Available online at www.sciencedirect.com
Remote Sensing of Environment 112 (2008) 156 – 172
www.elsevier.com/locate/rse
Normalized difference spectral indices for estimating photosynthetic
efficiency and capacity at a canopy scale derived from
hyperspectral and CO2 flux measurements in rice
Y. Inoue a,⁎, J. Peñuelas b , A. Miyata a , M. Mano a
a
b
National Institute for Agro-Environmental Sciences, Tsukuba, Ibaraki, 305-8604, Japan
Center for Ecological Research and Forestry Applications, Universitat Autònoma de Barcelona, 08193 Bellaterra, Spain
Received 30 December 2006; received in revised form 30 March 2007; accepted 21 April 2007
Abstract
We explored simple and useful spectral indices for estimating photosynthetic variables (radiation use efficiency and photosynthetic
capacity) at a canopy scale based on seasonal measurements of hyperspectral reflectance, ecosystem CO2 flux, and plant and micrometeorological variables. An experimental study was conducted over the simple and homogenous ecosystem of an irrigated rice field.
Photosynthetically active radiation absorbed by the canopy (APAR), the canopy absorptivity of APAR (fAPAR), net ecosystem exchange of
CO2 (NEECO2) gross primary productivity (GPP), photosynthetic capacity at the saturating APAR (Pmax), and three parameters of radiation use
efficiency (εN: NEECO2/APAR; εG: GPP/APAR; φ: quantum efficiency) were derived from the data set. Based on the statistical analysis of
relationships between these ecophysiological variables and reflectance indicators such as normalized difference spectral indices (NDSI[i,j])
using all combinations of two wavelengths (i and j nm), we found several new indices that would were more effective than conventional
spectral indices such as photochemical reflectance index (PRI) and normalized difference vegetation index (NDVI = NDSI[near-infrared, red]).
εG was correlated well with NDSI[710, 410], NDSI[710, 520], and NDSI[530, 550] derived from nadir measurements. φ was best correlated
with NDSI[450, 1330]. NDSI[550, 410] and NDSI[720, 420] had a consistent linear relationships with fAPAR throughout the growing season,
whereas conventional indices such as NDVI showed very different relationships before and after heading. Off-nadir measurements were more
closely related to the efficiency parameters than nadir measurements. Our results provide useful insights for assessing plant productivity and
ecosystem CO2 exchange, using a wide range of available spectral data as well as useful information for designing future sensors for ecosystem
observations.
© 2007 Elsevier Inc. All rights reserved.
Keywords: CO2 flux; Eddy covariance; fAPAR; GPP; Hyperspectra; Radiation use efficiency; Spectral index
1. Introduction
Assessment of photosynthetic functioning is one of the
most important bases for the diagnosis and prediction of plant
growth, as well as carbon exchange between ecosystems and
the atmosphere. A great deal of research effort has been
directed towards the assessment of crop yield and ecophysiological variables using remote sensing in optical, thermal,
and microwave wavelength domains together with modeling
⁎ Corresponding author. Tel.: +81 298 38 8220; fax: +81 298 38 8199.
E-mail address: [email protected] (Y. Inoue).
0034-4257/$ - see front matter © 2007 Elsevier Inc. All rights reserved.
doi:10.1016/j.rse.2007.04.011
approaches (e.g., reviews by Moran et al., 1997; Inoue, 2003;
Olioso et al., 2005). Remotely sensed optical signatures have
proved useful for estimating ecological variables such as leaf
area index (LAI) and the absorptivity of photosynthetically
active radiation (fAPAR; Asrar et al., 1984; Daughtry et al.,
1983; Huete, 1988) that affect photosynthetic capacity. The
relationship between fAPAR and spectral indices such as
normalized difference vegetation index (NDVI) have been
examined in detail with theoretical and experimental analyses
(Asrar et al., 1989; Baret & Guyot, 1991; Myneni &
Williams, 1994). Several important physiological variables,
such as chlorophyll and nitrogen concentration, that affect
Y. Inoue et al. / Remote Sensing of Environment 112 (2008) 156–172
photosynthetic efficiency are also related to reflectance
signatures (e.g., Inada, 1985; Inoue et al., 1998; Gitelson &
Merzlyak, 1997; Yoder & Pettigrew-Crosby, 1995). It is
obvious that efficiency and capacity are related to each other
during long growth periods. However, in the context of
ecosystem monitoring at high temporal resolution, we use the
term “photosynthetic-efficiency variable” to indicate photosynthetic rates per unit of absorbed radiation and the term
“photosynthetic-capacity variable” for potential photosynthesis at canopy scales, such as fAPAR and the canopy
photosynthesis at the saturating APAR (Pmax). In other
words, the former is related to “rate”, whereas the latter is
related mainly to “size”. Furthermore, stomatal or canopy
conductance and canopy transpiration are also estimated
using remotely sensed canopy temperatures with energy
balance models (e.g., Inoue et al., 1990, 1994; Olioso et al.,
1999; Taconet et al., 1995). These variables strongly control
photosynthetic efficiency. Nevertheless, direct estimation of
photosynthetic efficiency and capacity is an important
research target for remote sensing.
In the past few decades, remote sensing has become more
and more crucial in the study of biogeochemical cycles such
as carbon, water, and energy exchange between ecosystems
and the atmosphere in the context of global climate change
(e.g., Grace, 2005; Roy & Saugier, 2001). One recent effort
for systematic data acquisition and analysis is the Spectral
Network (SpecNet) proposed by Gamon et al. (2006).
Advanced concepts, methodologies, or achievements in
agricultural remote sensing have been applied to various
types of natural and managed ecosystems. For example, the
use of fAPAR in a simple mechanistic model (e.g.,
Monteith, 1977) has been employed in a range of
applications for rough assessment of net primary productivity (NPP) at various scales using remote sensing (e.g., Goetz
et al., 1999; Maisongrande et al., 1995; Potter et al., 1993;
Ruimy et al., 1994; Turner et al., 2002; Veroustraete et al.,
1996).
This simple approach using fAPAR and radiation (or light)
use efficiency (ε) has been widely used to estimate crop
biomass, yield, or NPP (e.g., Choudhury, 2000). fAPAR is
estimated from spectral indices such as NDVI (= [RNIR − Rred] /
[RNIR + Rred]) or a simple ratio (SR = RNIR / Rred]; e.g., Asrar
et al., 1984; Baret & Guyot, 1991; Kumar & Monteith, 1982)
and applied to a model (e.g., Nouvellon et al., 2000; Ruimy
et al., 1994), where RNIR and Rred are reflectance at nearinfrared and red wavelength regions, respectively (hereafter,
the term Rx indicates spectral reflectance at a wavelength x
nm). The relationship is little affected by pixel heterogeneity,
LAI, or variation in leaf orientation and optical properties
(e.g., Pinter et al., 1985), but is affected by background,
atmospheric, and bidirectional effects (Myneni & Williams,
1994) and by phenological stages and senescence. For
instance, the NDVI-fAPAR relationship is largely different
between before and after heading (anthesis) in most crops
(e.g., Asrar et al., 1989; Choudhury, 2000; Daughtry et al.,
1983; Inoue & Iwasaki, 1991; Inoue et al., 1998). In this
approach, ε is first defined as the ratio of dry matter produc-
157
tion (dDM) to APAR for season-long periods (e.g., Shibles &
Weber, 1966). ε may be assumed to be constant under nonstressed conditions, but it is affected by stresses, phenological
stages, and the physical environment (Choudhury, 2000;
Kiniry et al., 1989; Sinclair, 1994). Hence, it may be
inappropriate to assume that ε is constant and that the
fAPAR-NDVI relationship is consistent for entire growth
periods, especially when the model is applied to the assessment
of canopy carbon exchange at a short (e.g., daily) time resolution over different phenological stages.
The photochemical reflectance index (PRI = [R531 − R570] /
[R531 + R570]) is closely related to photosynthetic radiation use
efficiency of plant leaves (Gamon et al., 1992; Peñuelas et al.,
1995). R531 is presumed to detect the photochemical reaction
in the xanthophyll cycle that dissipates excess light to protect
the photosynthetic apparatus, while R570 is used as a reference
assumed not to be affected by changes in short-term stress
events (Peñuelas & Filella, 1998). Its simplicity, using only two
wavelengths, is an attractive feature since it may meet the
specifications for a range of available airborne and satellite
sensors; thus a number of studies have examined the usefulness
of PRI using various datasets and approaches (Barton & North,
2001; Drolet et al., 2005; Filella et al., 1996; Gamon et al.,
1997; Inoue & Peñuelas, 2006; Rahman et al., 2001; Thenot
et al., 2002; Trotter et al., 2002). Nevertheless, its applicability
at canopy or ecosystem scales is not well known, and the use of
remotely sensed signatures in the assessment of primary productivity (NPP or gross primary productivity, GPP) and net
ecosystem CO2 flux (NEECO2) awaits further methodological
innovations. Synergy of remote sensing signatures in ecosystem
functioning models that include photosynthetic processes is
promising (e.g., Inoue & Olioso, 2006), but simple and
direct means are still a very attractive approach. Nevertheless,
canopy-scale investigations of PRI and other spectral indices
based on in situ measurements of CO2 flux and hyperspectral
measurements over homogenous ecosystems are still lacking
(e.g., Peñuelas & Inoue, 2000), so that useful indices may yet be
undiscovered.
Therefore, the objective of this study was to explore the
significant and/or consistent relationships for remote
sensing of photosynthetic-efficiency and-capacity variables
such as å and fAPAR. We made seasonal measurements of
hyperspectral reflectance, ecosystem CO2 flux, and plant
and micrometeorological variables over an irrigated rice
field. The dataset from this simple and homogenous
ecosystem facilitates efficient extraction of more significant
relationships. In general, hyperspectral measurements allow
various analyses such as multiple linear regression,
principal component regression, partial least squares regression, and derivatives (e.g., Adams et al., 1999; Grossman
et al., 1996; Takahashi et al., 2000), or assimilation
techniques using a radiative transfer model such as
PROSPECT + SAIL (e.g., Jacquemoud, 1993). Nevertheless,
here we focus mainly on the comparative usefulness of the
simple normalized difference spectral indices (NDSIs) using
narrow band reflectance, instead of the whole spectrum for
wider applications.
158
Y. Inoue et al. / Remote Sensing of Environment 112 (2008) 156–172
2. Materials and methods
2.1. Experimental site
The experimental site was a typical paddy field area in
Tsukuba, central part of Japan (36°03′ N, 140° 01′ E, 15 m
above sea level). Measurements were made in an irrigated rice paddy field (100 × 50 m) in 2003 and 2004. The
field was surrounded by similar rice fields extending the
area to 1.5 × 1.0 km. The soil was clay loam that was classified to grey lowland paddy soil (Cultivated Soil Classification Committee of Japan, 1995) and typic endoaquepts
(Soil Survey Staff, 1992). Rice seedlings (Oryza sativa L.,
cv. Koshihikari) were transplanted in early May and harvested in mid-September. The paddy field was managed
based on standard management practices for the region.
The field was flooded all the time except for a mid-summer
drainage period of about a week and the pre-harvest period,
which is the most common practice for rice cultivation in
Japan.
2.2. Hyperspectral reflectance measurements
Canopy reflectance spectra were obtained under clear sky
conditions near midday using a portable spectro-radiometer
(FieldSpec-FR, ASD). Spectral range and the field of view
of the sensor were 350–2500 nm and 25°, respectively.
Spectral resolution (full-width-half-maximum; FWHM) was
3 nm for the region of 350–1000 nm and 10 nm for 1000–
2500 nm, while the sampling interval was 1.4 nm and 2 nm
for each region, respectively. These signatures were used to
derive spectra at 1 nm interval using cubic spline interpolation functions. We used these spectral data at 1-nm
intervals in this study because they could carry the maximum
information at narrow wavelengths. The spectrometer was
equipped with a 2-m optical fiber to allow measurements
from higher positions. Reflectance measurements were taken
at nadir-looking and off-nadir angles (40–45°) from 2 m
above the canopy. The off-nadir measurements were added
because we presumed that they would carry more physiological information on photosynthetic efficiency since little
background is viewed within the field of view (Takebe et al.,
1990) and because the increasing number of sensors would
allow directional measurements (Chen et al., 2003). More
than 30 measurements were made each time, moving over
the canopy, and averaged for spectral analyses. Spectral
reflectance was derived as the ratio of reflected radiance to
incident radiance estimated by a calibrated white reference
(Spectralon, Labsphere).
2.3. Flux measurements
Net ecosystem exchange of CO2 (NEECO2) was measured
using an open-path eddy covariance system, which consisted
of a sonic anemometer (DA-600, Kaijo) and an infrared gas
analyzer (LI-7500, LI-COR). The sensors were mounted
3.0 m above the ground. Half-hour flux densities of CO2,
water vapor, sensible heat, and momentum were calculated
from the covariance between the vertical wind velocity, w,
and the respective quantities. The effect of air density
fluctuations on the measurement of CO2 and water vapor
fluxes was corrected following Webb et al. (1980). Quality
control of the fluxes was based on the standard methodology
as described in a previous report (Saito et al., 2005). We
assumed that the measured flux was representative for
the field since all surrounding rice fields were managed by
similar standard practices. It was also confirmed that the
majority of the footprint was within the field based on a
preliminary assessment using a footprint model (Fan et al.,
1992; Kljun et al., 2004) with in situ measurements of
micrometeorological data and remotely-sensed images
(Inoue et al., 2005).
2.4. Micrometeorological and plant measurements
Basic micrometeorological data were recorded throughout
the growing season: global solar radiation and net radiation
(CNR1, Kipp & Zonen), incident and reflected photosynthetically active radiation (PAR; LI-190, LI-COR), transmitted
PAR (LI-191, LI-COR), air temperature and relative humidity
(HMP-45A, Vaisala), soil heat flux density (MF-81, Eko), and
volumetric soil water content (TDR100, Campbell). The 1-m
long PAR sensor (LI-191) was placed across three rows
below the canopy to obtain spatially representative values.
These data were sampled every 5 s, and 15-min averages of the
sampled data were recorded using a data logger (CR23X,
Campbell). The fraction of absorbed photosynthetically
active radiation (fAPAR) was derived from PAR measurements as [1 − rc − (1 −r0)τc], where rc and τc are the reflectance
and transmittance of the canopy and r0 is the reflectance of the
ground.
Ten stocks of rice plants were sampled every 2 weeks during
the growing season. The rice samples were divided into roots,
stems, leaves (green or dead), and panicles, and the green leaf
area was measured using an optical area meter (AAM-7,
Hayashi Denkoh) to determine the LAI. All plant parts were
dried at 70 °C for 2 weeks to determine dry biomass.
3. Results and discussion
3.1. NEECO2 and canopy photosynthetic variables
Fig. 1 shows the seasonal changes in NEECO2 in the rice
field. The NEECO2 was low for several weeks after transplantation, but CO2 uptake during midday became obvious
even at low LAI. The midday peak value of NEECO2 increased with increasing LAI and decreased rapidly during
the late ripening period, although it is was strongly affected
by incoming PAR intensity. The photosynthetic rate was
much lower during the ripening stage than during the early
vegetative stage, with similar or larger values of green LAI.
This large difference is presumably due to senescence,
increasing respiration, and decreasing PAR intensity at the
active leaves. In general, CO2 flux from the water surface
Y. Inoue et al. / Remote Sensing of Environment 112 (2008) 156–172
159
beneath the rice canopy is low compared to plant CO2 flux,
since soil microbial respiration is strongly suppressed by the
water layer (Miyata et al., 2000). Nighttime NEECO2, i.e.,
ecosystem respiration (Re), increased across growth stages
and was highly correlated with air temperature (r = 0.81 in
2003 and 0.80 in 2004). Re during the night was expressed as
an exponential function of air temperature (ta):
Re ¼ a expðb ta Þ
ð1Þ
where a and b are parameters.
Since the Re is the sum of plant and microbial respiration,
by definition, and it is not proportional to photosynthesis,
the photosynthetic activity of ecosystems may be represented
better by GPP than by NEECO2, NPP, or other variables. Hence,
we derived GPP values from NEECO2 using the following
equation, assuming that the relationship derived above (Eq. (1))
could be applied during the day:
GPP ¼ NEECO2 þ Re
ð2Þ
Here, NEECO2 is positive when CO2 is emitted from the
ecosystem into the atmosphere, while GPP and Re are both positive.
Fig. 2. Seasonal change of APAR-GPP relationship. Data points represent halfhourly measurements available for each week in 2003.
We derived two indicators of canopy radiation use efficiency
at the time of remote sensing measurements from half-hourly
data.
Fig. 1. Seasonal change of net ecosystem CO2 exchange (NEECO2) in a rice
field measured by eddy covariance method in 2004. Negative NEECO2 values
indicate the flux from the atmosphere into the ecosystem. Solid lines indicate
the incident PAR. Solid lines indicate the incident PAR. Dates for
transplanting, heading, and harvesting were DOY123 (May 2), DOY208
(July 28), and DOY253 (September 9), respectively.
eN ¼ NEECO2 =APAR
ð3Þ
eG ¼ GPP=APAR
ð4Þ
To infer the photosynthetic functioning of the canopy, we
also derived quantum efficiency and maximum GPP by
160
Y. Inoue et al. / Remote Sensing of Environment 112 (2008) 156–172
Fig. 3. Seasonal change of photosynthetic capacity (Pmax), quantum efficiency (φ), and radiation use efficiency (εG) for the rice canopy. These
parameters were determined for weekly sets of GPP and APAR data in
2003.
applying a light-photosynthesis curve to the APAR-GPP
relationship. It was obvious that the APAR-GPP relationship
was not linear in either daily or weekly scatter plots (Fig. 2).
Although both linear (e.g., Campbell et al., 2001) and
hyperbolic formulae (e.g., Wall & Kanemasu, 1990) have
been used to represent the APAR-GPP relationship, the linear
and hyperbolic formulae do not express the lower part of the
relationship very well. Hence, we applied the asymptotic
exponential equation as follows:
GPP ¼ Pmax ð1 exp½uAPAR=Pmax Þ
various weather conditions including cloudy and rainy
conditions. Under cloudy conditions, for example, the
radiation use efficiency could be higher because a higher
percentage of irradiance is diffusive and more effectively
penetrates the canopy and because leaf photosynthesis is
more efficient at weaker light intensity (Campbell et al.,
2001). Hence, in general, the relationship of GPP (or εG) at
different time scales largely depends upon variability in
environmental conditions (such as PAR and water availability) as well as the response of efficiency to such conditions
(e.g., Sims et al., 2005). Nevertheless, our results strongly
suggest that one midday measurement would be a good
indicator for inferring the potential values of radiation use
efficiency or those for longer periods using ancillary data
such as daily PAR.
3.2. Relationship of photosynthetic-efficiency variables with
reflectance spectra
Fig. 5 shows the seasonal change in reflectance spectra
obtained over the rice canopy with different levels of LAI,
biomass, Pmax, and φ. Wavelength windows around 1430 nm
and 1910 nm were eliminated due to strong disturbance by
atmospheric vapor. The spectra just after transplanting (05/12)
ð5Þ
where φ is the initial slope of the curve (quantum efficiency)
and Pmax is the photosynthetic capacity at the saturating
APAR. The coefficient of determination for all curve fitting
was as high as 0.98, so both daily and weekly parameters
were derived accordingly. The overall patterns of seasonal
change were similar for Pmax and LAI, while both φ and εG
were high at the early vegetative stage but decreased
drastically with plant growth (Fig. 3). The seasonally
integrated value of radiation use efficiency is consistent for
each crop species, especially when there are no environmental stresses (Monteith, 1977; Sinclair, 1994), and it is
often assumed to be constant in growth models. Nevertheless, it is obvious that the photosynthetic-efficiency variables
do change and are affected by growth stage and environmental conditions.
We examined the effect of time scale on the estimates of
radiation use efficiency at different integration periods to
assess whether the instantaneous (often near midday) remote
sensing measurements could provide robust information on
the photosynthetic status of vegetation over longer terms.
Fig. 4 compares εG values derived for midday 30 min
(εG_30 m midday), 1 day (εG_day), and 1 week (εG_week). We
found significant linear relationships among all three values.
However, it should be noted that εG_day was 4% lower than
εG_30 m midday, and εG_week was 22% lower than εG_day.
These differences are mainly due to diurnal or daily changes
in PAR intensity. The values of εG_30 m midday represent the
efficiency at midday under clear sky conditions, whereas
both εG_day and εG_week values incorporate all data under
Fig. 4. Relationship between the values of radiation use efficiency εG at the
time scale of 30 min (εG_30 m_midday), one day (εG_day), and one week
(εG_week).
Y. Inoue et al. / Remote Sensing of Environment 112 (2008) 156–172
were similar to those of shallow rivers or ponds but already
showed a slight trough at the chlorophyll absorption
wavelength (670 nm). The reflectance in the red wavelengths decreased with plant growth until the time of
heading (near maximum LAI), but decreased again during
the ripening period until harvesting. A similar seasonal
trend was found in the blue wavelengths (420–540 nm). In
contrast, green (540–580 nm) and near-infrared (700–
1350 nm) wavelengths increased throughout the growth
period until harvesting. The range of change was much
higher in the near-infrared region than in the visible region
during the vegetative stage (i.e., before heading), whereas it
was inverse during the ripening stage (i.e., from heading to
harvesting). The post-harvest surface (09/27) showed
typical spectra for crop residues. In accordance with the
above trends, the spectrum of the post-harvest surface (09/
27) was largely different from the spectra during the
ripening stage in red and blue regions, while it was slightly
different in the green and near-infrared regions. Reflectance
in shortwave-infrared regions showed a trend similar to the
near-infrared, but was less sensitive to vegetation change
during the vegetative stage (e.g., see lines for 06/14, 07/05,
and 08/03). This trend was more obvious at longer
wavelengths (2000–2300 nm). These spectra are determined
largely by the spectral features of photosynthetic pigment,
water, and lignin/cellulose, as well as the background surface (Curran, 1989; Jacquemoud, 1993). Nevertheless, we
assume these spectra also carry a number of minor spectral
responses to the ecophysiological or biochemical functioning
of vegetation.
Arrows in Fig. 5 indicate the wavelengths used to
calculate the majority of previously reported spectral
indices such as NDVI, SAVI, PRI, WI, SIPI, NPQI, GRI,
SWWI, RVI, WDVI, MSAVI, EVI, GEMI, NDMI1,
NDMI2, and BI (Peñuelas & Filella, 1998; definitions of
161
Fig. 6. Correlation coefficients between reflectance at single wavelength
and a) three efficiency parameters εN, εG and φ, and b) four capacity
parameters fAPAR, NEECO2, GPP and Pmax . Data are from nadir
measurements.
these indices are given in Tables 1 and 2). The spectral
widths for broad-band indices were determined based on
the specifications of Landsat-TM. We examined the
relationship of three efficiency parameters, εN, εG, and φ,
with these conventional indices, with the reflectance (R i) at
Fig. 5. Seasonal change in reflectance spectra of a rice canopy with different LAI and GPP. Date numbers indicate month/day in 2003. Wavelengths indicated with
arrows were used to derive conventional spectral indices. Data are from nadir measurements.
162
Y. Inoue et al. / Remote Sensing of Environment 112 (2008) 156–172
Fig. 7. A contour map of coefficient of determination (r2) between εG and the normalized difference spectral indices (NDSIs) using two separate wavelengths on x and
y axes. A significant portion has been expanded out of a map for the entire wavelength regions to depict the details. Data are for nadir measurements.
a single wavelength (i nm), and with NDSIs using all
combinations of two separate wavelengths (i and j nm),
NDSI½i; j ¼ ðRi Rj Þ=ðRi þ Rj Þ
ð6Þ
In theory, NDSI[i, j] = − NDSI[j, i] and the range of NDSIs
is from − 1 to 1. This normalization is effective at canceling
atmospheric disturbance or other error sources as well as to
enhancing and standardizing the spectral response to
observed targets (e.g., Qi et al., 1994). For example, NDVI
(i.e., a specific index NDSI[830, 660] in our generalized
expression) has been widely used in scientific and operational
applications (e.g., see reviews by Inoue & Olioso, 2006;
Moran et al., 1997). Other well known indices such as PRI are
in a similar formulation (Tables 1 and 2). In principle, NDSIs
using consecutive wavelengths have a role similar to the first
derivative.
The correlation coefficients between reflectance at a
single wavelength and the three radiation use efficiency
parameters are shown in Fig. 6a. Both εN and φ were highly
correlated with near-infrared wavelengths (r = − 0.75 − − 0.8),
while εG was best correlated with R710 (r = 0.7) and less
correlated with longer wavelengths. There was an obvious
peak at around 550 nm, but the visible region was not highly
correlated in general. The line for NEECO2 looks nearly
inverse to the others simply because the signs of NEECO2
and the other parameters are inverse (negative NEECO2
means ecosystem CO2 uptake); nevertheless, it is interesting
that a peak at around 710 nm (the so-called red-edge) was
found only in εG and not in εN or φ. The response in nearinfrared regions was very flat with two minor peaks (970
and 1130 nm).
Fig. 7 shows a contour map of the coefficient of
determination (r2 ) between εG and NDSIs with the two
wavelengths on the x and y axes. The map provides an
overview of the statistical significance of NDSIs for all
combination of two wavelengths. It allows efficient extraction
of significant peak-wavelengths as well as the extent of the
effective regions for assessment of each target variable. There
are several narrow peaks (reddish) and deep troughs (bluish).
Some steep walls are also found (e.g., around the 530 vs.
560 nm region). Therefore, selection of effective wavelengths
should be made carefully so as not to span such steep-wall
regions, since even a small bias could largely reduce the
potential significance.
Table 1 summarizes the statistical significance of the
conventional indices and selected NDSIs in the estimation of
canopy photosynthetic variables. The NDSIs were selected
on the basis of the contour map, a part of which is shown in
Fig. 7. εN was best correlated with NDSI[542, 550]
(r 2 = 0.660), followed by NPQI using R415 and R435
(r 2 = 0.512), while all other indices were poorly correlated.
Y. Inoue et al. / Remote Sensing of Environment 112 (2008) 156–172
163
Table 1
Coefficient of determination (r2) between efficiency parameters and selected spectral indices derived from nadir measurements
Spectral indices
(Nadir)
εN
εG
φ
fAPAR
Pmax
NDSI[542, 550]
Δ [11, 7]
0.660
–
–
–
–
NDSI[410, 550]
NDSI[410, 710]
NDSI[520, 710]
NDSI[530, 550]
NDSI[940, 1122]
NDSI[1050, 1122]
Δ [8, 15]
Δ [10, 10]
Δ [20, 10]
Δ [3, 3]
Δ [3, 3]
Δ [10, 3]
–
–
–
–
–
–
0.704
0.715
0.711
0.711
0.675
0.706
–
–
–
–
–
–
–
–
–
–
–
–
–
–
–
–
–
–
NDSI[403, 830]
NDSI[420, 970]
NDSI[450, 1330]
NDSI[933, 940]
NDSI[933, 948]
NDSI[1053, 1058]
Δ [5, 80]
Δ [5, 7]
Δ [80, 20]
Δ [3, 3]
Δ [3, 3]
Δ [3, 3]
–
–
–
–
–
–
–
–
–
–
–
–
0.760
0.759
0.773
0.762
0.759
0.755
–
–
–
–
–
–
–
–
NDSI[550, 410]
NDSI[720, 420]
Δ [16, 10]
Δ [6, 40]
–
–
–
–
–
–
0.931
0.926
–
–
NDSI[518, 676]
NDSI[620, 623]
NDSI[620, 637]
NDSI[750, 761]
Δ [5, 7]
Δ [5, 5]
Δ [5, 3]
Δ [5, 7]
–
–
–
–
–
–
–
–
–
–
–
–
–
–
–
–
0.651
0.664
0.655
0.665
0.311
0.163
0.026
0.512
0.004
0.063
0.016
0.105
0.001
0.150
0.094
0.080
0.149
0.061
0.071
0.429
0.018
0.623
0.097
0.681
0.271
0.188
0.403
0.539
0.192
0.547
0.484
0.489
0.575
0.128
0.273
0.228
0.031
0.737
0.407
0.448
0.374
0.371
0.630
0.625
0.254
0.569
0.538
0.585
0.627
0.254
0.116
0.060
0.000
0.705
0.213
0.772
0.502
0.145
0.629
0.802
0.386
0.801
0.754
0.745
0.834
0.072
0.365
0.255
0.312
0.210
0.455
0.105
0.662
0.004
0.522
0.522
0.667
0.505
0.594
0.571
0.468
0.024
0.416
0.042
PRI: [R531 − R570] / [R531 + R570]
WI: R900 / R970
SIPI: [R800 − R445] / [R800 − R680]
NPQI: [R415 − R435] / [R415 + R435]
GRI: R830 / R550
SWWI: R1650 / R850
NDVI: [R830 − R660] / [R830 + R660]
SAVI: 1.5[R830 − R660] / [R830 + R660 + 0.5]
RVI: R830 / R660
WDVI: R830 − 1.06R660
MSAVI: (1 + 0.5)[R830 − R660] / [R830 + R660 + 0.5]
EVI: 2.0[R830 − R660] / [1 + R830 + 6R660 − 7.5R460]
GEMI: η(1 − 0.25η) − (R660 − 0.125) / (1 − R660)
NDMI1: [R1650 − R830] / [R1650 − R850]
NDMI2: [R2200 − R1100] / [R2200 + R11100]
BI: 6 [R460 / R660]
–
–
–
Notes: NDSI[i, j] is the normalized difference spectral index using the wavelengths i and j nm. Δ[i, j] indicates the spectral width for i and j nm. The NDSIs in this
Table were selected on the basis of the contour map a part of which was shown in Figs. 7 and 14. Values (r2) larger than 0.6 are shown for the new indices, and all are
indicated for conventional indices for comparison. Relatively high values are indicated in bold. Wavelength widths for conventional indices were determined based on
Landsat-TM bands, or original definition of each index; the NDSI-type indices using narrow bands can be identified in the contour maps. L = 1 − 2.12 NDVI × WDVI.
η = [2 (R8302 − R6602) + 1.5 R830 + 0.5 R660] / (R830 + R660 + 0.5).
Since εN is the ratio of NEECO2 to APAR, the relationship
between photosynthesis and spectral reflectance may be
weakened due to the effects of plant respiration and microbial
respiration. On the other hand, εG was correlated with NDSI
[710, 410] (r 2 = 0.715), NDSI[710, 530] (r2 = 0.711), and
NDSI[530, 550] (r 2 = 0.711; Fig. 8). R550, R710, and R1122
were selected for multiple combinations because they may
have some specific role in ecophysiological processes that
directly or indirectly affect photosynthetic efficiency. R550
normalized by near-infrared wavelengths (e.g., 830 nm) is
best related to leaf nitrogen content (Inada, 1985) and rice
canopy (Inoue et al., 1998; Takebe et al., 1990). Since
nitrogen content is closely related to chlorophyll content and
thus photosynthesis, R550 may play an essential role in the
estimation of photosynthetic efficiency. R710 is positioned
around the so-called red-edge wavelengths, which represent
the large jump from the bottom of chlorophyll absorption
(660–670 nm) to the near-infrared plateau in typical
vegetation spectra (Fig. 5). Therefore, it may have a role in
normalizing the amount of vegetation when it is used with
other visible wavelengths such as 410 or 530 nm. Considering the effective width and spectral resolution of the data,
R530 may have a role similar to that of 531 nm, which is used
for PRI. Since the PRI using R531 and R570 was poorly
correlated, but NDSI [530,550] was highly correlated, the use
of 550 nm instead of 570 nm as a reference wavelength may
significantly improve the estimation of photosynthetic
efficiency.
164
Y. Inoue et al. / Remote Sensing of Environment 112 (2008) 156–172
The performance of PRI was surprisingly poor, although a
number of papers have suggested that PRI would be useful for
assessing photosynthetic radiation use efficiency (see review in
the Introduction). Several attempts to apply PRI to ecosystem
scales (e.g., Rahman et al., 2001), but its performance is not
approved yet (Gamon et al., 2006; Sims et al., 2006). These
reports suggest that PRI would work, to some extent, under
“normal” rainfall conditions, but would work poorly for wider
conditions including drought. These results coincide with
similar results in other drought-prone systems where canopies
undergo large structural changes (Filella et al., 2004; Gamon
et al., 1992). Inoue and Peñuelas (2006) also confirmed a close
relationship between PRI and photosynthetic radiation use
efficiency (the ratio of net photosynthesis to incident PAR) at
the leaf scale, but also found that the relationship was expressed
as a function of soil water content. Hence, application of PRI to
canopy or ecosystem scales may currently have critical limitations. More comprehensive investigations based on precise
measurements at canopy scales would yield better spectral
indices for assessment of photosynthetic activity or other ecophysiological variables. It is also obvious, on the other hand,
that we need more systematic approaches to use remotelysensed data in combination with ecophysiological process
models for dynamic and accurate assessment of carbon, water,
and biomass of plant ecosystems (e.g., Inoue & Olioso, 2006).
As for R1122, we have long recognized that nearby
wavelengths are often selected when estimating the amount
of canopy nitrogen using seasonal reflectance measurements
(e.g., Inoue et al., 1998). We presume that R1122 is related to
the amount of canopy nitrogen because 1120 nm is the
wavelength of lignin absorption (Curran, 1989), and lignin/
cellulose content is inversely related to nitrogen content during
the growing season.
The quantum efficiency φ was best correlated with NDSI
[450, 1330] (r2 = 0.773). It is interesting that NDSI[403, 830]
(r2 = 0.760) and NDSI[420, 970] (r2 = 0.759) using very far
wavelengths were also highly correlated with φ, whereas some
other NDSIs use nearby wavelengths such as NDSI[933, 940]
(r2 = 0.762), NDSI[933, 948] (r2 = 0.759) and NDSI[1053,
1058] (r2 = 0.755). The bandwidth is relatively broad for the
former NDSIs (5–80 nm) and narrow for the latter NDSIs
(3 nm) that have similar information as derivatives. The WI
using R970, a weak absorption peak of water, has proved useful
for detecting leaf water status (Peñuelas & Inoue, 1999). This
may explain the high correlations of WI with φ and εG, since
leaf water status strongly affects photosynthetic efficiency.
However, there is no literature that suggests the role of 933, and
940, or 948 nm, although our results suggest the 930–950
region may have potential when estimating φ.
In general, the newly explored hyperspectral indices were
highly correlated with these efficiency variables, while most
conventional indices were less well correlated. Only NPQI
and WI had relatively high correlations with efficiency
variables (r 2 = 0.51–73). This may be because most conventional indices were examined using measurements in a
few broad bands such as those for Landsat-TM, while
hyperspectral measurements allow optimum selection of the
Fig. 8. Relationship of εG with NDSI[710,410], NDSI[710,530] and NDSI
[530,550] by nadir measurements.
most significant wavelengths. The spectral width for selected NDSIs was fairly narrow (3–10 nm), which requires a
high signal/noise ratio, but future image sensors should
enable detection of low energy signatures at high spectral
resolution.
Y. Inoue et al. / Remote Sensing of Environment 112 (2008) 156–172
A similar analysis was conducted for off-nadir measurements since we presumed that they would carry more
physiological information on photosynthetic efficiency, as
little background is within the field of view. Fig. 9 zooms in on
a part of the contour map of r 2 between εG and NDSIs derived
from off-nadir measurements. The region consisting of 400–
440 nm vs. 400–500 nm has rather high r2 values (reddish),
while there is a deep trough around 410–530 nm vs. 500–
580 nm region (bluish). It is interesting that the 400–420 nm
regions showed a high correlation ridge along R422. Fig. 10
shows a close-up view of two significant parts in the contour
map of r2 between φ and NDSIs derived from off-nadir
measurements. A ridge of high r 2 was found along 413 nm,
while the other regions showed poor correlations (Fig. 10a). In
contrast, relatively large regions had high r2 values around the
1000–1120 nm vs. 1100–1130 nm and the 1130–1150 nm vs.
1160–1200 nm regions. There are deep troughs in other parts
around the 1000–1130 nm vs. 1150–1170 nm and the 1020–
1040 nm vs. 1020–1050 nm regions.
Table 2 summarizes the statistical significance of the
conventional indices and some selected NDSIs in assessment
of photosynthetic-efficiency variables. As expected, among the
conventional indices, PRI and NPQI were better correlated with
εN (r2 = 0.636 and 0.708, respectively), εG (r2 = 0.391 and
0.601) and φ (r2 = 0.262 and 0.500) than NDSIs from nadir
measurements. Nevertheless, a closer relationship was found at
165
NDSI[442, 435] for εN (r2 = 0.747; Fig. 11), NDSI[442, 416]
for εG (r2 = 0.716; Fig. 12), and NDSI[1107, 1110] for φ
(r2 = 0.766; Fig. 13). The spectral width at these peaks was as
narrow as 1–3 nm in most indices. These results suggest that
physiological information on photosynthetic efficiency may be
better estimated by directional measurement than nadir measurement, as suggested by previous reports (e.g., Takebe et al.,
1990). Since multi-angular measurements provide more
accurate or extra information on vegetation canopies (Asner
et al., 1998; Bicheron & Leroy, 1999), photosyntheticefficiency variables may be included in the targets of such
directional sensors. In case of off-nadir measurements, both εN
and εG were related to NDSIs using visible wavelengths only,
whereas NDSIs using water-related wavelengths (such as 970
and 1122 nm) were selected in nadir-measurements. This may
be because the background was only slightly included, if at all,
in the field of view in the off-nadir measurements. φ might be
related to NDSIs using such water-related wavelengths since
quantum efficiency represents the initial slope at weak light
intensity and may be sensitive to leaf water status. The reflectance
spectrum of the background surface was consistent during flooded
conditions; nevertheless, water conditions in both background
surfaces and plant leaves might have affected the results. Analyses
using process models such as PROSPECT +SAIL with additional
measurements would allow more detailed investigation of the
effects of water in the background and in plants.
Fig. 9. A contour map of coefficient of determination (r2) between εG and the normalized difference spectral indices (NDSIs) using the two separate wavelengths on x
and y axes. A significant portion has been expanded out of a map for the entire wavelength regions to depict the details. Data are for off-nadir measurements.
166
Y. Inoue et al. / Remote Sensing of Environment 112 (2008) 156–172
Fig. 10. A contour map of coefficient of determination (r2) between φ and the normalized difference spectral indices (NDSIs) using the two separate wavelengths on x
and y axes. A significant portion has been expanded out of a map for the entire wavelength regions to depict the details. Data are for off-nadir measurements.
Y. Inoue et al. / Remote Sensing of Environment 112 (2008) 156–172
167
Table 2
Coefficient of determination (r2) between efficiency parameters and selected spectral indices derived from off-nadir measurements
Spectral indices
(Off-nadir)
εN
εG
φ
NDSI[442, 435]
NDSI[442, 438]
NDSI[543, 548]
Δ [2, 2]
Δ [2, 2]
Δ [3, 2]
0.747
0.745
0.720
–
–
–
–
–
–
NDSI[422, 406]
NDSI[422, 416]
NDSI[422, 419]
Δ [2, 1]
Δ [2, 2]
Δ [2, 2]
–
–
–
0.704
0.716
0.711
–
–
–
NDSI[413, 416]
NDSI[962, 964]
NDSI[971, 973]
NDSI[1060, 1118]
NDSI[1107, 1110]
Δ
Δ
Δ
Δ
Δ
–
–
–
–
–
–
–
–
–
–
0.722
0.708
0.710
0.708
0.766
0.636
0.037
0.550
0.708
0.304
0.077
0.495
0.042
0.376
0.000
0.066
0.057
0.001
0.126
0.111
0.495
0.391
0.029
0.312
0.601
0.130
0.131
0.228
0.001
0.108
0.032
0.000
0.000
0.028
0.193
0.220
0.323
0.262
0.020
0.139
0.500
0.016
0.199
0.099
0.032
0.069
0.096
0.018
0.024
0.086
0.261
0.287
0.141
[1, 2]
[1,1]
[1, 1]
[20, 3]
[2, 2]
PRI: [R531 − R570] / [R531 + R570]
WI: R900 / R970
SIPI: [R800 − R445] / [R800 − R680]
NPQI: [R415 − R435] / [R415 + R435]
GRI: R830 / R550
SWWI: R1650 / R850
NDVI: [R830 − R660] / [R830 + R660]
SAVI: 15[R830 − R660] / [R830 + R660 + 0.5]
RVI: R830 / R660
WDVI: R830 − 1.06R660
MSAVI: (1 + L)[R830 − R660] / [R830 + R660 + L]
EVI: 2.0[R830 − R660] / [1 + R830 + 6R660 − 7.5R460]
GEMI: η(1 − 0.25η) − (R660 − 0.125) / (1 − R660)
NDMI1: [R1650 − R830] / [R1650 − R850]
NDMI2: [R2200 − R1100] / [R2200 + R11100]
BI: 6 [R460 / R660]
Notes: The NDSIs in this Table were selected on the basis of the contour map a part of which was shown in Figs. 9 and 10. Other notes are same as in Table 1.
3.3. Relationship of photosynthetic-capacity variables with
reflectance spectra
Next, we investigated the relationship between hyperspectral
reflectance and photosynthetic-capacity variables such as
fAPAR and Pmax. The correlation coefficients between
reflectance at a single wavelength and canopy photosynthesis
parameters are shown in Fig. 6b. The response curves were
quite similar for fAPAR, NEECO2, GPP, and Pmax; as explained
for the efficiency variables in Fig. 6a, the line for NEECO2 is
nearly inverse to the others due to the inverse sign of NEECO2.
In general, near-infrared wavelengths were highly correlated
Fig. 11. Relationship between εN and NDSI[442, 435] by off-nadir measurements.
Fig. 12. Relationship between εG and NDSI[422, 416] by off-nadir measurements.
168
Y. Inoue et al. / Remote Sensing of Environment 112 (2008) 156–172
Fig. 13. Relationship of φ with NDSI[1107,1110] and NDSI[413,416] by offnadir measurements.
with all variables (|r| = 0.7–0.9). The highest correlation was
found at around 720 nm for fAPAR, although the curve was
rather flat over the near-infrared wavelengths. Peaks were found
at around 550 and 670 nm, but the blue region (400–500 nm)
was also relatively highly correlated with all capacity variables.
These wavelengths may have significant roles in the assessment
of capacity variables, although they must be normalized in
some way.
Fig. 14 shows a contour map of r2 between fAPAR and
NDSIs for all combinations of two wavelengths from nadir
measurements. In general, regions with high r2 values are
relatively broad compared to those for the photosyntheticefficiency variables examined in the previous section. One
obvious feature is the overall usefulness of the blue wavelengths
(400–500 nm), although the red-edge region (around 700 nm)
had a very deep trough (bluish) when combined with 400–
670 nm wavelengths. Three peaks were also found over the
near-infrared regions along 750 nm in combination with 970,
1120, and 1250 nm wavelength regions. These peak-regions
extended from 740 nm toward longer wavelengths up to
900 nm, depicting the typical spectral response in green
vegetation. On the other hand, 970 and 1120 nm might be
selected due to their response to water content, as discussed
before, while no potential feature was identified for 1250 nm in
the literature.
Another peak was in the region around 1075 nm vs. 1100 nm.
The most commonly used combination of red region (650–
670 nm) and near-infrared region (800–1300 nm), i.e., NDVI,
also showed a ridge of high r2, although the significance seems
secondary compared to some other peaks. According to the
summary of useful NDSIs and conventional indices in the
assessment of fAPAR (Table 1), fAPAR was well correlated with
some conventional indices such as GEMI (r2 = 0.834), SAVI
(r2 = 0.801), and WDVI (r2 = 0.801), while NDVI was less well
correlated (r2 = 0.629). Considering that these four indices use the
same two bands (red and near-infrared wavelengths only), the
adjustment employed in the former three indices seems effective
in the assessment of fAPAR. Nevertheless, a much higher
correlation was found at NDSI[550, 410] (r2 = 0.931; Fig. 15b)
and NDSI[720, 420] (r2 = 0.926; Fig. 15c).
The NDVI-fAPAR relationship is very different in the
vegetative and ripening stages (Asrar et al., 1989; Choudhury,
2000; Daughtry et al., 1983; Inoue & Iwasaki, 1991; Inoue
et al., 1998). The lower correlation with NDVI in this analysis
could be explained by the dependence of the relationship on
phenology (Fig. 15a). In contrast, new indices using blue
(410 and 420 nm), green (550 nm), and near-infrared
(720 nm) wavelengths showed highly linear relationships
with fAPAR for the entire growing period, with no
phenology dependence. This feature should be attractive for
extending the applicability of the fAPAR approach, since
previous studies have focused on the vegetative stage only
(Choudhury, 2000; Kiniry et al., 1989) or the phenology
dependence has been otherwise neglected. Adjusting for
background effects (like SAVI or MSAVI) to 410, 420, 550,
and 720 nm wavelengths may bring further improvement in
the assessment of fAPAR.
Pmax was moderately correlated with RVI (r2 = 0.667), GRI
2
(r = 0.662), MSAVI (r2 = 0.594), and four new indices, NDSI
[518, 676], NDSI[620, 623], NDSI[620, 637], and NDSI[750,
761] (r2 = 0.65–0.66). In principle, Pmax may not be highly
correlated with actual measurements since Pmax is the potential
photosynthesis at the highest incident radiation without any
stresses. Nevertheless, it is noteworthy that these wavelengths
are all in the chlorophyll a and b absorption region. These
results suggest that conventional indices using red and nearinfrared wavelength regions (e.g., RVI, GRI, and MSAVI) could
be useful for a rough assessment of Pmax. Carter (1998)
showed that R701 was best correlated (r2 = 0.41) with the
“photosynthetic capacity” (observed maximum) in pine canopies, assuming that it could be estimated from photosynthetic
measurements at the leaf level using a leaf chamber. It is
reasonable that the significant wavelengths should be around
the red-edge region, but should not necessarily agree with our
results since the “capacity” variable of Carter (1998) is neither
the efficiency nor capacity variable used in our study but a
combination of both. Accordingly, it is also reasonable that
Y. Inoue et al. / Remote Sensing of Environment 112 (2008) 156–172
169
Fig. 14. A contour map of coefficient of determination (r2) between fAPAR and the normalized difference indices (NDSIs) using the two separate wavelengths on x
and y axes. Data are for nadir measurements.
Carter (1998) found no significant relationship between PRI and
the “capacity“ variable.
4. Conclusions
Based on season-long measurements of CO2 flux and
hyperspectral reflectance in a homogenous ecosystem of flooded
rice, we derived new spectral indices that will be useful for
estimating canopy photosynthetic variables. NDSIs using narrow
spectral bands were highly correlated with radiation use
efficiency (εN, εG, φ) and capacity-related variables (fAPAR,
Pmax). Our selected indices were much more effective than
conventional indices. εG was best estimated by NDSI[410, 710],
and fAPAR was best estimated by NDSI[550,410]. NDSI[550,
410] and NDSI[720, 420] had linear relationships with fAPAR
throughout the growing period with no phenology dependence,
whereas the phenology dependence of NDVI and similar indices
has been a critical issue for the fAPAR approach. Some NDSIs,
such as NDSI[933, 940], use wavelengths that are near each
other, while several NDSIs (e.g., NDSI[450, 1330]) use widely
spaced wavelengths. Also, some NDSIs, such as NDSI[1050,
1222], were a combination of only near-infrared wavelengths,
while the photosynthetic pigments are directly related to visible
wavelengths (400–700 nm). The contribution of some wavelengths in the near-infrared or shortwave-infrared regions (e.g.,
970, 1050, 1120, 1222, and 1330 nm) may be attributed to the
effects of water and lignin/cellulose to photosynthetic efficiency
and capacity. It is also interesting that some NDSIs, such as NDSI
[422, 406], had very sharp, significant peaks (1 to 3 nm),
although several NDSIs, (e.g., NDSI[450, 1330]), were stable
over relatively broad wavelength ranges (40 to 80 nm).
This experimental study was conducted in a rice paddy
ecosystem that was rather simple and homogenous, with few
confounding factors. Therefore, our results include potential
ecophysiological relationships that might be common for various
types of vegetation. These results provide useful insights for the
assessment of radiation use efficiency and photosynthetic
capacity in other ecosystems using a wide range of available
spectral data. This information will also be useful for selection of
sensors and for designing future sensors for ecosystem observations. Nevertheless, the underlying mechanisms for the NDSIs
are not well identified at the moment, except for the well-known
absorption features of major pigments and materials such as
chlorophyll, xanthophyll, carotenoids, and water (Chappelle et
al., 1992; Gamon et al., 1992; Inoue et al., 1993; Peñuelas &
Filella, 1998). Further studies are needed to examine their
applicability to other ecosystems and to investigate the
ecophysiological and photochemical mechanisms to systematically account for the effects of biotic and abiotic stresses.
Acknowledgment
The authors would like to thank Drs. A. Olioso and F. Baret,
INRA-CSE (Avignon, France) for their encouraging suggestions
and information. We are also grateful to the useful comments
from the anonymous reviewers. This study was supported in part
170
Y. Inoue et al. / Remote Sensing of Environment 112 (2008) 156–172
Fig. 15. Relationship of fAPAR with NDVI, NDSI[550, 410] and NDSI[720,
420] by nadir measurements.
by the Global Environment Research Fund, Ministry of
Environment of Japan.
References
Adams, M. L., Philpot, W. D., & Norvell, W. A. (1999). Yellowness index: An
application of spectral second derivatives to estimate chlorosis of leaves in
stressed vegetation. International Journal of Remote Sensing, 20, 3663−3675.
Asner, G. P., Braswell, B. H., Schimel, D. S., & Wessman, C. A. (1998).
Ecological research needs from multi-angle remote sensing data. Remote
Sensing of Environment, 63, 155−165.
Asrar, G., Fuchs, M., Kanemasu, E. T., & Hatfield, J. L. (1984). Estimating
absorbed photosynthetic radiation and leaf area index from spectral
reflectance in wheat. Agronomy Journal, 76, 300−306.
Asrar, G., Myneni, R. B., & Kanemasu, E. T. (1989). Estimation of plant-canopy
attributes from spectral reflectance measurements. In G. Asrar (Ed.), Theory
and application of optical remote sensing (pp. 252−296). New York: Wiley
Interscience.
Baret, F., & Guyot, G. (1991). Potentials and limits of vegetation indices for LAI
and APAR assessment. Remote Sensing of Environment, 35, 161−173.
Barton, C. V. M., & North, P. R. J. (2001). Remote sensing of canopy light use
efficiency using the photochemical reflectance index — Model and
sensitivity analysis. Remote Sensing of Environment, 78, 264−273.
Bicheron, P., & Leroy, M. (1999). A method of biophysical parameter retrieval
at global scale by inversion of a vegetation reflectance model. Remote
Sensing of Environment, 67, 251−266.
Campbell, C. S., Heilman, J. L., McInnes, K. J., Wilson, L. T., Medley, J. C.,
Wu, G., et al. (2001). Seasonal variation in radiation use efficiency of
irrigated rice. Agricultural and Forest Meteorology, 110, 45−54.
Carter, G. A. (1998). Reflectance wavebands and indices for remote estimation
of photosynthesis and stomatal conductance in pine canopies. Remote
Sensing of Environment, 63, 61−72.
Chappelle, E. W., Kim, M. S., & Mcmurtrey, J. E., III (1992). Ratio analysis of
reflectance spectra (RARS): An algorithm for the remote estimation of the
concentrations of Chlorophyll a, Chlorophyll b, and carotenoids in soybean
leaves. Remote Sensing of Environment, 39, 239−247.
Chen, J. M., Liu, J., Leblanc, S. G., Lacaze, R., & Roujean, J. L. (2003). Multiangular optical remote sensing for assessing vegetation structure and carbon
absorption. Remote Sensing of Environment, 84, 516−525.
Choudhury, B. J. (2000). A sensitivity analysis of the radiation use efficiency for
gross photosynthesis and net carbon accumulation by wheat. Agricultural
and Forest Meteorology, 101, 217−234.
Cultivated Soil Classification Committee (1995). Classification of Cultivated
Soils in Japan: Third Approximation. Miscellaneous Publication of the
National Institute of Agro-Environmental Sciences, 17, 79.
Curran, P. J. (1989). Remote sensing of foliar chemistry. Remote Sensing of
Environment, 30, 271−278.
Daughtry, C. S. T., Gallo, K. P., & Bauer, M. E. (1983). Spectral estimates of solar
radiation intercepted by corn canopies. Agronomy Journal, 75, 527−531.
Drolet, G. G., Huemmrich, K. F., Hall, F. G., Middleton, E. M., Black, T. A., Barr, A. G.,
et al. (2005). A MODIS-derived photochemical reflectance index to detect
inter-annual variations in the photosynthetic light-use efficiency of a boreal
deciduous forest. Remote Sensing of Environment, 98, 212−224.
Fan, S. M., Wofsy, S. C., Bakwin, P. B., Jacob, D. J., Anderson, S. M.,
Kebabian, P. L., et al. (1992). Micrometeorological measurements of CH4
and CO2 exchange between the atmosphere and the subarctic tundra.
Journal of Geophysical Research D, Atmospheres, 97, 16627−16643.
Filella, I., Amano, T., Araus, J. L., & Peñuelas, J. (1996). Relationship between
photosynthetic radiation-use efficiency of barley canopies and the photochemical reflectance index (PRI). Physiologia Plantarum, 96, 211−216.
Filella, I., Peñuelas, J., Llorens, L., & Estiarte, M. (2004). Reflectance assessment
of seasonal and annual changes in biomass and CO2 uptake of a Mediterranean
shrubland submitted to experimental warming and drought. Remote Sensing of
Environment, 90, 308−318.
Gamon, J. A., Peñuelas, J., & Field, C. B. (1992). A narrow-waveband spectral
index that tracks diurnal changes in photosynthetic efficiency. Remote
Sensing of Environment, 41, 35−44.
Gamon, J. A., Rahman, A. F., Duncan, J. L., Schildhauer, M., & Huemmrich, K. F.
(2006). Spectral Network (SpecNet) — What is it and why do we need it?
Remote Sensing of Environment, 103, 227−235.
Gamon, J. A., Serrano, L., & Surfus, J. S. (1997). The photosynthetic reflectance
index: An optical indicator of photosynthetic radiation use efficiency across
species, functional types, and nutrient levels. Oeocologia, 112, 492−501.
Gitelson, A. A., & Merzlyak, M. N. (1997). Remote estimation of chlorophyll
content in higher plant leaves. International Journal of Remote Sensing, 18,
2691−2697.
Y. Inoue et al. / Remote Sensing of Environment 112 (2008) 156–172
Goetz, S. J., Prince, S. D., Goward, S. N., Thawley, M. M., & Small, J. (1999).
Satellite remote sensing of primary production: An improved production
efficiency modeling approach. Ecological Modeling, 122, 239−255.
Grace, J. (2005). Role of forest biomes in the global carbon balance. In H. Griffiths,
& P. G. Jarvis (Eds.), The carbon balance of forest biomes (pp. 19−46).
New York: Taylor & Francis Group.
Grossman, Y. L., Ustin, S. L., Jacquemoud, S., Sanderson, E. W., Schmuck, G.,
& Verdebout, J. (1996). Critique of stepwise multiple linear regression for
the extraction of leaf biochemistry information from leaf reflectance data.
Remote Sensing of Environment, 56, 182−193.
Huete, A. R. (1988). A soil-adjusted vegetation index (SAVI). Remote Sensing
of Environment, 25, 89−105.
Inada, K. (1985). Spectral ratio of reflectance for estimating chlorophyll content
of leaf. Japanese Journal of Crop Science, 154, 261−265.
Inoue, Y. (2003). Synergy of remote sensing and modeling for estimating
ecophysiological processes in plant production. Plant Production Science,
6, 3−16.
Inoue, Y., & Iwasaki, K. (1991). Spectral estimation of radiation absorptance
and leaf area index in corn canopies as affected by canopy architecture and
growth stage. Japanese Journal of Crop Science, 60, 578−580.
Inoue, Y., Kimball, B. A., Jackson, R. D., Pinter, P. J., & Reginato, R. J.
(1990). Remote estimation of leaf transpiration rate and stomatal resistance
based on infrared thermometry. Agricultural and Forest Meteorology, 51,
21−33.
Inoue, Y., Miyata, A., Mano, M., Nagai, H., & Olioso, A. (2005). Spatial
analysis of footprint effects on CO2 flux measurements by eddy covariance
method — A case study in rice paddy ecosystem. Proceedings of the 38th
Conference of the Remote Sensing Society of Japan (pp. 71−72).
(in Japanese with English summary).
Inoue, Y., Moran, M. S., & Horie, T. (1998). Analysis of spectral measurements
in rice paddies for predicting rice growth and yield based on a simple crop
simulation model. Plant Production Science, 1, 269−279.
Inoue, Y., Morinaga, S., & Shibayama, M. (1993). Non-destructive estimation of
water status of intact crop leaves based on spectral reflectance measurements. Japanese Journal of Crop Science, 62, 462−469.
Inoue, Y., & Olioso, A. (2006). Estimating dynamics of ecosystem CO2 flux and
biomass production in agricultural field by synergy of process model and
remotely sensed signature. Journal of Geophysical Research D, Atmospheres,
111, D24S91. doi:10.1029/2006JD007469
Inoue, Y., & Peñuelas, J. (2006). Relationship between light use efficiency and
photochemical reflectance index in soybean leaves as affected by soil water
content. International Journal of Remote Sensing, 27, 5249−5254.
Inoue, Y., Sakuratani, T., Shibayama, M., & Morinaga, S. (1994). Remote and
real-time sensing of canopy transpiration and conductance — Comparison
of remote and stem flow gauge methods in soybean canopies as affected by
soil water status. Japanese Journal of Crop Science, 63, 664−670.
Jacquemoud, S. (1993). Inversion of the PROSPECT + SAIL canopy reflectance
model from AVIRIS equivalent spectra: Theoretical study. Remote Sensing
of Environment, 44, 281−292.
Kiniry, J. R., Jones, C. A., O'Toole, J. C., Blanchet, R., Cabelguenne, M., &
Spanel, D. A. (1989). Radiation-use efficiency in biomass accumulation
prior to grain-filling for five grain-crop species. Field Crops Research, 20,
51−64.
Kljun, N., Kastner-Klein, P., Fedorovich, E., & Rotach, M. W. (2004). Evaluation
of Lagrangian footprint model using data from wind tunnel convective
boundary layer. Agricultural and Forest Meteorology, 127, 189−201.
Kumar, M., & Monteith, J. L. (1982). Remote sensing of crop growth. In H.
Smith (Ed.), Plants and the daylight spectrum (pp. 133−144). London:
Academic Press.
Maisongrande, P., Ruimy, A., Dedieu, G., & Saugier, B. (1995). Monitoring seasonal
and interannual variations of gross primary productivity, net primary productivity
and net ecosystem productivity using a diagnostic model and remotely-sensed
data. Tellus, 47B, 178−190.
Miyata, A., Leuning, R., Denmead, O. T., Kim, J., & Harazono, Y. (2000).
Carbon dioxide and methane fluxes from an intermittently flooded paddy
field. Agricultural and Forest Meteorology, 102, 287−303.
Monteith, J. L. (1977). Climate and the efficiency of crop production in Britain.
Philosophical Transactions of the Royal Society B, 281, 277−294.
171
Moran, M. S., Inoue, Y., & Barnes, E. M. (1997). Opportunities and limitations
for image-based remote sensing in precision crop management. Remote
Sensing of Environment, 61, 319−346.
Myneni, R. B., & Williams, D. L. (1994). On the relationship between FAPAR
and NDVI. Remote Sensing of Environment, 49, 200−211.
Nouvellon, Y., Seen, D. L., Rambal, S., Bégué, A., Moran, M. S., Kerr, Y., et al.
(2000). Time course of radiation use efficiency in a shortgrass ecosystem:
Consequences for remotely sensed estimation of primary production. Remote
Sensing of Environment, 71, 43−55.
Olioso, A., Chauki, H., Courault, D., & Wigneron, J. P. (1999). Estimation of
evapotranspiration and photosynthesis by assimilation of remote sensing
data into SVAT models. Remote Sensing of Environment, 68, 341−356.
Olioso, A., Inoue, Y., Ortega-Farias, S., Demarty, J., Wigneron, J. P., Braud, I., et al.
(2005). Future directions for advanced evapotranspiration modeling: Assimilation of remote sensing data into crop simulation models and SVAT models.
Irrigation and Drainage Systems, 19, 377−412.
Peñuelas, J., & Filella, I. (1998). Visible and near-infrared reflectance techniques
for diagnosing plant physiological status. Trends in Plant Science, 3, 151−156.
Peñuelas, J., Filella, I., & Gamon, J. A. (1995). Assessment of plant photosynthetic
radiation-use efficiency with spectral reflectance. New Phytologist, 131,
291−296.
Peñuelas, J., & Inoue, Y. (1999). Reflectance indices indicative of changes in
water and pigment content of peanut and wheat leaves. Photosynthetica, 36,
355−360.
Peñuelas, J., & Inoue, Y. (2000). Reflectance assessment of canopy CO2 uptake.
International Journal of Remote Sensing, 21, 3353−3356.
Pinter, P. J., Jackson, R. D., Ezra, C. E., & Gausman, H. W. (1985). Sun angle
and canopy architecture effects on the reflectance of six wheat cultivars.
International Journal of Remote Sensing, 6, 1813−1825.
Potter, C. P., Randerson, J. T., Field, C. B., Matson, P. A., Vitousek, P. M.,
Mooney, H. A., et al. (1993). Terrestrial ecosystem production: A process
model based on global satellite and surface data. Global Biogeochemical
Cycles, 7, 811−841.
Qi, J., Chehbouni, A., Huete, A. R., Kerr, Y. H., & Sorooshian, S. (1994). A
modified soil adjusted vegetation index. Remote Sensing of Environment, 48,
119−126.
Rahman, A. F., Gamon, J. A., Fuentes, D. A., Roberts, D. A., & Prentiss, D. (2001).
Modeling spatially distributed ecosystem flux of boreal forest using
hyperspectral indices from AVIRIS imagery. Journal of Geophysical Research,
106(D24), 33579−33592.
Roy, J., & Saugier, B. (2001). Terrestrial primary productivity: Definitions and
milestones. In J. Roy, B. Saugier, & H. A. Mooney (Eds.), Terrestrial global
productivity (pp. 1−6). New York: Academic Press.
Ruimy, A., Saugier, B., & Dedieu, G. (1994). Methodology for the estimation of
terrestrial net primary production from remotely sensed data. Journal of
Geophysical Research, 99, 5263−5283.
Saito, M., Miyata, A., Nagai, H., & Yamada, T. (2005). Seasonal variation of
carbon dioxide exchange in rice paddy field in Japan. Agricultural and
Forest Meteorology, 135, 93−109.
Shibles, R. M., & Weber, R. C. (1966). Interception of solar radiation and dry matter
production by various soybean planting patterns. Crop Science, 6, 55−59.
Sims, D. A., Luo, H., Hastings, S., Oechel, W. C., Rahman, A. F., & Gamon, J. A.
(2006). Parallel adjustments in vegetation greenness and ecosystem CO2
exchange in response to drought in a Southern California chaparral ecosystem.
Remote Sensing of Environment, 103, 289−303.
Sims, D. A., Rahman, A. F., Cordova, V. D., Baldocchi, D. D., Flanagan, L. B.,
Goldstein, A. H., et al. (2005). Midday values of gross CO2 flux and light
use efficiency during satellite overpasses can be used to directly estimate
eight-day mean flux. Agricultural and Forest Meteorology, 131, 1−12.
Sinclair, T. R. (1994). Limits to crop yield? In K. J. Boote, J. M. Bennet, T. R.
Sinclair, & G. M. Paulsen (Eds.), Physiology and determination of crop
yield (pp. 509−532). Madison, WI: American Society of Agronomy.
Soil Survey Staff (1992). Keys to Soil Taxonomy. SMSS Technical Monograph,
19 (5th ed.) VA, USA: Pocahontas Press Inc.
Taconet, O., Olioso, A., Ben Mehrez, M., & Brisson, N. (1995). Seasonal
estimation of evaporation and stomatal conductance over a soybean field
using surface infrared temperature. Agricultural and Forest Meteorology,
73, 321−337.
172
Y. Inoue et al. / Remote Sensing of Environment 112 (2008) 156–172
Takahashi, W., Nguyen-Cong, V., Kawaguchi, S., & Minamiyama, M. (2000).
Statistical models for prediction of dry weight and nitrogen accumulation
based on visible and near-infrared hyper-spectral reflectance of rice
canopies. Plant Production Science, 3, 377−386.
Takebe, M., Yoneyama, T., Inada, K., & Murakami, T. (1990). Spectral
reflectance of rice canopy for estimating crop nitrogen status. Plant and Soil,
122, 295−297.
Thenot, F., Méthy, M., & Winkel, T. (2002). The photochemical reflectance index
(PRI) as a water-stress index. International Journal of Remote Sensing, 23,
5135−5139.
Trotter, G. M., Whitehead, D., & Pinkney, E. J. (2002). The photochemical
reflectance index as a measure of photosynthetic light use efficiency for
plants with varying foliar nitrogen contents. International Journal of
Remote Sensing, 23, 1207−1212.
Turner, D. P., Gower, S. T., Cohen, W. B., Gregory, M., & Maiersperger, T. K.
(2002). Effects of spatial variability in light use efficiency on satellite-based
NPP monitoring. Remote Sensing of Environment, 80, 397−405.
Veroustraete, F., Patyn, J., & Myneni, R. B. (1996). Estimating net ecosystem
exchange of carbon using the normalized difference vegetation index and an
ecosystem model. Remote Sensing of Environment, 58, 115−130.
Wall, G. W., & Kanemasu, E. T. (1990). Carbon dioxide exchange rates in wheat
canopies. Part 1. Influence of canopy geometry on trends in leaf area index,
light interception and instantaneous exchange rates. Agricultural and Forest
Meteorology, 49, 81−102.
Webb, E. K., Pearman, G. I., & Leuning, R. (1980). Correction for flux
measurements for density effects due to heat and water vapor transfer.
Quarterly Journal of Royal Meteorological Society, 106, 85−100.
Yoder, B. J., & Pettigrew-Crosby, R. E. (1995). Predicting nitrogen and
chlorophyll content and concentrations form reflectance spectra (400–
2500 nm) at leaf and canopy scales. Remote Sensing of Environment, 53,
199−211.