The empirical calibration of Granier-type heat dissipation sap flow probes
that relate temperature difference (Delta T) to sap velocity (nu) was reeva
luated in stems of three tropical tree species. The original calibration wa
s confirmed when the entire heated probe was in contact with conducting xyl
em, but mean nu was underestimated when part of the probe was in contact wi
th nonconducting xylem or bark. Analysis of the effects of nonuniform sap v
elocity profiles on heat dissipation estimates showed that errors increased
as nu and the proportion of the probe in nonconducting wood increased. If
half of a 20-mm probe is in sapwood with a nu of 0.15 mm s(-1) and the othe
r half is in nonconducting wood, then mean nu for the whole probe can be un
derestimated by as much as 50%. A correction was developed that can be used
if the proportion of the probe in nonconducting wood is known. Even with t
he entire heated probe in contact with conducting xylem, nu would be undere
stimated when radial velocity gradients are present, In this case, the erro
r would be smaller except when velocity gradients are very steep, as can oc
cur in species with ring-porous wood anatomy. Errors occur because the rela
tionship between Delta T and nu is nonlinear. Mean Delta T along the probe
is therefore not a measure of mean nu, and users of heat dissipation probes
should not assume that nu is integrated along the length of the probe. The
same type of error can occur when Delta T is averaged through time while n
u is changing, but the error is small unless there are sudden, step changes
between zero and high sap velocity. It is recommended that relatively shor
t probes (20 mm or less) be used and that probes longer than the depth of c
onducting sapwood be avoided. Multiple probes inserted to a range of depths
should be used in situations where steep gradients in v are expected. If t
hese conditions are met, heat dissipation probes remain useful and widely a
pplicable for measuring sap flow in woody stems.