You are on page 1of 16

This article was downloaded by: [Sardar Patel University]

On: 14 March 2013, At: 03:49


Publisher: Taylor & Francis
Informa Ltd Registered in England and Wales Registered Number: 1072954 Registered
office: Mortimer House, 37-41 Mortimer Street, London W1T 3JH, UK

International Journal of Remote


Sensing
Publication details, including instructions for authors and
subscription information:
http://www.tandfonline.com/loi/tres20

Canopy level estimations of chlorophyll


and LAI for two tropical species (teak
and bamboo) from Hyperion (EO1) data
a a a
Dhaval Vyas , Binal Christian & N. S.R. Krishnayya
a
Ecology Lab, Department of Botany, Faculty of Science, M.S.
University of Baroda, Baroda, 390 002, India
Version of record first published: 02 Nov 2012.

To cite this article: Dhaval Vyas , Binal Christian & N. S.R. Krishnayya (2013): Canopy level
estimations of chlorophyll and LAI for two tropical species (teak and bamboo) from Hyperion (EO1)
data, International Journal of Remote Sensing, 34:5, 1676-1690

To link to this article: http://dx.doi.org/10.1080/01431161.2012.725484

PLEASE SCROLL DOWN FOR ARTICLE

Full terms and conditions of use: http://www.tandfonline.com/page/terms-and-


conditions

This article may be used for research, teaching, and private study purposes. Any
substantial or systematic reproduction, redistribution, reselling, loan, sub-licensing,
systematic supply, or distribution in any form to anyone is expressly forbidden.

The publisher does not give any warranty express or implied or make any representation
that the contents will be complete or accurate or up to date. The accuracy of any
instructions, formulae, and drug doses should be independently verified with primary
sources. The publisher shall not be liable for any loss, actions, claims, proceedings,
demand, or costs or damages whatsoever or howsoever caused arising directly or
indirectly in connection with or arising out of the use of this material.
International Journal of Remote Sensing
Vol. 34, No. 5, 10 March 2013, 1676–1690

Canopy level estimations of chlorophyll and LAI for two tropical


species (teak and bamboo) from Hyperion (EO1) data
Dhaval Vyas, Binal Christian, and N.S.R. Krishnayya*

Ecology Lab, Department of Botany, Faculty of Science, M.S. University of Baroda, Baroda
390 002, India
(Received 14 July 2010; accepted 14 May 2012)
Downloaded by [Sardar Patel University] at 03:49 14 March 2013

Current economic development in tropical regions (especially in India, China, and


Brazil) is putting tremendous pressure on tropical forest cover. Some of the domi-
nant and economically important species are planted at large scale in these countries.
Teak and bamboo are two important species of tropical regions because of their com-
mercial and conservation values. Accurate estimates of foliar chemistry can help in
evaluating the health status of vegetation in these regions. An attempt has been made
to derive canopy level estimation of chlorophyll and leaf area index (LAI) for these
species utilizing Hyperion data. Partial least square (PLS) regression analysis was car-
ried out to identify the correlation between measured parameters (chlorophyll and LAI)
and Hyperion reflectance spectra. PLS regression identified 600–750 nm as a sensi-
tive spectral region for chlorophyll and 1000–1507 nm for LAI. The PLS regression
model tested in this study worked well for the estimation of chlorophyll (R2 = 0.90,
root mean square error (RMSE) = 0.182 for teak and R2 = 0.84, RMSE = 0.113 for
bamboo) and for the estimation of LAI (R2 = 0.87, RMSE = 0.425). The lower predic-
tive error obtained indicates the robustness of the data set and also of the applicability
of the PLS regression analysis. Wavelengths recognized by the PLS regression model
were utilized for the development of vegetation indices for estimating chlorophyll and
LAI. Predictive performances of the developed simple ratios (SRs) were evaluated using
the cross-validation method. SR 743/692 gave the best results for the prediction of
chlorophyll with the leave-one-out cross-validation (LOO-CV) method (R2 = 0.73,
RMSE = 0.28 for teak and R2 = 0.71, RMSE = 0.15 for bamboo). The normalized
difference ratio (ND 1457/1084) gave the best results for the prediction of LAI with
LOO-CV (R2 = 0.66, RMSE = 0.57). Ratios developed here can be tested for teak and
bamboo cover spread in tropical regions with similar environmental conditions.

1. Introduction
Estimates of the foliar chemistry of vegetation canopies allow a better understanding of
ecosystem functioning since many biochemical processes such as photosynthesis, respi-
ration, and litter decomposition are related to the foliar chemistry of plants (Huber et al.
2008). Among vegetation characteristics, canopy chlorophyll content and leaf area index
(LAI) are of prime importance (Chen et al. 2002; Hansen and Schjoerring 2003; Bacour
et al. 2006; Houborg, Soegaard, and Boegh 2007). Both of these are essential to understand
plant and whole ecosystem functions. Blackburn and Ferwerda (2008) have mentioned
that information concerning the spatial and temporal dynamics of leaf chlorophyll is

*Corresponding author. Email: krish14@gmail.com

ISSN 0143-1161 print/ISSN 1366-5901 online


© 2013 Taylor & Francis
http://dx.doi.org/10.1080/01431161.2012.725484
http://www.tandfonline.com
International Journal of Remote Sensing 1677

of considerable value from a scientific viewpoint, particularly in investigating plant–


environment interactions. Estimating chlorophyll at larger spatial scales is a difficult task.
With the advent of imaging spectroscopy, also known as hyperspectral remote sensing,
significant advancement was achieved in monitoring quantitative characteristics of forest
ecosystems (Wessman et al. 1988; Ollinger et al. 2002; Asner, Scurlock, and Hicke 2003;
Ustin et al. 2004). Researchers have shown the significance of hyperspectral data for the
estimation of chlorophyll (Sims and Gamon 2002; Zhao et al. 2005; Haboudane et al. 2008;
Wu et al. 2010a). Many of the studies were mainly concentrated on coniferous tree species
largely having structurally homogeneous distribution. Studies from temperate regions can-
not be extrapolated for tropical regions because of the differences in the distribution and
biophysical characteristics of vegetation. Therefore, it is essential to develop unique and
appropriate models for chlorophyll estimation.
LAI (the total one-sided leaf area per ground surface area) is considered to be a key
biophysical parameter of ecosystem processes (Asner 2008). Various ecophysiological pro-
Downloaded by [Sardar Patel University] at 03:49 14 March 2013

cesses of a forest ecosystem such as interception of light (Vargas, Anderson, and Jenson
2002), precipitation (Van dijk and Bruijnzeel 2001), and transpiration (Granier et al. 2000)
are controlled by LAI. It is an important biophysical characteristic of the forest stand, show-
ing a strong spatial variability from stand to regional scales (Davi et al. 2006). Many reports
have indicated the utilization of hyperspectral data for LAI estimation (Durbha, King, and
Younan 2007; Zhao et al. 2007; Darvishzadeh et al. 2008; Delalieux et al. 2008). Most of
these studies were carried out either at the laboratory or at the field level. A spaceborne
satellite sensor provides an easy and timely opportunity for regional and global evalua-
tion of canopy characteristics. Vegetation indices proposed from these observations will be
helpful to produce continuous products of terrestrial variables for global climate change
research (Wu et al. 2010a). Very few researchers have made an attempt to estimate dif-
ferent parameters for tropical forests using canopy level spectra (Kalacska et al. 2004;
Arroyo-Mora, Kalascska, and BenjaminArroyo-Mora 2008; Asner and Martin 2008). It is
important to have canopy level estimates of LAI specific to tropical systems.
Most of the canopy level studies have relied on regression analysis with hyperspectral
data or selecting narrow spectral bands for developing indices to estimate individual param-
eters of interest (Peñuelas et al. 1994; Curran, Kuoiec, and Smith 1997; Serrano et al.
2000; Ollinger et al. 2002; Smith et al. 2003). Vegetation indices are computationally fast
and require little expertise (Haboudane et al. 2008). Developing indices for tropical cover
is very important. Indices developed using hyperspectral remote-sensing data are useful
in monitoring tropical forest covers at a larger scale. An advanced hyperspectral remote-
sensing system covers the wavelength regions from 400 to 2500 nm at a nominal spectral
resolution of 10 nm. However, identification of appropriate wavelengths from hyperspectral
data is a difficult task. Different types of analytical techniques have been tested to select
appropriate wavelengths. Partial least square (PLS) regression analysis has been proved
to be a successful technique for accurate estimation of foliar biochemical properties from
canopy spectral data (Ollinger et al. 2002; Smith et al. 2003, 2002; Asner and Martin 2008;
Martin et al. 2008; Song et al. 2011). PLS regression uses the entire spectrum as a single
measurement rather than band-by-band analysis. The values generated by PLS calculation
relate the features of the spectra to the constituents analysed (Haaland and Thomas 1988).
Current human interventions are modifying tropical landscapes. In many regions,
important species are being planted under social forestry programmes. Teak (Tectona
grandis Linn.) and bamboo (Dendrocalamus strictus Nees.) are two important species
of tropical regions. In tropical countries like India, teak occurs across the region (Kaul,
Mohren, and Dadhwal 2010). It is a fine quality timber-yielding deciduous species, suitable
for rapid production of large volumes of timber, poles, and fuel wood (Kaul, Mohren,
1678 D. Vyas et al.

and Dadhwal 2010). Similarly, bamboo occurs in many types of forests in this continent.
Bamboo species are widely distributed in dry deciduous forests and grow rapidly in all cli-
matic conditions across India (Reddy 2006). The wider distribution and larger utility value
of teak and bamboo make it necessary to monitor them at larger spatial scales.
Considering the importance of these species, the present study was carried out to esti-
mate chlorophyll and LAI of teak and bamboo at canopy level using Hyperion (EO1) data.
PLS regression analysis was tested to look at its versatility in identifying spectral regions
sensitive to chlorophyll and LAI. It was also tested to identify precise bands to develop
indices.

2. Materials and methods


2.1. Study site
The study was carried out in the Shoolpaneshwar Wildlife Sanctuary (SWS) located at
Downloaded by [Sardar Patel University] at 03:49 14 March 2013

21◦ 29 –21◦ 52 N latitude and 73◦ 29 –73◦ 54 E longitude, in Narmada district, Gujarat,
India (Figure 1). SWS is an important protected area supporting sizeable biota. It occu-
pies an area of 675 km2 . The topography of the area is undulating with both continuous
and discontinuous hilly tracts up to an elevation of ∼800 m intermingled with valleys,
streams, and sporadic clearings for agriculture. Annual rainfall of the area is in the range of
900–1200 mm. Rainfall is restricted to the months of June–October. Minimum and max-
imum annual mean temperatures are 8◦ C and 42◦ C, respectively. Vegetation is of a dry
deciduous type. Large tracts of land (≈100 m × 100 m), covered with pure patches of teak
and bamboo, are spread across the sanctuary. These come from the plantation activity of
the forest department carried out over the past 50 years. Some of the patches come from
natural occurrence. The rest of the study area is occupied by other tropical tree species.
Parts of areas uncovered by tree species have ephemeral herbaceous vegetation as ground
cover. The tribal population is dependent on the produce of these plantations for firewood,
timber, and hutments. At times, these plantations are utilized for revenue generation.

2.2. Field data collection


Field data were collected in 2006 during the months of October and November. A total of
80 quadrats (40 in teak and 40 in bamboo) of 30 m × 30 m size were marked in differ-
ent parts of the study area (67.5 km2 ). Quadrat size coincided with the spatial resolution
of Hyperion data. Quadrats were marked in such a manner that they fell in 50 m × 50 m
windows covered by the same type of vegetation cover. GPS locations of all of the quadrats
were recorded within an error of ±4 m. All of the quadrats were purely occupied by either
teak or bamboo. Quadrats with a heterogeneous distribution were not included in the sam-
pling. The minimum area of occupancy (by teak or bamboo) in each quadrat was ≥60%.
The rest of the area was covered with ground cover (sparse to moderate). Phytosociological
data such as DBH (diameter at breast height), height, density, and spread of canopy cover
were recorded from each quadrat. Specific leaf area (SLA) (leaf area per unit leaf dry mass)
was calculated for both species. A total of 50 leaves (for each species) were picked up from
different individuals. The leaves were plucked from both sunlit and shaded sides of the
canopy. Leaf area and leaf dry weight were measured. SLA was calculated. Averages were
obtained from these values.

2.3. LAI measurements


The leaf area of both species was measured by the litter trap method (for details refer Vyas
et al. 2010). Ten plots were laid for each cover and in each plot 5–8 (1 m × 1 m) quadrats
International Journal of Remote Sensing 1679
Downloaded by [Sardar Patel University] at 03:49 14 March 2013

Figure 1. Map showing the location of the study area (21◦ 29 –21◦ 52 N latitude and 73◦ 29 –73◦
54 E longitude) and Hyperion image subset.

were laid. Quadrats were randomly laid on the forest floor in the areas purely covered by
teak or bamboo. More than 95% of the fallen litter comprised dried leaves. Leaves fallen in
these areas were picked up, oven-dried (at 70◦ C for 48 h), and their dry weights measured.
Extreme values (more than twice the mean and/or half of the mean) were discarded and
the data were pooled. Leaf litter dry mass was converted into leaf area by multiplying the
dry weights thus obtained with SLA of that species. The leaf area of all of the individuals
(of either teak or bamboo) present in a 30 m × 30 m quadrat was obtained. The ground
area of a species was calculated from the mean radius values of the canopy spread of each
individual. LAI for each quadrat was calculated by dividing the leaf area of the quadrat
with the ground area occupied by the species.

2.4. Chlorophyll measurements


Mature leaf samples from different sides of the canopy were collected (sunlit and shaded
areas) from each individual member of teak and bamboo species. A total of 5–8 individuals
1680 D. Vyas et al.

present in each quadrat were covered in this sampling. Foliar samples were washed in
running tap water (to remove dust) and 1 cm diameter samples were cut. These fresh sam-
ples were ground in dimethyl sulphoxide (DMSO). Extractants were centrifuged and the
precipitants were again extracted in DMSO. This process was repeated until the samples
became pale in colour. Supernatants from each individual member were pooled in a test
tube. Spectrophotometer readings were taken from the samples of all individuals to deter-
mine chlorophyll content (Lichtenthaler 1987). Mean chlorophyll values (g m−2 ) for each
species were calculated. They were extrapolated to get the chlorophyll content of a quadrat
(mean chlorophyll per unit area × total leaf area of the quadrat).

2.5. Hyperion data


Hyperion data were obtained on 21 October 2006. At the time of data acquisition, cloud
cover was less than 25%. The spatial resolution of the sensor was 30 m and the spectral
Downloaded by [Sardar Patel University] at 03:49 14 March 2013

resolution was 10 nm with a range of 427–2345 nm. Atmospheric correction was carried
out using ACORN 1.5 software (ImSpec LLC, Palmdale, CA, USA). Atmospheric cor-
rection was done on 196 bands after removing 46 non-calibrated and overlapping bands.
Zero-value bands were deleted and the final image contained 165 bands (for details refer to
Christian and Krishnayya 2009). The image was georegistered with the WGS-84 Geodetic
datum using ERDAS Imagine V.8.7 (Intergraph Corporation, Huntsville, AL, USA) (root
mean square error (RMSE) = 0.1 pixel). A subset was extracted from this image cover-
ing an area of 67.5 km2 . The subset area extracted coincided exactly with that covered in
the field survey. Subset extraction and image processing were performed using ENVI V.4.6
(Exelis Visual Information Solutions, Boulder, CO, USA) software.

2.6. PLS regression analysis


PLS regression analysis was carried out to determine the relative contribution of each foliar
property to the Hyperion reflectance spectra of teak and bamboo. PLS was found to be a
successful approach for the prediction of foliar properties from airborne or spaceborne
hyperspectral data (Smith et al. 2002, 2003; Asner and Martin 2008; Martin et al. 2008).
PLS regression analysis was performed using Unscrambler X software (CAMO AS, Oslo,
Norway). The optimum number of factors was estimated by RMSE generated by leave-one-
out cross-validation (LOO-CV). PLS regression was performed using the full Hyperion
reflectance spectra (427–2375 nm) coming from 165 bands. Subsequently, all of the wave-
lengths above 1507 nm were excluded from analysis because of their lesser sensitivity to
chlorophyll and LAI and higher sensitivity towards foliage moisture content, lignin, nitro-
gen, and cellulose (Curran 1989; Thenkabail et al. 2004; Martin et al. 2008). Weighted
coefficients generated by PLS regression for each wavelength (between 427 and 1507 nm)
showed their relative sensitivity towards chlorophyll and LAI. PLS was performed for
the entire range of chlorophyll concentrations found in the quadrats of teak and bamboo
independently. Estimated chlorophyll values were divided into low- and high-concentration
ranges. PLS was performed on these data sets to check the sensitivity of chlorophyll con-
centration in identifying specific wavelengths. PLS was performed on the entire range of
LAI found in teak and bamboo quadrats independently. This analysis did not give desir-
able coefficient values. This is because, unlike chlorophyll, LAI values of bamboo and teak
showed a skewed distribution. To make the distribution of LAI data normal, data values of
both the species were combined and PLS analysis was carried out. A data set of LAI was
not divided into low and high ranges of values. PLS analysis was also performed across
narrow ranges of spectral wavelengths (600–750 and 750–950 nm for chlorophyll and
International Journal of Remote Sensing 1681

1000–1140 and 1140–1507 nm for LAI) to find a suitable spectral range for each parame-
ter. The prediction error (divergence between predicted and measured parameters) for each
PLS model was measured in terms of RMSE values. For all PLS regression models, RMSE
values were generated by LOO-CV.
Highly sensitive wavelengths from the best performing PLS regression model for
chlorophyll and LAI were selected for vegetation index development. Simple and normal-
ized difference (ND) vegetation indices were developed. The LOO-CV technique was used
for validation of developed models. Vegetation indices that gave the best results were recog-
nized for the development of linear regression models for the quantification of chlorophyll
and LAI. Prediction errors of vegetation indices from the present study were compared with
prediction errors obtained from other studies.

3. Results and discussion


Downloaded by [Sardar Patel University] at 03:49 14 March 2013

3.1. Hyperion reflectance spectra and measured parameters


Maximum, minimum, and mean Hyperion reflectance spectra for teak and bamboo quadrats
with homogeneous occupancy are shown in Figure 2. The figure clearly shows variations
in the magnitude of reflectance at different spectral regions. Reflectance spectra in the vis-
ible region (400–700 nm) are associated with leaf pigmentation (Sims and Gamon 2002).
Spectra in this region correlated with chlorophyll values recorded in this study. The large
variation observed in reflectance in near-infrared (NIR) and shortwave-infrared (SWIR)
regions can be explained by large differences in LAI values. Thenkabail et al. (2004)

(a) 0.50 Maximum


Minimum
0.45
Mean
0.40
0.35
Reflectance

0.30
0.25
0.20
0.15
0.10
0.05
0.00
400 900 1400 1900 2400
Wavlength (nm)
(b) 0.40
Maximum
0.35 Minimum
Mean
0.30

0.25
Reflectance

0.20

0.15

0.10

0.05

0.00
400 900 1400 1900 2400
Wavelength (nm)

Figure 2. Mean, minimum, and maximum of Hyperion reflectance spectra for (a) teak and
(b) bamboo.
1682 D. Vyas et al.

Table 1. Measured parameters for teak and bamboo.

Number of Standard
Parameters observation deviation Minimum Maximum
−2
Teak total chlorophyll (g m ) 40 ± 0.57 0.77 2.64
Bamboo total chlorophyll (g m−2 ) 40 ± 0.27 1.22 2.38
LAI (teak + bamboo)* 46 ± 0.98 2.38 6.63

Note: * Results for LAI are from the combined data set of teak and bamboo.

suggested that there is a strong relationship between LAI and reflectance in NIR and
SWIR regions. The results of this study indicate the same. The measured chlorophyll values
ranged from 0.77 to 2.64 g m−2 for teak and from 1.22 to 2.38 g m−2 for bamboo, whereas
the measured LAI values (together for both species) ranged from 1.22 to 6.63 (Table 1).
The range of chlorophyll values tested in this study is similar to that published by others
Downloaded by [Sardar Patel University] at 03:49 14 March 2013

for temperate/tropical forests and grasslands (Asner and Martin 2008; Darvishzadeh et al.
2008; Le Maire et al. 2008; Wu et al. 2010a, 2010b). The range of LAI values recorded
here is comparable with other global level compilations (Soudani et al. 2006; Yang et al.
2006; Ganguly et al. 2008).

3.2. PLS regression analysis


Coefficients generated by PLS regression analysis for chlorophyll (427–1326 nm) and LAI
(427–1507 nm) are shown in Figures 3(a) and (b). Coefficients generated by the best
performing PLS regression model for prediction of chlorophyll and LAI are shown in
Figures 3(c) and (d). The optimum number of factors in PLS regression models is estimated
by RMSE generated by LOO-CV. The number of factors in the final PLS model was two
in LAI estimation and three in chlorophyll estimation. Earlier, Darvishzadeh et al. (2008)
used the same number of factors for LAI estimation and a higher number for chlorophyll
estimation using PLS regression analysis.

3.2.1. Prediction of chlorophyll by PLS regression models


LOO-CV prediction of chlorophyll of both the species is shown in Figures 4 and 5. PLS
regression performed on a narrow spectral range of 600–750 nm gave the best results
for the prediction of chlorophyll (R2 = 0.90, RMSE = 0.182 for teak and R2 = 0.84,
RMSE = 0.113 for bamboo), with minimum prediction errors (11.5% of mean for teak
and 7.15% of mean for bamboo) (Table 3a and Figures 4(b) and 5(b)). Earlier, Asner and
Martin (2008) predicted chlorophyll a and b of several tropical species with the help of
PLS regression analysis of airborne hyperspectral data (RMSE of 15% and 13% of mean,
respectively). The results of this study suggest that the 600–750 nm region contributes the
most for the prediction of chlorophyll. Sykioti et al. (2011) found high sensitivity of the
550–750 nm spectral region towards stand level chlorophyll content. Asner and Martin
(2008) found that chlorophyll contributes more in the spectral region of 510–730 nm.
Our results reiterate the importance of this spectral region for chlorophyll estimation of
two important tropical species. Our results also reaffirmed the importance of the red edge
region (680–750 nm) for estimation of chlorophyll. They are also comparable with the
findings of other studies (Gitelson and Merzlyak 1997, 2003; Blackburn and Ferwerda
2008; Huber et al. 2008). A PLS regression model for prediction of chlorophyll showed
a maximum negative coefficient value at 692 nm and a maximum positive coefficient
value at 743 nm for both teak and bamboo. Earlier, Cho, Skidmore, and Atzberger (2008)
Downloaded by [Sardar Patel University] at 03:49 14 March 2013

(a) 0.6
(b) 1.0
0.8
0.4 0.6

0.4
0.2
0.2
Teak 0.0 Combined data
0.0
Bamboo 400 600 800 1000 1200 1400 1600
400 500 600 700 800 900 1000 1100 1200 1300 1400 –0.2
–0.2 –0.4

PLS coefficient (weighted)

PLS coefficient (weighted)


–0.6
–0.4
–0.8

–0.6 –1.0
Wavelength (nm) Wavelength (nm)

(d) 1.0
(c)
1.0 0.8

0.6
0.8
0.4
0.6
0.2
0.4
0.0 Combined data
0.2 Teak 1000 1100 1200 1300 1400 1500 1600
International Journal of Remote Sensing

Bamboo –0.2
0.0
600 620 640 660 680 700 720 740 –0.4
PLS coefficient (weighted)

–0.2
–0.6

PLSR coefficient (weighted)


–0.4 –0.8
–0.6 –1.0
Wavelength (nm) Wavelength (nm)

Figure 3. PLS regression coefficients (weighted). (a) Chlorophyll 427–1326 nm, (b) LAI 427–1507 nm, (c) chlorophyll spectral subset 600–750 nm, and
(d) LAI spectral subset 1000–1507 nm.
1683
1684 D. Vyas et al.

(a) Teak chlorophyll (427–1326 nm) (b) Teak chlorophyll (600–750 nm)
3 3
R2 = 0.78, RMSE = 0.269 R2 = 0.90, RMSE = 0.182
2.5 2.5
Measured (g m–2)

Measured (g m–2)
2 2

1.5 1.5

1 1

0.5 0.5
0.5 1 1.5 2 2.5 3 0.5 1 1.5 2 2.5 3
Predicted (g m–2) Predicted (g m–2)
Downloaded by [Sardar Patel University] at 03:49 14 March 2013

Figure 4. Relationships between predicted and measured chlorophyll (LOO-CV) for teak using
Hyperion reflectance spectra in partial least square regression models. (a) Reflectance spectra
427–1326 nm and (b) spectral subset 600–750 nm.

(a) (b)
Bamboo chlorophyll (427–1326 nm) Bamboo chlorophyll (600–750 nm)
2.4 2.4

R2 = 0.814, RMSE = 0.118 R2 = 0.837, RMSE = 0.113


2.2 2.2
Measured (g m–2)

Measured (g m–2)

2 2

1.8 1.8

1.6 1.6

1.4 1.4

1.2 1.2
1.2 1.4 1.6 1.8 2 2.2 2.4 1.2 1.4 1.6 1.8 2 2.2 2.4
Predicted (g m–2) Predicted (g m–2)

Figure 5. Relationships between predicted and measured chlorophyll (LOO-CV) for bamboo using
Hyperion reflectance spectra in partial least square regression models. (a) Reflectance spectra
427–1326 nm and (b) spectral subset 600–750 nm.

suggested that wavebands at 680, 694, 724, and 760 nm have the potential for maximally
explaining variations in leaf chlorophyll content with minimal effects of leaf and canopy
biophysical confounders such as LAI. These two wavelengths (692 and 743 nm) were
selected for developing ratios for chlorophyll estimation (Table 2). A simple ratio (SR
743/692) gave the best results for the prediction of chlorophyll with LOO-CV (R2 = 0.73,
RMSE = 0.28 for teak and R2 = 0.71, RMSE = 0.15 for bamboo) (Figures 7(a) and (b)).
Wu et al. (2010a) evaluated Hyperion data for chlorophyll content estimation with a range
of vegetation indices. The lowest reported in their study was 30.53% of mean. RMSE
values of this study are lower (Table 4) for both species, indicating the higher accuracy of
the developed ratio in chlorophyll estimation. Predictive errors (for chlorophyll estimation)
obtained for the two tropical species are similar to those reported by Le Maire et al. (2008)
in temperate forest stands using Hyperion data.

3.2.2. Prediction of LAI by PLS regression models


Cross-validated prediction of LAI by all PLS regression models is shown in Figure 6.
PLS regression performed on the spectral region of 1000–1507 nm gave the best results
International Journal of Remote Sensing 1685

Table 2. Developed vegetation indices for measured parameters.

Index Data set Parameter

SR 743/692 Teak Chlorophyll (g m−2 )


SR 743/692 Bamboo Chlorophyll (g m−2 )
ND 743/692 Teak Chlorophyll (g m−2 )
ND 743/692 Bamboo Chlorophyll (g m−2 )
SR 1457/1084 (Teak + bamboo) LAI
ND 1457/1084 (Teak + bamboo) LAI

(a) (b)
LAI (427–1507 nm) LAI (1000–1507 nm)
7 7

R2 = 0.72, RMSE = 0.560 R2 = 0.87, RMSE = 0.425


6 6
Downloaded by [Sardar Patel University] at 03:49 14 March 2013

Measured LAI

Measured LAI
5 5

4 4

3 3

2 2
2 3 4 5 6 7 2 3 4 5 6 7
Predicted LAI Predicted LAI

Figure 6. Relationships between predicted and measured LAI (LOO-CV) for combined data using
Hyperion reflectance spectra in partial least square regression models. (a) Reflectance spectra
427–1507 nm and (b) spectral subset 1000–1507 nm.

Table 3. R2 values for calibration (R2c ) and cross-validation (R2cv ) methods and RMSECV for cross-
validation (for the best performing PLS models).

Plant name R2c R2cv RMSECV RMSECV % per mean value


(a) Chlorophyll
Teak 0.96 0.9 0.18 11.50
Bamboo 0.87 0.84 0.11 7.15
(b) LAI
Teak + bamboo* 0.9 0.87 0.43 10.11

Note: * Results for LAI are from the combined data set of teak and bamboo.

(R2 = 0.87, RMSE = 0.425) with a minimum prediction error (10.11% of the mean)
(Table 3b and Figure 6(b)). Further separation of the spectral subset into narrow spectral
regions and subsequent PLS analysis failed in making any more improvement of predic-
tion error for LAI. Earlier, Darvishzadeh et al. (2008) predicted LAI using PLS regression
analysis at canopy level using airborne spectra (RMSE = 11.59% of the mean). Our results
showed a lesser prediction error for LAI. The results suggested that the broad spectral
region of 1000–1507 nm is sensitive towards LAI. This confirmed the findings of previous
studies suggesting a strong relationship between reflectance values at NIR to SWIR bands
and LAI (Brown et al. 2000; Cohen and Goward 2004; Lee et al. 2004; Thenkabail et al.
2004; Schlerf, Atzberger, and Hill 2005). The best PLS regression model for prediction
1686 D. Vyas et al.

(a) SR 743/692 (teak) (b) SR 743/692 (bamboo)


3 2.6
R2 = 0.73, RMSE = 0.282 2.4 R2 = 0.71, RMSE = 0.148
2.5
Measured. Chl. (g m–2)

Measured. Chl. (g m–2)


2.2
2 2

1.5 1.8

1.6
1
1.4

0.5 1.2
0.5 1 1.5 2 2.5 3 1.2 1.4 1.6 1.8 2 2.2 2.4 2.6
Predicted Chl. (g m–2) Predicted Chl. (g m–2)
Downloaded by [Sardar Patel University] at 03:49 14 March 2013

(c) ND 1457/1084 (teak + bamboo)


7

6 R2 = 0.66, RMSE = 0.571


Measured LAI

2
2 3 4 5 6 7
Predicted LAI

Figure 7. Cross-validated prediction of chlorophyll and LAI by the leave-one-out method using the
best performing developed indices. (a) LOO-CV for the developed vegetation index 743/692 for
teak, (b) LOO-CV for the developed vegetation index 743/692 for bamboo, and (c) LOO-CV for the
developed vegetation index 1457/1084.

of LAI showed a maximum negative coefficient value at 1457 nm and a maximum posi-
tive coefficient value at 1084 nm. This coincides with the findings of Schlerf et al. (2004)
and Schlerf, Atzberger, and Hill (2005), mentioning the importance of 1088 nm for the
estimation of LAI. These two wavelengths (1084 and 1457 nm) were selected to develop
ratios for LAI estimation. Table 2 shows the developed vegetation indices for LAI. The
normalized difference ratio (ND 1457/1084) gave the best results for the prediction of LAI
with the LOO-CV method (R2 = 0.66, RMSE = 0.57) (Figure 7(c)). Wu et al. (2010a)
evaluated Hyperion data for LAI estimation using various vegetation indices. The lowest
RMSE reported in their study was 32.73% of mean. In this study, predicted LAI with the
help of vegetation indices resulted in a lower RMSE value of 13.57% of the mean (Table 4).
Prediction errors obtained in this study are comparable to those reported earlier by Wu et al.
(2010a) and Le Maire et al. (2008) (Table 4).

4. Conclusions
This study reaffirms the ability of Hyperion (EO1) data for estimation of chlorophyll
and LAI of two important species of tropical forests. PLS regression performed well
International Journal of Remote Sensing 1687

Table 4. Comparison of the best performing developed indices with other indices.

Remote- Lowest %
Developed sensing RMSE (per
indices Data set Parameter data mean value) References

SR 743/692 Teak Chlorophyll Spaceborne 16.57 Present study


SR 743/692 Bamboo Chlorophyll Spaceborne 9.49 Present study
ND 1457/1084 Teak + bamboo LAI Spaceborne 13.57 Present study
MCARI/OSAVI Temperate Chlorophyll Spaceborne 30.53 Wu et al. (2010a)
750,705* vegetation
stands
MCARI II Temperate LAI Spaceborne 32.73 Wu et al. (2010a)
750,705† vegetation
stands
ND 925/710 Temperate Chlorophyll Spaceborne 17.33 Le Mairy et al.
vegetation (2008)
Downloaded by [Sardar Patel University] at 03:49 14 March 2013

stands
Derivative index Temperate LAI Spaceborne 31.19 Le Mairy et al.
1725–970 vegetation (2008)
stands
SR 753/710 Pinus stands Chlorophyll Airborne 19.37 Zarco-Tejada
et al. (2004)
PVI 1088/1148‡ Temperate LAI Airborne 21.29 Schlerf et al.
vegetation (2005)
stands
ND 1141/1150 Heterogeneous Chlorophyll Airborne 40.12 Darvishzadeh
grassland et al. (2008)
SR 750/445 Temperate tree Chlorophyll Leaf level 4.28 Sims and Gamon
species (2002)

Notes: SR, simple ratio; ND, normalized difference.


* Modified chlorophyll absorption ratio index/optimized soil-adjusted vegetation index (Wu et al. 2010a).
†Modified chlorophyll absorption ratio index II (Wu et al. 2010a).
‡Perpendicular vegetation index.

in the accurate estimation of chlorophyll and LAI at stand level (r > 0.84) with lesser
predictive errors (7.15–11.50% per mean value). These values are highly appropriate as
they come from spaceborne data. SR 743/692 gave good results for chlorophyll and ND
1457/1084 worked well for LAI. Chlorophyll and LAI estimates developed here are sim-
ple markers for evaluating the healthy status of two important tropical vegetation covers.
Ratios developed here can be tested for teak and bamboo covers spread in tropical regions
with similar environmental conditions.

Acknowledgements
The authors thank the Space Applications Centre (Indian Space Research Organization), Ahmedabad,
for sponsoring the project. They are thankful to the reviewers for their valuable suggestions.

References
Arroyo-Mora, J., M. Kalascska, and L. Benjamin. 2008. “Spectral Expression of Gender: A Pilot
Study with Two Neotropical Tree Species.” In Hyperspectral Remote Sensing of Tropical and
Sub-Tropical Forest, edited by M. Kalascska, and G. Arturo Sanchez-Azofeifa, 125–35. Boca
Raton, FL: CRC Press Taylor & Francis Group.
Asner, G. P. 2008. “Hyperspectral Remote Sensing of Canopy Chemistry, Physiology, and
Biodiversity in Tropical Rain Forests.” In Hyperspectral Remote Sensing of Tropical and
1688 D. Vyas et al.

Sub-Tropical Forest, edited by M. Kalascska, and G. Arturo Sanchez-Azofeifa, 161–96. Boca


Raton, FL: CRC Press Taylor & Francis Group.
Asner, G. P., and R. E. Martin. 2008. “Spectral and Chemical Analysis of Tropical Forests: Scaling
from Leaf to Canopy Levels.” Remote Sensing of Environment 12: 3958–70.
Asner, G. P., J. M. O. Scurlock, and J. A. Hicke. 2003. “Global Synthesis of Leaf Area Index
Observations: Implications for Ecological and Remote Sensing Studies.” Global Ecology and
Biogeography 12: 191–205.
Bacour, C., F. Baret, D. Beal, M. Weiss, and K. Pavageau. 2006. “Neural Network Estimation of
LAI, fPAR, fCover and LAIxCAB from Top of Canopy MERIS Reflectance Data: Principles and
Validation.” Remote Sensing of Environment 105: 313–25.
Blackburn, G. A., and J. G. Ferwerda. 2008. “Retrieval of Chlorophyll Concentration from Leaf
Reflectance Spectra Using Wavelet Analysis.” Remote Sensing of Environment 112: 1614–32.
Brown, L., J. M. Chen, S. G. Leblanc, and J. Cihlar. 2000. “A Short Wave Infrared Modification to
the Simple Ratio for LAI Retrieval in Boreal Forests: An Image and Model Analysis.” Remote
Sensing of Environment 71: 16–25.
Chen, J. M., G. Pavlic, L. Brown, J. Cihlar, S. G. Leblanc, H. P. White, R. J. Hall, D. R. Peddle, D. J.
Downloaded by [Sardar Patel University] at 03:49 14 March 2013

King, and J. A. Trofymow. 2002. “Derivation and Validation of Canada-Wide Course-Resolution


Leaf Area Index Maps Using High-Resolution Satellite Imagery and Ground Measurements.”
Remote Sensing of Environment 80: 165–84.
Cho, M. A., A. K. Skidmore, and C. Atzberger. 2008. “Towards Red-Edge Positions Less Sensitive
to Canopy Biophysical Parameters for Leaf Chlorophyll Estimation Using Properties Optique
Spectrales Des Feuilles (PROSPECT) and Scattering by Arbitrarily Inclined Leaves (SAILH)
Simulated Data.” International Journal of Remote Sensing 29: 2241–55.
Christian, B., and N. S. R. Krishnayya. 2009. “Classification of Tropical Trees Growing in a Sanctuary
Using Hyperion (EO-1) and SAM Algorithm.” Current Science 96: 1062–606.
Cohen, W. B., and S. N. Goward. 2004. “Landsat’s Role in Ecological Application of Remote
Sensing.” BioScience 54: 535–45.
Curran, P. J. 1989. “Remote Sensing of Foliar Chemistry.” Remote Sensing of Environment 30:
271–78.
Curran, P. J., J. A. Kuoiec, and G. M. Smith. 1997. “Remote Sensing the Biochemical Composition
of a Slash Pine Canopy.” IEEE Transactions on Geoscience and Remote Sensing 35:415–20.
Darvishzadeh, R., A. Skidmorea, M. Schlerf, C. Atzberger, F. Corsia, and M. Choa. 2008. “LAI
and Chlorophyll Estimation for Heterogeneous Grassland Using Hyperspectral Measurements.”
ISPRS Journal of Photogrammetry and Remote Sensing 63: 409–26.
Davi, H., O. Bouriaud, E. Dufrene, K. Soudani, J. Y. Pontailler, and G. Le Maire. 2006. “Effect of
Aggregating Spatial Parameters on Modelling Forest Carbon and Water Fluxes.” Agricultural
and Forest Meteorology 139: 269–87.
Delalieux, S., B. Somers, S. Hereijgers, W. W. Verstraeten, W. Keulemans, and P. Coppin. 2008.
“A Near-Infrared Narrow-Waveband Ratio to Determine Leaf Area Index in Orchards.” Remote
Sensing of Environment 112: 3762–72.
Durbha, S. S., R. L. King, and N. H. Younan. 2007. “Support Vector Machines Regression for
Retrieval of Leaf Area Index from Multiangle Imaging Spectroradiometer.” Remote Sensing of
Environment 107: 348–61.
Ganguly, S., M. A. Schull, A. Samanta, N. V. Shabanov, M. Milesi, R. R. Nemani, Y. Knyazikhin,
and R. B. Myneni. 2008. “Generating Vegetation Leaf Area Index Earth System Data Record
from Multiple Sensors. Part 2: Implementation, Analysis and Validation.” Remote Sensing of
Environment 112: 4318–32.
Gitelson, A., and M. N. Merzlyak. 1997. “Remote Estimation of Chlorophyll Content in Higher Plant
Leaves.” International Journal of Remote Sensing 18: 2691–7.
Gitelson, A., and M. N. Merzlyak. 2003. “Relationships between Leaf Chlorophyll Content and
Spectral Reflectance and Algorithms for Non-Destructive Chlorophyll Assessment in Higher
Plant Leaves.” Journal of Plant Physiology 18: 271–82.
Granier, A., E. Ceshia, C. Damesin, E. Dufrene, D. Epron, and P. Gross. 2000. “The Carbon Balance
of a Young Beech Forest.” Functional Ecology 14: 312–25.
Haaland, D. M., and E. V. Thomas. 1988. “Partial Least-Squares Methods for Spectral Analyses.
1. Relation to Other Quantitative Calibration Methods and the Extraction of Qualitative
Information.” Analytical Chemistry 60: 1193–202.
International Journal of Remote Sensing 1689

Haboudane, D., N. Tremlay, J. R. Miller, and P. Vigneaault. 2008. “Remote Estimation of


Crop Chlorophyll Content Using Spectral Indices Derived from Hyperspectral Data.” IEEE
Transactions on Geosciences and Remote Sensing 46: 423–37.
Hansen, P. M., and J. K. Schjoerring. 2003. “Reflectance Measurements of Canopy Biomass and
Nitrogen Status in Wheat Crops Using Normalized Difference Vegetation Indices and Partial
Least Squares Regression.” Remote Sensing of Environment 86: 542–53.
Houborg, R., H. Soegaard, and E. Boegh. 2007. “Combining Vegetation Index and Model Inversion
Methods for the Extraction of Key Vegetation Biophysical Parameters Using Terra and Aqua
MODIS Reflectance Data.” Remote Sensing of Environment 106: 39–58.
Huber, S., M. Kneubuhler, A. Psomas, K. Ltten, and N. E. Zimmermann. 2008. “Estimating
Foliar Biochemistry from Hyperspectral Data in Mixed Forest Canopy.” Forest Ecology and
Management 256: 491–501.
Kalacska, M., G. A. Sanchez-Azofeifa, B. Rivard, J. C. Calvo-Alvarado, A. R. P. Journet, J. P. Arroyo-
Mora, and D. Ortiz-Ortiz. 2004. “Leaf Area Index Measurements in a Tropical Moist Forest: A
Case Study from Costa Rica.” Remote Sensing of Environment 91: 134–52.
Kaul, M., G. M. J. Mohren, and V. K. Dadhwal. 2010. “Carbon Storage and Sequestration Potential
Downloaded by [Sardar Patel University] at 03:49 14 March 2013

of Selected Tree Species in India.” Mitigation and Adaptation Strategies for Global Change 15:
489–510.
Lee, K. S., W. B. Cohen, R. E. Kennedy, T. K. Maiersperger, and S. T. Gowar. 2004. “Hyperspectral
Versus Multispectral Data for Estimating Leaf Area Index in Four Different Biomes.” Remote
Sensing of Environment 91: 508–20.
Le Maire, G., C. Francois, K. Saoudani, D. Berveiller, J. Pontailler, N. Breda, H. Genet, H. Davi,
and E. Dufrene. 2008. “Calibration and Validation of Hyperspectral Indices for the Estimation
of Broadleaved Forest Leaf Chlorophyll Content, Leaf Biomass per Area and Leaf Canopy
Biomass.” Remote Sensing of Environment 112: 3846–64.
Lichtenthaler, H. K. 1987. “Chlorophyll and Carotenoids: Pigments of Photosynthetic
Biomembranes.” Methods in Enzymology 148: 331–82.
Martin, M. E., L. C. Plourde, S. V. Ollinger, M. L. Smith, and B. E. McNeil. 2008. “A Generalizable
Method for Remote Sensing of Canopy Nitrogen across a Wide Range of Forest Ecosystems.”
Remote Sensing of Environment 112: 3511–19.
Ollinger, S. V., M. L. Smith, M. E. Martin, R. A. Hallett, C. L. Goodale, and J. D. Aber. 2002.
“Regional Variation in Foliar Chemistry and N Cycling among Forests of Diverse History and
Composition.” Ecology 83: 339–55.
Peñuelas J., F. Baret, and I. Filella. 1994. “Semi-Empirical Indices to Assess Carotenoids/
Chlorophyll-a Ratio from Leaf Spectral Reflectance.” Photosynthetica 31: 221–30.
Reddy, G. M. 2006. “Clonal Propagation of Bamboo (Dendrocalamus strictus).” Current Science 91:
1462–3.
Schlerf, M., C. Atzberger, and J. Hill. 2005. “Remote Sensing of Forest Biophysical Variables Using
HyMap Imaging Spectrometer Data.” Remote Sensing of Environment 95: 177–94.
Schlerf, M., C. Atzberger, M. Vohland, H. Buddenbaum, S. Seeling, and J. Hill. 2004. “Derivation
of Forest Leaf Area Index from Multi and Hyperspectral Remote Sensing Data.” EARSeL
eProceedings 3: 405–13.
Serrano, L., S. L. Ustin, D. A. Roberts, J. A. Gamon, and J. Penueles. 2000. “Deriving Water Content
of Chaparral Vegetation from AVIRIS Data.” Remote Sensing of Environment 74: 570–81.
Sims, D. A., and J. A. Gamon. 2002. “Relationship between Leaf Pigment Content and Spectral
Reflectance across a Wide Range of Species, Leaf Structures and Developmental Stages.” Remote
Sensing of Environment 81: 337–54.
Smith, M. L., M. E. Martin, L. Plourde, and S. V. Ollinger. 2003. “Analysis of Hyperspectral Data
for Estimation of Temperate Forest Canopy Nitrogen Concentration: Comparison between an
Airborne (AVIRIS) and a Space-Borne (Hyperion) Sensor.” IEEE Transactions on Geoscience
and Remote Sensing 41: 1332–7.
Smith, M. L., S. V. Ollinger, M. E. Martin, J. D. Aber, R. A. Hallett, and C. L. Goodale. 2002.
“Direct Estimation of Aboveground Forest Productivity through Hyperspectral Remote Sensing
of Canopy Nitrogen.” Ecological Applications 12: 1286–302.
Song, S., W. Gong, B. Zhu, and X. Huang. 2011. “Wavelength Selection and Spectral Discrimination
for Paddy Rice, with Laboratory Measurements of Hyperspectral Leaf Reflectance.” ISPRS
Journal of Photogrammetry and Remote Sensing 66: 672–82.
1690 D. Vyas et al.

Soudani, K., C. Francois, G. Maire, V. Dantec, and E. Dufrene. 2006. “Comparative Analysis of
IKONOS, SPOT, and ETM+ Data for Leaf Area Index Estimation in Temperate Coniferous and
Deciduous Forest Stands.” Remote Sensing of Environment 102: 161–75.
Sykioti, O., D. Paronis, S. Stagakis, and A. Kypariss. 2011. “Band Depth Analysis of CHRIS/PROBA
Data for the Study of a Mediterranean Natural Ecosystem. Correlation with Leaf Optical
Properties of Ecophysiological Parameters.” Remote Sensing of Environment 115: 752–66.
Thenkabail, P. S., E. A. Anclona, M. S. Ashton, and B. Van Der Meer. 2004. “Accuracy Assessments
of Hyperspectral Waveband Performance for Vegetation Analysis Applications.” Remote Sensing
of Environment 91: 354–76.
Ustin, S. L., D. A. Roberts, J. A. Gamon, G. P. Asner, and R. O. Green. 2004. “Using Imaging
Spectroscopy to Study Ecosystem Processes and Properties.” BioScience 54: 523–34.
Van Dijk, A., and L. A. Bruijnzeel. 2001. “Modeling Rainfall Interception by Vegetation of Variable
Density Using an Adopted Analytical Method: Part 1. Model description.” Journal of Hydrology
35: 79–93.
Vargas, L. A., M. N. Anderson, C. R. Jenson, and U. Jorgensen. 2002. “Estimation of Leaf Area Index,
Light Interception and Biomass Accumulation of Miscanthus sinensis ‘Goliath’ from Radiation
Downloaded by [Sardar Patel University] at 03:49 14 March 2013

Measurements.” Biomass and Bioenergy 22: 1–14.


Vyas, D., N. Mehta, J. Dinakaran, and N. S. R. Krishnayya. 2010. “Allometric Equations for
Estimating Leaf Area Index (LAI) of Two Important Tropical Species (Tectona grandis and
Dendrocalamus strictus).” Journal of Forestry Research 21: 197–200.
Wessman, C. A., J. D. Aber, D. L. Peterson, and J. M. Melillo. 1988. “Remote Sensing of Canopy
Chemistry and Nitrogen Cycling in Temperate Forest Ecosystems.” Nature 335: 154–6.
Wu, C., X. Han, Z. Niu, and J. Dong. 2010a. “An Evaluation of EO-1 Hyperspectral Hyperion Data for
Chlorophyll Content and Leaf Area Index Estimation.” International Journal of Remote Sensing
31: 1079–86.
Wu, C., L. Wang, Z. Niu, S. Gaob, and M. Wu. 2010b. “Nondestructive Estimation of Canopy
Chlorophyll Content Using Hyperion and Landsat/TM Images.” International Journal of Remote
Sensing 31: 2159–67.
Yang, W., B. Tan, D. Huang, M. Rautiainen, N. V. Shabanov, Y. Wang, J. L. Privette, K. F. Huemmrich,
R. Fensholt, I. Sandholt, M. Weiss, D. E. Ahl, S. T. Gower, R. R. Nemani, Y. Knyazikhin, and R. B.
Myneni. 2006. “MODIS Leaf Area Index Products: From Validation to Algorithm Improvement.”
IEEE Transactions on Geosciences and Remote Sensing 44: 1885–98.
Zarco-Tejada, P. J., J. R. Miller, A. Morales, A. Berjon, and J. Agura. 2004. “Hyperspectral Indices
and Model Simulation for Chlorophyll Estimation in Open-Canopy Tree Crops.” Remote Sensing
of Environment 90: 463–76.
Zhao, D., L. Huang, J. Li, and J. Qi. 2007. “A Comparative Analysis of Broadband and Narrowband
Derived Vegetation Indices in Predicting LAI and CCD of a Cotton Canopy.” ISPRS Journal of
Photogrammetry and Remote Sensing 62: 25–33.
Zhao, D., R. Reddy, V. G. Kakani, J. J. Read, and S. Koti. 2005. “Selection of Optimum Ratios for
Estimating Leaf Nitrogen and Chlorophyll Concentration of Field Grown Cotton.” Agronomy
Journal 97: 89–98.

You might also like