In situ water absorption by tree fine roots measured in real time using miniature sap-flow gauges


  • H. Coners,

    1. Plant Ecology, Albrecht-von-Haller-Institute for Plant Sciences, University of Göttingen, Untere Karspüle 2, 37073 Göttingen, Germany
    Search for more papers by this author
  • C. Leuschner

    Corresponding author
    1. Plant Ecology, Albrecht-von-Haller-Institute for Plant Sciences, University of Göttingen, Untere Karspüle 2, 37073 Göttingen, Germany
      *Author to whom correspondence should be addressed. E-mail:
    Search for more papers by this author

*Author to whom correspondence should be addressed. E-mail:


  • 1Root water uptake is a key process in the circulation of water in forest ecosystems. Until recently, water absorption by tree fine roots could not be measured in situ in undisturbed soil.
  • 2We present a new technique that allows continuous recording of the water absorption of fine root endings in mature stands without altering soil structure, hydrology or mycorrhizal infection.
  • 3The approach combines miniature sap-flow gauges mounted on small-diameter tree roots (3–4 mm) with a complete extraction and visual surface analysis of the adjacent absorbing fine root endings. This technique yields continuous data on water absorption per fine root surface area, and allows analysis of the spatial heterogeneity of root water uptake in the rhizosphere of forests.
  • 4We present the results of laboratory and field calibration experiments with Fagus sylvatica L. roots (3–4 mm), which show a good agreement between gauge flow data and synchronous gravimetric flow measurements for flows between 2 and >50 g h−1. Gauge readings were unreliable during low flows (<2 g h−1) at night. In these periods, which cover ≈10% of daily flow, we used an empirically derived linear relationship between root temperature difference and flow.
  • 5Measurements on F. sylvatica root endings during 10 summer days showed daily water absorption maxima ranging between 0·20 (rainy days) and 0·58 mmol m−2 root surface area s−1 (bright or overcast days). The corresponding daily maxima of leaf transpiration rate were ≈10 times higher (2–4 mmol m−2 leaf area s−1).
  • 6The combination of miniature sap-flow gauges and determination of fine root surface area provides a promising tool for analysing water absorption by tree root systems in situ.


Roots and leaves are the principal interfaces in the soil–plant–atmosphere continuum of water flow in ecosystems. Over the past 30 years, techniques including porometers, leaf cuvettes and eddy covariance systems have been developed for measuring leaf and canopy transpiration in the laboratory and in the field.

Our extensive knowledge of the water-loss side of the soil–plant–atmosphere continuum contrasts sharply with our poor understanding of the uptake side. The functions of isolated roots in water uptake have been studied directly in laboratory experiments using nuclear magnetic resonance (MacFall et al. 1991). However, this approach is not suited to field application in undisturbed soils. Moreover, laboratory measurements of water flow in individual roots cannot simply be transferred to the extensive root systems of mature trees in forests, where most fine root tips are mycorrhizal, and roots occur at high densities and compete for water and nutrients.

Nevertheless, water uptake by roots is a key process in forest hydrology as it represents the only source term that may determine the water supply of the trees. Until recently, only indirect methods were available for estimating the water absorption of tree roots in situ. Recent developments in thermoelectric sap-flow measurement have allowed water flow in single large-diameter tree roots to be measured directly in the field, offering new opportunities for understanding water uptake and redistribution in forest soils (Burgess et al. 1998; Green & Clothier 1988; Green & Clothier 1995; Howard et al. 1997; Lott et al. 1996). For example, sap-flow measurements on roots in a cave allowed the contribution to water uptake by deep roots of Juniperus ashei trees to be quantified (Jackson et al. 2000). By applying this technique to coarse roots of Grevillea robusta, Smith et al. (1999) demonstrated downward siphoning of water in vertical roots of a tropical tree. These approaches concentrated on large roots with diameters of ≥2 cm that serve mainly as conducting elements, but do not absorb water. By applying a miniaturized gauge design, Senock & Leuschner (1999) measured sap flow in coarse roots of Eucalyptus saligna (diameter: 3 mm) that are closer to sites of water absorption. However, these studies have not related root sap flow to fine root surface area, and so they cannot localize water uptake precisely or quantify water absorption per root surface area.

Here we present an advanced technique that uses miniature gauges (Senock & Ham 1993; Senock & Leuschner 1999) and combines them with an optical analysis of fine root surface area. This technique monitors sap flow in tree roots, and also measures water absorption by the fine root endings in situ on a surface-area basis. In a study on the root system of mature Fagus sylvatica trees we continuously measured water absorption of single fine root endings in different parts of the root system. This paper describes the methodology used, and presents the results of several calibration experiments which tested the reliability of this method under laboratory and field conditions.


Study Site

Field experiments were conducted in an old-growth Fagus sylvatica L./Quercus petraea (Matt.) Liebl. mixed forest (site no. OB 5) in the north-west German state of Lower Saxony. The site is located at 115 m asl west of Unterlüss in the Lüneburger Heide area (52°45′ N, 10°30′ E). The stand is on highly acidic, nutrient-poor, sandy soils [pH(KCl), 2·6–2·8; Spodo-dystric Cambisols] derived from fluvioglacial sand of the penultimate ice age; the profile is covered by periglacial drift sand (Leuschner et al. 1993). The closed stand reaches a maximum height of 31 m (average 28 m) and has a density of 222 stems per hectare. About 90% of the stems belong to Fagus (90–110 years old, mean diameter 37 cm); 10% to Quercus (180–200 years old, mean diameter 52 cm). Shrub and field layers are absent. A 9–11 cm thick organic profile (Hemimor to Hemihumimor with L, Of and Oh horizons) above the mineral soil contains a large proportion of the fine roots, revealing the shallow rooting patterns in this forest (Büttner & Leuschner 1994). The density of fine roots decreases rapidly with depth from the nutrient-rich organic Of and Oh horizons (>2000 g dm m−3) to the mineral soil at 50 cm depth (<100 g dm m−3). Details on fine and coarse root distribution are given by Leuschner et al. (2001). The climate is cool-temperate and humid suboceanic (800 mm, 8 °C). During summer, rainless periods of up to 30 days occur periodically, causing low water contents in the sandy soil profile (Backes & Leuschner 2000; Leuschner 1993). The groundwater table is far below the rooting horizon.

Measurements of Root Sap Flow and Water Absorption Per Root Surface Area

All field measurements on root sap flow were conducted on roots of three 100-year-old F. sylvatica trees that were representative for the stand with respect to tree height and diameter. On the forest floor, segments of coarse roots were uncovered in small soil pits (<0·4 m wide, 0·1–0·6 m deep) by use of compressed air (0·2–0·4 MPa) to allow the mounting of miniature sap-flow gauges on the roots (Fig. 1). Ten roots, 3–4 mm in diameter at a distance of 1–2 m from the stem, were selected for study. The root sections used for gauge mounting had a brownish, suberized periderm and were not branched. During installation, any damage to the measured root or other exposed roots in the soil pit was carefully avoided. Duckboards were used to avoid trampling the soil. After installation the pits were covered with a wooden board and aluminium foil to minimize temperature fluctuations in the gauge surroundings. Sap flow was recorded continuously for periods of 7–60 days for all 10 roots in the summers of 1997 and 1998. Root growth during the measuring period was not taken into account because growth observations in soil chambers had shown branch root increments of ≤4 cm year−1, which is negligible in a total fine root surface area per root ending of about 1000 cm2 (H.C., unpublished results).

Figure 1.

Installation of miniature sap-flow gauges in the field. Root segments inside the pit were uncovered with compressed air prior to measurement, as was the soil volume with the fine root endings after measurement. The inlet shows the arrangement of thermocouples and thermopile relative to the heater position in the gauge.

After measurement, the distal part of the root with all appending branch roots, rootlets and root tips was uncovered by removing the soil with compressed air (Fig. 1). The root was harvested quantitatively by horizon (10 cm depth intervals in the organic layers and mineral soil) and transported to the laboratory in sealed plastic bags. The roots had endings distal to the gauge mounting point that were between 0·3 and 2·5 m long.

In the laboratory, the fresh root material was rinsed and separated into live and dead root mass under the stereomicroscope by colour, root elasticity and the degree of cohesion of cortex and stele (Hertel 1999; Leuschner et al. 2001). A dark cortex and stele; or a white but non-turgid cortex and stele; or the complete loss of stele and cortex, were used as indicators of root death. The live root mass was classed for diameter (0–1, 1–2, 2–5, >5 mm; Böhm 1979). An image-processing unit (WinRhizo, Régent, Quebec, Canada) was used to determine the total root surface area per diameter class. This procedure allowed us to express the measured flow (J) in terms of water uptake per surface area (g water m−2 root surface h−1 or mmol water m−2 root surface s−1) of the adjacent root endings. We also calculated root sap-flow density at the measuring point (g water m−2 root cross-sectional area h−1).

Miniature Sap-Flow Gauges

Measurements of root sap-flow rates were conducted with the constant power heat-balance method. The basic design described by Sakuratani (1981) and Baker & van Bavel (1987) was modified substantially to accommodate the small diameters and irregular shape of the tree roots, following Senock & Ham (1993); Senock & Leuschner (1999). The gauges consist of a 2 mm thick cork–neoprene jacket with two pairs of thermocouples and a thermopile mounted on it. A Kapton film-resistance heater (Heater Designs Inc., Bloomington, CA), supplied with a constant power of 0·04–0·07 W, was placed between the thermocouple junctions. For most roots selected, the flexible jacket enabled good contact of heater, thermocouples and thermopile to the root surface, which is crucial for accurate measurements. The gauges were wrapped with additional polyurethane foam insulation (5 mm) and fixed on the root with a metal clip. Solar-powered CR10 data loggers and AM416 multiplexers (Campbell, Cambridge, UK) were used to sample gauge signals every 15 s and to calculate 15 min averages.

Water-flow measurements with miniature sap-flow gauges are based on the steady-state energy balance of a heated portion of a root, defined as:

image(eqn 1)

where each term is a heat flux (W) calculated from equations describing heat flow within the system (Sakuratani 1981). Q is the energy supplied, Qr the radial conduction of energy into the insulation measured by the thermopile, Qv the apical and basal heat energy transferred by conduction along the root axis, Qf the heat energy transported by the mass flow of water, and S the rate of change in heat storage in the root segment.

The energy transported by the sap flow, Qf, is defined as:

image(eqn 2)

where J is the sap-flow rate (g s−1), c the heat capacity of the xylem water (4·187 J g−1 K−1), and Tso − Tsi the temperature difference of the water flowing into and out of the heated root segment. Tso − Tsi is recorded by the two pairs of thermocouples, assuming that the xylem temperature is represented by the root peridermal surface temperature. This assumption seems to be justified given that the ratio of the gauge heater width (10 mm) to the root diameter (3–4 mm) is large, and so large interior cross-sectional temperature gradients are unlikely (Sakuratani 1981; Senock & Ham 1995). The thermocouple junctions were placed at distances of 3 and 6 mm from the heater.

The portion of heat transported along the root axis (Qv) is the product of the root cross-sectional area (A), the thermal heat conductivity (Kr), and the axial temperature gradient (ΔT) measured with the two pairs of thermocouples (Qv = A KrΔTx). Kr has been calculated for woody and herbaceous tissues by Sakuratani (1984) and is 0·42 W m−2 for woody species.

The radial heat flux (Qr) is given by:

image(eqn 3)

where E is the thermopile output (V), and Kg the gauge conductance (W µV−1). Kg is determined at zero flow by combining equations 1 and 3, and setting Qf to zero:

image(eqn 4)

Zero flow can be obtained by excision of the root segment at the end of the measurement (Baker & van Bavel 1987). In longer measurement periods (in this study, 3–6 weeks), changes in temperature and humidity inside the pit can lead to changes in Kg so that it has to be reset more frequently. We recalculated Kg daily by assuming that sap flow in F. sylvatica trees approaches zero typically at about 03:00 h, when leaf water potentials are at their predawn maximum. On dry and windy summer nights, when sap flow was substantially larger than zero, we used the Kg values of earlier nights with sap flow minima close to zero.

Empirical data have shown that the rate of change in heat storage in the root segment (S in equation 1) typically contributes <3% to the energy balance of small-diameter stems, and can usually be ignored in measurements of small stems or roots (Senock & Ham 1993).

To obtain sap flow, J, equations 1 and 2 are combined to give:

image(eqn 5)

Sources of Measurement Errors in Miniature Sap-Flow Gauges

When applying the miniature sap-flow technique to small-diameter roots with low flows, errors may originate from the specific position of the thermocouples relative to the heater. Principally, the temperature difference Tso − Tsi above and below the heated zone can be measured in sap-flow systems in two different ways, with the thermocouple junctions arranged either equidistantly or asymmetrically from the heater coil. In the latter system, one junction is placed directly at the end of the heater, while the second is positioned at a safe distance upstream from the heater so that it will record the baseline temperature of the unheated xylem. Equation 5 was originally based on this thermocouple arrangement (Sakuratani 1981), where zero or low night-flow rates maximize Tso − Tsi. Increasing sap-flow rates lead to lower temperature differences because part of the heat is carried away with the moving xylem sap and the heated segment cools down.

The application of equation 5 to a symmetrical arrangement of the thermocouple junctions, as in the systems used by Sakuratani (1981) and in this study, has two consequences. First, Tso − Tsi is zero at zero flow because the heat is distributed homogeneously around the heater. This case, however, is not defined by equation 5. Second, no uniform negative correlation between flow rate and Tso − Tsi exists over the whole range of flows. Instead, Tso − Tsi initially increases with increasing flow rates to a maximum when the flow is sufficient to move the zone of highest temperature in the xylem toward the Tso junction. Higher flow rates are characterized by a negative correlation between J and Tso − Tsi, as predicted by the basic theory of sap-flow measurement.

Calibration of Flow Data Obtained with Miniature Gauges

We tested the accuracy of gauge measurements on small-diameter roots in two types of calibration experiments where root water uptake and sap flow were determined precisely by gravimetry or volumetry.

Laboratory Calibration Experiment

To test the accuracy of the miniature sap-flow gauges at a broad range of flow rates, we developed a laboratory setup with excised segments of F. sylvatica coarse roots (3–4 mm diameter). The root segments were sampled in the Lüneburger Heide, close to the field study plot.

A sap-flow gauge was mounted on a 10 cm long root segment and the flow rate recorded as described above. One end of the root segment was attached to a vacuum pump that sucked water through the root by applying suctions between 0 and 0·09 MPa. The opposite end of the root extracted water from a 50 cm3 reservoir placed on an analytical balance (precision 0·1 mg) to obtain flow rates through the segment. Because the root was fixed to a holder, buoyancy forces had no influence on the measurement. Balance and gauge signals were read synchronously at 1 min intervals with a Campbell CR10 data logger. To minimize temperature gradients and air turbulence, the experimental setup was enclosed in a box. Evaporative water losses from the reservoir were determined for 15 min before and after each experiment, and were subtracted from the balance data.

The laboratory calibration experiments were conducted on six root segments with flow rates in the range 0–15 g h−1 to cover the range found in the field. Special interest was paid to flows between 0 and 2 g h−1, where the largest relative measurement error was expected.

Field Calibration Experiment

On August 6 1997, three coarse roots (3–6 mm in diameter) of the 100-year-old F. sylvatica trees at the Lüneburger Heide site were selected for measuring water uptake synchronously by heat balance and volumetric techniques. The root segments were excavated and the gauges mounted as described above. After measuring night-time sap flow for determining Kg the following morning, two of the root segments were placed in a 50 cm3 basin of water inside the soil pit. Both roots were cut under water, and the water absorption of the decapitated root endings was measured for 5 h by refilling the basin to a marker at 15 min intervals. The third root served as a control for sap-flow measurement with the heat-balance method on intact roots. This experiment gave synchronous recordings of root sap flow by heat-balance and volumetric techniques for flow rates between 3 and 50 g h−1.


Accuracy of Gauge Data at Medium to High Flow Rates

Flow rates <2 g h−1 can result in erroneous values in sap flow if equation 5 is used. As these errors are caused mainly by inadequate interpretation of the temperature difference Tso − Tsi at low flow rates, we analysed the accuracy of the gauges separately for flows >2 and <2 g h−1.

For flows between 2 and 15 g h−1, laboratory experiments with six F. sylvatica roots 3–4 mm in diameter showed good agreement (r = 0·94) between heat-balance data and synchronous gravimetric data (Fig. 2). The slope of the linear regression indicates a slight overestimation of flow by the gauge data.

Figure 2.

Correlation between measured (gravimetric data) and calculated flow rates (miniature sap-flow gauges) in 3–4 mm thick roots of Fagus sylvatica. Data are based on six laboratory experiments with excised root segments. The regression was obtained by the major axis method.

Even better agreement was found in the field calibration experiments, where fine root endings of F. sylvatica were cut to measure flows by volumetric and heat-balance methods synchronously in situ in the soil (Fig. 3). Continuous measurements on the two roots over 5 h on August 7 1997 gave a close correlation between the results of the two independent techniques at flows between 3 and 42 g h−1 (r = 0·99). In both roots, sap flow increased more than fivefold at the onset of the experiment because roots were cut at 08.30 h and placed in a water reservoir to measure uptake volumetrically in the period 10:00–15:30 h. Both techniques recorded the subsequent relaxation of flow.

Figure 3.

Daily course of sap flow in three Fagus sylvatica roots measured either volumetrically (data points) or by flow gauges (lines) at the Lüneburger Heide site on August 7 1997. Roots 1 and 2 were cut under water at 08.30 h, and the subsequent water uptake from the reservoir was measured at 15 min intervals (root 3, uncut control). Gauge data were read every 15 s to obtain 15 min means.

Accuracy of Gauge Data at Low Flows

Sap flow in small-diameter roots of F. sylvatica reaches flow rates of 20–50 g h−1 only if root endings are cut and supplied with unlimited water. Sap flow in 3–4 mm roots in undisturbed soil rarely exceeds 6–8 g h−1 at its daily maximum, and typical daytime flows at the Lüneburger Heide site were between 2 and 5 g h−1. Calibration experiments for flows <2 g h−1 were conducted in the laboratory on excised root segments using a high-precision electronic balance. Flow rates <2 g h−1 typically had large measurement errors of the gauges (>200%). Consequently, night flows often were overestimated.

As predicted by theory, the relative error increased when approaching zero flow (Fig. 2) because Tso − Tsi decreases in parallel to flow rate in gauges with a symmetrical thermocouple arrangement. We obtained a positive correlation between sap flow (gravimetric data) and measured temperature difference Tso − Tsi at flow rates <2 g h−1 in our calibration experiments (Fig. 4). These data were recalculated by subtracting a baseline of Tso − Tsi values to account for gauge-specific differences in the (Tso − Tsi)/flow relationship which result from small differences in the relative position of the thermocouples in the three instruments.

Figure 4.

Correlation between the temperature difference across the heated zone (Tso − Tsi) of a mounted gauge and water flow in root segments of Fagus sylvatica recorded gravimetrically with an analytical balance (based on six laboratory experiments with three sap-flow gauges mounted consecutively on two root segments). The regression was obtained by the major axis method.

A positive relationship between flow rate and Tso −Tsi for small flow rates contrasts sharply with the negative correlation of flow and temperature difference at higher flow rates. In a laboratory experiment with a root segment of F. sylvatica, flow rates were decreased successively from 11 to 0 g h−1 by reducing suction stepwise over 10 h (Fig. 5). For flow rates between 11 and ≈ 3 g h−1, a weak negative relationship between flow and Tso − Tsi occurred, as would be expected from equation 5. A further reduction of flow rate below 2 g h−1, however, decreased Tso − Tsi rather than increasing it. Consequently, calculating flow rates from Tso − Tsi with equation 5 induces large errors for flows <2 g h−1.

Figure 5.

Water flow (as determined gravimetrically) in a root segment of Fagus sylvatica and temperature difference across the heated zone (Tso − Tsi) of a mounted gauge during a laboratory experiment where flow was decreased stepwise from 11 to 0 g h−1 by altering suction over 10 h.

A Modified Calculation Procedure for Low Flows

Based on the initially positive relationship between flow rate and Tso − Tsi (Fig. 4), we developed a procedure to estimate J for low flows from gauge measurements of Tso − Tsi, independently of equation 5. This alternative method gives J (hereafter termed J*) from:

image(eqn 6)

with β being 1/slope in Fig. 4 (1·028 g h−1 K−1). Independent measurements with three gauges at three root segments of F. sylvatica indicated only small variation in β (r = 0·92 for the relation between flow rate and Tso − Tsi). Therefore a constant β value of 1·028 was applied for all gauges and experiments done with F. sylvatica roots.

A composite method was then used to calculate daily courses and totals of sap flow in small-diameter roots: (i) daytime flows (J) were calculated according to equation 5 based on the negative correlation between temperature difference and flow; and (ii) in periods with assumed low flows (at dawn, dusk and night) J* was calculated using β and the measured temperature difference Tso − Tsi, as in equation 6. J was replaced by J* if the following criteria were met: J < 2 g h−1 and E < 0·995Enight, where E is the thermopile output (µV) which is highest at night (Enight) when sap flow is close to zero and most heat is dispersed by radial conduction.

With this procedure, corrected values of J were obtained for daily courses and totals that exclude night-time data with a possible bias. The possible error that is introduced by using the less reliable J* is small, as <10% of daily root sap flow typically occurs when J is <2 g h−1: on 10 days in July 1998, on average, 8·3 ± 3·0% of daily flow occurred in periods with low flow. To test the composite calculation method in a laboratory calibration experiment with variable flow rates in the range 11–0 g h−1, J and J* were calculated for high or low flows, respectively, and the data were checked against the gravimetric control (Fig. 6). A good agreement was found for gauge and gravimetric data.

Figure 6.

Water flow determined gravimetrically and from gauge data in a root segment of Fagus sylvatica during a laboratory experiment where flow was decreased stepwise from 11 to 0 g h−1 by altering suction over a period of 10 h (cf. Fig. 5). The gauge data were calculated from equation 5 for flows >2 g h−1, and according to the empirical relationship between Tso − Tsi and flow rate (J*) for very low flows (<2 g h−1, see text).

Based on the composite flow calculation method and a subsequent surface analysis of the adjacent root endings, water absorption per root surface area was calculated for a 3·5 mm F. sylvatica root at the field site for 10 consecutive days in July 1998. The computation of flow with equation 5 gives high absorption rates in the nights of July 16/17 and 17/18 which probably represent measurement errors (Fig. 7a). In Fig. 7(b) the biased night-time data are replaced by J* values. Daily maxima of root water absorption as calculated by this method ranged between 25 and 38 g m−2 surface area h−1 on bright or overcast days (0·39–0·58 mmol m−2 s−1). On a rainy day (as on July 16) the maximum did not exceed 13 g m−2 h−1 (0·20 mmol m−2 s−1). When compared with the transpiration rate (porometer data of the sun canopy at 24 m height), root water uptake was ≈ 10 times smaller than leaf water loss on an area basis (daily maxima, 2–4 mmol m−2 leaf area s−1).

Figure 7.

Water absorption per root surface area for a Fagus sylvatica root ending (diameter at measuring point, 3·5 mm) during 10 days in July 1998 at the Lüneburger Heide field site. (a) Absorption rate was calculated using equation 5 (J, solid line), and during low flow using equation 6 (J*, dotted line). Flows were related to root surface area after complete excavation and visual surface analysis. (b) Absorption rate calculated with the composite method using J for periods with flow >2 g h−1, and J* for flows <2 g h−1.


The recent approach to measuring water uptake of tree roots with miniature gauges represents a major methodological advance because measured flow is related to root surface area and thus root water absorption can be calculated. This is possible only for tree roots with diameters <3 or 5 mm where the fine root endings distal to the measuring point are typically shorter than 1 m and can be completely excavated with reasonable effort in sandy soil. With our approach, water absorption by fine roots can be monitored continuously under natural conditions of soil moisture, density and mycorrhizal infection. The sensors are cheap (<US$20 each), which allows many measurements to be made simultaneously.

The precise localization of water uptake in fine and coarse tree roots is still a matter of dispute (Escamilla & Comerford 1998; Peterson & Cholewa 1998); this question cannot be answered precisely with the recently introduced gauge technique because uptake rates are averaged over branch roots 0·3–2·5 m in length. A number of experimental studies found transport of water and ions through peridermal woody roots (e.g. Chung & Kramer 1975; MacFall et al. 1991). However, it is still an open question whether the dead peridermal cork cells are permeable enough to account for this water flow, or whether passage occurs through breaks in the periderm (McKenzie & Peterson 1995). Anatomical and chemical investigation of F. sylvatica fine roots from the Lüneburger Heide site showed that a suberized periderm covers the root surface from a few millimetres behind the terminal tip onwards (C.L. and co-workers, unpublished results). Thus at least part of the water absorbed by the beech rootlets must have passed through the multilayered periderm.

Our field and laboratory calibrations showed that miniature sap-flow gauges can give reliable data on water flow in tree roots with diameters as small as 3 mm. The fact that calibration by volumetry is much easier in small-diameter than in large-diameter roots allows a check of gauge data already in the field. Possible measurement errors during low flows at night have been identified as the major disadvantage when measuring small-diameter roots. This type of error will bias daily sap-flow totals on rainy summer days and winter days but, according to our data, is of minor relevance during bright or overcast summer periods. The problem has been partly solved by calculating low flows on the basis of a positive linear relationship between measured temperature difference and flow, instead of using the negative correlation applicable during higher flows. This procedure is not entirely satisfactory because the shift from a negative to a positive J/(Tso − Tsi) relationship, which occurs with decreasing flow, is influenced by the gauge design and requires testing of each gauge model. Future developments of miniature sap-flow gauges should consider asymmetrical spacing of thermocouples as a possible alternative to the equidistant arrangement used in our gauges. A disadvantage of gauges with an asymmetrical design may be the greater distance between the two thermocouple positions, which can introduce additional errors due to environmentally induced temperature gradients along the root.

Erroneous night-time values are not only a problem of miniature gauges, but also occur in large sap-flow systems. These gauges often measure daytime flows as high as 100 g h−1 (Allen & Grime 1995; Dugas et al. 1994; Kjelgaard et al. 1997), which is more than 10-fold larger than flows recorded with the miniature gauges in this study. To solve these problems, Grime et al. (1995b) applied a low-flow filter because they calculated unrealistically high night flows >100 g h−1 where daytime maxima were ≈200–250 g h−1.

Other technical improvements have been proposed to overcome problems with erroneous night-time readings in large-scale gauges. They include control of the heater input (Ishida et al. 1991; Weibel & Boersma 1995), an improved stem-heater contact (Weibel & de Vos 1994), and a reduction in heat conduction along sensor leads (Groot & King 1992). Moreover, heat storage in plant tissue should be included in the calculation when measuring large stem or root diameters (Grime et al. 1995a). Most of these sources of error, however, are not serious when applying miniature gauges to small-diameter roots in the soil because (i) the gauges are well isolated from temperature fluctuations; (ii) the flexible cork jacket provides a good contact to the root surface; and (iii) heat storage in the root is of minor importance.

We conclude that by combining the miniature sap-flow technique with determination of the root surface, promising opportunities are opened up for experimental research on water flow in tree root systems. The miniature sap-flow technique may also be useful for assessing the role of shallow and deep roots in water uptake, and for analysing water use by competing root systems in mixed forests or agroforestry systems.


We thank the German Science Foundation (DFG) for supporting this study by a grant to C.L. We are very grateful to Randy Senock (Hilo) for introducing us to the miniature sap-flow technique and for giving advice during the study. Two anonymous referees gave valuable comments on an earlier draft of the paper.