Irrig Sci (2012) 30:511–522
DOI 10.1007/s00271-012-0382-9
ORIGINAL PAPER
Assessment of vineyard water status variability by thermal
and multispectral imagery using an unmanned aerial vehicle
(UAV)
Javier Baluja • Maria P. Diago • Pedro Balda •
Roberto Zorer • Franco Meggio • Fermin Morales
Javier Tardaguila
•
Received: 14 August 2011 / Accepted: 4 January 2012 / Published online: 21 August 2012
Ó Springer-Verlag 2012
Abstract The goal of this study was to assess the water
status variability of a commercial rain-fed Tempranillo
vineyard (Vitis vinifera L.) by thermal and multispectral
imagery using an unmanned aerial vehicle (UAV). The
relationships between aerial temperatures or indices
derived from the imagery and leaf stomatal conductance
(gs) and stem water potential (Wstem) were determined.
Aerial temperature was significantly correlated with gs
(R2 = 0.68, p \ 0.01) and Wstem (R2 = 0.50, p \ 0.05).
Furthermore, the thermal indices derived from aerial
imagery were also strongly correlated with Wstem and gs.
Moreover, different spectral indices were related to vineyard water status, although NDVI (normalized difference
vegetation index) and TCARI/OSAVI (ratio between
transformed chlorophyll absorption in reflectance and
optimized soil-adjusted vegetation index) showed the
highest coefficient of determination with Wstem (R2 = 0.68,
p \ 0.05) and gs (R2 = 0.84, p \ 0.05), respectively.
While the relationship with thermal imagery and water
status parameters could be considered as a short-term
response, NDVI and TCARI/OSAVI indices were probably
reflecting the result of cumulative water deficits, hence a
long-term response. In conclusion, thermal and multispectral imagery using an UAV allowed assessing and
mapping spatial variability of water status within the
vineyard.
Communicated by S. Ortega-Farias.
Introduction
J. Baluja M. P. Diago P. Balda J. Tardaguila (&)
Instituto de Ciencias de la Vid y del Vino, University of La
Rioja, CSIC, Gobierno de La Rioja, Madre de Dios, 51,
26006 Logroño, Spain
e-mail: [email protected]
R. Zorer
GIS and Remote Sensing Unit, Biodiversity and Molecular
Ecology Department—DBEM, IASMA Research and Innovation
Centre, Fondazione Edmund Mach, Via E. Mach 1,
38010 San Michele all’Adige, TN, Italy
e-mail: [email protected]
F. Meggio
Department of Environmental Agronomy and Crop Science,
University of Padova, AGRIPOLIS—Viale dell’Università 16,
35020 Legnaro, PD, Italy
e-mail: [email protected]
F. Morales
Department of Plant Nutrition, Experimental Station of Aula
Dei, CSIC, Apdo. 13034, 50080 Zaragoza, Spain
e-mail: [email protected]
Drought is a common stressor in many wine-producing
countries located in areas with scarce rainfall and high
evaporative demand. Water stress may negatively affect
vegetative growth, yield, grape composition and wine
sensorial attributes (Chapman et al. 2005; Chaves et al.
2007; Ojeda et al. 2002). Therefore, optimizing irrigation
should be a key issue of research and a priority for most
wine countries, and it could be summarized as ‘better
grapes per drop’.
The usefulness of different physiological parameters
and their applicability for water stress detection and irrigation management in grapevines was reviewed by different authors (Acevedo-Opazo et al. 2008b; Cifre et al.
2005; Jones 2004a). Atmospheric monitoring of weather
data to calculate evapotranspiration and sensors for
monitoring soil water content have been used for scheduling irrigation in agriculture (Allen et al. 1999).
However, Jones (2004b) suggested that greater precision
123
512
in the application of irrigation can potentially be obtained
using plant-based responses rather than measurements of
soil water status. The main plant-based techniques for
irrigation scheduling are the following: sap flow meters
(Escalona et al. 2002), trunk growth variations by dendrometers (Intrigliolo and Castel 2007), stem water
potential (Acevedo-Opazo et al. 2008a; Girona et al.
2006), canopy temperature, assessed by thermal imaging
(Costa et al. 2010), reflectance indices (De Bei et al.
2011) and chlorophyll fluorescence (Flexas et al. 2002).
Although very reliable and informative, many of these
tools monitor only a single plant in the field and/or are
time-consuming; therefore, they are unsuited for detecting
spatial variation in water status within a vineyard.
Canopy temperature has long been recognized as an
indicator of plant water status and as a potential tool for
irrigation scheduling (Costa et al. 2010; Jones 1999).
Thermal imaging allows the visualization of differences in
surface temperature from emitted infrared radiation. This
technique relies on the fact that when water is lost through
the stomata, leaf temperature decreases, but once stomata
close transpiration no longer occurs, the temperature of
sunlit leaves increases (Costa et al. 2010). Thus, leaf or
canopy temperatures can be considered as suitable indicators of stomatal conductance and hence canopy stress
(Fuchs 1990; Jones et al. 2002).
Thermal stress indices such as crop water stress index
(Idso et al. 1981) and the stomatal conductance index
(Jones 1999) have been developed to reduce the impact of
environmental variation. Significant correlations between
canopy temperatures (or thermal indices) and stomatal
conductance or stem water potential were observed in a
commercial Tempranillo vineyard under different water
regimes (Ochagavı́a et al. 2011). Similar results were found
in grapevines under other conditions (Grant et al. 2007;
Jones et al. 2002; Möller et al. 2007).
The water status of cotton (Cohen et al. 2005) and olive
trees (Ben-Gal et al. 2009) has been assessed by thermal
imagery from mobile or aerial platforms in order to expand
the viewing area. Both RGB (Red, Green, Blue) and thermal cameras were mounted on a truck-crane about 15 m
above the canopy to assess vineyard water status in Israel
(Möller et al. 2007). Jones et al. (2009) obtained thermal
and visible images from a balloon and observed a strong
relationship between ground canopy temperature and aerial
temperature. A vineyard is typically a discontinuous crop,
as vines are planted in rows, and canopy width is usually
within the range of 30–50 cm for a vertically shoot-positioned trellis system. Most published data from thermal
imaging of grapevine relate to images taken facing the
rows (Grant et al. 2007; Jones et al. 2002). However,
recently Ochagavı́a et al. (2011) have showed that zenithal
123
Irrig Sci (2012) 30:511–522
temperatures were similarly indicative of vine water status
compared to lateral imaging. All these results suggested
that aerial imaging might be suitable for monitoring vineyard water status.
Precision viticulture has developed in the last few years
for the assessment of intra- and inter-field spatial variability and to implement site-specific management systems (Arnó et al. 2009; Bramley 2010). Implementation of
precision irrigation systems could contribute to reducing
water consumption in viticulture. Site-specific irrigation
may be defined as an irrigation regime that is matched, in
timing and amount, to the actual crop need at the smallest
manageable scale, to achieve the desired crop responses
(Cohen et al. 2005). The methods to characterize the
spatial variability of the vine water status at the vineyard
scale were reviewed by Acevedo-Opazo et al. (2008a).
Research efforts in precision viticulture demonstrate the
feasibility of remote sensing as a consistent method to
study the vineyard spatial variability (Hall et al. 2002) and
can be used to assess the water status spatial variability of
the vineyard. The usefulness of high-spatial resolution
information provided by aerial imagery to define plant
water restriction zones within vineyards was tested by
Acevedo-Opazo et al. (2008b). They observed that individual NDVI (normalized difference vegetation index)
defined zones displaying large differences in vigour, yield
and vine water status, so this information could be used
for the assessment of the vineyard spatial variability.
Moreover, the images obtained by remote sensing can be
used to map vineyard water status supporting irrigation
scheduling decisions.
The use of airborne remote sensing is currently
increasing due to superior temporal flexibility and the
acquisition of higher resolution imagery, as compared with
satellite platforms. Recently, miniaturized digital sensors
onboard UAVs have been used, providing very high
spectral and spatial resolutions. Research work conducted
to detect water and nutritional stress in crops has been
successfully conducted using UAVs (Berni et al. 2009;
Zarco-Tejada et al. 2008), showing its feasibility for precision agriculture applications. Sepulcre-Cantó et al.
(2009) used an UAV for mapping growth spatial variability
of a commercial vineyard.
The goal of this study was to assess the water status of
a commercial rain-fed vineyard of Tempranillo cv. (Vitis
vinifera L.) by thermal and multispectral imagery using
an unmanned aerial vehicle (UAV). The relationships
between aerial temperatures or indices derived from the
imagery and stomatal conductance and water potential
were therefore determined. In addition, spatial variability of vineyard water status was evaluated and
mapped.
Irrig Sci (2012) 30:511–522
513
Materials and methods
Airborne campaign: thermal and multispectral indices
Study site description
Imagery was acquired from the Tempranillo vineyard on 5
August 2008 (veraison) in collaboration with Quantalab
(IAS, CSIC, Spain) using an unmanned aerial vehicle
(UAV) equipped with a multispectral sensor Multiple
Camera Array (MCA-6, Tetracam, Inc., California, USA)
and a thermal camera Thermovision A40 M (FLIR, USA).
The flight height of 200 m AGL (above ground level) for
multispectral and thermal acquisition allowed to obtaining
imagery at 10-cm spatial resolution in the VIS–NIR (visible–near-infrared) region and at 30-cm spatial resolution in
the thermal region, respectively. The MCA sensor was
configured to comprise six bands in the VIS–NIR spectral
range at 530, 550, 570, 670, 700 and 800 nm, respectively.
Imagery was acquired at 13:00 for VIS/NIR and 14:00 for
thermal in two different flights (with 1 h of time acquisition, approximately, for each one). The weather conditions
during the flight were satisfactory, with clear sky, atmosphere stability, moderate wind (mean wind velocity = 0.0 m/s, maximum wind velocity = 0.9 m/s), 50 %
of air relative humidity and a relatively stable air temperature (mean air temperature = 27.5 °C, maximum air
temperature = 29.9 °C and minimum air temperature =
26.1 °C). The geometric corrections and image calibrations
for both multispectral and thermal datasets were performed
as described by Berni et al. (2009).
The airborne campaign was conducted over the vineyard
in order to compute image-derived spectral and thermal
indices (Table 1). Three thermal indices were tested as
indicators of water status in vines: crop water stress index
(CWSI) (Idso et al. 1981), stomatal conductance index (Ig)
(Jones 1999) and a second formulation of the stomatal
conductance index (I3) (Jones 1999). Indices were calculated from temperatures obtained from the aerial image
after atmospheric correction as follows:
The study was conducted in summer 2008 in a 5-ha commercial vineyard located in Logroño (428280 400 ’N,
28290 3900 W, at 490 m), in Rioja wine-producing region
(Spain), under Mediterranean–continental climate. Tempranillo (Vitis vinifera L.) vines, grafted on 110R rootstock,
were trained to a vertically shoot-positioned trellis system
(VSP) on a double-cordon Royat and spur-pruned to 10–12
buds per vine. Vine spacing was 2.8 m between rows and
1.2 m in the row. No irrigation was applied during the
growing season (rain-fed vineyard). All water shoots were
pulled out in May, and shoot trimming was performed in
July. Standard management practices in Rioja Appellation
were applied in this vineyard, without the presence of
cover crop between the rows. The vineyard presented a
7 % linear concave slope facing southeast, and the soil
showed spatial variability. The variation of the vineyard
soil properties was previously determined and described in
detail by Tardaguila et al. (2011) in a companion study.
Global radiation, wind speed, air temperature and
humidity were acquired by a meteorological station (Siar,
La Rioja Government) located in Logroño (42°260 1600 N,
2°300 5400 W, at 465 m).
Physiological measurements
Stomatal conductance was measured with a leaf gas
exchange system (Model GFS-3000, Heinz Walz GMBH,
Effeltrich, Germany). The measurements were conducted
at full ambient light with a mean temperature inside the
cuvette of 28.7 °C (r, standard deviation of 1.8 °C), a
mean leaf temperature of 32.9 °C (r = 2.85 °C) and a
mean VPD (vapour pressure deficit) of 3.16 kPa
(r = 0.52 kPa). Air and leaf temperature were recorded
with a hand-held thermal gun (Model Optris LS, Mesurex
SL, Berlin, Germany). Both stomatal conductance and
canopy temperature were measured between 10:00 and
12:00. A Scholander pressure bomb (Model 600, PMS
Instruments Co., Albany, USA) was used to assess stem
water potential, of the previous measured leaves, at midday
(14:00, solar noon), after 2 h of dark adaptation in
aluminium foil bags.
All physiological measurements were conducted in 10
blocks, consisting of three adjacent vines, randomly
selected in the Tempranillo vineyard, at veraison, on 5
August 2008, simultaneous with the unmanned aerial
vehicle overflight and thermal and multispectral image
acquisition. All physiological parameters were measured
on two sun-exposed fully mature main leaves per vine at
each block.
CWSI ¼ Tcanopy Twet
Tdry Twet
Ig ¼ Tdry Tcanopy
Tcanopy Twet
I3 ¼ ðTcanopy Twet Þ Tdry Tcanopy
ð1Þ
ð2Þ
ð3Þ
where Tcanopy (°C) was the canopy temperature obtained
from the thermal images, and Tdry (°C) and Twet (°C) were
the lower and upper boundary temperatures, corresponding
to a fully transpiring leaf with open stomata and nontranspiring leaf with closed stomata, respectively. Note that
Tdry and Twet are equivalent to Tbase and Tmax in the original
formulation of CWSI as described by Idso et al. (1981).
In the present study, in the absence of a well-watered
reference vine, Twet was computed as the pixel with minimum value extracted by a moving window of 3 9 3 pixels
within the studied vines, after masking the rows, and then
123
514
Irrig Sci (2012) 30:511–522
Table 1 Spectral indices used in the present study calculated using MCA sensor set of bands
Index
Equation
References
Chlorophyll absorption ratio
ðR550ððR700R500Þ670Þ550Þ
pffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
ffi
CAR ¼ ½ðR700R500Þ670þR670þ
2
Broge and Leblanc (2001)
Chlorophyll absorption ratio
CAR1 ¼ CAR* R700
R670
Kim et al. (1994); Broge
and Leblanc (2001)
Greenness index
GI ¼ R550
R670
Zarco-Tejada et al. (2005a, b)
Green normalized difference
vegetation index
R800R550
GNDVI ¼ R800þR550
Gitelson and Merzlyak (1998)
Modified chlorophyll
absorption in reflectance
Modified chlorophyll
absorption in reflectance
MCARI ¼ ½ðR700 R670Þ 0:2 ðR700 R550Þ ðR700=R670Þ
Daughtry et al. (2000)
MCARI1 ¼ 1:2 ½2:5 ðR800 R670Þ 1:3 ðR800 R550Þ
Haboudane et al. (2004)
½2:5ðR800R670Þ1:3ðR800R550Þ
pffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
ffi
MCARI2 ¼ 1:2
2
Haboudane et al. (2004)
qffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
MSAVI ¼ 12 2 R800 þ 1 ð2 R800 þ 1Þ2 8 ðR800 R670Þ
Qi et al. (1994)
Þ1
ffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
MSR ¼ pðR800=R670
Chen (1996)
MTVI3 ¼ 1:2 ½1:2 ðR800 R550Þ 2:5 ðR670 R550Þ
Rodrı́guez-Pérez et al. (2007)
ððR700R500Þ670Þ
Modified chlorophyll
absorption in reflectance
Improved SAVI
(soil-adjusted VI)
with self-adjustment factor L
Modified simple ratio
ð2R800þ1Þ 6ðR8005R670Þ0:5
ðR800=R670Þþ1
Modified triangular VI
R800R670
R800þR670
Normalized difference
vegetation index
NDVI ¼
Optimized soil-adjusted
vegetation index
OSAVI ¼ ð1 þ 0:16Þ ðR800 R670Þ=ðR800 þ R670 þ 0:16Þ
Rondeaux et al. (1996)
Simple ratio index
SRI ¼ R800
R550
Jordan (1969)
Photochemical reflectance index
PRI ¼ R530R570
R530þR570
Fuentes et al. (2001);
Gamon and Surfus (1999)
Renormalized difference VI
ffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
RDVI ¼ pR800R670
R800þR670
Reujean and Breon (1995)
Transformed chlorophyll
absorption in reflectance
TCARI ¼ 3 ½ðR700 R670Þ 0:2 ðR700 R550Þ ðR700=R670Þ
Haboudane et al. (2002)
TCARI/OSAVI
Þ0:2ðR700R550ÞðR700=R670Þ
TCARI/OSAVI ¼ 3½ððR700R670
1þ0:16ÞðR800R670Þ=ðR800þR670þ0:16Þ
Haboudane et al. (2002)
Rouse et al. (1974)
on pure canopy pixels, based on the method proposed by
Alchanatis et al. (2010). Tdry was computed as Tair ? 5 °C
as proposed by Cohen et al. (2005) and Möller et al. (2007).
Alternative estimates of Twet and Tdry based on energy
balance model were also tested (Jones 1999). In particular,
Twet was derived from the following equation:
Twet ¼ Ta þ
rRH raW cRni
rHR VPD
qa Cp ðcraW þ DrHR Þ craW þ DrHR
ð4Þ
where Ta is the air temperature (°C), rHR is the resistance to
the radiative heat transfer (s m-1) (Jones 1992, 1999),
based on the characteristic leaf dimension of 0.1 m; raW is
the boundary layer resistance for water vapour (s m-1)
(Jones 1992), c is the psychrometric constant (Pa K-1), Rni
is the net radiation (W m-2), qa is the density of the air
(kg m-3), Cp is the specific heat of air (J kg-1 K-1), and D
is the slope of the saturation vapour pressure curve
(Pa K-1). Tdry (°C) was also theoretically estimated on
the assumption of isothermal radiation as suggested by
Jones (1999):
123
Tdry ¼ Ta þ
rHR Rni
:
qa Cp
ð5Þ
Net radiation estimation
In this study, net isothermal radiation was assumed to be
equal to the absorbed short-wave radiation (Jones 1999;
Leinonen et al. 2006). Furthermore, a continuous surface
for net radiation was computed by means of r.sun algorithm (Hofierka 1997, Hofierka and Súri 2002), available in
GRASS GIS (Neteler and Mitasova 2007). This algorithm
requires elevation, slope or aspect of the terrain as inputs
that were computed from a 1-m resolution digital elevation
model by r.slope.aspect algorithm (Hofierka et al. 2009).
Additionally, the algorithm requires a raster map or a
constant value of Linke turbidity factor (Kasten 1996) and
albedo. Linke turbidity factor was retrieved by the solar
application data (SODA) (http://www.soda-is.com/eng/
services/linke_turbidity_info.html), which computes the
Linke turbidity factor based on several global information
Irrig Sci (2012) 30:511–522
from satellites like global sky radiation (SRB), precipitable
water vapour (NVAP), aerosol optical depth (pathfinder)
and ground information about turbidity from radiation or
aerosol measurements (AERONET), for more details see
Rigollier et al. (2000) and Holben et al. (2001). Albedo, in
the present study, was assumed as a constant value of 0.2
based on the results of Pieri and Gaudillére (2003) and in
the absence of net radiation measurements.
515
values) in terms of root mean square error (RMSE) and
standardized root mean square error (sRMSE). For interpolation, anisotropy was taken into account to avoid the
strange features caused by the linear rows pattern. Finally,
stomatal conductance and water potential were mapped
applying the regression lines to these interpolated surfaces,
trying to minimize the effect of other kind of land covers in
the prediction.
Pure canopy pixels extraction
Results
Rows extraction was carried out to obtain pure vine pixels
by means of the watershed algorithm available in GRASS
GIS called r.watershed (Metz et al. 2011). Watershed
algorithm has been demonstrated as a good technique to
extract canopy information in palm orchard (Cohen et al.
2011) from thermal images and in combination with a
second procedure based on a threshold applied to the histogram of the final accumulation image (Karantzalos and
Argialas 2006; Safren et al. 2007).
In the present study, the watershed algorithm was
applied to the thermal image but also to an inverse NDVI
computed image (1—NDVI) obtaining consistent results.
As a second step to extract canopy pixels, a threshold based
on the Otsu method (Otsu 1979) was applied on the
watershed algorithm resulting flow accumulation map.
The extracted canopy pixels were reclassified to a binary
image (1, 0), which was used as a mask, to extract pixel
values from the multispectral and thermal images, because
of the different spatial resolution of the two sources of
information, two masks were used, one for multispectral
and another one for thermal images.
Statistical analysis and mapping
Pearson correlations and linear regressions were computed
between extracted thermal and multispectral information
and physiological variables measured at field level using
Statistica 8.0 (Tulsa, Oklahoma, USA). In the absence of
an independent dataset for the validation of the models, due
to the limited number of samples available (n = 10), a
variation of the k-fold cross-validation was performed,
called leave-one-out cross-validation (LOOCV). This
method uses a single observation from the original sample
as the validation data, and the remaining observations as
the training data. This is repeated for the entire dataset.
The final discontinuous information provided by the
mask built through watershed algorithm was interpolated
to a surface by means of Spline with tension technique
(Mitasova et al. 2005), available in GRASS 5.0 GIS, as this
method conserves the original pixel values and provided, in
the present study, the best LOOCV results from the
different exact methods (methods that conserve original
Pure canopy pixels extraction
A bimodal histogram was found for temperature image and
NDVI (Fig. 1d). This allowed to performing histogram
thresholding techniques, which presented the problem of a
large loss of information with the more restrictive threshold
(Fig. 1b–c) or the inclusion of a large number of soil pixels. The use of watershed algorithm allowed to obtain a
more continuous information within the rows (Fig. 1b–c),
with less problems of soil inclusion than the application of
low restrictive thresholds.
Vine water status assessment by aerial image
temperature and thermal indices
Relationships between canopy physiological parameters
and data retrieved from the imagery were performed.
Results are presented from the comparison between
physiological parameters such as leaf temperature
(Fig. 2), stem water potential (Wstem) and leaf stomatal
conductance (gs) (Fig. 3) with canopy temperature
retrieved from the thermal imagery. In both cases, the
relationships were obtained using vine pixels filtered by
watershed method.
Results presented in Fig. 2 showed a significant positive
correlation, with a coefficient of determination (R2) of 0.65
between leaf temperature measurements taken at ground
level and concurrent estimates of canopy temperature by
thermal imagery. The relationship found through linear
regression with thermal temperature yielded a consistent
RMSE of 0.56 °C, computed by the LOOCV method, in
leaf temperature estimation despite the presence of some
noise in data at high temperatures (T [ 32 °C) due to an
overall overestimation made by aerial imagery.
Results presented in Fig. 3 showed a negative correlation between stem water potential (R2 = 0.50; p value
\ 0.05) and leaf temperature computed by the thermal
imagery acquired by the UAV. Furthermore, a negative
correlation was found between stomatal conductance
(R2 = 0.68; p value \ 0.01) and leaf temperature computed by the thermal imagery acquired by the UAV.
123
516
Irrig Sci (2012) 30:511–522
Fig. 1 Example of image
processing for canopy pixel
extraction. a Pseudocolour
composition with 10-cm image
resolution acquired by UAV,
with the area shown at
b remarked by a green polygon.
b Rows extraction by watershed
algorithm (red) and by threshold
applied based on NDVI
histogram (NDVI [ 0.75).
c Rows extraction by watershed
algorithm (red) and based on a
threshold applied based on
NDVI histogram
(NDVI [ 0.75) for the entire
image. d NDVI histogram for
the entire area shown at (a)
Fig. 2 Linear regressions obtained between ground leaf temperature
measured using a thermal gun and the temperature estimated from
aerial imagery (n = 10) with intercept (solid line) and with intercept
fixed to 0 (dotted line). Where ** was significant at p value\0.01 and
* was significant at p value \0.05. The root mean square error
(RMSE) has been computed by means of leave-one-out crossvalidation technique
Linear regressions between the thermal indices, calculated from canopy temperatures extracted from aerial
thermal imagery and different Twet and Tdry estimated
values, and Wstem and gs are shown in Table 2. Coefficients
of determination found for linear regression revealed consistent results for canopy and air temperature difference
(Tc - Ta) with values of R2 = 0.50 and 0.69 for Wstem and
gs, respectively. Significant correlations were encountered
for most of the thermal indices and Wstem with coefficients
of determination ranging from 0.46 up to 0.54. Furthermore, R2 between 0.54 and 0.73 were found for gs. From
the three different methods tested for the estimation of Tdry
123
Fig. 3 Linear regressions obtained with temperature estimated from
aerial imagery with stomatal conductance (a) and stem water potential
(b) at veraison (n = 10). Where the regressions that show ** were
significant at p value \0.01 and with * when p value \0.05
and Twet, all of them seemed to be equally suitable for the
computation of the thermal indices, as their coefficients
of determination with Wstem and gs were very similar.
Irrig Sci (2012) 30:511–522
517
However, energy balance estimation for the upper boundary temperature tended to underestimate the Tdry reference
as compared with the same reference temperature estimated as Tair ? 5 °C; furthermore, the Twet was overestimated by the same method as compared with the Twet value
estimated from the imagery.
Vine water status assessment by multispectral indices
The coefficients of determination between some vegetation
indices, which are related to vine vigour and physiological
status, and Wstem and gs were computed through linear
regression and their values are shown in Table 3.
Wstem was better related with multispectral indices,
linked to vine vigour. In particular, NDVI, MSR and SRI
provided the best coefficient of determination values with
R2 = 0.68, 0.66 and 0.64, respectively, all of them being
significant at p value \ 0.01. Other widely used indices
like GI, GNDVI, TCARI and TCARI/OSAVI showed
significant (p value \ 0.05) coefficients of determination
around *0.50. Regarding the linear regressions with gs, in
particular TCARI/OSAVI, TCARI, MSR, SRI and NDVI
indices yielded the best results with coefficients of determination within the range 0.75–0.84, all significant at
p \ 0.01. Furthermore, PRI and GI yielded R2 = 0.46 and
0.63, respectively, when correlated with gs.
Mapping stomatal conductance and water potential
The quantitative results, as shown in Table 3, allowed the
use of NDVI for the spatial computation of both Wstem
and gs maps since it proved to be a feasible indicator
of vineyard physiological status (Fig. 4). Linear relationships through NDVI with Wstem [Wstem = 4.802
NDVI—5.30 Mpa; R2 = 0.68] and gs [gs = 1,400
NDVI—1,125 mmolm-2 s-1; R2 = 0.75] were used for
spatial maps computation. The more stressed vines were
found at the west part the plot, with a trend from the
southwest towards the northeast. Furthermore, a larger
amount of variability was found for stomatal conductance,
when quantile maps were compared, these results being
associated with the different scale of measurement for the
two variables.
Calibration of vine water status relationships through
thermal and multispectral indices
Vine water status estimates were compared to assess consistency of relationships proposed in Tables 2 and 3
through thermal and multispectral indices, respectively
(Fig. 5). Physiological parameters Wstem and gs were estimated using the best relationships presented in Tables 2
Table 2 Coefficients of determination for linear regression and sign for Pearson’s correlations (r) found between thermal indices derived from
aerial imagery with stem water potential (Wstem) and stomatal conductance (gs)
Indices
gs (mmol m-2 s-1)
Wstem (MPa)
Sign (r)
R2
Sign (r)
R2
Tc - Ta
-
0.50*
-
0.69**
(Tc - Ta)/NDVI
?
0.52*
?
0.69**
CWSIa
-
0.50*
-
0.68**
CWSIb
-
0.52*
-
0.70**
CWSIc
-
0.50*
-
0.70**
a
Ig
?
0.50*
?
0.72**
Igb
?
0.54*
?
0.73**
Igc
?
0.49*
?
0.72**
I3a
-
0.42*
-
0.50*
I3b
-
0.43*
-
0.54*
c
-
0.46*
-
0.72**
I3
The significance of coefficients and coefficients of determination, found for linear regressions, is indicated as follows: *p value \ 0.10;
**p value \ 0.05; n = 10
a
Tdry and Twet estimated by energy balance model
b
Tdry as Tair ? 5 °C and Twet by energy balance model
c
Tdry estimated by energy balance model and Twet from imagery
CWSI ¼ ðTcanopy Twet Þ=ðTdry Twet Þ
Ig ¼ ðTdry Tcanopy Þ=ðTcanopy Twet Þ
I3 ¼ ðTcanopy Twet Þ=ðTdry Tcanopy Þ
123
518
Irrig Sci (2012) 30:511–522
Table 3 Coefficients of determination for linear regression and sign for Pearson0 s correlations (r) found between spectral indices derived from
aerial imagery (calculated as Table 1) with stem water potential (Wstem) and stomatal conductance (gs)
Indices
gs (mmol m-2 s-1)
Wstem (MPa)
Sign (r)
R2
Sign (r)
R2
CAR
?
0.33
?
0.22
CAR1
?
\0.00
?
0.03
GI
GNDVI
?
?
0.54*
0.58*
?
?
0.63**
0.70**
MCARI
?
0.01
?
0.32
MCARI1
-
0.21
-
0.02
MCARI2
?
0.00
?
0.20
MSAVI
?
0.11
?
0.01
MSR
?
0.66**
?
0.78**
MTVI3
-
0.01
-
0.32
NDVI
?
0.68**
?
0.75**
OSAVI
?
0.35
?
0.13
PRI
-
0.25
-
0.46*
RDVI
?
0.10
?
0.10
SRI
?
0.64**
?
0.78**
TCARI
?
0.45*
?
0.80**
TCARI/OSAVI
-
0.58*
-
0.84**
The significance of coefficients of determination, found for linear regressions, is indicated as follows: *p value \ 0.10; **p value \ 0.05; n = 10
Discussion
Fig. 4 Water potential (Wstem) and stomatal conductance (gs) maps at
veraison obtained through NDVI index computed from 10-cm image
resolution acquired by UAV
and 3 in terms of coefficient of determination. Estimates
through NDVI and Ig2 yielded a coefficient of determination of 0.68 and 0.54, respectively. The relationships found
for gs with multispectral [gs = -834.8 TCARI/OSAVI ? 229.8 mmolm-2 s-1; R2 = 0 0.84, p \ 0.05] and
thermal indices [gs = 145.6Ig2—13.22 mmol m-2 s-1;
R2 = 0.73, p \ 0.05]. were, in general, better than such
provided for water potential, in terms of explained
variance.
123
One of the problems detected on the thresholding techniques used to extract the canopy pure pixels was the loss
of information or the inclusion of soil. The differences
between vines in terms of vigour made difficult the thresholding selection based on NDVI or temperature. Using the
more restrictive thresholds, a large amount of vines disappeared into the final mask in the less vigour parts of the
vineyard. Furthermore, when a less restrictive threshold
was applied, a large number of soil pixels were included
into the mask, especially in the areas with more vigour. The
use of watershed algorithm allowed an automatic way for
canopy extraction, because the number of soil pixels in the
zone with more vigour was similar to the more restrictive
NDVI threshold and there were no missing vines in the
areas with less vigour. The algorithm takes into account the
neighbourhood relations between pixels, the flow from
bigger to lower values is searched within the image, and
this allows finding the purest pixels as an increasing
function to the row centre. Except for death vines, always
pure canopy pixels were found, and single high-value
pixels between the rows were filtered by the algorithm.
Thermal images can be processed directly in watershed
algorithm, as soil pixels should show the higher temperature values within the rows; on the contrary for multispectral products, such as NDVI, an inverse image should
be computed before applying the watershed algorithm, to
Irrig Sci (2012) 30:511–522
Fig. 5 Scatter-plot between stem water potential (a) measured in the
marked vines and estimated by NDVI (white dots) and Ig2 (black
dots), Scatter-plot between stomatal conductance (b) measured in the
marked vines and estimated by TCARI/OSAVI (white dots) and Ig2
(black dots). The estimated values and root mean square errors
(RMSE) have been computed by means of leave-one-out crossvalidation technique
set the higher values to soil pixels. The development of
these methodologies could allow a more rapid and suitable
way for structural canopy parameters extraction where
usually soil effect is not always treated properly.
The relationships obtained between aerial image temperature and physiological parameters demonstrated the
feasibility of assessing vine water status using thermal
imagery acquired by an UAV in rain-fed vineyards. Results
presented in Fig. 2 highlighted the reliability of canopy
temperature estimation and the feasibility of estimating
Wstem and gs using canopy temperature computed by
thermal imagery.
Coefficients of determination resulting from the linear
regressions between physiological data and remote sensing
indices, derived from multispectral and thermal imagery,
enabled the feasibility of the application of multispectral
and thermal methods for vine physiology assessment. In
particular, vegetation indices widely utilized for vegetative
vigour and leaf chlorophyll content retrieval in grapevines
(Meggio et al. 2008, 2010; Zarco-Tejada et al. 2005a)
yielded consistent results for NDVI and TCARI/OSAVI
indices. It allowed the computation of distribution maps in
519
five levels of Wstem and gs, which showed the within-field
variability potentially linked to vine water status.
Regarding the different methods employed for computing the reference temperatures, Twet and Tdry needed for
calculating the thermal indices, all methods proved to be
consistent and reliable when correlated with Wstem and gs.
However, indices computed using Tdry as Tair ? 5 °C and
Twet by energy balance yielded the highest coefficients of
determination for both physiological parameters.
Leave-one-out cross-validation estimated values for
Wstem and gs through multispectral and thermal imagery
were compared to the original ones. Both methods exhibited a near perfect 1:1 relationship. When comparing Wstem
estimation through NDVI and Ig2 by means of RMSE,
NDVI performed better, whereas a better estimation of gs
was obtained from TCARI/OSAVI as compared to Ig2.
Relationships calibration allowed to assess their estimation
capability for vine water status assessment and revealed a
general good agreement between multispectral and thermal
imagery. Multispectral and thermal indices relationships
showed slight underestimates of Wstem between -1.2 and
-1.4 MPa, which, even if affecting the estimate reliability,
remained in good agreement with ground measurements. Ig2
tended to underestimate water potential highest values while
tended to a little overestimation of stomatal conductance
smallest values. TCARI/OSAVI showed a consistent reliability to estimate gs over the entire range. Both methods,
multispectral and thermal, showed a low RMSE for Wstem
and gs estimation.
The relationship between aerial image temperature, or
thermal indices, and vine water status confirmed the
capability of thermal remote sensing for vineyard water
status assessment in the short term (Acevedo-Opazo et al.
2010; Kazmierski et al. 2011). Moreover, the correlation
between vine water status and NDVI or TCARI/OSAVI
indices (widely used in the literature for vigour and leaf
chlorophyll detection purposes), showed in the present
study, revealed that multispectral vegetation indices could
be also used for water status assessment in vineyards
undergoing high and continuous water stress. Moderate to
high water stress in the rain-fed vineyard during the whole
season was mainly caused by scarce annual precipitation
registered (data not shown) in the study area. Vegetative
growth reduction, caused by the recurrence of water stress
conditions, related to soil properties in this vineyard as
described by Tardaguila et al. (2011), in the absence of
irrigation support resulted to great differences in NDVI and
TCARI/OSAVI indices. As a result, high determination
coefficients have been found for linear regression between
those spectral indices and Wstem and gs, revealing a longterm water stress effect on vine vegetative growth. This
cumulative effect was confirmed by results shown in
Fig. 5, enabling the assessment of the within-field
123
520
variability in terms of vegetative growth reduction and the
creation of maps, which can be used as a decision support
tool for irrigation scheduling.
The applicability of best irrigation strategies needs the
development of consistent models capable of explaining
processes occurring during the day. While NDVI and
TCARI/OSAVI indices do not reveal detectable daily changes as instead plant physiology processes do, that is, photosynthetic activity and transpiration, the present work suggests
that temperature represents a short-term indicator of even
daily stress onsets. For these reasons, aerial temperature and
thermal indices estimates are recommended for water status
assessment in a rain-fed vineyard and for the development of
irrigation models based on UAV technology.
Conclusions
The use of both multispectral and thermal cameras onboard
of the UAV allowed taking high-resolution images, to
assess vine water status. In fact, high correlations were
found between indices based on thermal and on multispectral images with vineyard water status. An UAV could
be used to assess vine water status and to map withinvineyard variability, which could be useful for irrigation
practices.
While the relationship with thermal imagery and derived
indices can be considered as a short-term response, a kind of
instantaneous ‘‘shot’’ of the plant water status situation,
spectral vegetation indices are probably the result of cumulative water deficits. Both approaches allowed assessing the
water status spatial variability within the vineyard, which can
be useful to create maps or to optimize the number and the
placement of automated irrigation stations in representative
parts of the vineyard, and in the near future to develop new
water scheduling models based on proximal or remote
sensing data.
Acknowledgments This work was supported by the Agencia de
Desarrollo Económico de La Rioja (ADER) with the Project TELEVITIS 2008-I-ID-00123. The authors want to thank Domecq Bodegas
allowing the execution of this study in their commercial vineyard.
Gratefulness also to Dr. Pablo Zarco-Tejada and Dr. Guadalupe
Sepulcre-Cantó for their collaboration and advice on this study. Also,
the authors wish to acknowledge the Quantalab (IAS, CSIC) participation in the UAV flights. Also, the authors want to thank Markus
Metz for the support provided with r.watershed algorithm and the
PGIS team (FEM, IASMA), particularly Markus Neteler, for
the support with GRASS GIS. Fermı́n Morales thanks Gobierno de
Aragón (A03 research group) for financial support.
References
Acevedo-Opazo C, Tysseire B, Taylor J, Ojeda H, Guillaume S
(2010) Spatial prediction model of the vine (Vitis vinifera L.)
123
Irrig Sci (2012) 30:511–522
water status using high resolution ancillary information. Precis
Agr 11:358–378
Acevedo-Opazo C, Tisseyre B, Ojeda H, Ortega-Farias S, Guillaume
S (2008a) Is it possible to assess the spatial variability of vine
water status? J Int Sci Vigne Vin 42:203–219
Acevedo-Opazo C, Tisseyre B, Guillaume S, Ojeda H (2008b) The
potential of high resolution information to define withinvineyard zones related to vine water status. Precis Agr 9:
285–302
Alchanatis V, Cohen Y, Cohen S, Moller M, Sprinstin M, Meron M,
Tsipris J, Saranga Y, Sela E (2010) Evaluation of different
approaches for estimating and mapping crop water status in
cotton with thermal imaging. Precis Agric 11:27–41
Allen RG, Pereira LS, Raes D, Smith M (1999) Crop evapotranspiration. FAO irrigation and drainage paper 56. Rome: FAO
Arnó J, Martı́nez Casanovas J, Ribes-Dasi M, Rosell JR (2009)
Review. Precision viticulture. Research topics, challenges and
opportunities in site-specific vineyard management. Spanish J
Agric Res 7:779–790
Ben-Gal A, Agam N, Alchanatis V, Cohen Y, Yermiyahu U, Zipori I,
Presnov E, Sprintsin M, Dag A (2009) Evaluating water stress in
irrigated olives: correlation of soil water status, tree water status,
and thermal imagery. Irrig Sci 27:367–376
Berni JAJ, Zarco-Tejada PJ, Suarez L, Fereres E (2009) Thermal and
narrow-band multispectral remote sensing for vegetation monitoring from an unmanned aerial vehicle. IEEE Trans Geosci
Rem Sens 47:722–738
Bramley RGV (2010) Precision viticulture: Managing vineyard
variability for improved quality outcomes. Chapter 12. In:
Reynolds AG (ed) Understanding and managing wine quality
and safety. Woodhead Publishing, Cambridge, pp 445–480
Broge NH, Leblanc E (2001) Comparing prediction power and
stability of broadband and hyperspectral vegetation indices for
estimation of green leaf area index and canopy chlorophyll
density. Rem Sens Environ 76:156–172
Chapman DM, Roby G, Ebeler SE, Guinard JX, Matthews MA (2005)
Sensory attributes of Cabernet Sauvignon wines made from
vines with different water status. Aust J Grape Wine Res 11:
339–347
Chaves MM, Santos TP, Souza CR, Ortuño MF, Rodrigues ML,
Lopes CM, Maroco JP, Pereira JS (2007) Deficit irrigation in
grapevine improves water-use efficiency while controlling
vigour and production quality. Ann Appl Biol 150:237–252
Chen J (1996) Evaluation of vegetation indices and modified simple
ratio for boreal applications. Can J Remote Sens 22:229–242
Cifre J, Bota J, Escalona JM, Medrano H, Flexas J (2005)
Physiological tools for irrigation scheduling in grapevine (Vitis
vinifera L.) An open gate to improve water-use efficiency? Agr
Ecosyst Environ 106:159–170
Cohen Y, Alchanatis V, Meron M, Saranga S, Tsipris J (2005)
Estimation of leaf water potential by thermal imagery and spatial
analysis. J Exp Bot 56:1843–1852
Cohen Y, Alchanatis V, Prigojin A, Levi A, Soroker V, Cohen Y
(2011) Use of aerial thermal imaging to estimate water status on
palm trees. Precis Agric. doi:10.1007/s11119-011-9232-7
Costa JM, Grant OM, Chaves MM (2010) Use of Thermal Imaging in
Viticulture: Current Application and Future Prospects. In: Delrot
S, Medrano H, Or E, Bavaresco L, Grando S (eds) Methodologies and results in grapevine research. Springer, New York,
pp 135–150
Daughtry CST, Walthall CL, Kim MS, Brown de Colstoun E,
McMurtrey JE III (2000) Estimating corn leaf chlorophyll
concentration from leaf and canopy reflectance. Rem Sens
Environ 74:229–239
De Bei R, Cozzolino D, Sullivan W, Cynkar W, Fuentes S, Dambergs
R, Pech J, Tyerman S (2011) Non-destructive measurement of
Irrig Sci (2012) 30:511–522
grapevine water potential using near infrared spectroscopy. Aust
J Grape Wine Res 17:62–71
Escalona JM, Flexas J, Medrano H (2002) Drought effects on water
flow, photosynthesis and growth of potted grapevines. Vitis
41:57–62
Flexas J, Escalona JM, Evain S, Gulias J, Moya I, Osmond CB,
Medrano H (2002) Steady-state chlorophyll fluorescence (Fs)
measurements as a tool to follow variations of net CO2
assimilation and stomatal conductance during water-stress in
C3 plants. Physiol Plant 114:231–240
Fuchs M (1990) Infrared measurement of canopy temperature and
detection of plant water stress. Theor Appl Climatol 42:253–261
Fuentes DA, Gamon JA, Qiu HL, Sims DA, Roberts DA (2001)
Mapping Canadian boreal forest vegetation using pigment and
water absorption features derived from the AVIRIS sensor.
J Geophys Res 106:33565–33577
Gamon JA, Surfus JS (1999) Assessing leaf pigment content and
activity with a reflectometer. New Phytol 143:105–117
Girona J, Mata AM, Del Campo J, Arbone A, Bartra E, Marsal J
(2006) The use of midday leaf water potential for scheduling
deficit irrigation in vineyards. Irrig Sci 24:115–117
Gitelson AA, Merzlyak MN (1998) Remote sensing of chlorophyll
concentration in higher plant leaves. Adv Space Res 22:689–692
Grant OM, Tronina L, Jones HG, Chaves MM (2007) Exploring
thermal imaging variables for the detection of stress responses in
grapevine under different irrigation regimes. J Exp Bot 58:
815–825
Haboudane D, Miller JR, Tremblay N, Zarco-Tejada PJ, Dextraze L
(2002) Integrated narrow-band vegetation indices for prediction
of crop chlorophyll content for application to precision agriculture. Rem Sens Environ 84:416–426
Haboudane D, Miller JR, Pattey E, Zarco-Tejada PJ, Strachan IB
(2004) Hyperspectral vegetation indices and novel algorithms for
predicting green LAI of crop canopies: modeling and validation
in the context of precision agriculture. Rem Sens Environ
90:337–352
Hall A, Lamb DW, Holzapfel B, Louis J (2002) Optical remote
sensing applications in viticulture: a review. Aust J Grape Wine
Res 8:36–47
Hofierka J (1997) Direct solar radiation modelling within an open GIS
environment. Conference, Vienna, Austry
Hofierka J, Súri M (2002) The solar radiation model for open source
GIS:implementation and applications. Open source GIS-GRASS
users conference. Trento, Italy
Hofierka J, Mitasova H, Neteler M (2009) Chapter 17: geomorphometry in GRASS GIS. Dev Soil Sci 33:387–410
Holben BN, Tanré D, Smirnov A, Eck TF, Slutsker I, Abuhassan N,
Newcomb WW, Schafer JS, Chatenet B, Lavenu F, Kaufman YJ,
Vande-Castle J, Setzer A, Markham B, Clark D, Frouin R,
Halthore R, Karneli A, O’Neill NT, Pietras C, Pinker RT, Voss
K, Zibordi G (2001) An emerging ground-based aerosol
climatology. Aerosol optical depth from AERONET. J Geophys
Res 106:12067–12097
Idso SB, Jackson RD, Pinter PJ, Reginato RJ, Hatfield JL (1981)
Normalizing the stress degree- day parameter for environmental
variability. Agr Forest Meteorol 24:45–55
Intrigliolo DS, Castel JR (2007) Evaluation of grapevine water status
from trunk diameter variations. Irrig Sci 26:49–59
Jones HG (1992) Plants and microclimate. Cambridge University
Press, Cambridge
Jones HG (1999) Use of infrared thermometry for estimation of
stomatal conductance as a possible aid to irrigation scheduling.
Agr Forest Meteorol 95:139–149
Jones HG (2004a) Irrigation scheduling: advantages and pitfalls of
plant based methods. J Exp Bot 55:2427–2436
521
Jones HG (2004b) Application of thermal imaging and infrared
sensing in plant physiology and ecophysiology. Adv Bot Res
41:107–163
Jones HG, Stoll M, Santos T, De Sousa C, Chaves MM, Grant OM
(2002) Use of infrared thermography for monitoring stomatal
closure in the field: application to grapevine. J Exp Bot 53:
2249–2260
Jones HG, Serraj R, Loveys BR, Xiong L, Wheaton A, Price AH
(2009) Thermal infrared imaging of crop canopies for the remote
diagnosis and quantification of plant responses to water stress in
the field. Funct Plant Biol 36:978–989
Jordan CF (1969) Derivation of leaf area index from quality of light
on the forest floor. Ecology 50:663–666
Karantzalos K, Argialas D (2006) Improving edge detection and
watershed segmentation with anisotropic diffusion and morphological levelings. Int J Rem Sens 27:5427–5434
Kasten F (1996) The linke turbidity factor based on improved values
of the integral Rayleigh optical thickness. Sol Energy 56:
239–244
Kazmierski M, Glemas P, Rousseau J, Tisseyre B (2011) Temporal
stability of within-field patterns of NDVI in non irrigated
Mediterranean vineyards. J Int Sci Vigne Vin 45:61–73
Kim MS, Daughtry CST, Chappelle EW, McMurtrey JE (1994) The
use of high spectral resolution bands for estimating absorbed
photosynthetically active radiation (APAR) In Proceedings for
ISPRS’94 299–306.Val d’Isere, France
Leinonen I, Grant OM, Tagliavia CPP, Chaves MM, Jones HG (2006)
Estimating stomatal conductance with thermal imagery. Plant
Cell Environ 29:1508–1518
Meggio F, Zarco-Tejada PJ, Miller JR, Martı́n P, González MR,
Berjón A (2008) Row orientation and viewing geometry effects
on row-structured vine crops for chlorophyll content estimation.
Can J Rem Sens 34(3):220–234
Meggio F, Zarco-Tejada PJ, Núñez LC, Sepulcre-Cantó G, González
MR, Martı́n P (2010) Grape quality assessment in vineyards
affected by iron deficiency chlorosis using narrow-band physiological remote sensing indices. Rem Sens Environ 114:1968–1986
Metz M, Mitasova H, Harmon RS (2011) Accurate stream extraction
from large, radar-based elevation models. Hydrol Earth Syst Sci
15:667–678. doi:10.5194/hess-15-667-2011
Mitasova H, Mitas L, Harmon RS (2005) Simultaneous spline
approximation and topographic analysis for lidar elevation data
in open-source GIS. Geosci Rem Sens Lett IEEE 2:375–379
Möller M, Alchanatis V, Cohen Y, Meron M, Tsipris J, Naor A,
Ostrovsky V, Sprintsin M, Cohen S (2007) Use of thermal and
visible imagery for estimating crop water status of irrigated
grapevine. J Exp Bot 58:827–838
Neteler M, Mitasova H (2007) Open GIS: a grass GIS approach.
Springer, New York
Ochagavı́a H, Grant OM, Baluja J, Diago MP, Tardáguila J (2011)
Exploring zenithal and lateral thermography for the assessment
of vineyard water status. Proceedings of 17th International
Symposium of GiESCO. Asti–Alba (Italia)
Ojeda H, Andary C, Kraeva E, Carbonneau A, Deloire A (2002)
Influence of pre and postveraison water deficit on synthesis and
concentration of skin phenolic compounds during berry growth
of Vitis vinifera cv Shiraz. Am J Enol Vitic 53:261–267
Otsu N (1979) Threshold selection method from gray-level histograms. IEEE Trans Syst Man Cybern 9:62–66
Pieri P, Gaudillére JP (2003) Sensitivity to training system parameters
and soil surface albedo of radiation intercepter by vine rows.
Vitis 42:77–82
Qi J, Chehbouni A, Huete AR, Kerr Y (1994) A modified soil
adjusted vegetation index (MSAVI). Rem Sens Environ 48:
119–126
123
522
Reujean J, Breon F (1995) Estimating PAR absorbed by vegetation
from bidirectional reflectance measurements. Rem Sens Environ
51:375–380
Rigollier C, Bauer O, Wald L (2000) On the clear sky model of the
ESRA with respect to the heliosat method. Sol Energy 68:33–48
Rodrı́guez-Pérez JR, Riaño D, Carlisle E, Ustin S, Smart DR (2007)
Evaluation of hyperspectral reflectance indexes to detect grapevine water status in vineyards. Am J Enol Vitic 58:302–317
Rondeaux G, Steven M, Baret F (1996) Optimization of soil-adjusted
vegetation indices. Rem Sens Environ 55:95–107
Rouse J, Haas RH, Schell JA, Deering DW, Harlan JC (1974)
Monitoring the vernal advancement and retrogradation (Greenwave effect) of natural vegetation. RS Center, A Texas, GSF
Center—1974—Texas A & M University, Remote Sensing Center
Safren O, Alchanatis V, Ostrovsky V, Levi O (2007) Detection of
green apples in hyperspectralimages of apple-tree foliage using
machine vision. Trans ASABE 50:2303–2313
Sepulcre-Cantó G, Diago MP, Balda P, Martı́nez de Toda F, Morales
F, Tardaguila J (2009) Monitoring vineyard spatial variability of
123
Irrig Sci (2012) 30:511–522
vegetative growth and physiological status using an unmanned
aerial vehicle (UAV). In: Proceedings of GIESCO symposium.
Davis, USA
Tardaguila J, Baluja J, Arpon L, Balda P, Oliveira M (2011)
Variations of soil properties affect the vegetative growth and
yield components of ‘‘Tempranillo’’ grapevines. Precis Agr 12:
762–773
Zarco-Tejada PJ, Berjón A, López-Lozano R, Miller JR, Martı́n P,
Cachorro V, González MR, de Frutos A (2005a) Assessing
vineyard condition with hyperspectral indices: leaf and canopy
reflectance simulation in a row-structured discontinuous canopy.
Rem Sens Environ 99:271–287
Zarco-Tejada PJ, Ustin SL, Whiting ML (2005b) Temporal and
spatial relationships between within-field yield variability in
cotton and high-spatial hyperspectral remote sensing imagery.
Agron J 97:641–653
Zarco-Tejada PJ, Berni JAJ, Suárez L, Fereres E (2008) A new era in
remote sensing of crops with unmanned robots. SPIE Newsroom.
doi:10.1117/2.1200812.1438