(This is a sample cover image for this issue. The actual cover is not yet available at this time.)
This article appeared in a journal published by Elsevier. The attached
copy is furnished to the author for internal non-commercial research
and education use, including for instruction at the authors institution
and sharing with colleagues.
Other uses, including reproduction and distribution, or selling or
licensing copies, or posting to personal, institutional or third party
websites are prohibited.
In most cases authors are permitted to post their version of the
article (e.g. in Word or Tex form) to their personal website or
institutional repository. Authors requiring further information
regarding Elsevier’s archiving and manuscript policies are
encouraged to visit:
http://www.elsevier.com/copyright
Author's personal copy
Computers and Electronics in Agriculture 85 (2012) 24–32
Contents lists available at SciVerse ScienceDirect
Computers and Electronics in Agriculture
journal homepage: www.elsevier.com/locate/compag
Prediction of leaf area index in almonds by vegetation indexes
Jose L. Zarate-Valdez a,b,⇑, Michael L. Whiting b, Bruce D. Lampinen c, Samuel Metcalf c,
Susan L. Ustin b, Patrick H. Brown c
a
Centro Regional Universitario del Noroeste, Universidad Autonoma Chapingo, Colima 163 Norte, Cd. Obregon, Sonora, Mexico
Center for Spatial Technologies and Remote Sensing (CSTARS), Department of Land, Air, and Water Resources, University of California, Davis, CA 95616, USA
c
Department of Plant Sciences, University of California, Davis, CA 95616, USA
b
a r t i c l e
i n f o
Article history:
Received 15 June 2011
Received in revised form 5 March 2012
Accepted 11 March 2012
Keywords:
Leaf area index
Vegetation indices
Multispectral indices
Canopy light interception
a b s t r a c t
Three levels of scale for determining leaf area index (LAI) were explored within an almond orchard of alternating rows of Nonpareil and Monterey varieties using hemispherical photography and mule lightbar
(MLB) at ground level up to airborne and satellite imagery. We compared LAI estimates of 56 fisheye photos
strategically placed in the orchard to validate 500,000 MLB point scans of a small portion of the aisles
between tree rows to water and vegetation indexes of MASTER (MODIS/ASTER simulator) and Landsat 5
imagery. The high correlation of fisheye photo LAI to MLB LAI estimates establishes this new method
against the measurement standard within the plant community while significantly increasing sample size.
MLB LAI and MASTER vegetation indexes, such as NDWI (normalized difference water index), GMI (Gitelson–Merzlyak index) and NDVI (normalized difference vegetation index), were highly correlated
(r2 = 0.90). In addition, a high correlation (r2 = 0.80) between the MLB measured LAI and selected Landsat
derived vegetation indexes (VI) was found. This scaling and validation of LAI estimate expands the spatial
area and frequency of determination for time series analysis of crop phenology studies.
Ó 2012 Elsevier B.V. All rights reserved.
1. Introduction
1.1. LAI and measurement methods
Leaf area index (LAI) is defined as the total, one-sided area of
leaves per unit ground surface area (Watson, 1947). This parameter
is essential to modeling the processes occurring in the soil–
plant–atmosphere continuum, such as evapotranspiration, CO2
assimilation, earth surface light extinction, and for estimating primary productivity of natural and managed ecosystems (Baret and
Guyot, 1991; Turner et al., 2004). Bréda (2003) and Jonckheere
et al. (2004) provide a comprehensive review of the methods available for LAI determination. Direct methods for LAI determination
are time consuming and destructive, and alternatively, indirect
methods such as point quadrant, allometric and non-contact meth-
Abbreviations: EVI, enhanced vegetation index; fPAR, fractional PAR (photosynthetically active radiation) intercepted by the canopy; GMI, Gitelson–Merzlyak
index; LADP, leaf angle distribution parameter; LAI, leaf area index; MASTER,
MODIS/ASTER simulator; MCARI, modified chlorophyll absorption reflectance
index; MLB, mule lightbar; NDVI, normalized difference vegetation index; NDWI,
normalized difference water index; RMSE, root mean squared error; SR, simple
ratio; VI, vegetation index(es).
⇑ Corresponding author at: Centro Regional Universitario del Noroeste, Universidad Autonoma Chapingo, Colima 163 Norte, Cd. Obregon, Sonora, Mexico. Tel./fax:
+52 644 413 7171.
E-mail address: [email protected] (J.L. Zarate-Valdez).
0168-1699/$ - see front matter Ó 2012 Elsevier B.V. All rights reserved.
http://dx.doi.org/10.1016/j.compag.2012.03.009
ods, have been developed. In the last two decades these non-contact, optical methods, have gained popularity through their
reliability and ease of operation, although they were conceptualized
early in the 1950s (Martens et al., 1993; Bréda, 2003).
Optical methods measure transmitted and non-intercepted
light through the canopy within part or all the photosynthetically
active radiation (PAR) spectral region between 400 and 700 nm
due to photosynthetic absorption, while reflecting the majority of
longer near infrared wavelengths for heat control (Jensen, 2007).
These techniques are based on the analysis of either the sky gap
fraction or the gap size distribution of light transmitted through
the canopy. The AccuPAR LP-80 (Decagon Devices, Inc., Pullman,
WA, USA), LAI-2000 plant canopy analyzer (Li-COR Bioscience, Lincoln, NE, USA) and hemispherical (fisheye lens) canopy photography are examples of optical methods based on gap fraction
analysis. The Tracing and Architecture of Canopies (TRAC, ThirdWave Engineering, Ottawa, Canada) and Multiband Vegetation Imager (MVI, Spectrasource Instruments, Westlake Village, Canada)
are based on the analysis of gap size distribution below the canopy.
The AccuPAR LP-80 ceptometer measures sky gap fraction by
comparing the intensity of PAR above to that below the canopy
to assess canopy light interception and leaf distribution. LAI is
determined instantaneously by the instrument when sun zenith
angle, sunbeam fraction and leaf angle distribution are specified.
The two main disadvantages of this instrument, as compared with
other devices based on gap fraction analysis, are the poor
Author's personal copy
J.L. Zarate-Valdez et al. / Computers and Electronics in Agriculture 85 (2012) 24–32
estimation of LAI in coniferous forests (Jonckheere et al., 2004;
Weiss et al., 2004; Garrigues et al., 2008) and the need of taking
multiple observations to obtain a reliable result (Bréda, 2003). This
latter can be easily overcome with highly intensive ceptometer
measurements using a mule lightbar (Lampinen et al., 2012).
In the LAI-2000, the sensor measures diffuse blue light (320–
490 nm) at five different zenith angles (7°, 23°, 38°, 53° and 68°)
to determine the gap fraction for each of the zenith angles and calculates LAI instantaneously by analyzing the remaining portion of
this highly absorbed light region that is transmitted through the
canopy compared to that measured above canopy.
Hemispherical photography is based on the estimation of position, size, density and distribution of canopy gaps. High resolution
digital cameras with 180° fisheye lenses acquire photos quickly
from beneath the canopy, and are rapidly analyzed with computer
software based on algorithms of zenith angle, light attenuation and
contrast between sky and canopy elements. In addition to LAI values, fisheye photography provides a record for characterizing canopy structure, below-canopy radiation microclimate and solar
radiation indices (Bréda, 2003).
1.2. Imagery spectral indexes and LAI
Airborne and satellite spectral estimates of LAI have used the
principle of incoming radiation (I) is either absorbed (A), transmitted (T) or reflected (R) by the canopy.
I ¼AþT þR
25
rows and pixel ground spatial distance caused by constructive
and destructive phase resonance that appears to enhance and reduce albedo within the pixel area (Meggio et al., 2008).
Beyond the estimation of LAI, some studies have aimed to predict crop yield with spectral measurements, in some cases with a
high degree of accuracy (Asrar et al., 1984; Christensen and
Goudriaan, 1993). Yield is directly related to plant cover in many
crops, mainly annual crops (Ferencz et al., 2004). Zarco-Tejada
et al. (2005) associated NDVI and 44 other VI to yield in cotton
due to ground cover, and Maas et al. (1999) demonstrated that area
covered by cotton (ground cover) was highly predictive of yield. In
the perennial crops of almond, walnut and peach, Lampinen et al.
(2012) demonstrated that yield potential is directly proportional
to light interception by the canopy. Water consumption, nutrient
status, and other precision farming-related inputs are directly
linked to LAI and plant biomass (Broge and Mortensen, 2002;
Thenkabail, 2003). Early determination of LAI and prediction of
yield potential would provide better distribution and utilization
of crop resources.
The objective of this study was to compare the effectiveness of
using MLB spatially intensive sampling measurements under orchard canopy in calibrating airborne and satellite image vegetation
indexes for determining LAI over entire orchards or multiple orchards. These spatially intensive MLB measurements were validated
from point samples using hemispherical photography.
ð1Þ
The amount of visible light in the PAR region absorbed by plant
pigments is directly related to LAI. This relationship was initially
demonstrated in cereals and other crops and natural vegetation
with sensors mounted on terrestrial platforms (Asrar et al., 1984;
Sellers, 1985; Tucker and Sellers, 1986; Christensen and Goudriaan, 1993), and then scaled up to airborne and satellite imagery
in many reports (e.g. Prince and Astle, 1986; Kerr and Ostrovsky,
2003; Petorelli et al., 2005; Glenn et al., 2008). The association of
specific wavelengths absorption or reflectance to specific pigments
and plant constituents, such as water, lignin, cellulose, starch and
protein, and with water and soil, has led to the creation of a large
number of vegetation indexes (VI) calibrated to specific biophysical
conditions. These include many VI developed to overcome limitations due to soil background, leaf inclination angle, leaf optical
properties and atmospheric conditions. While some authors consider VI are poor estimators of LAI because of only moderate,
non-linear correlation with canopy attributes, including LAI (Baret
and Guyot, 1991), and recommend using VI to predict canopy light
absorption (Glenn et al., 2008), others have found a good correlation between green LAI and chlorophyll related VIs (Viña et al.,
2011).
In these techniques, the difficulty of scaling a few point measurements from mixed components of plant canopy, non-photosynthetic vegetation (NPV) and soil within the pixel area may
reduce the accuracy of the broad view acquired with airborne
and satellite imagery. There are several sources of error contributing to low correlations between point sampled LAI and image VI,
including variation in the component mixture within each image
pixel, some misregistration of image to ground point measurement
locations, as well as inadequate number of LAI samples and measurement errors. Aspinall et al. (2002) suggest that while increasing pixel area reduces georegistration errors, large pixels also
require a substantially greater number of ground reference sites
to represent the pixel mixture of reflectance. In agricultural applications, the structure and repeated regularity of alternating row
canopy with soil in exposed furrows and clean orchard aisles presents a significant problem in determining the actual variability
within the component mixtures with misalignment between crop
2. Materials and methods
2.1. Site description
This study combines canopy data collected during the mid-season of 2009 in two experiments in separate commercial almond
(Prunus dulcis (Miller)) orchards. The orchards, of approximately
66 and 84 ha, are located near the town of Lost Hills, Kern County,
in southern San Joaquin Valley of California (Belridge 35.51°,
119.67° and Spur Dynamics 35.60°, 119.67°) and are owned
by the Paramount Farming Corp. (Los Angeles, CA). The location,
age and general characteristics of the two sites are described in
Table 1. These orchards are on soils of fine sandy loam surface,
mixed mineralogy, superactive, calcareous, Thermic Typic
Torriorthents (Kimberlina) and sandy loam surface, mixed mineralogy, superactive, calcareous, Thermic Typic Haplargids (Milham),
well drained, formed on nearly level Quaternary alluvium.
However, in portions of both orchards, zinc deficiency leaf
chlorosis was diagnosed from visual appearance and leaf analysis.
Nonpareil trees were planted in alternating rows running north
and south with Monterey and with Monterey and Wood Colony
in Belridge and Spur Dynamics sites. Before harvest, the orchards
floor was mowed and the aisles were scraped smooth. In the orchard given the name of its locality Belridge, fertilizer was injected at
various application rates into two irrigation system trials of small
emitter types (fan jet micro-sprinkler and drip). The range of nitrogen rates was 140–392 kg ha1 in two soluble forms, urea and calcium nitrate, as well as various potassium rates. This ‘‘fertigation’’
study also included an increased water application by 20% more
than the anticipated crop evapotranspiration (ETc) commonly used
by local growers to schedule irrigations. In the second orchard
named for its experiment Spur Dynamics, fertilizer with microsprinkler irrigation were consistent with general grower practices
of the area to evaluate the dynamics of almond spur bearing as related to previous bearing as well as leaf area and sunlight exposure
(Tombesi et al., 2011). These important treatment distinctions between orchards presented broad variations in canopy age and density to evaluate the robustness of LAI models. Almond trees in this
Author's personal copy
26
J.L. Zarate-Valdez et al. / Computers and Electronics in Agriculture 85 (2012) 24–32
Table 1
General characteristics of the almond orchards studied.
Orchard
Location
0
Belridge
Dynamic Spur
00
35° 30 36 N
119° 400 30 0 W
35° 360 180 0 N
119° 400 390 0 W
Area (ha)
Year planted
Irrigation system
Varieties
Trees per hectare
65.6
1999
Fan-Jet and drip
Nonpareil (50%) and Monterey (50%)
212
83.5
1996
Fan-Jet
Nonpareil (50%), Monterey (25%) and Wood colony (25%)
212
region flower and leaf out in February and March with harvest
occurring in August through October.
2.2. MLB PAR interception measurements
Natural light in the PAR region was intensively measured above
and below the canopy, on both sides of selected rows of Nonpareil
almond trees using a set of AccuPAR LP-80 ceptometers (Decagon
Devices, Inc., Pullman WA) mounted in series across the front of
a Kawasaki mule four-wheel drive vehicle (Fig. 1). The mule lightbar (MLB) consists of 18 ceptometer segments that recorded the
PAR beneath the canopy (PARbelow) as the mule ran the length of
the tree row, and recorded the full sun PAR at each end of the rows
(PARabove). The fractional PAR intercepted by the canopy (fPAR) is
calculated as
fPAR ¼ 1 PARbelow
PARabove
ð2Þ
The lightbar maintained a level PAR measurement of approximately 40 cm above the soil surface, and spanned the approximate
seven meter distance between rows for both orchards. Within
2 hours of solar noon, the mule sped north or south along the tree
rows at a nearly constant 10 km h1recording 10 measurements
per second. At this rate, readings were acquired every 0.3–0.4 m.
Geopositions were attained by accurately tracking known initial
and ending positions of the rows, while mule speed was determined by radar measured ground pass-rate. During the collection
time, the full sun PAR was nearly constant at 2000 lmol s1 m2
with a typical fluctuation of 0.5% in the reference (above canopy)
PAR used for the fPAR calculations. For a detailed description of
the MLB refer to Lampinen et al. (2012).
A sample of 10 Nonpareil rows in the Belridge orchard were
measured with the MLB on July 21, 2009 and 12 rows in Spur
Dynamics orchard were measured on July 4, 2009. Two million
fPAR measurements were collected in the two orchards within
the 4 hours of these 2 days. The fPAR measurements were between
0 (no interception) and 1 (full interception). Individual segment
measurements were aggregated into single band, rasterized
images using ENVI image processing software (ITT Visual Information Solutions, Boulder, CO). Only the Belridge fPAR image was georegistered to a georeferenced 10 cm resolution airborne color
infrared image (13 April 2009, SmartImage, Beltsville, VA), while
the Spur Dynamics image did not need further georegistration.
The fPAR sampled rows were further divided each into four
200 m sections to solve a ground registration accuracy issue of
the MLB. The resulting 40 and 48 sections for the Belridge and Spur
Dynamics orchards were the sample units used for comparison between LAI and fPAR values and vegetation indexes derived from
airborne and satellite imagery. LAI values were derived from fPAR
values computed on a pixel basis using the inverted formula for the
prediction of scattered and transmitted PAR under a canopy,
according to Eq. (3) (Norman and Campbell, 1989; Decagon
Devices, 2008).
LAI ¼
1 ½ð1 2K
Þ fb 1 ln s
A ð1 0:47 fb Þ
ð3Þ
In which K is the canopy extinction coefficient and is calculated
as a function of the leaf angle distribution parameter (LADP), and
measured zenith angle of the sun; fb is the sun beam fraction of
incident radiation, with a value of 0.98, independently collected;
A is the leaf absorptivity in the PAR band and is assumed constant
with a value of 0.86071 for a wide range of green plants (Ross,
1975; Decagon Devices, 2008), and s is the transmittance of the
canopy in the PAR range, calculated as the ratio of PAR measurements beneath the canopy and outside canopy, and which is equivalent to 1-fPAR.
Fig. 1. Mule lightbar and dimensions of the 18 PAR sensing elements.
Author's personal copy
J.L. Zarate-Valdez et al. / Computers and Electronics in Agriculture 85 (2012) 24–32
In the LAI computation, the LADP (v or chi) value was set to 1.0,
in assuming a spherical leaf distribution. However, to evaluate the
impact of this parameter on the correlation between LAI and vegetation indexes, we also set the chi values at 1.3, 1.7, 2.3 and 3.0.
2.3. Hemispherical photography LAI measurements
In addition to the MLB measurements, LAI was estimated for selected trees at the Belridge orchard using hemispherical lens (fisheye) photography. Digital cameras fitted with fisheye lenses
(Coolpix 4300, FC-E8 fisheye lens, Nikon Corp., Tokyo, Japan) captured hemispherical pictures beneath the tree canopies in the required indirect lighting conditions within the half hour before
and after sunrise or sunset. During the 5 day field campaign at
Belridge, 14 Nonpareil almond trees were photographed within
the restricted time available. Canopies were photographed at four
positions around the trees, 2 m from the base of the tree at the four
cardinal points (Fig. 2). At this distance from the tree trunk, the
photos clearly captured the open canopy above the aisles, and
the closed canopy shared between trees in the row. Fisheye photos
were not taken at the Spur Dynamics orchard. The photos were
analyzed at three thresholds of contrast with Hemiview software
(Delta-T Devices, Ltd., Cambridge, UK).
2.4. Airborne and satellite imagery
A MODIS/ASTER airborne simulator (MASTER) image including
both Belridge and Spur Dynamics orchards was acquired on 24 July
2009, 1 hour before solar noon. A general description of the MASTER flight for the image analyzed in this paper can be found in
http://www.masterweb.jpl.nasa.gov/data/MissionView.htm?Missionid=100758. MASTER imagery includes 50 bands, 25 in the visible to short wave infrared and 25 in the mid and thermal infrared
portions of the electromagnetic spectrum (Hook et al., 2001). Only
the bands in the visible, near- and short-wave infrared regions
were used for this study (see Table 2). The flight altitude was in
average 3328 m above ground level (3425 m above sea level), with
a resulting image ground spatial distance (GSD, or pixel width) of
7.2 m. The acquired image was atmospherically corrected in ENVI
using the FLAASH procedure. Color infrared composites of the
bands 3, 5, and 9 of the MASTER image for these orchards are
shown in Fig. 3. The images were transformed into the spectral indexes listed in Table 3 using band-math. The VI studied included
those based on the differential reflectance due to leaf structural
and density (NDVI, Green NDVI, simple ratio (SR), optimized soiladjusted VI (OSAVI), and enhanced VI), VI based on light absorption
by pigments and chlorophyll (Gitelson and Merzlyak index (GMI),
structural independent pigment index (SIPI), green blue ratio, simple ratio pigment index (SRPI) and modified chlorophyll absorption
27
reflectance index (MCARI)), as well as indexes based on light
absorption by leaf water (normalized difference water index
(NDWI) and Zygielbaum water stress index (ZWSI)). For comparison of MLB LAI and image VIs among orchard sections, the mean
index values were calculated from the corresponding 7.2 m wide
pixels of the MASTER image within the sections.
Landsat TM L1T processed scene from path 42 and row 35 acquired on 28 July, 2009 was downloaded from USGS site (http://
www.glovis.usgs.gov/, last verified on May 24, 2010). The 30 m
pixel image was converted to radiance and then FLAASH reflectance corrected, clipped for each orchard, and further georegistered
using the Master image as the base image. Color infrared composites of the bands 2, 3, and 4 of the Landsat image for these orchards
are shown in Fig. 3. NDVI and NDWI were the only VIs calculated
from Landsat imagery using the band math tool in ENVI. The pixels
with mixed road reflectance at the ends of each section were eliminated from the orchard sections. The mean index values for each
of the 40 or 48 sections of Belridge and Spur Dynamics orchards
were calculated for comparison to the mean reference data from
sampled trees within the sections.
2.5. Statistical Analysis
The mean values of LAI and fPAR from the MLB measurements
within each of 40 and 48 sections were calculated for exploring
relationships directly to image VI. The pixel values for the 14 vegetation indexes defined in Table 3 in the MASTER image were averaged over the same orchard sections using ENVI software.
Statistical analyses of correlation between LAI, fPAR, and VIs were
conducted in JMP (ver. 8.0, SAS Institute Inc., Cary, NC) using the
bivariate scatter plot regression procedure. The statistical coefficients of determination (r2) were calculated for the regression:
yi ¼ b0 þ
X
b1 xi þ ei
ð4Þ
where the prediction of y, e.g. LAI, at any pixel i was based on the
intercept of b0 and coefficient b1 times the predictor x, e.g. NDVI,
associated with the pixel i. The error e at sample i was used to calculate the overall error, root mean squared error of prediction
(RMSE) according to Eq. (5):
sffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
P 2
ðei Þ
RMSE ¼
n
ð5Þ
where n is the total of samples evaluated.
3. Results
3.1. Comparison of hemispherical photography LAI to MLB LAI
a
b
Fig. 2. Fisheye photo positions around the tree (a) and fisheye photo taken 2 m
north from the tree trunk (b). P in (a) represents the position of the fisheye picture
taken 2 m from the tree trunk (T).
The fisheye and MLB LAI values ranged from 1.65 to 2.48 and 3.86
to 4.90, respectively. Sampled rows of Belridge had in general higher
MLB fPAR and LAI than those of Spur Dynamics. The wide range in
LAI values is the result of a combination of different almond varieties, ages, irrigation amounts and methods, fertilizer treatments,
nutrient status and management practices. Due to a lack of coordination between the two experiments, only seven fisheye photographed trees were in the Nonpareil rows measured with the
MLB. The photo LAI mean value for each tree was compared to the
mean of the MLB LAI measurements within an 8 m radius from
the tree center. The two LAI methods were well correlated
(r2 = 0.67), as seen in Fig. 4. The LAI values between the two methods
are offset, though consistent predictors of each other.
Author's personal copy
28
J.L. Zarate-Valdez et al. / Computers and Electronics in Agriculture 85 (2012) 24–32
Table 2
Spectral characteristics of the MASTER bands used in this study.
Band
Band center (lm)
FWHM (lm)
Band
Band center (lm)
FWHM (lm)
1
2
3
4
5
6
7
8
9
10
11
12
0.4640
0.5040
0.5420
0.5840
0.6560
0.7140
0.7540
0.8040
0.8700
0.9100
0.9520
1.6100
0.0455
0.0455
0.0453
0.0438
0.0605
0.0442
0.0433
0.0436
0.0439
0.0432
0.0435
0.0555
13
14
15
16
17
18
19
20
21
22
23
24
1.6660
1.7200
1.7760
1.8280
1.9280
1.9800
2.0800
2.1620
2.2120
2.2620
2.3240
2.3900
0.0549
0.0519
0.0521
0.0518
0.0519
0.0492
0.0494
0.0485
0.0503
0.0486
0.0703
0.0623
a
b
c
d
0 50 100
200
Meters
Fig. 3. Color infrared composites of MASTER, GSD 7.2 m, (a, c) and Landsat, GSD 30 m, (b, d) images of the Belridge (a, b) and Spur Dynamics (c, d) orchards.
3.2. MLB LAI and MASTER and Landsat vegetation indexes
Within the Belridge and Spur Dynamics orchard sections, fPAR
and LAI mean values derived from the MLB were regressed against
mean values for 14 common spectral vegetation and water indexes
derived from the MASTER image pixels. The resulting r2 values are
shown in Table 4, and an illustration of the fit of MLB fPAR and LAI
with three MASTER and two Landsat VI is shown in Figs. 5 and 6.
Of the VI predictions of LAI evaluated, only MCARI1 gave r2 values smaller than 0.83, and Green/Blue and EVI gave r2 values
between 0.83 and 0.87. The remaining indexes were better predictors of LAI with r2 values greater than 0.88 to more than 0.90,
including many water (NDWI, ZWSI) and chlorophyll/pigment
(GMI and SR) absorption indexes. Conversely, the r2 values for
these VIs in predicting fPAR were lower, ranging from 0.56 to
0.83, and were quadratic, exhibiting the saturation tendency at
high fPAR values. Nevertheless, the ability of the vegetation indexes to predict either fPAR or LAI was in general acceptable; the
range in RMSE values was between 0.19 and 0.26 for LAI and
0.027 and 0.038 for fPAR for average values of 3.35 and 0.699 of
Author's personal copy
29
J.L. Zarate-Valdez et al. / Computers and Electronics in Agriculture 85 (2012) 24–32
Table 3
Spectral indexes calculated from the MASTER image.
Vegetation
index
Name
Formula
MASTER bands
Reference
NDWI
(R860 R1240)/(R860 + R1240)
(Rb9 Rb12)/(Rb9 + Rb12)
Gao, (1996)
GMI
Modified normalized difference
water index
Gitelson and Merzlyak index
R750/R550
Rb7/Rb3
SR
Simple ratio index
RNIR/RRed
Rb9/Rb5
ZWSI
Zygielbaum water stress index
R720/R520
Rb6/Rb3
GNDVI (b7)
Green normalized difference
vegetation index
Green normalized difference
vegetation index
Structural independent pigment
index
Normalized difference
vegetation index
Optimized soil-adjusted
vegetation index
Normalized difference
vegetation index
(RNIR RGreen)/(RNIR + RGreen)
(Rb7 Rb3)/(Rb7 + Rb3)
(RNIR RGreen)/(RNIR + RGreen)
(Rb9 Rb3)/(Rb9 + Rb3)
(R880 R470)/(R880 R674)
(Rb8 Rb1)/(Rb8 + Rb5)
(RNIR RRed)/(RNIR + RRed)
(Rb9 Rb5)/(Rb9 + Rb5)
1.16 (R800 R670)/(R800 + R670+0.16)
1.16 (Rb8 Rb5)/(Rb8 + Rb5 + 0.16)
(RNIR RRed)/(RNIR + RRed)
(Rb7 Rb5)/(Rb7 + Rb5)
RGreen/RBlue
2.5 (R880 R674)/
(1 + R880 + 6 R674 7.5 R470)
R430/R680
Rb3/Rb1
2.5 (Rb9 Rb5)/
(1 + Rb9 + 6 Rb5 7.5 Rb1)
Rb1/Rb5
1.2 [2.5 (R800 R670) 1.3 (R800 R550)]
1.2 [2.5 (Rb8 Rb5) 1.3 (Rb8 Rb3)]
Gitelson and
Merzlyak, (1997)
Rouse et al.,
(1974)
Zygielbaum et al.,
(2009)
Gitelson et al.,
(1996)
Gitelson et al.,
(1996)
Peñuelas et al.,
(1995)
Rouse et al.,
(1974)
Rondeaux et al.,
(1996)
Rouse et al.,
(1974)
This study
Liu and Huete,
1995
Peñuelas et al.,
(1995)
Haboudane et al.,
(2004)
(RNIR RRed)/(RNIR + RRed)
(Rb4 Rb3)/(Rb4 + Rb3)*
GNDVI (b9)
SIPI
NDVI (b9)
OSAVI
NDVI (b7)
Green/Blue
EVI
Enhanced vegetation index
SRPI
Simple ratio pigment index
MCARI1
Modified chlorophyll
absorption in reflectance index
1
Normalized difference
vegetation index
Modified normalized difference
vegetation index
Landsat NDVI
July 28 2009
Landsat NDWI
July 28 2009
*
Rouse et al.,
(1974)
Rouse et al.,
(1974)
*
(RNIR RSWIR1)/(RNIR + RSWIR1)
(Rb4 Rb5)/(Rb4 + Rb5)
Landsat band reflectances.
2.7
Table 4
Regression parameters for the linear fit of MLB LAI or fPAR (PAR interception by the
canopy) as predicted by different vegetation indexes. Belridge and Spur Dynamics
orchards data included.
Fisheye LAI
2.5
2.3
Vegetation index (predictor)
Predicted variable
LAIa
2.1
2
1.9
R² = 0.6704
1.7
1.5
NDWI
GMI
SR
ZWSI
GNDVI (b7)
GNDVI (b9)
SIPI
NDVI (b9)
OSAVI
NDVI (b7)
Green/Blue
EVI
SRPI
Landsat NDWI July 28, 2009
Landsat NDVI July 28, 2009
MCARI1
3.8
4.0
4.2
4.4
4.6
4.8
5.0
MLB LAI
Fig. 4. Correlation between fisheye LAI and MLB LAI for selected almond trees in the
Belridge orchard. Fisheye LAI is the average value calculated from the N–S and E–W
photos taken at 2 m from the trees. MLB LAI is the average value for an 8 m buffer
from the tree trunk.
these variables, respectively. Accordingly, Landsat NDVI and NDWI
were also highly correlated with MLB measurements (r2 values of
0.71–0.80).
During the time of aerial and satellite image acquisition in the
two experimental sites, the ground was sparsely covered with
grass and forbs due to low light conditions, but the open areas between tree rows (aisles) remained bare due to the floor scraping
done before harvest and to the low soil humidity in these sunexposed areas of micro-sprinkler and drip irrigated orchards.
Therefore, the reflectance in the visible and near infrared wavelengths attributable to vegetation is essentially due to almond
canopy, and the MASTER and Landsat VI are expected to be highly
a
fPARa
R
RMSE
R2
RMSE
0.9175
0.9054
0.9032
0.8999
0.8983
0.8972
0.8942
0.8843
0.8843
0.8810
0.8748
0.8472
0.8320
0.8001
0.7837
0.6701
0.1865
0.1997
0.2020
0.2054
0.2070
0.2081
0.2112
0.2208
0.2208
0.2240
0.2297
0.2537
0.2660
0.2980
0.3094
0.3728
0.8409
0.7014
0.7438
0.7050
0.7341
0.7274
0.7685
0.7808
0.7805
0.7851
0.6550
0.7800
0.5721
0.7909
0.7110
0.6779
0.0267
0.0359
0.0339
0.0364
0.0346
0.0350
0.0323
0.0314
0.0314
0.0311
0.0394
0.0314
0.0438
0.0350
0.0367
0.0380
Mean LAI = 3.351; mean fPAR = 0.6985.
correlated with the canopy fPAR and LAI measurements, as we
found here.
3.3. Comparison of hemispherical photo LAI to MASTER imagery VIs
The mean LAI values determined from the four fisheye
photographs taken around individual trees were well correlated
Author's personal copy
J.L. Zarate-Valdez et al. / Computers and Electronics in Agriculture 85 (2012) 24–32
Mule lightbar LAI
5.0
4.5
4.0
3.5
3.0
R² = 0.918
2.5
2.0
0.15
4.5
0.20
0.25
0.30
NDWI
0.35
b
0.75
0.70
0.65
0.60
0.55
R² = 0.8666
0.50
0.15
0.40
0.80
c
Mule lightbar fPAR
5.0
Mule lightbar LAI
0.80
a
Mule lightbar fPAR
30
4.0
3.5
3.0
2.5
R² = 0.884
2.0
0.40
0.50
0.60
0.20
0.65
0.60
0.55
R² = 0.7928
0.50
Mule lightbar fPAR
Mule lightbar LAI
0.80
4.0
3.5
3.0
2.5
R² = 0.9032
2.0
1.5
3.0
4.0
5.0
NIR/Red
0.60
0.70
NDVI
e
4.5
0.40
0.70
NDVI
5.0
0.35
d
0.75
0.50
0.40
0.70
0.25
0.30
NDWI
6.0
0.70
0.65
0.60
R² = 0.7839
0.55
0.50
7.0
f
0.75
3.0
4.0
5.0
NIR/Red
6.0
7.0
Fig. 5. MASTER NDWI (a, b), NDVI (c, d) and simple ratio VI (e, f) as predictors of mule lightbar LAI (a, c, e) and light interception by the canopy, fPAR (b, d, f), in almond
orchards.
4.50
4.00
Mule lightbar LAI
Mule lightbar LAI
4.00
4.50
a
3.50
3.00
2.50
1.50
0.15
0.2
0.25
0.3
0.35
Landsat NDWI
0.4
3.50
3.00
2.50
R² = 0.784
R² = 0.800
2.00
b
2.00
0.45
1.50
0.50
0.55
0.60
0.65
0.70
Landsat NDVI
0.75
0.80
Fig. 6. Landsat NDWI (a) and NDVI (b) as predictors of mule lightbar LAI in almond orchards.
(r2 = 0.77) with the four closest pixels from the MASTER NDWI image for the corresponding trees (Fig. 7). The other VI evaluated were
less correlated to the fisheye LAI (Table 5).
Fisheye pictures taken 2 m from the trees are adequate for trees
spaced 7 m from each other, because they capture the canopy gaps
above the aisles. If LAI is measured for a group of trees rather than
for a specific individual tree, fisheye photos should be taken at
equal distances between neighboring trees. Photos within the drive
row gave, in general, higher LAI values than those taken in the tree
rows (north and south).
Analyzing the NDWI images obtained from the MASTER and
Landsat 5 scenes for the Belridge orchard, we observed that due
Author's personal copy
J.L. Zarate-Valdez et al. / Computers and Electronics in Agriculture 85 (2012) 24–32
2.70
Fisheye LAI
2.50
2.30
2.10
1.90
1.70
0.28
R² = 0.7719
0.30
0.32
0.34
0.36
MASTER NDWI
Fig. 7. Correlation between fisheye LAI and MASTER NDWI for selected almond
trees in Belridge orchard. LAI is the average value calculated from the N–S and E–W
photos taken at 2 m from the trees. NDWI values correspond to the four
neighboring pixels to the respective tree.
Table 5
Linear fit correlation between hemispherical photography LAI and selected MASTER
VI in the Belridge orchard.
Vegetation index (predictor)
NDWI
OSAVI
NDVI
EVI
Green/Blue
GMI
MCARI1
a
Fisheye LAIa
R2
RMSE
0.7702
0.6108
0.6104
0.5456
0.4750
0.4508
0.3947
0.1212
0.1564
0.1564
0.1689
0.1816
0.1857
0.1950
Mean LAI = 2.17.
to its homogeneity and subtle differences in growth of the alternated Nonpareil and Monterey rows, we detected no striping in
either of the two images. In the case of the Spur Dynamics orchard,
however, row striping is evident in the Master NDWI image due
mainly to (a) the differential growth of the alternated row varieties
and (b) the offset between pixel size (7.2 m) and inter-row distance
(6.8–7.0 m); this striping is not reflected in the Landsat NDWI image, however.
4. Discussion and conclusions
LAI is a key variable commonly used in a wide variety of models
to describe vegetation-atmosphere interactions in both natural and
managed ecosystems; it is also a variable used in the prediction of
biomass, yield and agronomic practices (van Gardingen et al.,
1999; Bréda, 2003; Hyer and Goetz, 2004; Jonckheere et al.,
2004; Tewolde et al., 2005). However, the issue of scaling up from
point measurements in a highly regular feature pattern has been
addressed in this study to verify LAI and VIs relationships. The image pixel value is also an average of the mixture of reflectance in
that section of the orchard, and this technique of scanning entire
regions of the orchard more closely matches the mixed pixel, than
a few or even dozen of points averaged within the pixel.
Since the mule fPAR cannot be separated into specific bands of
the airborne and satellite instruments, a common calibration of
these instruments was sought through comparisons between VIs
and LAI measurements. By searching many VI that have been
shown in the literature as related to LAI, we demonstrated with
great precision and within a narrow range of regression values that
the indexes represent LAI with a high level of accuracy. The highly
intensive MLB measurements illustrated the advantage of scanning
beneath the canopy with MLB over the fewer photographs of
31
hemispherical lens technique. The mean fisheye photo LAI values
and mean LAI and fPAR values from the MLB were compared. In
the unique environment of regular row spacing and high degree
of tree branching, the MLB characterizes canopy light interception
and LAI with high resolution.
A high correlation between MLB LAI and vegetation indexes
such as GMI, a chlorophyll density proxy, and NDWI, an indicator
of water content in the canopy, derived from the MASTER image
was found. The LAI relationship indicates that the variation in index values is due mainly to LAI variation. Chlorophyll concentration, estimated with calibrated SPAD measurements, was not
correlated with MASTER GMI, SIPI, SRPI, MCARI or NDVI
(r2 < 0.10) and the trees in these orchards are constantly irrigated
to keep up with the crop evaporative demand. We do not know,
however, whether the trees were water stressed at the time of image acquisition. Similar results to the ones presented here were
found when comparing AVIRIS NDVI and canopy water content
with LAI in natural vegetation (Roberts et al., 2004).
LAI calculated from the MLB fPAR measurements was better
correlated with most of the vegetation indexes tested than fPAR
alone. This contrasts with what some authors believe that variables
such as LAI often exhibit only moderate, non-linear relationships
with vegetation indexes (Marsden et al., 2010; Viña et al., 2011),
and therefore VIs should not be used to estimate LAI, but to predict
canopy light absorption, instead (Gallo et al., 1985; Myneni et al.,
1995; Glenn et al., 2008,). Unlike many reports of LAI vs. VI functions (Sellers, 1985; Carlson and Ripley, 1997), the VIs we evaluated did not saturate at LAI values greater than four. MLB LAI
values can be over or under estimated by varying leaf angle distribution parameter (LADP); a spherical distribution of leaf angles
where v = 1.0 provided strong correlations to VIs. While higher
LADP between 2.0 and 3.0 produced smaller LAI values, and similar
to those obtained with fisheye photography, these higher LADP
values did not reduce the correlation with VIs.
The correlation of fisheye photography LAI with MLB LAI was
lower than that with MASTER VIs in part due to the small number
of pairwise comparison (seven in the former, as compared to 14 in
the latter); on the other hand, while hemispherical photography
was well correlated with MLB LAI and MASTER VI, the contrast
thresholding step in the Hemiview software that differentiates between foliage and sky, is a key factor in producing reliable LAI values, and this step should be done carefully, and repeatedly.
Similarly, other studies report underestimation of LAI by the fisheye technique, what has been attributed to leaf clumping, a problem derived from the non-random distribution of leaves; one way
to overcome the leaf clumping problem is by determining the
clumping index of the canopy (Bréda, 2003). Other kinds of errors
that this technique can introduce are reviewed by Rich (1988) and
Jonckheere et al. (2004). Nevertheless, some authors found hemispherical photography to be a good estimator of LAI and has been
therefore used as a standard technique for LAI determination (van
Gardingen et al., 1999; Malone et al., 2002; Bréda, 2003; Hyer and
Goetz, 2004; Jonckheere et al., 2004).
The high correlation of MLB fPAR or its derived variable LAI,
with multispectral vegetation indexes gives the possibility of predicting fPAR and LAI from aerial and satellite VI within a reasonable degree of accuracy and for a wide range of LAI values
common to commercial orchards. Based on this, the generated
model is expected to accurately predict almond LAI under different
fraction canopy covers or canopy leaf development during the year,
especially from mid February to mid November.
Acknowledgements
Authors wish to extend our appreciation for research funding
from the Almond Board of California and the USDA-National Insti-
Author's personal copy
32
J.L. Zarate-Valdez et al. / Computers and Electronics in Agriculture 85 (2012) 24–32
tute of Food and Agriculture, Specialty Crop Research Initiative
(Grant No: 2008-51180-19563). We gratefully acknowledge the
collaboration with the NASA Student Airborne Research Program
and sharing of MASTER imagery and student assistance in the field
data collection.
References
Aspinall, R.J., Marcus, W.A., Boardman, J.W., 2002. Considerations in collecting,
processing, and analyzing high spatial resolution hyperspectral data for
environmental investigations. J. Geogr. Syst. 4, 15–29.
Asrar, G., Fuchs, M., Kanemasu, E.T., Hatfield, J.L., 1984. Estimating absorbed
photosynthetic radiation and leaf area index from spectral reflectance in wheat.
Agron. J. 76, 300–306.
Baret, F., Guyot, G., 1991. Potentials and limits of vegetation indices for LAI and
APAR assessment. Remote Sens. Environ. 35, 161–173.
Bréda, N.J.J., 2003. Ground-based measurements of leaf area index: a review of
methods, instruments and current controversies. J. Exp. Bot. 54 (392), 2403–
2417.
Broge, N.H., Mortensen, J.V., 2002. Deriving green crop area index and canopy
chlorophyll density of winter wheat from spectral reflectance data. Remote
Sens. Environ. 81, 45–57.
Carlson, T.N., Ripley, D.A., 1997. On the relation between NDVI, fractional vegetation
cover, and leaf area index. Remote Sens. Environ. 62, 241–252.
Christensen, S., Goudriaan, J., 1993. Deriving light interception and biomass from
spectral reflectance ratio. Remote Sens. Environ. 43, 87–95.
Decagon Devices, Inc., 2008. AccuPAR PAR/LAI ceptometer model LP-80. Operator’s
Manual, Version 7. Decagon Devices, Inc., Pullman, WA.
Ferencz, C., Bognar, P., Lichtenberger, J., Hamar, D., Tarcsai, G., Tima, G., Molnar, G.,
Pasztor, S., Steinbach, P., Szekely, B., Ferencz, O.E., Ferencz-Arkos, I., 2004. Crop
yield estimation by satellite remote sensing. Int. J. Remote Sens. 25 (20), 4113–
4149.
Gallo, K.P., Daughtry, C.S.T., Bauer, M.E., 1985. Spectral estimation of absorbed
photosynthetically active radiation in corn canopies. Remote Sens. Environ. 17,
221–232.
Gao, B.C., 1996. NDWI – A normalized difference water index for remote sensing of
vegetation liquid water from space. Remote Sens. Environ. 58, 257–266.
Garrigues, S., Shabanov, N.V., Swanson, K., Morisette, J.T., Baret, F., Myneni, R.B.,
2008. Intercomparison and sensitivity analysis of leaf area index retrievals from
LAI-2000, AccuPAR and digital hemispherical photography over croplands.
Agric. For. Meteorol. 148, 1193–1209.
Gitelson, A.A., Kaufman, Y., Merzlyak, M.N., 1996. Use of a green channel in remote
sensing of global vegetation from EOS-MODIS. Remote Sens. Environ. 58, 289–
298.
Gitelson, A.A., Merzlyak, M.N., 1997. Remote estimation of chlorophyll content in
higher plant leaves. Int. J. Remote Sens. 18, 2691–2697.
Glenn, E., Huete, A.R., Nagler, P.L., Nelson, S.G., 2008. Relationship between
remotely-sensed vegetation indices, canopy attributes and plant physiological
processes: what vegetation indices can and cannot tell us about the landscape.
Sensors 8, 2136–2160.
Haboudane, D., Millera, J., Patteyc, E., Zarco-Tejada, P., Strachan, I., 2004.
Hyperspectral vegetation indices and novel algorithms for predicting green
LAI of crop canopies: modeling and validation in the context of precision
agriculture. Remote Sens. Environ. 90, 337–352.
Hook, S.J., Myers, J.J., Thome, K.J., Fitzgerald, M., Kahle, A.B., 2001. The MODIS/ASTER
airborne simulator (MASTER) – A new instrument for earth science studies.
Remote Sens. Environ. 76, 93–102.
Hyer, E.J., Goetz, S.J., 2004. Comparison and sensitivity analysis of instruments and
radiometric methods for LAI estimation: assessments from a boreal forest site.
Agric. For. Meteorol. 122, 157–174.
Jensen, J.R., 2007. Remote Sensing of the Environment: An Earth Resource
Perspective, Second ed. Prentice Hall, Upper Saddle River, NJ.
Jonckheere, I., Fleck, S., Nackaerts, K., Muysa, B., Coppin, P., Weiss, M., Baret, F., 2004.
Review of methods for in situ leaf area index determination. Part I. Theories,
sensors and hemispherical photography. Agric. For. Meteorol. 121, 19–35.
Kerr, J.T., Ostrovsky, M., 2003. From space to species: ecological applications for
remote sensing. Trends Ecol. Evol. 18 (6), 299–305.
Lampinen, B., Udompetaikul, V., Browne, G., Metcalf, S., Stewart, W., Contador, L.,
Negrón, C., Upadhyaya, S., 2012. A mobile platform for measuring canopy
photosynthetically active radiation interception in orchard systems. Hort.
Technol, in press.
Liu, H.Q., Huete, A.R., 1995. A feedback based modification of the NDVI to minimize
canopy background and atmospheric noise. IEEE T. Geosci. Remote 33, 457–465.
Maas, S.J., Fitzgerald, G.J., DeTar, W.R., 1999. Spatial and temporal correlations
between crop yield and remotely sensed plant canopy characteristics. In:
Proceedings of the 17th Biennial Workshop on Color Photography and
Videography in Resource Assessment. American Society of Photogrammetry
and Remote Sensing, Bethesda, MD, pp. 106–110.
Malone, S., Herbert Jr., D.A., Holshouser, D.L., 2002. Evaluation of the LAI-2000 plant
canopy analyzer to estimate leaf area in manually defoliated soybean. Agron. J.
94, 1012–1019.
Marsden, C., le Maire, G., Stape, J.L., Lo Seen, D., Roupsard, O., Cabral, O., Epron, E.,
Nascimento Lima, A.M., Nouvellon, Y., 2010. Relating MODIS vegetation index
time-series with structure, light absorption and stem production of fastgrowing Eucalyptus plantations. Forest Ecol. Manag. 259, 1741–1753.
Martens, S.N., Ustin, S.L., Rousseau, R.A., 1993. Estimation of tree canopy leaf area
index by gap fraction analysis. Forest Ecol. Manag. 61, 91–108.
Meggio, F., Zarco-Tejada, P.J., Miller, J.R., Martin, P., Gonzalez, M.R., Berjon, A., 2008.
Row orientation and viewing geometry effects of row-structured vine crops for
chlorophyll content estimation. Can. J. Remote Sens. 34 (3), 220–234.
Myneni, R.B., Hall, F.G., Sellers, P.J., Marshak, A.L., 1995. The interpretation of
spectral vegetation indexes. IEEE T. Geosci. Remote 33, 481–486.
Norman, J.M., Campbell, G.S., 1989. Canopy structure. In: Pearcy, R.W., Ehleringer,
J.R., Mooney, H.A., Rundel, P.W. (Eds.), Plant Physiological Ecology: Field
Methods and Instrumentation. Chapman and Hall, London, pp. 301–325.
Peñuelas, J., Baret, F., Filella, I., 1995. Semi-empirical indices to assess carotenoids/
chlorophyll a ratio from leaf spectral reflectance. Photosynthetica 31, 221–230.
Pettorelli, N., Vik, J.O., Mysterud, A., Gaillard, J.M., Tucker, C.J., Stenseth, N.C., 2005.
Using the satellite-derived NDVI to assess ecological responses to
environmental change. Trends Ecol. Evol. 20 (9), 503–510.
Prince, S.D., Astle, W.L., 1986. Satellite remote sensing of rangelands in Botswana. I.
Landsat MSS and herbaceous vegetation. Int. J. Remote Sens. 7 (11), 1533–1553.
Rich, P.M., 1988. Video image analysis of hemispherical canopy photography. In:
Mausel, P.W., (Ed.), Proceedings of the First Special Workshop on Videography.
American Society for Photogrammetry and Remote Sensing, Terre Haute, IN, 1920 May, pp. 84–95.
Roberts, D.A., Ustin, S.L., Ogunjemiyo, S., Greenberg, J., Dobrowski, S.Z., Chen, J.,
Hinckley, T.M., 2004. Spectral and structural measures of northwest forest
vegetation at leaf to landscape scales. Ecosystems 7, 545–562.
Rondeaux, G., Steven, M., Baret, F., 1996. Optimization of soil adjusted vegetation
indices. Remote Sens. Environ. 55, 95–107.
Ross, J., 1975. Radiative transfer in plant communities. In: Monteith, J.L. (Ed.),
Vegetation and the Atmosphere, vol. 1. Principles Academic Press, London, pp.
13–55.
Rouse, J.W., Haas, R.H., Schell, J.A., Deering, D.W., Harlan, J.C. 1974. Monitoring the
vernal advancement and retrogradation (green wave effect) of natural
vegetation. NASA/GSFC Type III, Final Report, Greenbelt, MD, USA.
Sellers, P.J., 1985. Canopy reflectance, photosynthesis and transpiration. Int. J.
Remote Sens. 6 (8), 1335–1372.
Tewolde, H., Sistani, K.R., Rowe, D.E., Adeli, A., Tsegaye, T., 2005. Estimating cotton
leaf area index non-destructively with a light sensor. Agron. J. 97, 1158–1163.
Thenkabail, P.S., 2003. Biophysical and yield information for precision farming from
near-real-time and historical Landsat TM images. Int. J. Remote Sens. 24 (14),
2879–2904.
Tombesi, S., Lampinen, B.D., Metcalf, S., DeJong, T.M., 2011. Relationships between
spur and orchard level fruit bearing in almond (Prunus dulcis). Tree Physiol. 31
(12), 1413–1421.
Tucker, C.J., Sellers, P.J., 1986. Satellite remote sensing of primary production. Int. J.
Remote Sens. 7 (11), 1395–1416.
Turner, D.P., Ollinger, S.V., Kimball, J.S., 2004. Integrating remote sensing and
ecosystem process models for landscape to regional scale analysis of the carbon
cycle. Bioscience 54, 573–584.
van Gardingen, P.R., Jackson, G.E., Hernandez-Daumas, S., Russell, G., Sharp, L., 1999.
Leaf area index estimates obtained for clumped canopies using hemispherical
photography. Agric. For. Meteorol. 94, 243–257.
Viña, A., Gitelson, A.A., Nguy-Robertson, A.L., Peng, Y., 2011. Comparison of different
vegetation indices for the remote assessment of green leaf area index of crops.
Remote Sens. Environ. 115, 3468–3478.
Watson, D.J., 1947. Comparative physiological studies in the growth of field crops. I.
Variation in net assimilation rate and leaf area between species and varieties,
and within and between years. Ann. Botany 11, 41–76.
Weiss, M., Baret, F., Smith, G.J., Jonckheere, I., Coppin, P., 2004. Review of methods
for in situ leaf area index (LAI) determination. Part II. Estimation of LAI, errors
and sampling. Agric. For. Meteorol. 121, 37–53.
Zarco-Tejada, P.J., Ustin, S.L., Whiting, M.L., 2005. Temporal and spatial relationships
between within-field yield variability in cotton and high-spatial hyperspectral
remote sensing imagery. Agron. J. 97, 641–653.
Zygielbaum, A.I., Gitelson, A.A., Arkebauer, T.J., Rundquist, D.C., 2009. Non-destructive
detection of water stress and estimation of relative water content in maize.
Geophys. Res. Lett. 36, L12403. http://dx.doi.org/10.1029/2009GL038906.