Available online at www.sciencedirect.

com

Advances in Space Research 41 (2008) 1764–1772

www.elsevier.com/locate/asr

Using the MIR bands in vegetation indices for the estimation

of grassland biophysical parameters from satellite remote
sensing in the Alps region of Trentino (Italy)
Loris Vescovo *, Damiano Gianelle
Centro di Ecologia Alpina, Viote del Monte Bondone, Trento, Italy

Received 12 January 2007; received in revised form 4 July 2007; accepted 30 July 2007

Abstract

Development of new methods for estimating biophysical parameters can be considered one of the most important targets for the
improvement of grassland parameters estimation at full canopy cover. In fact, accurate assessment methods of biophysical characteristics
of vegetation are needed in order to avoid the uncertainties of carbon terrestrial sinks.
Remote sensing is a valid tool for scaling up ecosystem measurements towards landscape levels serving a wide range of applications,
many of them being related to carbon-cycle models. The aim of this study was to test the suitability of satellite platform sensors in esti-
mating grassland biophysical parameters such as LAI, biomass, phytomass, and Green herbage ratio (GR). Also, we wanted to compare
some of the most common NIR and red/green-based vegetation indices with ones that also make use of the MIR band, in relation to
their ability to predict grassland biophysical parameters.
Ground-truth measurements were taken on July 2003 and 2004 on the Monte Bondone plateau (Italian Alps, Trento district) in grass-
lands varying in land use and management intensities. From satellite platforms, an IRS-1C-LISS III image (18/07/2003; 25 m resolution in
the visible-NIR and 70 m resolution in the MIR) and a SPOT 5 image (27/07/2004, 10 m resolution in the visible-NIR and MIR) were used.
LAI, biomass, and phytomass measurements showed logarithmic relationships with the investigated NIR and red/green-based indi-
ces. GreenNDVI showed the highest R2 values (0.59, IRS 2003; 0.60, SPOT 2004). Index saturation occurred above approximately 100–
150 g m2 of biomass (LAI 1.5–2). On the other hand, GR relationships were shown to be linear. MIR-based indices performed better
than NIR and red/green-based ones in estimating biophysical variables, with no saturation effect. Biomass showed a linear regression
with Canopy Index (MIR/green ratio) and with the Normalised Canopy Index (NCI) calculated as a normalised difference between
MIR and green bands (IRS: R2 = 0.91 and 0.90, respectively. SPOT: R2 = 0.63 and 0.64). Similar correlations could also be found
for LAI and phytomass, and GR predictability was shown to be higher than NDVI and GreenNDVI. According to these results obtained
in the investigated areas, phytomass, biomass, LAI, and GR are linearly correlated with the investigated MIR band indices and as a
result, these parameters could be estimated from the adopted satellite platforms with limited saturation problems.
 2007 COSPAR. Published by Elsevier Ltd. All rights reserved.

Keywords: MIR; Vegetation indices; Grassland; Biophysical parameters

1. Introduction Inversion of complex radiation transfer models is one of

the most promising methods and it is based on physical
Quantitative remote sensing tools follow different meth- processes description which allows drivers assessment.
ods to retrieve useful information from the observed vege- The numerical inversion of radiative transfer models on
tation targets. grasslands such as PROSPECT-SAIL has been demon-
strated by Zarco-Tejada and Ustin (2001), indicating that
*
Corresponding author. biophysical properties of grasslands could be theoretically
E-mail address: vescovo@cealp.it (L. Vescovo). acquired from one spectrum collected at nadir. Considering

0273-1177/$34.00  2007 COSPAR. Published by Elsevier Ltd. All rights reserved.

doi:10.1016/j.asr.2007.07.043
 L. Vescovo, D. Gianelle / Advances in Space Research 41 (2008) 1764–1772
the high structural and spatial variability of grasslands, at
the satellite level there is a need of more complex models in
order to incorporate the wide range of optical and struc-
tural parameters (Kuusk, 1998).
More empirical and simpler methods such as the spec-
tral vegetation indices approach have been used as a signif-
icant source of information. Indices analysis is carried out
using ground measurements, satellite-derived data or
model-generated datasets. When ground or satellite data-
sets are used, the indices approach is somewhat limited
because it focuses on homogeneous plant canopies param-
eters estimation (e.g., soybean crops, temperate grasslands,
and conifer forests) and so indices evaluation is valid only
for a specific vegetation type. On the other hand, the indi-
ces approach which uses simulated surface spectral data
allows one to extend the observations to a large class of ter-
restrial environments. This approach, in any case, implies
many assumptions (uniform distribution of leaf angle, no
interaction between soil and vegetation), consequently
models do not represent the full complexity of a real sur-
face (Leprieur et al., 1994).
Spectral vegetation indices calculated as a ratio or nor-
malized difference from Near Infrared (NIR) and visible
bands have been widely used to retrieve biophysical param-
eters for over three decades (Colwell, 1974; Tucker, 1979;
Sellers, 1985, 1987; Gitelson et al., 1996). Accurate esti-
mates of biophysical parameters at full canopy cover have
still not been fully achieved from satellite sensors as broad-
band spectral vegetation indices tend to saturate at full can-
opy cover. For some biophysical parameters expressed as a
ratio (Green Ratio, the relative amount of live and dead
material in herbage and fAPAR, the fraction of adsorbed
PAR) indices saturation seems not to be a major problem
(Di Bella et al., 2004; Vescovo and Gianelle, 2006), but
for many other biophysical characteristics such as LAI,
biomass (green herbage fraction), and phytomass (green
and brown herbage fractions), the relationships with vege-
tation indices are in fact non-linear (Myneni et al., 1995;
Huete et al., 2002).
Many indices have been introduced in order to linearize
the relationships with quantitative biophysical parameters
(LAI, phytomass, and biomass), such as NLI (Non-Linear
Index, Goel and Quin, 1994), Renormalised Difference
Index (RDI, Roujean and Breon, 1995), and Modified Sim-
ple Ratio (MSR, Chen, 1996). The theoretical approach
behind the design of optimal spectral indices has been
described in Verstraete and Pinty (1996) and applied in
Gobron et al. (2000). Thanks to the modelling approach,
optimal indices can be designed and optimised for specific
applications and particular sensors.
On the other hand, relationships between indices and
biophysical parameters expressed as a ratio, such as the
Green Ratio (GR, the relative amount of live and dead
material in herbage) and the vegetation fraction (VF, the
relative amount of area covered by vegetation and covered
by bare soil) are shown to be linear (Gitelson et al., 2002;
Vescovo and Gianelle, 2006). The possibility of detecting
1765
grassland curing and vegetation fraction allows for a good
estimation of grassland development and phenological sta-
tus with no problems related to saturation effect.
The uncertainties involved with parameters estimation
at full canopy cover can result in frequent errors in the
application of the results. For instance, some ecosystem
models used NDVI estimates as surrogates for biomass
(Box et al., 1989) and even though most of the recent mod-
els use the well- know relationship between FAPAR and
NDVI to derive biomass (Myneni and Williams, 1994)
some adjustments are necessary to include stress conditions
and phenology, because if carbon credit budgeting is to be
adopted, accurate assessment of biophysical characteristics
of vegetation will be needed in order to avoid uncertainties
of carbon terrestrial sinks. A significant underestimate of
carbon flux over regions of dense vegetation is expected
in high LAI forest, grassland, and crops ecosystems (Sell-
ers, 1985).
Many multi-spectral sensors not only provide visible
and Near Infrared bands but also Medium Infrared ones
(Landsat ETM, 1550–1750 nm; IRS-LISS III, 1550–
1700 nm; SPOT 5 1580–1750 nm, and Terra-ASTER
1500–1600 nm). The use of the medium-infrared (MIR)
region of the spectrum was proposed many years ago and
the good relationships obtained with biomass (Everitt
et al., 1989) suggest that indices using this spectral region
have probably been under-utilized. Such indices show that
the sensitivity of canopy reflectance to water content in the
1500–1750 nm region (a decrease in leaf water content
causes an increase in MIR reflectance) can be used to
develop indices which make use of greenness/chlorophyll-
based bands together with MIR bands. Huete and Liu
(1994) and Nemani et al. (1993) proposed to correct the
widely used Simple Ratio (SR; Jordan, 1969) and Norma-
lised Difference Vegetation Index (Rouse et al., 1974) by
means of MIR minimum and maximum reflectance values
detected in the ground-truth areas. In addition, Normalised
Difference Water Index (Hardisky et al., 1983) was intro-
duced to determine water content using the normalised dif-
ference between NIR and MIR bands. Specific Leaf Area
Vegetation Index (SLAVI, Lymburner et al., 2000) also
uses both NIR and MIR bands together with the red band.
Ceccato et al. (2002), using reflectance values simulated
with PROSPECT, performed a sensitivity analysis at leaf
level showing that the MIR and NIR regions together are
able to retrieve leaf equivalent water thickness, leaf internal
structure, and leaf dry matter content. On these bases, an
index was created to optimise retrieval of vegetation water
content.
The importance of using a water index for LAI estima-
tion has been underlined by Ustin et al. (2004). Water con-
tent determination, in fact, can be seen as a new approach
to estimate vegetation parameters from remote sensing
platforms, as canopy liquid water absorption has been
shown to be linearly related to LAI. Development of a
non-greeness/chlorophyll based method for estimating
LAI can be seen as complementary to other methods using
1766 L. Vescovo, D. Gianelle / Advances in Space Research 41 (2008) 1764–1772
other spectral regions and could be incorporated into eco-
system models.
According to Dozier (1989) and Hall et al. (1995), the
Normalised Difference Snow Index (NDSI) allows a spec-
tral discrimination between snow, soil, rock, and cloud
cover to be made. NDSI is an index calculated from the
normalised difference between the green and the MIR
bands. The use of similar indices obtained from the green
band (sensitive to chlorophyll content) and the MIR band
(sensitive to water content) is of particular interest to veg-
etation studies. For this reason, testing was conducted for
biophysical parameters estimation from IRS and SPOT
satellite imagery, within this study. Investigations were also
carried out on the MIR/Green ratio (Canopy Index, CI)
and the normalised difference between the MIR and the
Green Band (Normalised Canopy Index, NCI).
The objectives of this study were to test the capability of
different vegetation indices to retrieve grassland biophysi-
cal parameters such as LAI, phytomass, biomass, and
GR. Analysis was focused towards:
– traditional spectral vegetation indices (using the green,
red, and NIR bands, such as SR, NDVI, SAVI, NLI,
MSR, and RDI).
– spectral vegetation indices also using the MIR bands
such as SRc, NDVIc, NDWI, and SLAVI
– two new indices: the Canopy Index (CI = qmir  qgreen)
and the Normalised Canopy Index (NCI = (qmir  qgreen)/
(qmir + qgreen)
– their ability to obtain a linear relationship with grass-
land biophysical parameters.
2. Materials and methods
In order to calibrate spectral measurements, a data-
base was compiled, which included different grasslands
varying in land use, management intensities and produc-
tivity. Data was collected on the Monte Bondone plateau
(approximately 250 ha, 4600 0 N, 1102 0 E) located in the
Trento province (Italy) at 1550 m a.s.l. (Table 1). The
Viote of Monte Bondone grassland eddy covariance site,
located on the plateau, belongs to the SpecNet Spectral
Network (http://www.specnet.info/field_sites/monte_bon-
done.html). The mean annual air temperature is 5.5 C,
Table 1
Site localization, land use, and year of data collection
Location Landuse, management
Gervasi Abandoned
Marocche Pasture
Marocche Pasture
Pra de Elica Medium intensive, slightly fertilized, already cut
Pra de Elica Medium intensive, slightly fertilized, 1 cut per year
Piani Extensive, not fertilized, already cut
Torbiera Peat bog, slightly utilized
Piani Medium intensive, slightly fertilized, 1 cut per year
ranging from 2.7 C in January to 14.4 C in July.
The mean annual rainfall is 1189 mm, with peaks in June
(132 mm) and October (142 mm). Snow cover occurs
between November and April. The vegetation of Monte
Bondone is dominated by a Sieverso-Nardetum strictae
association, and the main species are Nardus stricta,
Trisetum flavescens, Festuca rubra, Brachypodium pinn-
atum, Koeleria piramidata, Avenella flexuosa, Trifolium
pratense, and Alchemilla vulgaris. Meadows represent
the principal land use on this plateau and they are tradi-
tionally managed with low mineral fertilisation and cut
once a year in mid-July. Biomass is commonly harvested
as hay. Pastures are located on the southern parts of this
area and grazed between June and September by cattle.
There are some areas that have been abandoned and
have been invaded by shrubs (Vaccinium sp., Rhododen-
drum sp., and Juniperus communis).
On July 8th, 2003, and July 17th, 2004, several different
study sites (10 · 10 m; see Table 1) were investigated in
order to assess spatial variability of biophysical parameters
such as phytomass, biomass, Leaf Area Index (LAI), and
Green Ratio (GR, the relative amount of live and dead
material in herbage). In each of the investigated ecosys-
tems, two 10 m long and 0.1 m wide strips (representing
1 m2) were cut to collect vegetation samples. The collected
samples of phytomass were divided manually into biomass
(green herbage) and necromass (brown herbage). Biomass
and necromass were both dried at 65 C for 48 h and
GR = (biomass/(biomass+necromass)) * 100 (Green Ratio,
GR) was calculated. Leaf Area Index was also measured
using a Sunscan Canopy Analyzer ceptometer at five points
for each investigated area.
For the satellite data, an IRS (Indian Remote Sensing
Satellite, LISS III sensor; spatial resolution: 25 m in the
visible-NIR and 70 m resolution in the MIR) image was
used in 2003 (July 18th), while a SPOT (Satellite Pour
l’Observation de la Terre, SPOT 5 sensor; spatial resolu-
tion: 10 m in the visible-NIR and MIR) image was used
in the 2004 campaign (July 27th). Both images were taken
on a clear day. Image-to-image registration was per-
formed, using Trentino orthophotos (1 m spatial resolu-
tion) as a reference.
Dark subtraction was applied in order to correct for
atmospheric scattering conditions. Dark pixel subtraction
uses ‘black bodies’ in the image to measure the radiance
Abbreviation Year
Aband 2003, 2004
Pasture 1 2003, 2004
Pasture 2 2003, 2004
Cut 2 2003, 2004
Extensive 1 2003, 2004
Cut 1 2003
Peatbog 2003
Extensive 2 2004
 L. Vescovo, D. Gianelle / Advances in Space Research 41 (2008) 1764–1772 1767

due to atmospheric scattering. The dark pixel value is then 3. Results

subtracted from each of the bands The method considers
brightness values in an area of shadow or for a very dark
object (such as a large clear lake) and determines the min-
imum value. The correction is applied by subtracting the
minimum observed value, determined for each specific
band, from all pixel values in each respective band. Since
scattering is wavelength dependent, minimum values vary
from band to band. No other corrections were applied,
considering the flat morphology and the limited dimensions
of the plateau (250 ha).
A 2 · 2 pixel region of interest was used to calculate the
mean reflectance values for the grassland sites. Vegetation
indices (Table 2) using the infrared and red/green bands
were tested, such as normalized spectral vegetation indices
(NDVI, GreenNDVI) non-normalized ratio (SR), non-lin-
ear (NLI, MSR, RDI), and soil adjusted indices (SAVI).
Also, tested were indices which use the MIR band such
as Corrected Simple Ratio (SRc), Corrected NDVI
(NDVIc), Normalised Difference Water Index (NDWI),
Specific Leaf Area Vegetation Index (SLAVI), together
with the proposed indices Canopy Index (CI), and Norma-
lised Canopy Index (NCI). Vegetation indices were calcu-
lated and correlated with investigated biophysical
parameters using Statistica software (release 7; http://
www.statsoft.it/).
The investigated grassland ground sites illustrated a
large range (Fig. 1) in the canopy structure (LAI 0.4–
5.3), botanical composition, phytomass (31–592 g m2),
and GR (24–93%) during the two year period. Phytomass
levels and LAIs were significantly higher in 2003 than in
2004. Also, necromass levels (and consequently Green
Ratio) varied considerably during 2003 and 2004, due to
the very dissimilar climatic conditions of the two summers
(the 2003 summer was hot and dry while in 2004 it was
much cooler and wetter). As a result, the Monte Bondone
plateau grasslands were more cured in 2003 than in 2004.
Interestingly, GR range was 24–74% in 2003 and 63–93%
in 2004. Necromass lelels were between 73 and 155 g m2
in 2003 and between 10 and 65 g m2 in 2004.
To develop a vegetation index, it is very important to
understand how biophysical parameters (such as biomass,
LAI, water, and nitrogen content) affect the signal mea-
sured at satellite level and which wavelengths can be used
for vegetation parameters retrieval. For this purpose,
reflectance values were measured on the different grassland
ecosystems using a ASD Field Spec Pro spectroradiometer,
and the linear correlation coefficient between reflectance
and the investigated parameters was calculated for all the
different wavelengths as shown in Fig. 2. It is possible to

Table 2
Summary of the characteristics of the vegetation indices used
Index Acronym
SR: Simple Ratio
NDVI: Normalized
Difference Vegetation
Index
SAVI: Soil Adjusted
Vegetation Index
NLI: Non-Linear Index
GNDVI: Green Normalized
Difference Vegetation
index
MSR: Modified Simple Ratio
RDI: Renormalized
Difference Index
SRc: Corrected Simple Ratio
NDVIc: Corrected NDVI
NDWI: Normalised
Difference Water Index
SLAVI: Specific Leaf Area
Vegetation Index
CI: Canopy Index
NCI: Normalized Canopy
Index
Algorithm
SR = qnir/qred
NDVI = (qnir  qred)/(qnir + qred)
SAVI: (1 + L) (qnir  qred)/
(qnir + qred + L)
NLI ¼ ðq2nir  qred Þ=ðq2nir þ qred Þ
NDVI = (qnir  qgreen)/(qnir + qgreen)
MSR = (qnir/qred  1)/((qnir/qred)1/2 + 1)
RDI = (qnir  qred)/((qnir + qred)1/2)
SRc = SR (1((qmir  qmir min)/
(qmir max  qmir min))
NDVIc = NDVI (1((qmir  qmir min)/
(qmir max  qmir min))
NDWI = (qnir  qmir)/(qnir + qmir)
SLAVI = qnir/(qred + qmir)
CI = qmir  qgreen
NCI = (qmir  qgreen)/(qmir + qgreen)
Description and use References
One of most simple indexes. Related to changes in Jordan (1969)
amount of green biomass, pigment content, and
concentration
One of most used indexes for biomass estimation. Rouse et al. (1974)
The values are normalized for the amount of
incident radiation, reducing the impact of factors
such as slope and aspect
L ranges from 0 for very high vegetation cover to Huete (1988)
1 for very low vegetation cover; minimizes soil
brightness-induced variations
Aims to linearize the relationships with surface Goel and Quin (1994)
parameters which tend to be non-linear
Uses the green channel and helps determine Gitelson et al. (1996)
nitrogen influences from the green colour of the
leaf
Aims to linearize the relationships between the Chen (1996)
index and biophysical parameters
Aims to linearize relationships with parameters Roujean and Breon
which tend to be non-linear (1995)
Aims to linearize relationships with parameters Brown et al. (2000)
using the MIR band
Aims to linearize relationships with parameters Nemani et al. (1993)
using the MIR band
Aims to determine vegetation water As NDII in Hardisky
et al. (1983); Gao (1996)
Aims to determine Specific Leaf Area Lymburner et al. (2000)
Aims to linearize relationships with parameters Proposed in this study
using the MIR and the green bands
Aims to linearize relationships with parameters Proposed in this study
using the MIR and the green bands
1768 L. Vescovo, D. Gianelle / Advances in Space Research 41 (2008) 1764–1772
Fig. 1. Grassland biophysical parameters measured during the ground campaigns of 2003 and 2004 summers.
see in Fig. 2 that water content shows high absolute r-val-
ues both in the green (fromj0.97j to j0.95j) and in the MIR
region (from j0.93j to j0.95j). Also, there is a good correla-
tion between those regions and other parameters such as
LAI and biomass. For these reasons, we were encouraged
to select indices such as NCI and CI, obtained from the
green band (sensitive to chlorophyll content) and the
MIR band (sensitive to water content).
From the satellite platforms, vegetation indices were sat-
urating at full canopy cover when quantitative biophysical
parameters (LAI, phytomass, and biomass) reached a given
threshold (biomass values higher than 150 g m2, corre-
sponding to LAI 1.8–2). This behaviour highlights the typ-
ical asymptotic trends that occurred in both years. Within
the investigated NIR and visible bands-based spectral veg-
etation indices, GreenNDVI demonstrated the highest log-
arithmic R2 values (0.59, IRS 2003; 0.78, SPOT 2004;
p < 0.05). R2 values of linear relationships between vegeta-
tion indices and biophysical parameters were considerably
lower for SR, NDVI, GNDVI, SAVI, NLI, MSR, and
RDI (the NIR-visible bands-based vegetation indices;
Fig. 3) because of the saturation tendency of the indices.
In 2003, only the Green-NDVI GR linear correlation was
significant (p < 0.05; R2 = 0.65) while in 2004 only the
Green-NDVI LAI linear correlation was significant
(p < 0.05; R2 = 0.68).
Quite surprisingly, considering that GR (the relative
amount of live and dead material in herbage) normally pro-
duces a strong linear relationship (Tucker, 1980; Vescovo
and Gianelle, 2006; Gianelle and Vescovo, 2007), GR did
not exhibit a significant linear correlation with any of the
NIR-visible investigated indices in 2004. This might be
related to the very narrow GR range observed in 2004,
and to a lack of sensitivity of the NIR and visible bands-
based indices.
The 2004 SPOT 5 higher spatial resolution and the sub-
sequent lower mixed-pixel effect (due to the high spatial
variability of grassland parameters) might play an
 L. Vescovo, D. Gianelle / Advances in Space Research 41 (2008) 1764–1772 1769

Fig. 2. Correlation coefficient between biophysical parameters and ground reflectance. R absolute values above 0.705 (indicated by the two horizontal
lines) are significant to p < 0.05.

Fig. 3. R2 obtained from linear regressions for the different vegetation indices (calculated using the NIR and visible bands) and biophysical parameters. R2
values in bold in the table are significant to p < 0.05.

important role for these indices as R2 values tended to be reflectance values within the 2 · 2 pixel region of interest,
generally higher in 2004 than in 2003 (Fig. 3). Higher spa- resulting in a more accurate index calculation for the inves-
tial resolution, in fact, allows a more precise calculation of tigated ground sites (Gianelle and Vescovo, 2007).
1770 L. Vescovo, D. Gianelle / Advances in Space Research 41 (2008) 1764–1772
Another important factor to consider is the fact that the
noticeably higher GR values were observed in 2004. Accord-
ing to the analysed data, vegetation indices and biophysical
parameters correlations tend to show higher R2 values in
2004, when GR values are considerably higher. Standing
dead necromass rates were much higher in 2003 (Fig. 4), so
a shading effect of the senescent leaves can be expected on
the indices values (as demonstrated by Di Bella et al., 2004)
and consequently on the correlation coefficients values.
Considering the vegetation indices that used the MIR
band (Fig. 4), CI and NCI showed very good linear corre-
lations with most of the investigated parameters. CI and
NCI correlation coefficients were generally higher in 2003
than in 2004 (despite the much lower spatial resolution of
the MIR band of IRS sensor vs. SPOT 5). This may again
indicate that the predictability of quantitative biophysical
parameters by CI and NCI is more efficient when the var-
iable ranges are higher. In 2003, very high correlations
(p < 0.01) were found with all the parameters investigated
(Fig. 4): phytomass (CI R2 = 0.91; NCI R2 = 0.92), bio-
mass (CI R2 = 0.91; NCI R2 = 0.90), LAI (CI R2 = 0.85;
NCI R2 = 0.88), and GR (CI R2 = 0.73; NCI R2 = 0.82).
In 2004, significant (p < 0.01) correlations were found
between CI and phytomass, biomass and GR (Fig. 4) while
NCI was correlated only with the two first biophysical
parameters (Fig. 4).
Fig. 4. R2 obtained from linear regressions for the different vegetation indices (calculated using also the MIR band) and biophysical parameters. R2 values
in bold in the table are significant to p < 0.05.
NDWI, SRc, and NDVIc did not produce satisfactory
results in 2003 but they were significantly-correlated with
LAI in 2004 data. No significant correlation was found
between SLAVI and any of the parameters investigated.
A marked shift in the 2003 and 2004 fit lines between CI
and biomass were observed (Fig. 5a), due to the higher
reflectance of 2003 ground points in the MIR band. This
was probably related to drier conditions in 2003, higher
necromass rates and lower water content. CI was well cor-
related with GR (Fig. 5b) even in 2004, when the GR range
was narrower. As seen for biomass, there is a clear shift in
the 2003 and 2004 fit lines probably due to the different
vegetation conditions and to the different MIR bands used
to calculated CI index.
4. Conclusion
Different results were obtained in 2003 and 2004 from
IRS and SPOT satellite sensors, which seemed to be mostly
related to the different phenology status (dealing to very
different GR values) and climatic conditions of the Monte
Bondone grasslands during these two years. For many indi-
ces, R2 values tended to be generally higher in 2004 than in
2003. SPOT higher spatial resolution in 2004 is one of the
important factors to consider to explain such inter-annual
variability, but another important factor to consider is the
 L. Vescovo, D. Gianelle / Advances in Space Research 41 (2008) 1764–1772
Fig. 5. Canopy Index relationships with biomass and Green Ratio in 2003 and 2004.
fact that the noticeably higher GR values were observed in
2004, so a shading effect of the senescent leaves can also
explain the differences between the 2003 and 2004 the cor-
relation coefficients values.
According to the investigated data set, spectral vegeta-
tion indices using the green, the red and the NIR bands
showed a limited capability to retrieve grassland quantita-
tive biophysical parameters at full canopy cover. This is
due to the well-known index saturation effect, which began
when biomass exceeded 150 g m2. In fact, although good
relationships between indices and LAI, phytomass, and
biomass could be found, they tended to be logarithmic.
In comparison, grassland GR did not show a saturation
effect.
By using the proposed indices CI and NCI, it was possi-
ble to obtain significant relationships with most of the
grassland studied parameters. These two indices, based on
green and MIR spectral reflectance, produced linear corre-
lations with the investigated quantitative parameters (LAI,
biomass, and phytomass) and allowed predictability of GR
in the summers of 2003 and 2004. The use of the other MIR-
including spectral vegetation indices such as SRc, NDVIc,
NDWI, and SLAVI did not produce satisfactory results
within 2004 data sets. Significant correlations could be
found between LAI and SRc, NDVIc, NDWI. The simulta-
neous use of the green band (sensitive to chlorophyll con-
tent) and the MIR band (sensitive to water content and to
LAI) appears to be a valid method for estimating grassland
quantitative biophysical parameters. Further studies are
necessary to test CI and NCI within a larger ground truth
data set. Also, evaluation of the indices on the basis of sim-
ulated data generated by computer models could help to
evaluate the indexes performances and to implement bio-
physical parameters estimation at the regional scale.
Acknowledgements
The authors gratefully acknowledge the SpecNet Spec-
tral Network (http://www.specnet.info) to which the Viote
of Monte Bondone site belongs to and the National Centre
1771
for Ecological Analysis & Synthesis (NCEAS, Santa
Barbara, California) for their valuable scientific and logis-
tic support.
References
Box, E.O., Holben, B.N., Kalb, V. Accuracy of the AVHRR vegetation
index as a predictor of biomass, primary productivity, and net CO2
flux. Vegetatio 80, 71–89, 1989.
Brown, L., Chen, J.M., Leblanc, S.G., Cihlar, J. A shortwave infrared
modification to the simple ratio for LAI retrieval in boreal forests: and
image and model analysis. Remote Sensing of Environment 71, 16–25,
2000.
Ceccato, P., Gobron, N., Flasse, S., Pinty, B., Tarantola, S. Designing a
spectral index to estimate vegetation water content from remote
sensing data: Part 1 – Theoretical approach. Remote Sensing of
Environment 82, 188–197, 2002.
Chen, J. Evaluation of vegetation indices and modified simple ratio for boreal
applications. Canadian Journal of Remote Sensing 22, 229–242, 1996.
Colwell, J.E. Vegetation canopy reflectance. Remote Sensing of Environ-
ment 3, 175–183, 1974.
Di Bella, C.M., Paruelo, J.M., Becerra, J.E., Bacour, C., Baret, F. Effect of
senescent leaves on NDVI-based estimates of APAR: experimental and
modelling evidences. International Journal of Remote Sensing 25 (23),
5415–5427, 2004.
Dozier, J. Spectral signature of Alpine snow cover from Landsat Thematic
Mapper. Remote Sensing of Environment 28, 9–22, 1989.
Everitt, J.H., Escobar, D.E., Richardson, A.J. Estimating grassland
phytomass production with near-infrared and mid infrared spectral
variables. Remote Sensing of Environment 30, 257–261, 1989.
Gao, B.C. NDWI – a normalized difference water index for remote sensing
of vegetation liquid water from space. Remote Sensing of Environment
58, 257–266, 1996.
Gianelle, D., Vescovo, L. The determination of green herbage ratio in
grasslands using spectral reflectance. Methods and ground measure-
ments. International Journal of Remote Sensing 28, 931–942, 2007.
Gitelson, A.A., Kaufman, Y.J., Merzlyak, M.N. Use of a green channel in
remote sensing of global vegetation from EOS-MODIS. Remote
Sensing of Environment 58, 289–298, 1996.
Gitelson, A.A., Kaufman, Y.J., Stark, R., Rundquist, D. Novel algo-
rithms for remote estimation of vegetation fraction. Remote Sensing of
Environment 80, 76–87, 2002.
Goel, N.S., Quin, W. Influences of canopy architecture on relationships
between various vegetation indexes and LAI and FPAR: a computer
simulation. Remote Sensing of Environment 10, 309–347, 1994.
Gobron, N., Pinty, B., Verstraete, M.M., Widlowski, J.L. Advanced
vegetation indices optimized for up-coming sensors: design, perfor-
1772 L. Vescovo, D. Gianelle / Advances in Space Research 41 (2008) 1764–1772
mance, and applications. IEEE Transactions on Geoscience and
Remote Sensing 38, 2489–2505, 2000.
Hall, D.K., Riggs, G.A., Salomonson, V.V. Development of methods for
mapping global snow cover using moderate resolution imaging
spectroradiometer (MODIS) data. Remote Sensing of Environment
54, 127–140, 1995.
Hardisky, M.A., Klemas, V., Smart, R.M. The influence of soil salinity,
growth form, and leaf moisture on spectral radiance of Spartina
alterniflora canopies. Photogrammetric Engineering and Remote
Sensing 49, 77–83, 1983.
Huete, A.R. A soil adjusted vegetation index SAVI. Remote Sensing of
Environment 25, 295–309, 1988.
Huete, A.R., Liu, H.Q. An error and sensitivity analysis of the
atmospheric and soil correcting variants of the normalized difference
vegetation index for the MODIS-EOS. IEEE Transactions on
Geoscience and Remote Sensing 32 (4), 897–905, 1994.
Huete, A.R., Didan, K., Miura, T., Rodriguez, E.U., Gao, X., Ferreira,
L.G. Overview of radiometric and biophysical performance of the
MODIS vegetation indices. Remote Sensing of Environment 83, 195–
213, 2002.
Jordan, C.F. Derivation of leaf area index from quality of light on the
forest floor. Ecology 50, 663–666, 1969.
Kuusk, A. Monitoring of vegetation parameters on large areas by the
inversion of a canopy reflectance model. International Journal of
Remote Sensing 19 (15), 2893–2905, 1998.
Leprieur, C., Verstraete, M.M., Pinty, B. Evaluation of the performance
of various vegetation indices to retrieve vegetation cover from
AVHRR data. Remote Sensing Reviews 10, 265–284, 1994.
Lymburner, L., Beggs, P.J., Jacobson, C.R. Estimation of canopy-average
surface-specific leaf area using Landsat TM data. Photogrammetric
Engineering and Remote Sensing 66 (2), 183–191, 2000.
Myneni, R.B., Williams, D.L. On the relationship between fAPAR and
NDVI. Remote Sensing and Environment 49, 200–211, 1994.
Myneni, R.B., Forrest, G.H., Piers, J.S., Marshack, A.L. The interpre-
tation of spectral vegetation indices. IEEE Transactions on Geoscience
and Remote Sensing 33 (2), 468–481, 1995.
Nemani, R., Pierce, L., Running, S. Forest ecosystem process at the
watershed scale: sensitivity to remotely-sensed leaf area index
estimates. International Journal of Remote Sensing 14, 2519–2539,
1993.
Roujean, J.L., Breon, F.M. Estimating PAR absorbed by vegetation from
bidirectional reflectance measurements. Remote Sensing of Environ-
ment 51, 375–384, 1995.
Rouse, J.W., Hass, R.H., Shell, J.A., Deering, D.W. Monitoring
vegetation systems in the great plains with ERTS-1. In: Proceed-
ings 3rd Earth Resources Technology Satellite Symposium 1, 309–
317, 1974.
Sellers, P.J. Canopy reflectance, photosynthesis and transpiration. Inter-
national Journal of Remote Sensing 6 (8), 1335–1372, 1985.
Sellers, P.J. Canopy reflectance, photosynthesis, and transpiration. 2. The
role of biophysics in the linearity of their interdependence. Remote
Sensing of Environment 21 (2), 143–183, 1987.
Tucker, C.J. Red and photographic infrared linear combination for
monitoring vegetation. Remote Sensing of Environment 8, 127–150,
1979.
Tucker, C.J. A spectral method for determining the percentage of green
herbage material in clipped sample. Remote Sensing of Environment 9,
175–181, 1980.
Ustin, S.L., Zarco-Tejada, P.J., Jacquemoud, S., Asner, G. Remote
Sensing of the Environment: state of the science and new directions.
Remote Sensing for Natural Resources, Management and environ-
mental monitoring. 3rd Ed., Vol. 4, 679–728, 2004.
Verstraete, M.M., Pinty, B. Designing optimal spectral indices for remote
sensing applications. IEEE Transactions on Geoscience and Remote
Sensing 34, 1254–1265, 1996.
Vescovo, L., Gianelle, D. Mapping the green herbage ration of grasslands
using both aerial and satellite-derived reflectance. Agriculture, Eco-
systems and Environment 115, 141–149, 2006.
Zarco-Tejada, P.J., Ustin, S.L. Modeling canopy water content for carbon
estimates from MODIS data at Land EOS validation sites. IEEE
Geoscience and Remote Sensing Symposium, IGARSS ’01. IEEE 2001
International, 1, 342–344, 2001.