Exploring thermal imaging variables for the detection of stress responses in grapevine under different irrigation regimes.

Abstract

Temperatures of leaves or canopies can be used as indicators of stomatal closure in response to soil water deficit. In 2 years of field experiments with grapevines (Vitis vinifera L., cvs Castelão and Aragonês), it was found that thermal imaging can distinguish between irrigated and non-irrigated canopies, and even between deficit irrigation treatments. Average canopy temperature was inversely correlated with stomatal conductance measured with a porometer. Variation of the distribution of temperatures within canopies was not found to be a reliable indicator of stress. A large degree of variation between images was found in reference 'wet' and 'dry' leaves used in the first year for the calculation of an index proportional to stomatal conductance. In the second year, fully irrigated (FI) (100% Et(c)) and non-irrigated (NI) canopies were used as alternatives to wet and dry leaves. A crop water stress index utilizing these FI and NI 'references', where stressed canopies have the highest values and non-stressed canopies have the lowest values, was found to be a suitable measure for detecting stress. It is suggested that the average temperatures of areas of canopies containing several leaves may be more useful for distinguishing between irrigation treatments than the temperatures of individual leaves. Average temperatures over several leaves per canopy may be expected to reduce the impact of variation in leaf angles. The results are discussed in relation to the application of thermal imaging to irrigation scheduling and monitoring crop performance.

Full text

Page 1

Exploring thermal imaging variables for the detection
of stress responses in grapevine under different
irrigation regimes
Olga M. Grant1,*,†, Łukasz Tronina1, Hamlyn G. Jones2and M. Manuela Chaves1,3
1Laborato ´rio de Ecofisiologia Molecular, Instituto de Tecnologia Quı ´ mica e Biolo ´gica, Apartado 127,
2781-901 Oeiras, Portugal
2Division of Environmental and Applied Biology, University of Dundee at SCRI, Invergowrie, Dundee DD2 5DA, UK
3Departamento Bota ˆnica e Engenharia Biolo ´gica, Instituto Superior de Agronomia, Universidade Te ´cnica de Lisboa,
Tapada da Ajuda, 1349-017 Lisboa, Portugal
Received 13 April 2006; Accepted 3 August 2006
Abstract
Temperatures of leaves or canopies can be used as
indicators of stomatal closure in response to soil water
deficit. In 2 years of field experiments with grapevines
(Vitis vinifera L., cvs Castela ˜o and Aragone ˆs), it was
found that thermal imaging can distinguish between
irrigated and non-irrigated canopies, and even between
deficit irrigation treatments. Average canopy temper-
ature was inversely correlated with stomatal conduc-
tance measured with a porometer. Variation of the
distribution of temperatures within canopies was not
found to be a reliable indicator of stress. A large degree
of variation between images was found in reference
‘wet’ and ‘dry’ leaves used in the first year for the
calculation of an index proportional to stomatal con-
ductance. In the second year, fully irrigated (FI) (100%
Etc) and non-irrigated (NI) canopies were used as
alternatives to wet and dry leaves. A crop water stress
index utilizing these FI and NI ‘references’, where
stressed canopies have the highest values and non-
stressed canopies have the lowest values, was found
to be a suitable measure for detecting stress. It is
suggested that the average temperatures of areas of
canopies containing several leaves may be more useful
for distinguishing between irrigation treatments than
the temperatures of individual leaves. Average temper-
atures over several leaves per canopy may be expected
to reduce the impact of variation in leaf angles. The
results are discussed in relation to the application of
thermal imaging to irrigation scheduling and monitor-
ing crop performance.
Key words: Leaf angle, leaf temperature, partial rootzone
drying, regulated deficit irrigation, stomatal conductance, ther-
mography, Vitis vinifera, water deficit.
Introduction
Mean global temperatures are expected to rise over the next
few decades, evaporation rates will increase, arid regions
will expand, and thus water availability will be a major
limitation to plant growth in the future (Houghton et al.,
2001; European Environment Agency, 2004). As a result,
irrigation will become an increasingly common practice.
Since water availability is already limited, an increase in the
area under irrigation will only be possible if the quantity of
water used per unit area is reduced, i.e. if plant water use
efficiency can be improved. Additionally, precise manipu-
lation of plant–water relations can be very important for
maximizing the quality of the product, particularly in
viticulture. Excessive application of water can reduce
colour and sugar content and produce acidity imbalances
in the wine (Bravdo et al., 1985; Esteban et al., 2001).
Conversely, insufficient water reduces grape yield and can
* To whom correspondence should be addressed. E-mail: olga.grant@emr.ac.uk
yPresent address: East Malling Research, New Road, East Malling, Kent, UK
Abbreviations: CWSI, crop water stress index; d13, carbon isotope discrimination; DI, deficit irrigation; Etc, crop evapotranspiration; FI, fully irrigated; gs,
stomatal conductance to water, NI, non-irrigated; PRD, partial rootzone drying; RDI, regulated deficit irrigation; T, temperature; W, leaf water potential.
ª The Author [2006]. Published by Oxford University Press [on behalf of the Society for Experimental Biology]. All rights reserved.
For Permissions, please e-mail: journals.permissions@oxfordjournals.org
Journal of Experimental Botany, Page 1 of 11
Imaging Stress Responses in Plants Special Issue
doi:10.1093/jxb/erl153

Journal of Experimental Botany Advance Access published October 10, 2006

Page 2

also adversely affect quality (Reynolds and Naylor, 1994;
dos Santos et al., 2003).
Deficit drip irrigation strategies have been used to save
water in viticulture and simultaneously to improve wine
quality. Regulated deficit irrigation (RDI) aims to manip-
ulate grapevine vegetative and reproductive growth by
withholding or applying less than the full vineyard water
use at specific periods of the growing season (Dry et al.,
2001). Partial rootzone drying (PRD) is an alternative
technique, currently of interest for a variety of crops
(Davies et al., 2000; Grant et al., 2004) including
grapevine (Dry et al., 2000; dos Santos et al., 2003), that
allows control of vegetative growth and transpiration
without the severe water stress periods that can occur in
RDI (Loveys et al., 1999). In PRD, part of the root system
is slowly dried and the remaining roots are exposed to wet
soil. Roots of the watered side maintain a favourable plant
water status, while dehydrating roots produce chemical
signals that are transported to the shoots via the xylem.
These signals are thought to control shoot vigour and
stomatal aperture (Dry and Loveys, 1999).
Leaf or stem water potential is a standard indicator
of stress, and is sometimes used in irrigation scheduling
(Smart et al., 2004). This method, however, is destructive
and time-consuming. Stomatal closure is known to be a
sensitive response to soil water deficit, occurring even in
the absence of any change in plant water status, as a result
of root signalling (Davies et al., 2000). It has potential as
an indicator of plant water stress and therefore could be
used in irrigation scheduling. Monitoring stomatal con-
ductance could be particularly useful to determine the timing
of irrigation (for example in RDI or PRD systems) where
a very precise regulation of water supply is required in the
production of high quality fruits, including grapes for wine
(Dry et al., 2001). However, the traditional methods of
measuring stomatal conductance (using porometers or infra-
red gas analysers) are time-consuming, labour-intensive,
and only give point measurements.
As stomata close under water deficits, leaf temperatures
rise. Thus leaf or canopy temperatures can be used as an
indicator of plant stress and stomatal closure. Thermal
imaging systems allow rapid and non-invasive collection of
data, integrated over the area of individual leaves or areas
of canopies. They may reveal spatial heterogeneity within
or between leaves, and can be used repeatedly on the same
leaves to monitor responses over time, without affecting the
natural behaviour of the leaves. The nature of grapevine
trellises, with plentiful leaves that are close to vertical
exposure, means that this crop may be particularly suited to
monitoring with a thermal imager which can be carried
along the rows.
The development of thermal imaging and the associated
image analysis software has overcome the problems ex-
perienced by researchers using infrared thermometry with
regard to the difficulty of separating leaf and non-leaf
(soil, sky, bark, etc.) temperatures. While application of
thermal imaging is more straightforward in the laboratory
(Chaerle et al., 1999; Lindenthal et al., 2005), researchers
have also applied the technique to the field (Jones et al.,
2002; Cohen et al., 2005). Nonetheless, rigorous testing
of thermal imaging against more traditional physiolog-
ical techniques under field conditions is still required for
different types of crops. Indices that relate leaf or canopy
temperatures to the temperatures of selected reference
surfaces allow for variation in air temperature, radiation,
and wind speed, thus removing the effect of environmental
variation so as to indicate increases or decreases in stomatal
conductance (Jones, 1999). An alternative possibility for
detecting stress in plant canopies is to analyse thermal
variation within the canopy. Leaf orientation plays a greater
role in the energy budget of leaves when stomatal aperture
is smaller, which may result in greater variation in temper-
atures within canopies that are more stressed, with lower
stomatal conductance, than in unstressed canopies with
very open stomata (Fuchs, 1990). Variability in temper-
atures between plants in the same management treatment
has been noted to increase with stress, for example, Gardner
et al. (1981), probably as a result of variation in soil
properties and root depth. Leaf orientation and canopy
geometry (row orientation, row spacing, plant height)
interact with environmental factors and stomatal conduc-
tance to determine the temperature of the plant canopy
(Boissard et al., 1990). As yet there has been little attempt
to analyse the impact of canopy architecture on the
application of thermal imaging. Leaf drooping during
wilting in stressed canopies, or altered leaf orientation or
inclination, may reduce the impact of stomatal closure on
leaf temperature.
The objectives of this work were to evaluate thermal
imaging as a tool for distinguishing between stressed and
unstressed plants, and to optimize thermal imaging for
determining plant responses to water deficits in the field.
Experiments were carried out to test whether thermal
imaging can be used to distinguish between irrigated and
water-limited grapevines, and between grapevines grow-
ing under different deficit irrigation systems. The relation-
shipbetweencanopyorleaftemperatures,orindicesderived
from these temperatures, and stomatal conductance as
measured with a porometer were explored. The influence
of leaf size, leaf orientation angle, and leaf inclination angle
on leaf and canopy temperatures was also investigated.
Materials and methods
Thermal imaging and stomatal conductance
All thermal images were obtained with a thermal imager (IR Snapshot
525, Infrared Solutions, Minneapolis, MN, USA) that operates in the
wavebands 8–12 lm, has a thermal resolution of 0.1 8C, and produces
pictures with spatial resolution of 1203120 pixels. Images were
analysed in SnapView Pro software (Infrared Solutions); all images
2 of 11 Grant et al.

Page 3

were corrected for spatial calibration drift by subtracting correspond-
ing reference images of an isothermal surface (Jones et al., 2002). For
each series of measurements, the background temperature was
determined as outlined in the imager manual as the temperature of
a crumpled sheet of aluminium foil in a similar position to the leaves
of interest. Emissivity for measurements of leaves and plant canopies
was set at 0.96 (see review by Jones, 2004). The areas of interest
for analysis in the imager’s software were outlined, manually, by
comparing thermal and normal digital images (Fig. 1). All thermal
images were taken with the thermal imager on a tripod perpendicular
to the area being imaged. Images of canopies and individual leaves
were taken ;1.5 and 0.9 m from the canopies and leaves, re-
spectively, capturing areas of ;50 cm350 cm and 29 cm329 cm,
respectively.
Where individual leaves were imaged in 2003, dry and wet
references were used to mimic leaves with fully closed and fully
open stomata, respectively (Jones et al., 2002). These references were
grapevine leaves, cut from the canopy prior to measurements and
placed close to the leaves of interest. Wet reference leaves were
sprayed with water on both sides, regularly, to maintain their
moisture. Dry reference leaves were covered in petroleum jelly
(Vaseline) on both sides. The temperatures of these references were
obtained (Tdryand Twet) and used in conjunction with leaf temper-
atures to obtain thermal indices. Stomatal conductance (gs) of the
same leaves used in thermography was measured with a steady-state
porometer (Li-Cor 1600, Li-Cor, Lincoln, NE, USA).
Where canopies rather than individual leaves were imaged,
reference leaves were not included. In 2004, images of non-irrigated
(NI) and fully irrigated (FI) canopies were used as indicators of low
and high stomatal conductance, respectively.
Experimental conditions
Field measurements were made in 2003 and 2004, in two different
commercial vineyards. Both are located in south-east Portugal, where
the climate is Mediterranean, with hot, dry summers and cool, wet
winters. Both of the cultivars of grapevine (Vitis vinifera L.) studied
(Castela ˜o and Aragone ˆs) are red varieties and were grafted on 1103
Paulsen rootstock, and trained on a bilateral Royat Cordon system.
The main characteristics of the vineyards are described in Table 1.
Crop evapotranspiration (Etc) was calculated from Class A pan
evaporation and using the crop coefficients proposed by Prichard
(1992). Irrigation was applied with drip emitters, two per vine,
positioned 25 cm from the vine trunk, one either side of the row.
Castela ˜o 2003: The cultivar Castela ˜o was subjected to the follow-
ing treatments: non-irrigated (rain-fed) (NI), partial rootzone drying
(PRD) where 50% of Etcwas supplied to only one side of the root
system, alternating sides every 15 d; deficit irrigation (DI), where
50% of Etcwas divided between the two sides of the row; and full
irrigation (FI), corresponding to 100% Etc. Each treatment was re-
plicated in each of four experimental rows, in a Latin square design,
with two guard rows between each pair of experimental rows.
Thermal images were taken and gsof the same leaves measured in
the morning and afternoon on different dates (Table 2). Four replicate
plants were used per treatment, one in each experimental row. The
porometer measurement was taken immediately after each thermal
infrared image. Additionally, on one date (6 August), eight replicates
were taken for thermal infrared images (two plants per treatment per
row). Thermal infrared images were also taken of areas of leaf
canopies (eight replicates per treatment, two plants per treatment per
row), in the morning and afternoon on different dates. Plants were
sampled along rows, so that the order of sampling of treatments was
randomized.
Aragone ˆs 2004: In 2004, measurements were conducted near
Estremoz using the cultivar Aragone ˆs. Three treatments were
imposed: PRD, DI, and regulated deficit irrigation (RDI). RDI plants
received more water than the other treatments at the start of the
growing season and less later in the growing season, with irrigation of
RDI plants being stopped on 10 August. Over the whole season, RDI
plants thus received the same total amount of water as PRD and DI
plants. Measurements were also conducted on adjacent NI vines and
FI vines. Thermal infrared images were taken of three vines per
treatment in each of three selected blocks and the same plants were
used throughout. Before measurements on each block, thermal
Fig. 1. An example of a thermal image and the corresponding digital image. The area of interest on the thermal image is outlined.
Table 1. Characteristics of two vineyards in south-east Portugal
where the experiments were conducted
Cultivar Castela ˜o Aragone ˆs
Year of experiment
First year of treatments
Year vines were grafted
Location
2003
2000
1995
Centro Experimental
de Pego ˜es
308389N
88399W
26 June 2003
Latin square,
12 rows
N–S
2004
2004
2000
Seis Reis, near
Estremoz
388489N
78299W
15 June 2004
8 blocks, 9 rows
per block
ENE–WSW
Latitude
Longitude
Start date for irrigation
Experimental design
Orientation of rows
Thermal imaging of stress in grapevines3 of 11

Page 4

infrared images were taken of one NI plant and one FI plant, to be
used as references. A set of measurements took ;1 h.
The main lateral veins of each individual marked leaf and of five
leaves in the selected sections of canopy were measured as an
indication of leaf area (Lopes and Pinto, 2000). The inclination from
horizontal of each individual marked leaf, and of five randomly
sampled representative leaves in two of the selected sections of
canopy per treatment per row were measured with a protractor
attached to a level, and the azimuths of the central vein of the same
leaves were measured relative to the orientation of the row with
a protractor, and then converted to absolute azimuths, where leaves
with the central blade facing directly north have an azimuth of 08.
To test the hypothesis that the effect of leaf drooping in stressed
grapevine canopies influences leaf temperature, leaves in the west–
south-west-facing side of nine NI vines were forced to stay in their
pre-stress position, using metal wire to hold the petiole and string to
maintain the distance between the petiole and the row. In adjacent
control plants, drooping of leaves was not prevented. Canopy thermal
images were taken and stomatal conductance was recorded.
Data analyses and statistics
Thermal indices: The index IGwas calculated from leaf temperatures:
IG=(Tdry–Tleaf)/(Tleaf–Twet). This index is theoretically proportional to
stomatal conductance (gs) (Jones, 1999). An index analogous to
Idso’s (1982) crop water stress index (CWSI) was also calculated,
where in this case CWSI=(Tdry–Tleaf)/(Tdry–Twet). Similar indices
were used in 2004 with TNI replacing Tdry and TFI replacing
Twet. These indices are called CWSINI/FIand INI/FIto distinguish
them from the more established indices CWSI and IG.
Temperature distribution in canopies: Images of areas of canopies in
SnapView Pro were exported to Excel, to obtain the temperature of
every pixel in the image. Canopies were outlined and the frequency
distributions of the temperatures of pixels in these areas were
calculated, together with the mean temperature, variance, skewness
(deviation of the distribution from symmetry), and kurtosis (deviation
of the distribution from the normal peak) as reported by Guiliani and
Flore (2000). A histogram-derived CWSI (HCWSI), based on the
approach of Bryant and Moran (1999), was also calculated as
a measure of the deviation of the shape of the histogram from
a normal curve with the same mean and variance. The observed
temperature frequency distribution was normalized by expressing
the frequency in any 0.1 K temperature range as a fraction of the
maximum frequency in any range to give fT. The corresponding
normal distribution for each range was calculated using the mean and
variance, and again normalized by expressing as a fraction of the
maximum to give distT. HCWSI was calculated as the sum of the
absolute differences for each temperature range:
HCWSI =
+
Tmax
t=Tmin
abs½ðfT? distTÞi?
where Tmaxand Tminare the maximum and minimum temperature
values of pixels in the image.
Variance, skewness, and kurtosis of thermal distributions were
calculated for images of canopies taken either year. Additionally,
indices of variation within images were calculated from each image
as (maximum temperature–minimum temperature)/maximum tem-
perature, and were averaged for each treatment. For indices of
variation within treatments, average canopy values were obtained
for each image and the index was calculated as (maximum aver-
age temperature–minimum average temperature)/maximum aver-
age temperature.
Statistical analyses
Data were tested for normality and homogeneity of variances using
Kolmogorav–Smirnov and Levene’s tests, respectively, in STATIS-
TICA (1995). The significance of relationships between IGand gswas
tested by Pearson-product or Spearman correlations. The effects of
treatments were analysed by analysis of variance (ANOVA), using
a Latin square design for 2003 data and two-factor ANOVA for
the 2004 data, with the factors being treatment and block. Co-
efficients of variation (=1003SD/mean) were calculated for
thermal measurements.
Results
Castela ˜o 2003
In 2003, air temperatures in the vineyard were very high
in July and August (Fig. 2); the average daily maximum
temperature recorded between 30 July and 27 August was
38 8C.
Nosignificant differences werefoundbetween treatments
in gs, as measured with the porometer, on the four dates of
measurement, perhaps due to small sample sizes as well as
variability between treatments. As a result, differences
between treatments in thermal variables might not have
been expected. However, lack of variation between treat-
ments in stomatal conductance was in contrast to predawn
leaf water potential, which was significantly lower in NI
and DI vines than in FI vines both at the end of July and in
mid-August (Table 3). Stomatal closure may have occurred
in all treatments at some time during the hot summer, but
evidently not for sufficient lengths of time to prevent the
development of differences in leaf water potentials.
Of the four dates on which the temperatures of individual
leaves were measured, only on one was a significant effect
of treatment observed (6 August; Fig. 3). The significant
effects on this date probably relate to greater sample sizes
(n=8 on 6 August compared with n=4 on the other dates)
Table 2. Dates on which different physiological variables were measured
VariableDates
20032004
31/07
am
05/08
am
06/08
am
13/08
am
14/08
am
26/08
pm
13/08
am
19/08
pm
24/08
am
25/08
am pm pmpm pmpm pmpmpm
gs
W
Tleaf
Tcanopy
O
O
O
OO
O
O
OOOOOOOOO
O
OOOOO
O
OOO
OOOOO
4 of 11 Grant et al.

Page 5

rather than any meteorological or other factor that might
differentiate this date from the others. Temperature differ-
ences were found both in the morning (in shaded leaves,
P=0.019) and in the afternoon (sunlit, P=0.049). At both
times, FI leaves were cooler than NI leaves. Correspond-
ingly, IGwas lower, both in the shade and in the sun, in NI
compared with FI leaves (Fig. 3B). In the morning (shade),
PRD canopies also showed significantly cooler temper-
atures and higher IGthan NI canopies. In the afternoon
(sun), all the irrigated canopies had significantly higher
IGthan NI.
Stomatal conductance as measured with the porometer
and IGshowed significant correlations (P <0.02)on 31 July
am, 13 August pm, and 14 August am and pm (example in
Fig. 4), indicating that individual vines with low leaf
temperatures showed high gs, and vice versa. The correla-
tion between gsand IGwas not significant on 31 July pm
or 13 August am. The significance of the correlations was
not related to the range of conductances, nor to exposure.
Some negative values of IGwere obtained, due to higher
values for Tleafthan Tdry.
When images of areas of canopies rather than individual
leaves were taken, there were significant treatment effects
on canopy temperature (P <0.02), both in the morning and
in the early afternoon (Fig. 5). FI canopies were cooler than
NI or DI canopies, whether viewing the sunlit or shaded
canopies. The HCWSI varied considerably within the same
treatment, and even between two canopies of the same
treatment imaged in quick succession. As a result, neither
HCWSI nor the other measures of temperature variation
varied significantly between treatments (Table 4). Thus no
increase in temperature variance was detected with greater
plant stress. In general, the frequency distributions of pixels
in NI and FI canopies were fairly similar. Indices of vari-
ation within images of individual vines were fairly high,
irrespective of treatment, with relatively low indices of
variation within treatments (i.e. between images of different
vines), when the maximum and minimum of mean image
temperatures is used in this calculation (Table 5).
Aragone ˆs 2004
Clear treatment effects on stomatal conductance (measured
by porometry) were found on all occasions studied in
August 2004 (P <0.03), with conductances consistently
increasing in the order: NI, RDI, FI, with the PRD and
DI treatments often approaching or equalling the FI value
(Fig. 6). RDI leaves had significantly lower predawn water
potentials than PRD or DI leaves at this time (Table 4).
For the afternoon of 13 August, a highly significant
effect of treatment was found on average canopy temper-
ature (P=0.0001), with post hoc tests showing that RDI
canopies were significantly hotter than PRD canopies or DI
canopies (Fig. 7C). Since in 2004 no reference wet and dry
leaves were included in thermal infrared images, alternative
reference temperatures were derived from the temperatures
of FI and NI canopies imaged at intervals: values were
interpolated between the three measurements of FI and
extrapolated to the time of the last measurement in any
given session. The same was done for NI measurements.
As a result, for every image of a PRD, DI, or RDI canopy,
corresponding FI and NI temperatures were obtained. Some
temperatures of PRD, DI, or RDI canopies fell outside the
range of the corresponding NI and FI canopy temperatures.
Canopy temperatures higher than the corresponding NI
temperature result in negative values of INI/FI[(TNI–Tleaf)
is negative], but canopy temperatures cooler than the cor-
responding FI also result in negative values of INI/FI
[(Tleaf–TFI) is negative] (examples in Table 6). Thus,
negative values of a INI/FIcould indicate either a very
stressed or a completely unstressed canopy. With the modi-
fied CWSI, on the other hand, values outside the range
Fig. 2. Spring and summer monthly precipitation (bars) and monthly
average temperature (circles) on average over the years 1954–1984
(shaded) and in the year 2003 (open) at the meteorological station in
Centro Experimental de Pego ˜es.
Table 3. Water potential and leaf area responses of the cultivar Castela ˜o to different irrigation schedules
WPDrefers to predawn leaf water potential and leaf area to total leaf area per vine. n=6 for WPDand 12 for total leaf area. Data are means 6SE. Different
letters (along rows) indicate significantly different means in Tukey tests following ANOVA.
PropertyCultivarDate Treatment
NI PRDDIRDI FI
WPD(MPa)Castela ˜o31/07/03
14/08/03
19/08/03
?0.4960.03 c
?0.5060.04 b
5.4860.52 b
?0.3460.01 b
?0.3360.02 a
8.4661.07 ab
?0.3660.02 b
?0.4960.02 b
8.0461.10 ab
?0.2660.02 a
?0.3060.01 a
10.0961.46 a
Leaf area (m2vine?1) Castela ˜o
Thermal imaging of stress in grapevines5 of 11

Page 6

0–1 are consistent with the idea of low values when
canopies are not stressed and high values when they are
stressed. Thus the temperatures of NI and FI canopies can
be seen not as absolute limits of possible canopy temper-
atures, but as indicator temperatures.
For CWSINI/FI, there was a highly significant effect of
treatment (P <0.0001), with RDI canopies showing higher
values than canopies receiving the other treatments
(Fig. 8A). Canopy temperature and CWSINI/FIwere sig-
nificantly negatively correlated with stomatal conductance
(P <0.02, r2=0.3; Fig. 9); stomatal conductance, however,
was measured for only one leaf within each canopy imaged.
A significant effect of treatment was also found on INI/FI
(P=0.012) (Fig. 8B), with lower INI/FI values for RDI
canopies than PRD canopies. The variance of the temper-
ature distribution was correlated with the average canopy
temperature (r2=0.37, P <0.001, Fig. 9C), but did not
significantly differ between treatments. Neither the kurtosis
nor skewness of the temperature of canopies was correlated
with the average canopy temperature. The index of
variation between canopies was highest for RDI (0.14),
a little lower for PRD (0.13), and lowest for DI (0.11).
With respect to thermal images of individual leaves
(rather than canopies) on 19 and 24 August, no significant
differences between treatments were found in leaf temper-
ature or CWSINI/FI. The only significant effect of treatment
(P=0.007) was found for INI/FIon the afternoon of 19
August, with the highest values being found for PRD leaves
and the lowest for RDI leaves (Fig. 8C).
The lengths of the two main lateral veins of the grapevine
leaves mostly fell between 6 cm and 15 cm (Fig. 10A). The
most frequent leaf orientations were between 1508 and 1808,
where 08 faces north, i.e. approximately perpendicular to
the direction of the row (Fig. 10B). Leaf inclination
angles were mostly distributed between 408 and 808
from horizontal (Fig. 10C). No significant effect of
treatment was found on the angle, orientation, or size
(average length of the two main lateral veins) of the
marked leaves. No significant correlation was found for
these leaf properties and either temperatures or CWSINI/FI
Fig. 3. Leaf temperature (A) and corresponding IG(B) values for 6
August 2003. Images were taken on the shaded side of vines in the
morning and on the sunlit side of canopies in the afternoon. n=8. Data are
means 6SE. Different letters indicate significantly different means in
Tukey tests following ANOVA.
Fig. 4. An example of the relationship between stomatal conductance
measured with a porometer and the index IGderived from thermal images
of grapevine leaves and wet and dry reference leaves, taken on 31 July
2003. n=16.
Fig. 5. Average canopy temperatures on different dates and times of day
in August. Measurements were taken of the shaded side of canopies on
the morning of 5 August and the sunlit side of canopies on the afternoons
of 5 and 26 August. n=8. Data are means 6SE. Different letters indicate
significantly different means in Tukey tests following ANOVA.
6 of 11 Grant et al.

Page 7

calculated from thermal infrared images of those leaves.
Equally, average angles, orientations, and vein lengths of
five replicate leaves per canopy (selected in the region of
the canopy thermal infrared images) also showed no cor-
relation with canopy temperatures or temperature indices. In
the experiment in which the leaves of some canopies were
allowed to droop as they became stressed, and others were
maintained with leaves in their prewilting positions, a sig-
nificant effect of leaf drooping on canopy temperatures
was not detected.
Discussion
Sensitivity to crop stress
While it is encouraging that thermal imaging could
consistently distinguish between NI and FI canopies in
the experiment in 2003, even more interesting is the ability
to distinguish between different irrigation treatments, as
seen in 2004. Thermal imaging distinguished RDI from
the other two treatments, PRD and DI (as observed on 13
August 2004), as did porometry on several dates. RDI
exhibited the highest temperature and lowest gs. Predawn
water potential followed the same pattern. A similar trend
was seen in carbon isotope discrimination of leaf material,
with less negative values for leaves from RDI and NI vines,
andmorenegativevaluesinPRD,DI, andFIvines(Table4).
In 2004, the same number of replicates was used for leaf
and canopy temperatures, so it is of interest that significant
differences were found between treatments that year in
canopy temperature but not in leaf temperature—more data
would be needed, however, in order to ascertain whether
canopy temperature is consistently more sensitive than the
temperature of individual leaves. Similar patterns in both
years between canopy thermal data and other indicators
of crop stress (stomatal conductance, water potential, and
carbon isotope discrimination) suggest that thermal imag-
ing is an effective method of detecting crop stress.
Table 4. Water potential and carbon isotope discrimination responses of the cultivar Aragone ˆs to different irrigation schedules
WPDrefers to predawn leaf water potential and d13C to carbon isotope discrimination of dried leaves, analysed at Mylnefield Research Services
(Invergowrie, Dundee, UK) and expressed as the percentage difference between the ratio13C/12C of the sample and that of the Pee Dee Belemnite
standard. n=12, except for NI and FI, which are each represented by one sample for d13C. Data are means 6SE. Different letters (along rows) indicate
significantly different means in Tukey tests following ANOVA.
PropertyCultivarDateTreatment
NI PRDDI RDIFI
WPD(MPa)
d13C
Aragone ˆs
Aragone ˆs
19/08/04
02/09/03
?0.3960.02 a
?28.3460.17 a
?0.4560.02 b
?28.4260.16 a
?0.6460.03 c
?27.7960.27 a
?27.25
?28.89
Table 5. The histogram crop water stress index (HCWSI) and variance, skewness, and kurtosis of temperature distributions of
canopies in different treatments, for Castela ˜o vines on two dates
Values are means 6SE. Within treatments, mean values are calculated from the maximum mean and minimum mean across images; absolute values are
calculated from the absolute maximum and minimum across images.
TreatmentDate Distribution measures Index of variation
HCWSI Variance SkewnessKurtosisWithin imagesWithin treatments
Mean Absolute
NI5 Aug
26 Aug
5 Aug
26 Aug
5 Aug
26 Aug
5 Aug
26 Aug
53.4768.02
67.3266.51
66.0865.15
72.4866.70
58.9967.79
66.1763.21
57.4865.09
64.2464.29
1.4460.28
2.4760.38
2.3560.34
3.0460.44
1.7960.24
2.5760.37
1.7660.13
3.3760.38
0.4160.21
1.1260.25
0.5560.18
1.2260.17
0.5260.25
1.0660.15
0.6260.21
1.0960.13
1.9761.27
3.4261.45
1.7060.84
1.3660.39
1.8360.89
1.9760.81
1.2160.42
0.4960.29
0.1960.03
0.3360.02
0.2560.04
0.3360.02
0.2160.03
0.3260.02
0.2160.01
0.3060.01
0.19
0.19
0.19
0.15
0.11
0.08
0.21
0.16
0.37
0.53
0.55
0.52
0.48
0.40
0.37
0.45
PRD
DI
FI
Table 6. Examples of IGand CWSI derived from temperatures
of the treatment canopies (Tcanopy) and interpolated NI and FI
canopy temperatures to correspond to each Tcanopymeasurement
(TNIand TFI., respectively), for Aragone ˆs grapevines
Data were collected on 13 August 2004.
Time
Tcanopy
TNI
TFI
IG
CWSI
08:39:36
08:40:56
08:55:00
09:03:54
13:29:48
13:30:31
13:31:04
13:32:40
13:33:31
13:34:03
19.06
20.63
19.8
23.53
31.4
30.45
29.20
31.27
29.51
30.88
20.43
20.53
21.63
22.32
35.82
35.81
35.81
35.80
35.79
35.79
18.06
18.29
20.20
20.81
29.82
29.83
29.82
29.83
29.83
29.83
1.37
?0.04
?4.59
?0.44
2.80
8.59
?10.53
3.14
?19.44
4.69
0.42
1.04
?0.28
1.80
0.26
0.10
?0.10
0.24
?0.05
0.18
Thermal imaging of stress in grapevines7 of 11

Page 8

Thermal indices
Significant differences between treatments in the absolute
temperatures of areas of canopy suggest that this may be an
effective method of distinguishing stressed from non-
stressed plants. However, in other situations, where there
are no randomized treatments to compare, such as moni-
toring a plant canopy over time for the purposes of
irrigation scheduling, it can be difficult to distinguish
increasing plant stress from an increase in air temperature.
The use of references is designed to eliminate such a
problem. Lower IG values in NI than FI leaves on 6
August 2003 reflected greater stress in the NI vines. IG,
however, often showed values below 0, resulting in an
inability to distinguish canopies with extremely low
conductance from canopies with very high conductance.
This problem does not occur with CWSI, for which
canopies with very high conductance should show very
low values of CWSI and canopies with very low conduc-
tance would always show relatively high values of CWSI.
Nonetheless, the individual wet and dry leaves used as
references to calculate these indices may not be good
references for whole canopies, whereas moving whole
branches around the vineyard to act as more suitable
references is not convenient. Different lengths of time
between spraying the ‘wet’ leaves and taking the image are
bound to lead to errors. Furthermore, previous work with
grapevine (Jones et al., 2002) and cotton (unpublished)
suggests a treatment effect on wet reference leaf temper-
atures, which would be possible if increased evaporation in
well-irrigated canopies affects the measured temperatures
of the wet references. For these reasons, it was decided to
explore an alternative to the use of wet and dry reference
leaves. Extrapolating between repeated measurements of
NI and FI canopies, as done in 2004, allowed the use of
indices similar to those currently used with reference
leaves, but without the associated problems listed above.
It is suggested that this system may be preferable to the use
of wet and dry leaves. This method allows easy detection
of areas within a field where vines are stressed, and could
be incorporated into vineyard management. It does not
require any additional meteorological data.
Canopy architecture
Water loss can be minimized by closing stomata, but also
by reducing light absorbance. Rolling leaves, wilted leaves
or steep leaf angles, or reduced canopy leaf area through
reduced growth and shedding of older leaves, are all
involved in minimizing water loss from plants (Ludlow
and Muchow, 1990; Chaves and Oliveira, 2004), and are
Fig. 6. Stomatal conductance (gs) of grapevine leavesin three treatments
(PRD, DI, and RDI) and reference canopies NI and FI at different times
and on different dates in August 2004. For the three treatments, n=9.
For NI and FI canopies, n=3. Data are means 6SE. Different letters
indicate significantly different means in Tukey tests following ANOVA.
Fig. 7. Average canopy temperature in FI (open circles), NI (filled
circles), PRD (open squares), RDI (open triangles), and DI (filled
squares) canopies in the morning (A) and afternoon (B, C) of 13 August
2004. Lines in (A) and (B) indicate approximate eye-fits to temperature
data for NI (dashed lines) and FI (solid lines). Data in (A) and (B)
represent individual plants, and in (C) are averages per treatment 6SE,
with different letters indicating significantly different means in Tukey
tests following ANOVA, n=9, except for NI and FI where n=3.
8 of 11 Grant et al.

Page 9

also important for preventing photoinhibition (Werner
et al., 2002). If leaf movements occur after stomata close,
they may contribute to canopy cooling, as intercepted
irradiance changes. Thus, with thermal imaging, a canopy
with closed stomata may not be distinguished from one
with open stomata, but different architecture. Greater
sensitivity of canopy temperatures than leaf temperatures
to irrigation would occur if variation in the angle of
individual leaves obscures differences relating to stomatal
conductance. These masking effects of individual leaf
angles may cancel out over whole canopies, if the dis-
tribution of leaf angles is similar in different canopies. It
had been considered that leaf angle may vary measurably
in different treatments, but did not find evidence to support
this. Additionally, variation in leaf angle was not correlated
with any temperature variables, and leaf drooping during
wilting did not affect canopy temperature. Nonetheless,
using a model to derive stomatal conductance from leaf
temperature and vice versa (Leinonen et al., 2006), the
range of orientations and angles found in the canopies
measured would be expected to have a large influence on
the relationship between conductance and temperature. The
inverse correlation of canopy temperatures with gsin 2004,
but lack of correlation between leaf temperature and gs,
suggests that individual leaf temperatures may bear less
relationship to gsthan temperatures of areas of canopies.
In the data collected in 2003, the possibility that
temperature differences between canopies might relate to
canopy density rather than stomatal conductance alone
cannot be ruled out. Irrigated plants had significantly
greater leaf area than non-irrigated vines (Table 3). A
reduced leaf area would result in a reduced area of
Fig. 8. A crop water stress index (CWSINI/FI, A) and INI/FI(B, C) as
calculated from treatment canopy temperatures and NI and FI canopy
temperatures for the afternoon of 13 August 2004 (A, B), and INI/FI
calculated from treatment leaf temperatures and NI and FI leaf
temperatures for the afternoon of 19 August 2004 (C). In all cases,
values for NI and FI were interpolated/extrapolated from three points
to the same time as the treatment measurement. n=9 in (A), 7–8 in (B),
and 4–6 in (C). Data are means 6SE. Different letters indicate signific-
antly different means in Tukey tests following ANOVA.
Fig. 9. Correlation between the stomatal conductance of representative
leaves in canopies measured on 13 August 2004 and the temperature
of those canopies (A), and CWSINI/FI(B), and between the variance
of canopy temperature distribution and the temperature of the canopy.
n=27 for A and B, and n=33 for C.
Thermal imaging of stress in grapevines9 of 11

Page 10

transpiring surface per area of canopy in a thermal image,
which may lead to a lower estimate of conductance, even
if the conductance per leaf is the same. Decreased leaf
density in non-irrigated vines means that the average
canopy temperature could be higher than the average of
a similarly transpiring but denser canopy. This effect would
accentuate differences between treatments in Tcanopy, but
not in Tleaf, and may partly explain why greater sensitivity
of canopy temperatures than leaf temperatures to irrigation
was found.
Temperature variability within canopies
It has been suggested that an alternative to using the
absolute temperatures of canopies to determine stress is to
use the variation in temperatures within a canopy (Gardner
et al., 1981; Fuchs, 1990; Leinonen and Jones, 2004). No
evidence was found to support this in Castela ˜o and
Aragone ˆs, with no greater variation of thermal distribution
within vines in stressed than well-irrigated canopies. This
may relate to the non-random distribution of leaf angles in
grapevine canopies, as the effect would only be expected in
a canopy with random leaf orientation (Fuchs, 1990), and
therefore should be investigated in other crops. Indeed, in
grapevines, leaf angles could become more uniform as the
vines become more stressed and leaves droop, with the
effect that temperature variability within canopies could
be greater in less stressed canopies. Furthermore, images
of dense canopies may contain a greater diversity of leaf
angles, which again could lead to greater temperature
variance within non-stressed than within stressed canopies.
However, temperature variability between rather than
within canopies of grapevines may be a better indicator
of stress, if some locations in a field become water deficient
before others. In these experiments, indices of within-
treatment variation were not consistently higher in stressed
canopies than in well-irrigated canopies, but analysis of
variation between canopies could aid in the detection of
individual stressed plants or areas of poor soil or faults in
irrigation systems.
Conclusions
It is suggested that the average temperatures of areas of
canopies containing several leaves are perhaps more useful
for distinguishing between irrigation treatments than the
temperatures of individual leaves. Average temperatures
over several leaves per canopy may be expected to reduce
the impact of variation in leaf angles. The effect of the
interaction of stomatal conductance and canopy architec-
ture on canopy temperature needs further investigation, but
it has been shown that thermal imaging can be a useful tool
for distinguishing between stressed and unstressed vines.
Temperature differences found between canopies under
two different irrigation regimes are encouraging for the
application of thermal imaging for irrigation scheduling.
While an estimation of stomatal conductance requires
additional meteorological data, the CWSI that was used
here, which requires no additional information, may be
sufficient for the detection of relative stress required for
irrigation scheduling. This CWSI using NI and FI canopies
as alternatives to wet and dry references removes problems
associated with wet and dry reference leaves. Since it does
not require any props or equipment other than the thermal
camera, it is also a more rapid and convenient approach,
and may be useful for commercial application. This needs
to be tested in experiments in which scheduling is de-
termined by different methods, one of these being thermal
imaging alone.
Fig. 10. Frequency distributions of the average length of the two main
lateral leaf veins (A), of the orientation angle (B), and of the inclination
angle (C) of leaves in all treatments in August 2004. n=100.
10 of 11 Grant et al.

Page 11

Acknowledgements
This work was largely funded by the European Union project
STRESSIMAGING; contract HPRN-CT-2002-00254 ‘Diagnosis
and analysis of plant stress using thermal and other imaging
techniques’, and the EU Project FP6-2002-INCO-WBC-509163.
WATERWEB (2004 experiment). OMG and LT benefited from EU
Training Network fellowships under STRESSIMAGING. We would
like to acknowledge Carlos Lopes, Filipe Barros, and Tiago dos
Santos for additional information regarding the field experiments,
and two anonymous reviewers for their helpful comments.
References
Boissard P, Guyot G, Jackson RD. 1990. Factors affecting the
radiative temperature of a vegetative canopy. In: Steven MD, Clark
JA, eds. Applications of remote sensing in agriculture. London:
Butterworths, 45–72.
Bravdo BA, Hepner Y, Loinger C, Cohen S, Tabacmen H. 1985.
Effect of irrigation and crop level on growth, yield and wine
quality of Cabernet Sauvignon. American Journal of Enology and
Viticulture 36, 132–139.
Bryant RB, Moran MS. 1999. Determining crop water stress
from crop temperature variability. In: Proceedings of the fourth
international airborne remote sensing conference and exhibition/
21st Canadian symposium on remote sensing, Ontario, Canada,
June1999.
AnnArbor,Michigan:
289–296.
Chaerle L, Van Caeneghem W, Messens E, Lambers H,
Van Montagu M, Van Der Straeten D. 1999. Presymp-
tomatic visualization of plant–virus interactions by thermography.
Nature Biotechnology 17, 813–816.
Chaves MM, Oliveira MM. 2004. Mechanisms underlying plant
resilience to water deficits: prospects for water-saving agriculture.
Journal of Experimental Botany 55, 2365–2384.
Cohen Y, Alchanatis V, Meron M, Saranga Y, Tsipris J. 2005.
Estimation of leaf water potential by thermal imagery and spatial
analysis. Journal of Experimental Botany 56, 1843–1852.
Davies WJ, Bacon MA, Thompson DS, Sobeih W, Rodrı ´guez LG.
2000. Regulation of leaf and fruit growth in plants growing in
drying soil: exploitation of the plants’ chemical signalling system
and hydraulic architecture to increase the efficiency of water use in
agriculture. Journal of Experimental Botany 51, 1617–1626.
dos Santos TP, Lopes CM, Rodriguez ML, de Souza CR,
Maroco JP, Pereira JS, Silva JR, Chaves MM. 2003. Partial
rootzone drying: effects on growth and fruit quality of field-
grown grapevines (Vitis vinifera). Functional Plant Biology 30,
663–671.
Dry PR, Loveys BR. 1999. Grapevine shoot growth and stomatal
conductance are reduced when part of the root system is dried.
Vitis 38, 151–154.
Dry PR, Loveys BR, Du ¨ring H. 2000. Partial drying of the rootzone
of grape. I. Transient changes in shoot growth and gas exchange.
Vitis 39, 3–7.
Dry PR, Loveys BR, McCarthy MG, Stoll M. 2001. Strategic
irrigation management in Australian vineyards. Journal Interna-
tional des Sciences de la Vigne et du Vin 35, 129–139.
Esteban MA, Villanueva MJ, Lissarrague JR. 2001. Effect of
irrigation on changes in the anthocyanin composition of the skin
of cv. Tempranillo (Vitis vinifera L.) grape berries during ripening.
Journal of the Science of Food and Agriculture 81, 409–420.
European Environment Agency. 2004. Impacts of Europe’s
changing climate. EEA report no. 2/2004.
ERIM International,
Fuchs M. 1990. Infrared measurement of canopy temperature and
detection of plant water stress. Theoretical and Applied Climatol-
ogy 42, 253–261.
Gardner BF, Blad BL, Watts DG. 1981. Plant and air temperatures
in differentially irrigated corn. Agricultural Meteorology 25,
207–217.
Grant OM, Stoll M, Jones HG. 2004. Partial rootzone drying
does not affect fruit yield of raspberries. Journal of Horticul-
tural Science and Biotechnology 79, 125–130.
Guiliani R, Flore JA. 2000. Potential use of infra-red thermometry
for the detection of water stress in apple trees. Acta Horticulturae
537, 383–392.
Houghton JT, Ding Y, Griggs DJ, Noguer M, van der Linden PJ,
Dai X, Maskell K, Johnson CA. 2001. Climate change 2001: the
scientific basis. Cambridge: Cambridge University Press.
Idso SB. 1982. Non-water stressed baselines: a key to measuring
and interpreting plant water stress. Agricultural Meteorology
27, 59–70.
Jones HG. 1999. Use of infrared thermometry for estimation of
stomatal conductance as a possible aid to irrigation scheduling.
Agricultural and Forest Meteorology 95, 139–149.
Jones HG. 2004. Application of thermal imaging and infrared
sensing in plant physiology and ecophysiology. Advances in
Botanical Research 41, 107–163.
Jones HG, Stoll M, Santos T, Sousa C, Chaves MM, Grant OM.
2002. Use of infrared thermography for monitoring stomatal
closure in the field: application to grapevine. Journal of Experi-
mental Botany 53, 2249–2260.
Leinonen I, Grant OM, Tagliavia CPP, Chaves MM, Jones HG.
2006. Estimating stomatal conductance with thermal imagery.
Plant, Cell and Environment 29, 1508–1518.
Leinonen I, Jones HG. 2004. Combining thermal and visible
imagery for estimating canopy temperature and identifying plant
stress. Journal of Experimental Botany 55, 1423–1431.
Lindenthal M, Steiner U, Dehne H-W, Oerke E-C. 2005. Effect
of downy mildew development on transpiration of cucumber lea-
ves visualized by digital infrared thermography. Phytopathology
95, 233–240.
Lopes CM, Pinto PA. 2000. Estimation de la surface foliaire
principale et secundaire d’un rameau de vigne. Progre `s Agricole
et Viticole 117, 160–166.
Loveys BR, Dry PR, McCarthy MG. 1999. Using plant physiology
to improve the water use efficiency of horticultural crops. Acta
Horticulturae 537, 187–199.
Ludlow MM, Muchow RC. 1990. A critical evaluation of traits for
improving crop yields in water-limited environments. Advances in
Agronomy 43, 107–153.
Prichard TL. 1992. A volume balance approach to quality wine
grape irrigation. In: Walker MA, Kliewer WM, eds. Viticultural
practices. Davis, CA: University of California, 12–23.
Reynolds AG, Naylor AP. 1994. ‘Pinot noir’ and ‘Riesling’ grape-
vines respond to water stress duration and soil water-holding
capacity. Hortscience 29, 1505–1510.
Smart DR, Taryn L, Bauerle CS, Eissenstat DM. 2004. Root
survivorship under deficit and dryland farming conditions for
1103P and 101-14MGT rootstocks in the Oakville region of the
Napa Valley. 7th International Symposium on Grapevine Physi-
ology and Biotechnology, June 21–25, 2004, University of
California, Davis, CA.
STATISTICA. 1995. STATISTICA for Windows, release 5.0. Tulsa:
StatSoft Inc.
Werner C, Correia O, Beyschlag W. 2002. Characteristic patterns
of chronic and dynamic photoinhibition of different functional
groups in a Mediterranean ecosystem. Functional Plant Biology
29, 999–1011.
Thermal imaging of stress in grapevines11 of 11