Evapotranspiration Measurement Methods

W. James Shuttleworth SAHRA, University of Arizona

n the United States and several
other countries, the term
evapotranspiration (ET) is used
when considering evaporation from
vegetation-covered ground. t descri!es
the total evaporation from the soil and wet
plants plus transpiration from dr" plants.
The two most common t"pes of direct-
measurement methods, water budget and
water vapor transfer measurements, are
descri!ed first in the ta!le !elow. #ater
!udget measurements deduce ET as a loss
of li$uid water !" measuring or estimating
all the other components in a water
!udget. Such methods are long-esta!lished
and have !een refined over the "ears.
#ater vapor transfer methods measure the
Brief Description Assumptions
Assumes relationship between measured evaporation from pans and actual
Evaporation pan
Directly measures change in water level over time for a sample of open
water in a pan with well-specified dimensions and siting.
evaporation from adjacent area can be calibrated and calibration is transfer-
able between locations and climates.
Water balance of
basin
!he unmeasured difference between other measured components of
the basin water balance including incoming precipitation surface and
groundwater outflow and soil water storage.
Assumes all other components of the basin water balance can be measured as
spatial averages with sufficient accuracy for evaporation to be reliably calcu-
lated as the difference between them.
Lysimetry
Measures change in weight of an isolated preferably undisturbed soil
sample with overlying vegetation "if present# while measuring precipitation
to and drainage from the sample.
Assumes the sample of soil and overlying vegetation on which measurements
are made are representative in terms of soil water content and vegetation
growth and vigor of the plot or field in $uestion.
Soil moisture
depletion
Bowen Ratio - En-
ergy Budget
Eddy correlation
(also called eddy
covariance
!ranspiration
measurement by
porometry or moni-
toring sap flow
Rainfall intercep-
tion loss from tall
vegetation
Soil evaporation
Scintillometer
measurements
Remote sensing
estimates
L"DAR (L"g#t Detec-
tion And Ranging
met#od
Measures change in water content of a representative sample of undis-
turbed soil and vegetation while measuring precipitation and run-on%runoff
and estimating deep drainage for the sample plot.
&alculates evaporation as latent heat from the surface energy budget
using the ratio of sensible to latent heat "'owen ratio# derived from the
ratio between atmospheric temperature and humidity gradients measured
a few meters above vegetation.
&alculates evaporation as ()- to *)-minute time averages from the
correlation coefficient between fluctuations in vertical windspeed and
atmospheric humidity measured at high fre$uency "+,) -.# at the same
location a few meters above vegetation.
/oromet r y0 measured from humidity increase in a chamber temporarily
enclosing transpiring leaves%shoots. 1 a p 2low0 measured from rate of sap
flow in trun3 branches or roots using heat as a tracer with an estimate of
the area of wood through which flow occurs.
Measured as difference between cumulative rainfall above%below tall
"usually forest# canopy. 4e$uires careful below-canopy sampling with
gauges%troughs that sample at spatial scale of canopy features preferably
randomly relocated after each measurement interval.
A small-scale shallow implementation of lysimetry or soil moisture
depletion methods for a near-surface soil sample below vegetation using
several microlysimeters or se$uential gravimetric multisampling.
5ses theoretical relationship between sensible and latent heat flu6es and
atmospheric scintillation introduced into a beam of electromagnetic radiation
between source and detector by temperature and humidity fluctuations.
Evaporation is deduced indirectly from the surface energy balance with
sensible heat calculated from the difference between air temperature and
the temperature of the evaporating surface along with an estimate of the
aerodynamic e6change resistance between these two.
!he local time-average vertical gradient of water vapor is sampled
remotely using 78DA4. 7ocal evaporation flu6 is calculated from this using
similarity theory and supplementary measurements of friction velocity and
atmospheric stability.
Assumes that soil water measuring devices "resistance bloc3s tensiometers
neutron probes time-domain reflectometers capacitance sensors# ade$uately
determine change in soil water the effects of deep roots and sensor placement
are small and deep drainage can be estimated ade$uately.
Assumes the turbulent diffusion coefficient for sensible heat and latent heat are
the same in the lower atmosphere in all conditions of atmospheric stability and
that plot-scale measurements of energy budget components "net radiation soil
heat# are representative of upwind conditions.
Assumes only turbulent transfer of water vapor at sample point and that cor-
rections for water vapor transfer in turbulence at time scales less than +).,
seconds or greater than the selected averaging time are acceptable.
/oromet r y assumes the enclosure of leaves and shoots in the chamber does
not significantly alter transpiration rate. 1 a p 2lo w assumes installation of
sensors does not alter sap flow rate and cross-sectional area over which flow
occurs can be determined accurately.
Assumes below-canopy sampling is ade$uate a re$uirement rarely met for
a typical ,-( wee3 measurement interval. 8t becomes feasible over several
measurement intervals if gauges are regularly and randomly relocated.
Assumes the average of all small soil samples regardless of their below-
canopy location are representative of the entire soil surface.
Applies strictly in an ideal turbulent field close but not too close to a surface
with uniform aerodynamic roughness. -owever field e6periments suggest a
worthwhile measurement is possible over a mi6ture of vegetation covers.
Assumes the aerodynamic surface temperature "that which controls sensible
heat transfer from the surface# is the same as "or can be estimated from#
the radiometric surface temperature "that which can be measured using an
airborne or satellite radiometer#.
Assumes Monin-9bu3ov similarity theory applies and the supplementary mea-
surements of friction velocity and atmospheric stability are locally applicable
within the measurement field of the 78DA4.

Water
Lar
ge-
sc
ale
ev
ap
ora
tio
n
$o
mp
on
ent
s
of
ev
ap
ora
tio
Wa
ter
bu
dg
et
me
as
ure
me
nts
%% : ;anuary%2ebruary ())< : 1outhwest -ydrology
flow of water vapor into the atmosphere
using meteorological sensors mounted
a!ove the surface. Sometimes these
sensors measure evaporation not in mass
terms, !ut in the conte%t of the surface-
energ" !alance as latent heat flux. This
is the flow of energ" that is transferred
with the water vapor and that leaves
the surface in the form of latent heat.
t can !e useful to measure the separate
contri!utions to ET& transpiration from
plants, rain or snowwater evaporated
from the plant canop", and evaporation
from the soil surface. Some of these
methods are descri!ed ne%t in the ta!le.
'ther recent ET measurement efforts
attempt to measure area-average ET.
E%amples of these also are included.
(n alternative to the direct measurement
methods descri!ed !elow is to model ET
rates using local climate data in empirical
and anal"tical e$uations. This approach is
not covered here.
ET measurement methods tend to have
their champions, individuals who are
convinced their method is !est. #hen
appraising the strengths, wea)nesses and
li)el" errors of the different methods,
have sought to !e impartial and
conservative, !ut the appraisal is to some
e%tent su!*ective and it is personal+
Contact i! Shuttleworth at shuttle"hw r #arizona#edu#
wea&nesses
S
c
al
e
o
f
'
e
a
-
s
u
r
e
m
e
n
t
(
A long-established and well-
recogni.ed method simple to
understand and implement and
reasonably
Error
=aries with
reliability and
relevance of
calibration factor
but ,)
ine6pensive> but because it
relies on the validity of an
e6trapolated calibration
factor previously defined
elsewhere is primarily used
for crop E! estimates
rather than heterogeneous
natural vegetation covers.
?ives an area-average
measurement for vegetation
covers for a hydrologically
significant region however
area-average measurement
of the other water balance
terms can be e6pensive and
difficult especially
groundwater flow and soil
moisture. &onse$uently only
longer time-average
estimates are possible.
8f the soil and vegetation
sample is truly
representative the
lysimeter is widely
accepted as being an
unparalleled standard
against which to compare
and validate other
evaporation
measurements and
models of crop
evaporation. Modern
high-precision lysimeters
are e6pensive
"+@A))))# and re$uire
e6pert supervision.
Most often used in crop-
covered plots.
Measurement is
reasonably ine6pensive
/lot
'asin
1ample
/lot
to ()B errors are possible for
crops with greater errors
li3ely for natural vegetation
because calibration may be
un3nown.
=aries with $uality of
implementation and si.e and
nature of basin but errors
as low as ,) to ()B may
be achievable in research
basins with persistent care.
1tate-of-the-art lysimeters can
low as ,) to
()B error.
Cell-established method.
4elatively ine6pensive
proprietary systems can be
purchased that wor3 for both
short crops and natural
vegetation. /roblematic over
tall vegetation when
atmospheric gradients are low.
9ften cannot be used near
dawn and dus3 when the
'owen ratio is minus one.
Errors
associated with
assumptions
and
representativen
ess plus errors
in re$uired
supplementary
sensors result in
overall errors of
around A to
,AB.
&urrently preferred method
for field-scale measurements
in research applications.
8mplemented using relatively
e6pensive proprietary logger
and colocated sensors but
prone to systematic
underestimation of flu6es.
/erhaps best used to
measure 'owen ratio with
evaporation deduced from
surface energy budget.
/oromet r y0 a manual
measurement that allows
determination of
environmental influences on
stomatal control at leaf level.
1 a p 2low0 allows routine
unsupervised measurement
of transpiration from whole
plants or plant components
over e6tended periods.
Allows separate identification of
wet canopy contribution to E! for
tall vegetation. 4arely if ever
attempted
2i
el
d
7ea
f-
to-
pla
nt>
pl
ot
s
c
al
e
w
it
h
m
ul
ti
pl
e
s
a
m
pl
in
g
1ystematic
underestimation
up to (AB can
occur in the basic
evaporation
measurement. 8f
sensible heat is
also measured to
determine 'owen
ratio and energy
balance is used to
calcu- late
evaporation error
can be reduced to
A to ,AB.
/oromet r y0 small
for leaves "+few
B#. 1 a p 2low0
errors as- sumed
to be A to ,AB for
individual plants.
'oth0 at plot scale
errors are strongly
determined by the
number of
samples
ta3en and the
variability in these
samples.
1trongly
depends on
below-canopy
sampling. 9ne-
gauge
arrangement
provides only
order of
magnitude
estimate but
for short vegetation and crops but
possible in principle.
/lot
time average
with many
gauge
relocations can
reduce error to
around A
to,)B.
A comparatively simple
and ine6pensive manual
measurement.
?ravimetric approach is
time-intensive and the
sample is destroyed
preventing repeated
measurement at same
place.
!he only
micrometerological method
that can be used to provide
an "albeit indirect#
measurement of the line-
average sensible and latent
heat over several
3ilometers.
/rovides opportunity for
instantaneous snapshots of
evaporation over large
areas in clear s3y condi-
tions but uncertainties in
the effective surface
emissivity and effective
aerodynamic e6change
resistance can give
systematic errorsDboth
being worst for sparse
canopies. !herefore
ground-truth evaporation
measurements are usually
re$uired.
?ives detailed and fre$uent
E-D mapping of the water-
vapor gradient valuable in
assessing variations over
areas with heterogeneous
evaporation. -owever
e$uipment costs are
e6tremely high and
indepen- dent E!
measurement is re$uired to
assess accuracy.
/l
ot
wi
th
m
ul
-
tip
le
sa
m
pli
ng
2
i
e
l
d

t
o

l
a
n
d
s
c
a
p
e
2i
el
d
t
o
r
e
gi
o
n
al
2
i
e
l
d

t
o

l
a
n
d
s
c
a
p
e
1trongly
depends on
below-canopy
sampling but
errors as low
as ,) to ()B
are possible
with many
samples and
care.
2ield
comparisons
between the
line-average
flu6 over
several types
of vegetation
and eddy-
correlation
measurements
for each
vegetation
type agree at
the ,) to ()B
level or better.
Cith ground-
truth
measurements
snapshot maps
of evapora- tion
in clear s3y
conditions may
be accurate to
,) to ()B
but time-
average
estimation from
these
snapshots
introduces
additional
uncertainty.
/rovides a
useful measure
of spatial E!
variations but
re$uires
independent
validation%calibr
ation.
*Scales: ,eaf-to-plant$ the size of
the basic canopy, typically s%uare
centi!eters to a few s%uare !eters&
Sample$ area of the soil and
vegetation sa!ple, typically a few
s%uare !eters& -lot$ typically a few to
tens of s%uare !eters& .ield$ typically a
few hundreds of s%uare !eters&
,andscape$ typically a few thousands of
s%uare !eters& /egional$ typically a few
s%uare !iles& 0asin$ varies fro!
landscape to regional scale and beyond#
;anuary%2ebruary ())< : 1outhwest -ydrology : %)