**Summary** Sapflow sensors were used to investigate
vari-ation in sapflow velocity at different positions in the sapwood
of three-year-old *Eucalyptus globulus ssp. globulus* Labill.
trees. Sapflow velocity was measured at 5-mm intervals across
the sapwood by moving two probe sets simultaneously on two
opposite radii. Another probe set placed in a fixed position at
right angles to the first two sets acted as a control. A sapflow
velocity ratio was defined as the velocity given by each moving
sensor divided by that given by the static sensor. Correlation
between observations of sapflow velocity at different positions
exceeded 95%, and the ratio of velocity between any pair of
sensors was constant.

We observed radial variation in sapflow velocity across the sapwood with the lowest velocities at the center of the tree. Variation due to sensor position was high implying the need for large numbers of sensors for accurate estimates of sap flux. To overcome this need, we used a correction coefficient, namely a simple weighted average of the sapflow ratios with depth in the sapwood, for each fixed sensor. We recommend the use of three probe sets to estimate the correction coefficient. Sub-sequently, two probe sets can be placed at two fixed positions for routine measurements of sap flux.

*Keywords: heat pulse velocity, sapflow sensor, sap flux, *
*tran-spiration, xylem.*

**Introduction**

Direct measurements of transpiration by heat pulse velocity
have been considerably improved by the development of
meth-ods for rapid and more accurate point estimates of sapflow
velocity. However, because the nature of the sapwood varies
widely with depth in a variety of tree species including
*Euca-lyptus *(Dunn and Connor 1993), estimates of sap flux for a tree
based on point samples may seriously depart from the true sap
flux through the xylem. As a result, the greatest potential
source of error in estimating stand transpiration with this

method is in the measurement of sap flux of individual trees
(Hatton et al. 1995). The aim of this study was to develop a
technique for measuring sap flux in *E. globulus* with a
mini-mum number of sensors.

The use of a heat pulse as a tracer to measure transpiration in sapwood was proposed by Huber (Huber 1932, Huber and Schmidt 1937). The compensation method of Marshall (1958) was the first quantitative treatment to account for the physical presence of implanted sensors in the sapwood. When used in conjunction with numerical models that correct for the effects of wounding (Swanson and Whitfield 1981), the compensation technique can provide accurate point estimates of sapflow velocity. However, because sapflow velocity is known to vary throughout the sapwood, with peak velocity occurring 10--20 mm in from the cambium (Swanson 1967, 1974), esti-mates of sap flux based on point estiesti-mates of sapflow velocity need to account for this variation.

Sampling at several depths in the xylem with integration
throughout the conducting area is one method for estimating
sap flux. A least-squares polynomial can be fitted to point
estimates of sapflow velocity and the function integrated
across the sapwood conducting area and around the bole to
obtain actual sap flux (Cohen et al. 1981, Green and Clothier
1988). An alternative method is based on a weighted average
of sapflow velocity with depth (Hatton et al. 1990, 1992). For
fast-growing *Pinus patula *Schlecht et Cham., particularly in
the earlier stages of growth, sapflow velocity shows a cyclical
pattern, with maxima and minima correlated with summer and
winter wood, respectively (Dye et al. 1991). The use of several
sensors to account for this variation during routine
measure-ments is not practical.

We have investigated the variation in sapflow velocity across
the radial profile of the sapwood of fast-growing *E. globulus.*
The results were used to develop a correction coefficient for
obtaining accurate estimates of sap flux from routine point
measurements of sapflow velocity using two sets of sensors.

**Variation of sapflow velocity in **

**Eucalyptus globulus**

**Eucalyptus globulus**

** with position in**

**sapwood and use of a correction coefficient**

### DAOQUN ZANG,

1### C. L. BEADLE

2### and D. A. WHITE

2**1**

* Chinese Academy of Forestry, Beijing 100091, P.R. China*

**2**

* CSIRO Division of Forestry and Cooperative Research Centre for Temperate Hardwood Forestry, Locked Bag No. 2, Sandy Bay, Tasmania 7005,*
*Australia*

Received March 6, 1995

**Materials and methods**

*Sensor and theory*

Measurements of heat pulse velocity were carried out with sapflow sensors (Model SF100 Greenspan Technology, Queensland, Australia). Three vertically aligned holes were drilled at each measurement position with a drilling jig to ensure that the holes were parallel. The probe sets, which included two stainless steel tubes located 10 mm above and 5 mm below a heater tube, were inserted into the holes. Each probe set included two thermistor pairs (Figure 1). One ther-mistor from each pair was housed in each stainless steel tube and the thermistors were spaced 5 mm apart within the tube. The thermistor pairs were connected in a Wheatstone bridge configuration and adjusted to zero output at intervals during the measurement period. The logger was programed to apply a heat pulse of 0.8 s duration.

Heat pulse velocity (*u*) was measured by the compensation
technique (Swanson and Whitfield 1981). Sapflow velocity
(*V*) was calculated after Swanson and Whitfield (1981),
Mar-shall (1958) and Dunlap (1912) as described by Olbrich
(1991).

*Experimental Procedure*

The measurements were made between February and May
1993 at an experimental plantation near Lewisham in southern
Tasmania, Australia (Honeysett et al. 1996). The trees were
from selected families of the King Island provenance of
*Euca-lyptus globulus ssp. globulus* Labill. and located in a plot that
was irrigated to maintain soil water deficits at approximately
30 mm below field capacity. At this soil water deficit, growth
was not limited by the availability of water. Five trees were

selected for this study, Trees 1, 2, and 3 for Experiment 1, and Trees 4 and 5 for Experiment 2. Diameters at 1.3 m (breast height) varied between 9.6 and 13.8 cm and bark thickness varied between 2 and 4 mm.

*Experiment 1* Heat pulse velocity was measured with two
probe sets in fixed positions. They were placed such that the
inner sensor of each set was 15 mm and 25 mm inside the
cambium on the east and west sides, respectively, and at 0.1 m
and 0.2 m above ground. The heat pulse interval was 30 min.
Data were collected over a 5--7 day periodthough only daytime
observations were used for analysis. The four observations
made at each interval were treated as one record. Data were
excluded if the time required to reach thermal balance between
the upper and lower thermistors of each probe exceeded 150 s.
Two cores were taken to the center of the stem of each tree for
determination of sapwood density and water content after data
collection was completed. The data were converted to sapflow
velocity as described. Regression and principal components
analyses were used to evaluate the relationship between
sap-flow velocities estimated by the four pairs of sensors in each
probe set.

*Experiment 2* Branches below 1 m were excised. Four probe
sets were placed in holes drilled in the four cardinal radii at
approximately 0.1 m intervals and between 0.5 and 1.0 m above
ground. Two probe sets were located on the northern and
southern radii, one initially at the periphery of the sapwood and
the second toward the center (Figure 1). For Trees 4 and 5, these
positions were, respectively, 5 and 60 mm, and 5 and 50 mm
inside the cambial ring. The profile of heat pulse velocity was
measured at 5-mm intervals across the sapwood by moving the
two probe sets simultaneously every 30 min (sometimes 15 or
45 min) for Tree 4 and every 10 min for Tree 5, 2 min before

the heat pulse. The heat pulse interval was 15 and 10 min for
Trees 4 and 5, respectively. The moving sensors were then
transferred to the eastern and western radii and the experiment
repeated. The second pair of probe sets was placed at right
angles to the moving pair in fixed positions 15 and 10 mm
inside the cambium. These were recorded at the same intervals
as for the moving sensors and were used to calculate the
sapflow velocity ratio (*R*, see below). For the north--south and
east--west profiles, the fixed sensor on the western and northern
radii, respectively, was used to calculate *R*. Cores were taken
to the centers of the trees at the end of the experiment and sap
flux calculated as for Experiment 1.

*Sapflow velocity ratio and correction coefficient*

Sapflow ratio (*R*) was calculated as the sapflow velocity
meas-ured by the moving sensors (*Vm*) divided by the sapflow

veloc-ity measured by the appropriate fixed sensor (*Vs*). For *i *=
that position. The average ratio at the *i*th position (*Ri*) is given

by:

A correction coefficient (*C*) to estimate the true sapflow
velocity at each fixed sensor was calculated from the weighted
average of the sapflow velocity ratios on the profile. The ratios
were weighted by the areas (*Ai*) represented at each position.

Because the distance (*d*) between positions on the profile was
constant, *Ai* is half the area of a ring of radius *rp,i* measured to

the center of the stem. Thus:

*Ai*= 0.5π((*rp*,*i*+0.5*d*)2−(*rp*,*i*− 0.5*d*)

2_{)}

=π*rp*,*id*

and the total conducting area *A *is:

*A*=

### ∑

*Ai*

The weighting factor (*wi*) for each position on the profile was

calculated as:

Equations 2 and 4 are combined and the correction coefficient
(*C*) of the static sensor expressed as:

*C*=

### ∑

(*Riwi*)

*i*= 1

*m*

. (5)

If *Vk* is the sapflow velocity observed at a static sensor *k* and

*n* is the number of static sensors, during routine measurements,
the average sap flux in the sapwood is:

*V*

The total sap flux is:

Sapflux=*AV*
__

. (7)

*Sources of variation*

The data for Trees 1, 2, and 3 in Experiment 1 were examined by analysis of variance using the model:

*yi,j*= µ+ *Vt,i* + *Vs,j*+ *ei,j* , (8)

where *yi,j* is the *i*th record of sapflow velocity at position *j*, µ is

the grand mean across all measurement times and sensors, and
*Vt,i*and *Vs,j* are fixed effects due to sampling time and sensor

pair, respectively, and *ei,j* is the error. Variation in sapwood

water content and density are not included in this model
because a single value was used to calculate all sapflow
veloci-ties. The fixed effect *Vs,j* is inclusive of the variation caused by

sensor position, errors in wound size measurement and
vari-ation in probe separvari-ation. All sources of random varivari-ation are
included in the error term. For each tree the total sums of
squares (*s*2) is given by:

*s*2=*st*2+*ss*2+*se*2, (9)

where *st*2, *ss*2 and *se*2 are the sums of squares for sampling time,

sensor pair and error, respectively.

The data for Trees 4 and 5 in Experiment 2 were examined using the model:

*yj*= µ + *Vs,j*+ *ej *, (10)

where *yj* is the average ratio (*R*) at position *j*, µ is the grand

mean, and *Vs,j* is the variation of average ratio (*R*) due to

position and *ej* is the error. For each tree the total sums of

squares (*s*2) is given by:

*s*2=*ss*

2_{+}

The number of sensors (*r*; i.e., the number of measurement
positions on the profile) required to attain a given accuracy for
the sapflow measurement was determined by:

*r*=*t*α2*ss*2*/*δ2, (12)

where *t*αis the value of the *t*-statistic at significance level α,*ss*

is the estimated standard error of either *Vs* (Experiment 1) or *R*

(Experiment 2) and δ is the desired accuracy. The value of *r*
was estimated for all the trees used in the experiments.

**Results and discussion**
*Experiment 1*

The first principal component accounted for 98.0--98.5% of the variance sums of squares (Table 1). The total of the three other components was less than 2% and can be ignored. Thus, the measurement of sapflow velocity by one sensor pair is expressed in the same terms as a linear combination of all four sensor pairs. The correlation coefficients for all combinations of the four sensor pairs on the two probe sets varied between 0.944 and 0.983 (Table 2), demonstrating that measurements by any one sensor pair can be used to estimate measurements made by another sensor pair.

Regression analyses showed that the relationship between
sapflow velocity (*V*) measured by a single sensor and sapflow
velocity estimated from the average of the three other sensors
was always linear (*r*2 = 0.92--0.97, Figure 2, Table 3).
Al-though the mean sapflow velocity estimated by each sensor
varied by a factor of > 2 in some instances, the intercepts did
not differ significantly from zero.

Thus, one point measurement is potentially a good predictor
of *V *at another point in the system or the mean of several
measurements. In a similar study with *Pinus radiata *D. Don,
Hatton and Vertessy (1989) found a high correlation between
sapflow velocities at any two depths within an individual tree.
Our results also confirm that the flow of water through stems
is well defined and that regularities in the distribution of *V*
should be amenable to a relatively simple analysis to determine
total sap flux.

*Experiment 2*

The sapflow ratios (*Ri*) were always lower in the middle of the

stem and higher near the periphery (Figure 3). For Tree 4,
variation in *Ri* in the northern, southern and western radii

showed a cyclic pattern with two pronounced peaks on either
side of the stem center. This pattern was not observed on the
eastern radius. In Tree 5, there was little evidence of a cyclic
pattern except on the eastern radius and maximum *Ri* was

observed 10--15 mm inside the cambium on all radii. Previous
studies have shown that variation in sapflow velocity across the
sapwood is a function of species, individuals within a species
and radii within the same stem (Swanson 1967, 1974, Hatton
and Vertessy 1989). Dye et al. (1991) concluded that the cyclic
patterns of *V* on a given radius were associated with the
production of summer and winter wood in *Pinus patula*. The
peaks observed in Tree 4 corresponded to the production of
summer wood in the second and third years of growth;
how-ever, the outer peaks also corresponded to the peak positions
of *V* that Swanson (1967, 1974) observed 10--20 mm inside the
cambium of the outer summer wood ring of a range of species.

*Correction coefficient*

The variation in sapflow velocity with position in the sapwood
limits both the accuracy with which *V *can be estimated and the
potential for using a single model to describe its radial
vari-ation. Dye et al. (1991) suggested random placement of
sen-sors as an effective approach to estimate *V* when confronted

Table 1. Principal component analysis for sapflow velocity measurements of four sensor pairs used in Trees 1, 2, and 3 in Experiment 1. The four components are denoted as C1, C2, C3 and C4.

Latent roots Variation (%)

Component Tree 1 Tree 2 Tree 3 Tree 1 Tree 2 Tree 3

C1 23401 39556 39368 97.97 98.45 98.16

C2 269 386 475 1.13 0.96 1.19

C3 165 172 198 0.69 0.43 0.49

C4 50 64 64 0.21 0.16 0.16

Table 2. Correlation matrix of sapflow velocity for Trees 1, 2, and 3 in Experiment 1. The four sensor pairs used in each tree are denoted as S1, S2, S3 and S4.

S4 0.971 0.967 0.970 1

*Tree 2*

S1 1

S2 0.967 1

S3 0.956 0.974 1

S4 0.944 0.961 0.980 1

*Tree 3*

S1 1

S2 0.979 1

S3 0.970 0.968 1

with such variation. Hatton et al (1995) found that stratified
sampling with depth and quadrant significantly reduced
sam-pling variance compared to random samsam-pling in *Eucalyptus*
*populnea* F. Muell when four sensors were used. Alternatively,

many sensors can be used. Swanson (1983) used 16 probes but
this is an expensive solution and may seriously interrupt flow
through the sapwood. Our finding that the ratio of sapflow
velocity between any two points in a profile remains constant
in *E. globulus *allowed the development of a system of
meas-urement based on only two probe sets positioned at right
angles to each other in the stem. Thus, accurate estimates of *V*
are based on the calculation of a correction coefficient (*C*)
obtained from measurements with three probe sets before
routine measurements commence. The average values of *R*
across the profiles and of *C* were higher for Tree 4 than for
Tree 5. There were also small differences in both variables
depending on which fixed sensor was used to estimate *R *(Table
4). Large tree to tree variation highlights the need for a simple
approach for accurately estimating sap flux in one tree if heat
pulse velocity is used to estimate canopy transpiration.

The method adopted here was similar to that of Hatton et al.
(1990), where information from each sensor was weighted by
the proportion of conducting area that each sensor represents.
Ideally, each sensor has equal areal weighting. The assumption
was made that measurements on a single radius were
repre-sentative of the annuli to which each sensor pertained. The
advantages of this approach are that relative values of *V* can be
investigated at frequent intervals (every 5 mm in this
investi-gation to give > 40 points on the profile of Trees 4 and 5),
heterogeneity of *V* with radial position is accounted for and, by
expressing *V* in terms of *R*, this variation can be incorporated
into a single factor *C *for use in routine measurements with
only two probe sets. For Trees 4 and 5, the weighted average
sapflow ratio expressed as *C *was always greater than the
average ratio expressed as *R* because sapflow was, on average,
lower in the inner part of the ring than in the outer part of the
ring (Figure 3).

*Sources of variance*

The ratios of *ss*2*/se*2 were 2.09, 2.86 and 2.91 for Trees 1, 2 and

3, respectively (Table 5). Thus, variation due to sensor pair, which includes sensor position, wound size measurement and variation in probe separation, was more important in determin-ing sapflow velocity than all other random errors considered together, suggesting that increased sensor numbers would im-prove the accuracy of sap flux estimated for each tree. Figure 2. Average sapflow velocity of three sensor pairs in

Experi-ment 1 as a function of that measured by a single sensor pair. Symbols S1, S2, S3 and S4 refer to the sensor pairs that were located 10, 15, 20 and 25 mm inside the cambium, respectively, on the two probe sets and the respective fitted least-squares regressions for each.

Table 3. Regression analysis of sapflow velocity for each sensor pair with average sapflow velocity for the other three sensor pairs as used in Experiment 1 and Figure 2. The four sensor pairs used in each tree are denoted as S1, S2, S3 and S4.

Variable Tree 1 Tree 2 Tree 3

S1 S2 S3 S4 S1 S2 S3 S4 S1 S2 S3 S4

*X* coefficient 1.51 1.37 0.71 0.69 2.08 2.07 0.78 0.41 1.73 1.86 0.84 0.45

SE of coefficient 0.02 0.02 0.01 0.01 0.05 0.04 0.01 0.01 0.02 0.03 0.01 0.01

Constant 0.44 0.06 0.06 0.37 0.29 0.65 0.07 0.57 0.54 0.71 0.15 0.19

SE of *Y* estimate 0.97 0.98 0.89 0.92 2.37 1.80 1.11 1.13 1.36 1.40 1.16 0.99

*r* 2 0.97 0.96 0.96 0.96 0.92 0.95 0.97 0.96 0.96 0.96 0.96 0.95

In Experiment 2, only one source of variation in sapflow
velocity, sensor position, was considered in detail. A decision
as to whether spatial variation warrants explicit consideration
in integrating sapflow velocity should be based on an analysis
of its contribution to total variation in sapflow velocity.
Sources of variation in sapflow velocity include temporal as
well as spatial variation (Dye et al. 1991), errors in estimation
of wound width (Olbrich 1991), variation in the water content
and density of the sapwood and systematic errors resulting
from the application of equations to calculate µ and *V*.

The number of sensors (*r*) required to estimate sapflow
velocity to a desired accuracy of 10% was as high as 72 for
Tree 2 in Experiment 1 at the 95% confidence interval: even
for a desired accuracy of 25%, up to 13 sensors were required

at the 90% confidence interval (Table 6). For the same range
of accuracies and confidence intervals, between four and 65
sensors would be required for Trees 4 and 5 (Table 7). Thus
acceptable accuracy demands measurement of sapflow at a
large number of points in *E. globulus *to allow for its large
variation with position in the sapwood. The use of a method
incorporating a correction coefficient as described is a simple
means of resolving this problem.

**Conclusions**

We observed cyclic variations in sapflow velocity (*V*) with
position in the sapwood of *Eucalyptus* *globulus. *The high
correlation between *V* at different points in the system enabled

Figure 3. Sapflow velocity ratio (*R*)
ex-pressed as sapflow velocity of the moving
sensor (*Vm*) divided by that of the static

sensor (*Vs*) as a function of position in

sap-wood for Trees 4 and 5 in Experiment 2. The symbols refer to the ratio calculated from each of the sensor pairs on the static sensor. The dotted line marks the center of the tree.

Table 4. Mean and standard error of sapflow velocity ratios (*R*) and correction coefficients (*C*) for the static sensors. The two sensor pairs on the
static sensors are denoted S1 and S2.

Variable Tree 4 Tree 5

E--W profile N--S profile E--W profile N--S profile

S1 S2 S1 S2 S1 S2 S1 S2

Mean *R* 0.853 0.787 0.992 0.870 0.555 0.576 0.504 0.502

Standard error 0.302 0.282 0.191 0.164 0.211 0.220 0.196 0.197

*C* 0.962 0.881 1.027 0.904 0.647 0.672 0.567 0.563

Table 5. Total sum of squares (*s*2) and sums of squares for sampling time (*st*2), sensor pair (*ss*2) and error (*se*2) for Trees 1, 2 and 3 in Experiment 1.

The means and standard errors (*s*s_) are those for sensors S1 to S4 in Table 3; *s*e is the standard error per observation.

*s*2 *st*2 *ss*2 *s*2*e* *ss*2*/se*2 Mean *s*s

_

*s*e

_

Tree 1 28154 21843 4270 2041 2.09 12.92 2.44 1.69

Tree 2 65910 31188 25731 8990 2.86 18.05 7.51 4.46

the development of a correction coefficient based on a
deter-mination of the ratio of sapflow velocity at different points on
a radial axis through the stem to that at a fixed position.
Differences in this ratio and the correction coefficient were
observed among trees. The correction coefficient applied to
data collected from two fixed positions provided a solution for
determining *V *of single trees to an acceptable level of accuracy
that would otherwise be possible only through the use of a
large number of sensors.

**Acknowledgments**

We thank the Australian Centre for International Agricultural Re-search for supporting this study and the senior author as visiting scientist, Mr. Dale Worledge for technical assistance, Messrs D. and P. Tinning for the plantation site and water supply for irrigation and Dr. P.J. Dye for sharing his ideas and enthusiasm for the work. Dr. P.J. Sands provided valuable comments on the manuscript.

**References**

Cohen, Y., M. Fuchs and G.C. Green. 1981. Improvement of the heat pulse method for determining sap flow in trees. Plant Cell Environ. 4:391--397.

Dunlap, F. 1912. The specific heat of wood. USDA For. Serv. Bull. No. 110.

Dunn, G.M. and D.J. Connor. 1993. An analysis of sap flow in
mountain ash (*Eucalyptus regnans*) forests of different age. Tree
Physiol. 13:321--336.

Dye, P.J., B.W. Olbrich and A.G.Poulter. 1991. The effect of growth
rings in *Pinus patula* on heat pulse velocities and sap flow
measure-ment. J. Exp. Bot. 42:867--870.

Green, S.R. and B.E. Clothier. 1988. Water use on kiwi fruit vines and apple trees by the heat pulse technique. J. Exp. Bot. 39:115--123. Hatton, T.J., E.A. Catchpole and R.A. Vertessy. 1990. Integration of

sapflow velocity to estimate plant water use. Tree Physiol. 6:201--209.

Hatton, T.J., D. Greenslade and W.R. Dawes. 1992. Integration of sapflow velocity in elliptical stems. Tree Physiol. 11:185--196. Hatton, T.J., S.J. Moore and P.H. Reece. 1995. Estimating stand

transpiration in a *Eucalyptus populnea* woodland with the heat
pulse method: measurement errors and sampling strategies. Tree
Physiol. 15:219--227.

Hatton, T.J. and R.A. Vertessy. 1989. Variability of sapflow in a *Pinus*
*radiata *plantation and the robust estimation of transpiration. *In*
Hydrology and Water Resources Symposium. Australian Institute
of Engineers, Christchurch, New Zealand, pp 6--10.

Honeysett, J.L., D.A. White, D. Worledge and C.L. Beadle. 1996.
Growth and water use of *Eucalyptus globulus ssp. globulus *and *E.*
*nitens* in irrigated and rainfed plantations. Aust. For. In press.
Huber, B. 1932. Beobachtung und messung pflanzliche saftströme.

Berl. Deutsche Bot. Ges. 50:89--109.

Huber, B. and E. Schmidt. 1937. Eine kompensationsmethode zu thermoelektrischen messung langsamer saftströme. Berl. Deutsche Bot. Ges. 55:514--529.

Marshall, D.C. 1958. Measurement of sap flow in conifers by heat transport. Plant Physiol. 33:385--396.

Olbrich, B.W. 1991. The verification of the heat pulse velocity
tech-nique for estimating sap flow in *Eucalyptus grandis*. Can. J. For.
Res. 21:836--841.

Swanson, R.H. 1967. Seasonal course of transpiration of Lodgeple
pine and Engelmann spruce. *In* Proc. Int. Symp. For. Hydrology*.*
Eds. W. E. Sopper and H. W. Lull. Pergamon Press, Oxford, pp
417--432.

Swanson, R.H. 1974. Velocity distribution patterns in ascending
xy-lem sap during transpiration. *In *Flow----Its measurement and
con-trol in science and industry. Ed. R. B. Dowdell. Instrument Society
of America, Pittsburgh, pp 1425--1430.

Swanson, R.H. 1983. Numerical and experimental analysis of heat pulse velocity theory. Ph.D. Diss., Univ. Alberta, Edmonton, Can-ada, 298 p.

Swanson, R.H. and D.W.A. Whitfield. 1981. A numerical analysis of
heat pulse velocity theory and practice. J. Exp. Bot. 32:221--239.
Table 6. The number of sensors (*r*) required to determine sap flux to accuracies (*S*) of 10 or 25% at confidence intervals (*P*) of 90 or 95% for Trees
1, 2 and 3 in Experiment 1. The means and standard errors (*s*s_) are those for sensors S1 to S4 in Table 5. The degrees of freedom = 30.

Mean *s*_s *P *= 95% *P *= 95% *P *= 90% *P *= 90%

*S *= 10% *S *= 25% *S *= 10% *S *= 25%

Tree 1 12.92 2.44 15 4 10 3

Tree 2 18.05 7.51 72 18 50 13

Tree 3 15.41 5.38 50 13 35 9

Table 7. The number of sensors (*r*) required to determine sap flux as in Table 6. The means and standard errors (*s*s_) are those for the sapflow ratios

for Trees 4 and 5 given in Table 4. The degrees of freedom were 21 and 19, respectively.

Mean *s*_s *P *= 95% *P *= 95% *P *= 90% *P *= 90%

*S *= 10% *S *= 25% *S *= 10% *S *= 25%

Tree 4 0.876 0.235 32 5 22 4