Botanical Studies (2007) 48: 71-77.
*
Corresponding author: E-mail: cct@gisfore.npust.edu.tw;
Tel: +886-87740301; Fax: +886-87740134.
INTRODUCTION
In recent decades, the development of remote sensing
methods to measure leaf chlorophyll content and surface
spectral reflectance has received much attention, since
variations in leaf chlorophyll content can provide
information concerning the physiological state of a leaf or
plant. Several vegetation indices estimated from remote
sensing data have been considered for assessing the status
of leaf chlorophyll content, plant biomass, production,
and vegetation health status. Destructive methods of leaf
chlorophyll content quantification include traditional
methods using extraction and spectrophotometric or HPLC
measurement, but they are considered time consuming and
expensive. In contrast, spectral reflectance measurements
are nondestructive, rapid, and can be applied across spatial
scales (Gamon and Qiu, 1999). Many theoretical models
have been developed for predicting leaf reflectance from
leaf chlorophyll, plant water content, and vegetation
structure variables (Dawson et al., 1998; Jacquemoud
et al., 1996). Without such extrapolation procedures, it
would be impossible to make landscale and ecosystem
assessments from leaf level analysis. Most large scale
research projects are now using remotely sensed data to
estimate the condition of ecosystems. However, most
relationships between leaf reflectance and chlorophyll
contents have been derived empirically derived.
Sims and Gamon (2002) have reported that spectral
indices provide relatively poor correlations with leaf
chlorophyll content when applied across a wide range
of species and plant functional types. They, therefore,
modified some vegetation indices to demonstrate the
application of spectral indices on species with widely
varying leaf structure. This strategy was applied in this
study and served as an impetus to further analyze leaf
chlorophyll content and surface spectral reflectance in
a wider range of vegetation, using modify vegetation
indices mNDVI (modify normalized difference) and
mSR (modify simple ratio) as test parameters. Normal
Leaf chlorophyll content and surface spectral reflectance
of tree species along a terrain gradient in Taiwan°¶s
Kenting National Park
Jan-Chang CHEN
1
, Chi-Ming YANG
2
,
Shou-Tsung Wu
3
, Yuh-Lurng CHUNG
4
, Albert Linton
CHARLES
1
,
and Chaur-Tzuhn CHEN
4,
*
1
Department of Tropical Agriculture and International Cooperation, National Pingtung University of Science and
Technology, Neipu, Pingtung, Taiwan
2
Research Center for Biodiversity, Academia Sinica, Taipei, Taiwan
3
Department of Tourism Management, Shih Chien University, Kaohsiung, Taiwan
4
Department of Forestry, National Pingtung University of Science and Technology, Neipu, Pingtung, Taiwan
(Received October 21, 2005; Accepted June 27, 2006)
ABSTRACT.
This study was conducted to investigate variations of leaf chlorophyll content and surface
spectral reflectance of different tree species across contrasting terrain in the Nanjenshan Reserve of Kenting
National Park, southern Taiwan. Tree species composition and forest types vary because of intense northeast
monsoons that frequent this area. In this study, we used several remote sensing technique indices°Xnormalized
difference vegetation index (NDVI), modified normalized difference vegetation index (mNDVI), simple ratio
(SR), and modified simple ratio (mSR)°Xto analyze the spectral reflectance data collected from portable
spectroradiometers and the GER 1500 and CM1000 chlorophyll meters to estimate leaf chlorophyll content.
The results showed that significant differences (P<0.01) arose only among the modified indices mSR 705 nm
and mNDVI 705 nm. The index mNDVI705 seemed more sensitive to detecting chlorophyll content in a wide
range of tree species across a terrain. Among the indices tested, the mNDVI consistently deviated from the
general relationship between chlorophyll content and spectral reflection in different vegetation. The findings
indicated that the modified indices were better at studying different tree species than normalized indices
across terrain.
Keywords: Leaf chlorophyll; mNDVI; Portable spectroradiometers; Spectral reflectance; Vegetation index.
PHYSIOLOGY
pg_0002
72
Botanical Studies, Vol. 48, 2007
vegetation indices are normally applied when investigating
vegetative differences among similar species, whereas
modified indices are specially designed to investigate
different species (Sims and Gamon, 2002).
Theoretically, terrain difference affects the growth of
vegetation and determines the chlorophyll content and
surface spectral reflectance. Our objective in this study
was to demonstrate how different leaf chlorophyll content
and surface spectral reflectance characteristics are affected
by terrain divergence in Kenting National Park, Taiwan.
MATERIALS AND METHODS
Study area
The experiment was
conducted on site at the
Nanjenshan Reserve of Kenting National Park, southern
Taiwan (Figure 1). In this area, forest vegetation
distribution is influenced by three major topographic
positions or terrains; windward slopes, valley and leeward
slopes (Hsieh and Hsieh, 1990). In the study area, the
winter precipitation and the intensity of northeast winds
Figure 1. Location of the study sites of Nanjenshan Reserve of Kenting National Park, southern Taiwan.
pg_0003
CHEN et al. °X Leaf chlorophyll and surface spectral reflectance
73
were found to correlate with the differentiation of forest
types. Those in the northeast district are evergreen
since the wind is rather, whereas, the thorny scrubs and
deciduous scrubs appear in the southwest district due to
severe dry winds. The high species diversity of Kenting
National Park is largely attributable to the heterogeneous
environment (Su and Su, 1988).
Leaf sampling
Measurement of leaf spectral reflectance was obtained
by randomly harvesting leaves from 20 tree species from
the three types of terrain. From each terrain type, one
sampling site from the leeward area and two sampling
sites from the windward and valley areas, respectively,
were selected. Fifty sample leaves taken from three to
nine dominant species were randomly collected from
each sampling site. Sampling was done at the top of the
canopies of each species and was carried out in late April
2005. Leaf samples were stored in plastic bags and kept
cool for further analysis. The leaf samples collected are
listed in Table 1.
Reflectance measurements
All spectral measurements were made with a field
portable spectrometer (GER-1500, SVC, Poughkeepsie,
NY, USA). GER-1500 has a nominal spectral range from
350 to 1050 nm with approximately 1.5 nm nominal
bandwidth. Leaf reflectance was measured with a
bifurcated fiber optic cable and the spectrometer collected
both leaf reflect once and reference reflect once data. Leaf
samples were illuminated by sunlight. All measurements
were carried out in triplicate and all tests in this study
were scaled to leaf level.
Quantification of chlorophyll
Leaf chlorophyll content was estimated by using a
portable chlorophyll meter (CM-1000, Spectrum Tech.,
Plainfield, Ill. USA). CM-1000 readings were made in
situ on the plants designated for harvesting, midway
along the youngest fully expanded leaf. The Minolta
SPAD 502 chlorophyll meter has become recognized as
a reliable substitute for total chlorophyll (Ommen et al.,
1999; Daughtry et al., 2000; Bauerle et al., 2004; Van den
Berg and Perkins, 2004; Netto et al., 2005). However, the
CM1000 chlorophyll meter was utilized and investigated
as a potential substitute for the SPAD meter in applications
with small leaves and where measurements were difficult
to make (i.e., turf). Four tree species (Michelia compressa;
Psychotria rubra; Daphniphyllum glaucescens; Gordonia
axillaris) were selected as test examples of the relationship
between the CM1000 chlorophyll meter and chlorophyll
content measurements, and the method for monitoring
chlorophyll content followed that described in Yang et
al. (1998). As the test results indicated (Figure 2), the
relationship between the CM1000 chlorophyll meter
and chlorophyll content is strong, in that the larger the
indices of the chlorophyll meter are, the higher is the total
chlorophyll content. Therefore, the CM1000 was selected
as the chlorophyll meter for use in this study.
Vegetation indices
The normalized difference vegetation index (NDVI)
(Eqn. 1) (Tucker, 1979) and simple ratio (SR) (Eqn. 2)
were used to calculate the vegetative indices obtained
from spectral reflectance measurements:
(1)
NDVI
705
=
R
750
°VR
705
R
750
+R
705
(2)
SR
705
=
R
750
R
705
where the wavelengths for NDVI and SR were 705 and
750 nm, respectively, and are based on the chlorophyll
index developed by Gitelson and Merzlyak (1994). R
705
and R
750
are the leaf sample spectral reflectance from
the GER-1500. Based on the results of Sims and Gamon
(2002), SR and NDVI indices may be affected by different
species leaf surfaces. They modified these two indices to
compensate for high leaf surface reflectance, which tends
to increase reflectance across the whole visible spectrum
of a wide range of species. Adding a constant to all
reflectance values reduces both SR and NDVI even when
there is no change in the absorptance of tissues below
Table 1. List of all the species used in this study
a
.
Species
Terrain
Aucuba chinensis Benth.
L
Bischofia javanica
L
Castanopsis carlesii
L
Champereia manillana
L
Cleyera japonica Thumb. var. japonica
L
Daphniphyllum glaucescens Blume subsp. var.
oldhamii (Hemsl.) Huang
L
Dendrocnide meyeniana
L
Gordonia axillaris
V
Ilex cochinchinensis
V
Illicium dunnianum
V
Machilus kusanoi
V
Michelia formosana
W
Neolitsea buisanensis
W
Neolitsea hiiranensis
W
Reevesia formosana Sprague
W
Schefflera octophylla
W
Syzygium euphlebium
W
a
The leaf samples measured for each species of the different
terrains (L= leeward; V= valley; W = windward).
pg_0004
74
Botanical Studies, Vol. 48, 2007
All statistical analyses were conducted using the
STATISTICA statistical software (Version 6.1, StatSoft
Inc. Tulsa, Oklahoma, USA, 2002). Coefficients of
determination (R
2
) were calculated for relationships
between various chlorophyll content values from
CM-1000 (independent variables) and mNDVI and mSR
values from GER-1500 (dependent variable). To test and
verify the relationship of chlorophyll content between
mNDVI and mSR, regression analyses were used in the
first data analysis step. For statistical reasons, we tried to
find out the best index to indicate the differences between
these four indices. The coefficient of variation (CV) was
chosen. To analyze the incidence of the two modified
indices, we used one way ANOVAs for each index,
comparing mNDVI and mSR indices with the CM-1000
index and different terrains separately.
RESULTS AND DISCUSSION
Vegetation indices and chlorophyll content
Results for the regression analysis showed that all of
the vegetation indices of this study positively correlated
with the CM1000 index. The coefficients of determination
(R
2
) calculated for relationships between this index
(independent variable) and the NDVI, SR, mNDVI and
mSR indices (dependent variables) for all samples (n =
117) showed that mSR (R
2
= 0.51) was slightly superior to
SR (R
2
= 0.34) in its correlation with the CM1000 index
the epidermis. They chose R445 as a measure of surface
reflectance and indicated that R
445
is a good reference for
all but the lowest chlorophyll content leaves. Besides, Le
Maire et al. (2004) also pointed out that the two modified
indices gave the best performance of universal broad leaf
chlorophyll indices. The modified indices of NDVI and
SR are as follows (Eqs. 3 and 4):
(3)
mND
705
=
R
750
°VR
705
R
750
+R
705
°V2R
445
(4)
mSR
705
=
R
750
°VR
445
R
750
°VR
445
Data analysis
The results of a Kolmogorov-Smirnov test (K-S test)
for normality of the CM1000 index were matched with the
normal distribution theory (Lilliefors, 1967). The indices
collected from the CM1000 meter, when fitted to the
normal distribution, all appeared to have the same index
value of 173 for the mean, mode and median statistical
parameters. Based on the above, we reclassified all the
samples (N=1136) into three groups (low, middle and
high) depending on the value of CM1000 indices. The
three groups were classified by the percentage of the total
samples 10%, 80% and 10%.
Figure 2. The relationship between leaf CM1000 meter readings and acetone extractable chlorophyll values measured from Michelia
compressa, Psychotria rubra, Gordonia axillaris and Daphniphyllum glaucescens.
pg_0005
CHEN et al. °X Leaf chlorophyll and surface spectral reflectance
75
(Figure 3), and mNDVI (R
2
= 0.59) was stronger than
NDVI (R
2
= 0.45).
Based on the analysis, our findings
were in agreement with Sims and Gamon (2002), in that
the modified indices mSR
705
and mNDVI
705
correlated
better with chlorophyll content. The results of CV showed
that the mNDVI had lowest CV (27%) of the four indices
(Table 2).
Terrain effect
Four indices in this study were computed for the
leaf samples in spectral reflectance measurement in the
three terrains. Each vegetation index (SR
705
, NDVI
705
,
mSR
705
and mNDVI
705
) was used in an analysis of
variance (ANOVA) to determine significant effects of
the three terrains (windward, valley, and leeward). The
results showed significant differences (P<0.01) among
the modified indices mSR
705
and mNDVI
705
, implying
the advantage of using modified vegetation indices over
normal indices on different terrains (Table 3).
The overall results showed that the mNDVI
705
was
successfully applied in this study. The confidence intervals
for leeward, windward, and valley were 0.084, 0.095,
and 0.105 respectively. These results may help us to
understand the variation of mND
705
on different terrains.
Based on the discrepancies of the confidence intervals
in the three terrains, the northeast wind was affected more
than the other environmental factors in the study area.
The degree to which terrain influenced forest composition
Figure 3. The normalized difference index and SR index with index wavelengths of 705 nm (a, SR
705
and b, NDVI
705
) and the
modified indices (c, mSR
705
and d, mNDVI
705
) as a function of the leaf chlorophyll content. As expected from the index analyses, the (c)
mSR index and (d) mNDVI were largely sensitive to chlorophyll content.
Table 2. The coefficient of variation of four vegetation indices
a
.
CM1000 D1
CM1000 D2
CM1000 D3
Mean
Index
Mean °” S.D. CV %
Mean °” S.D. CV %
Mean °” S.D. CV %
CV %
SR
705
1.244 °” 0.278 22.3 2.189 °” 0.700 32.0 1.781 °” 0.571 32.1
28.8
mSR
705
0.099 °” 0.082 82.2 0.346 °” 0.129 37.2 0.255 °” 0.131 51.2
56.9
NDVI
705
2.317 °” 1.284 55.4 5.176 °” 1.038 20.1 4.179 °” 1.293 30.9
35.5
mNDVI
705
0.335 °” 0.179 53.5 0.666 °” 0.062 9.4
0.59 °” 0.101 17.1
26.7
a
N=117. Data shows mean °” S.D.
pg_0006
76
Botanical Studies, Vol. 48, 2007
was quite clear. A marked difference in forest composition
existed among species in the windward and leeward
forests. In the understory of the windward forest, saplings
and small trees are more abundant than in the leeward
forest (Hsieh et al., 2000). In the study area, winter
precipitation and the intensity of the Northeast wind
positively correlated with differentiation of forest types
(Su and Su, 1988). Monsoon rainforests receive sufficient
rains brought about by the Northeast wind during the
winter, and this supports evergreen trees. Large quantities
of deciduous trees occur in the leeward southwest district
with semi-deciduous forest physiognomy (Su and Su,
1988). Chen (1999) used NDVI to detect seasonal
variation in vegetation greenness with four SPOT XS
images (Systeme Pour l°¶Observation de la Terre) in
the same study area and reported that leeward (0.438)
and windward (0.441) habitat had the highest degree of
vegetation greenness from June to September. However,
the NDVI between windward (0.241) and leeward (0.352)
from December to March was significantly different
(P>0.01). The study concluded that the northeastern
monsoon would decrease vegetation greenness. Similarly,
our analysis indicated that the northeast monsoon did
certainly affect the vegetation physiology in the study
area, and by using spectral characteristics detected the
differences of terrain variance and also spectral reflection
between vegetation species is possible.
CONCLUSION
We found a positive correlation between terrain and leaf
chlorophyll content, indicating that modified vegetation
indices, such as mNDVI, can help us to understand the
relationship between a native vegetation cover and its
terrain habitat. Among the indices tested, only the mNDVI
consistently deviated from the general relationships
between chlorophyll content and spectral reflection in
different vegetation. The ecological significance of this
study is based on the speed and efficiency by which
modified vegetation indices can detect chlorophyll content
of a complex forest vegetation. In this study, we have
shown that among the indices tested, mNDVI detects the
best in different terrain vegetation reflection. Nonetheless,
further investigative work remains to be carried out,
especially in leaf chlorophyll contents and remote sensing
applications.
LITERATURE CITED
Bauerle, W.L., D.J. Weston, J.D. Bowden, J.B. Dudley, and J.E.
Toler. 2004. Leaf absorptance of photosynthetically active
radiation in relation to chlorophyll meter estimates among
woody plant species. Sci. Hort. 101: 169-178.
Chen, C.T. 1999. Seasonal variation of vegetation greenness of
Nanjenshan forest ecosystem. Quarterly J. Chinese Forest
32: 53-66.
Daughtry, C.S .T., C.L. Walthall, M.S. Kim, E.B. de Colstoun,
a nd J .E. McMurt rey I II. 2000. Es ti ma tin g cor n le af
chlorophyll concentration from leaf and canopy reflectance.
Remote Sens. Environ. 74: 229-239.
Dawson, T.P., P.J. Curran, and S.E. Plummer. 1998. LIBERTY -
modeling the effects of leaf biochemical concentration on
reflectance spectra. Remote Sens. Environ. 65: 50-60.
Gam on, J.A. a nd H. Qiu. 1999. Ecological applications of
remote sensing at multiple scales. In F.I. Pugnaire, and F.
Valladares (eds.), Handbook of Functional Plant Ecology.
Marcel Dekker, New York, pp. 805-846.
Gitelson, A.A. and M.N. Merzlyak. 1994. Spectral reflectance
changes associate with autum n s enes cence of Aesculus
hippocastanum L. and Acer platanoides L. leaves. Spectral
features and relation to chlorophyll estim ation. J. Plant
Physiol. 143: 286-292.
Hsieh, T.H. and C.F. Hsieh. 1990. Woody Floristic Composition
a nd Di stri bution P att ern in the S ubtropi cal F ores t of
Nanjenshan Area. Ann. Taiwan Mus., pp. 121-146.
Table 3. The variance of analysis of four vegetation indices between three terrains.
Index
Terrain
Mean
sd
-95%
95%
Range
F
P
SR
705
Leeward
1.781 0.078
1.628
1.935
0.307
Valley
1.669 0.097
1.478
1.860
0.382
0.412 0.663
Windward 1.740 0.088
1.567
1.913
0.346
NDVI
705
Leeward
0.245 0.018
0.209
0.280
0.071
Valley
0.231 0.022
0.187
0.275
0.088
0.395 0.675
Windward 0.221 0.020
0.181
0.261
0.08
mSR
705
Leeward
4.346
a
0.191
3.968
4.723
0.755
Valley
3.311
bc
0.238
2.842
3.780
0.938
5.907 0.003**
Windward 3.790
ac
0.216
3.364
4.217
0.853
mNDVI
705
Leeward
0.587
a
0.021
0.545
0.629
0.084
Valley
0.477
b
0.026
0.425
0.530
0.105
6.304 0.002**
Windward 0.502
b
0.024
0.454
0.549
0.095
**Indicates significance different among the indices, P<0.01.
pg_0007
CHEN et al. °X Leaf chlorophyll and surface spectral reflectance
77
Hsieh, C.F., I.F. Sun, and C.C. Yang. 2000. Species composition
and veget ation pat tern of a lawla nd ra in forest at the
Nanjenshan LTER Site, Southern Taiwan. Taiwania 45(1):
107-119.
Jacquemoud, S., S.L. Ustin, J. Verdebout, G. Schmuck,
G. Andreoli , and B. Hos good. 1996. E st ima tin g lea f
biochemistry using the PROSPECT leaf optical properties
model. Remote Sens. Environ. 56: 194-202.
Lilliefors, H.W. 1967. On the Kolmorogrov-Smirnov test for
normality with mean and variance unknown. J. Am. Statist.
Assoc. 62: 399-402.
Le Maire, G., C. F rancois, and E. Dufrene. 2004. Towards
universal broad leaf chlorophyll indices using PROSPECT
s im u la t e d d at a ba s e a nd hy pe rs p e ct ra l re f le c ta n ce
measurements. Remote. Sens. Environ. 89: 1-28.
Netto, A.T., E. Campostrini, J.G. Oliveira, and R.E. Bressan-
S m i th . 20 05 . P ho t os y nt h et i c p ig m e nt s , ni t ro ge n ,
chlorophyll a fluores cence and S PAD-502 readings in
coffee leaves. Sci. Hort. 104: 199-209.
Ommen, O. E., A. Donnelly, S. Vanhoutvin, M. van Oijen, and R.
Manderscheid. 1999. Chlorophyll content of spring wheat
flag leaves grown under elevated CO
2
concentrations and
other environmental stresses within the °•ESPACE-wheat°¶
project. Europ. J. Agron. 10: 197-203.
Sims, D.A. and J.A. Gamon. 2002. Relationships between leaf
pigment content and spectral reflectance acros s a wide
range of species, leaf structures and developmental stages.
Remote Sens. Environ. 81: 337-354.
Su, H.J. and C.Y. Su. 1988. Multivariate analysis on the
vegetation of Kentin National Park. Quart. J. Chinese Forest
21: 17-32.
Tucker, C.J. 1979. Red and photographic infrared linear
combinations for monitoring vegetation. Remote Sens.
Environ. 8: 127-150.
Van den Berg, A.K. and T.D. P erkins. 2004. Evaluation of a
portable chlorophyll m eter to estimate chlorophyll and
nitrogen contents in sugar maple (Acer saccharum Marsh.)
leaves. Forest Ecol. Manag. 200: 113-117.
Yang, C.M., K.W. Chang, M.H. Yin, and H.M. Huang. 1998.
Methods for the determination of the chlorophylls and their
derivatives. Taiwania 43(2): 116-122.
Yang, C.M., J.C. Chen, L.L. Peng, J.S. Yang, and C.H. Chou.
20 02. Chi -C hi Ea rt hqu ak e-c a us ed L an ds li de : gre y
prediction model for pioneer vegetation recovery monitored
by satellite images. Bot. Bull. Acad. Sin. 43: 69-75.
pg_0008