finnish Centre of Excellence (FCoE) in Physics, Chemistry, Biology and Meteorology of Atmospheric Composition and Climate Change, Integrated Carbon Observation System (ICOS), Instrumentation for Measuring European Carbon Cycle (IMECC) and Academy of Finland project #140781 for the support for the work. T.H. was funded by the Academy of Finland (#1132561). Part of the work was carried out thanks to the Natural Environment Research Council (NERC) grant NE/I011749/1 to M.M.