stimate
on and
s than
not be
forest
e LAI
pet al
onship
ndition
ere the
> high.
remote
j were
1999;
dense
ite and
x the
fective
e very
along
se the
International Archives of the Photogrammetry, Remote Sensing and Spatial Information Sciences, Vol XXX V, Part B7. Istanbul 2004
relationship between the reflectance spectra and LAI in close
canopy situation. For this purpose, a laboratory experiment was
conducted to measure the reflectance spectra of the vegetation
sample that simulates various level of LAI using a portable
spectro-radiometer.
Vegetation samples of different LAI values were prepared by
stacking multiple-layers of evergreen broad leaves (Euonymus
Japonicus Thunb). As seen in Figure 1, the multiple-layers of
leaves fill the field of view (FOV) of the spectro-radiometer.
Since each leaf has approximately the same size, the total area
of leaves and the LAI can be easily calculated. Total of 15
samples were simulated and LAI value ranges from 1 to 6 with
an interval of approximately 0.25. Every sample was fully
covered by these leaves to simulate the close canopy situation
and there was no influence from the background soil.
Figure 1. A vegetation sample to simulate known LAI within
the FOV (black line) of the spectro-radiometer.
Reflectance spectra were measured using a portable spectro-
radiometer (GER 2600), which can measure spectral
reflectance over the wavelength region between 350nm and
2,500nm. Spectral reflectance were measured at 140cm height
with a 10 degree FOV lens. The actual size of the FOV for the
spectro-radiometer did not exactly correspond to simple
trigonometry calculation and it looks an ellipse shape with
diameters of 24cm and 18cm.
At each measurement, the spectro-radiometer actually provides
percent reflectance value for each of 612 continuous bands over
the wavelength from 350nm and 2,500nm. The simplest
statistical investigation was calculation of a band-by-band
correlation between spectral reflectance value and LAI of the
vegetation sample. Because each wavelength band represented
a different combination of spectral strengths and weaknesses,
discrepancies in correlation at particular wavelengths might
provide us particular spectral qualities for estimating LAL
We also compared sample LAI values with normalized
difference vegetation index (NDVI). Since spectral
measurements by the spectro-radiometer give us many adjacent
bands within the spectrum of red and near infrared wavelengths,
several combinations of NDVI calculation are possible.
However, as Teillet ct al. (1997) pointed out, NDVI is not very
sensitive to the location of any particular wavelength within the
red and near-infrared spectra. Two spectral reflectance
401
measurements at 655nm and 846nm were used to calculate
NDVI.
ETM+ Reflectance and Field-Measured LAI
From the laboratory experiment to compare reflectance spectra
and LAI in close canopy situation, further analysis was
conducted using actual multispectral image and field-measured
LAL The study area selected was a relatively small watershed
covering an area of approximately 500 km? of mixed
coniferous and deciduous forests in central part of the Korean
Peninsula. The temperate mixed forest has diverse group of
species composition and stand ages between 20 to 50 years old
and the canopy closure is over 80%. One third of the forest
lands are plantation pine stands (Pinus koraiensis, Pinus rigida,
and Larix leptolepis) and the remainin g two third of forests are
natural stands of mixed deciduous species.
During the growing season of 2003, 30 ground sample plots
were selected and species, LAI, stand density, and stand height
were measured (Figure 2). Each plot has an area of 20 x 20 m?
and includes five subplots for LAI measurement within it. All
subplot measurements were averaged to provide a single value
for the LAI at each plot. Plot locations were determined using a
differential global positioning system (GPS). LAI values were
measured using an optical device (Li-Cor LAI 2000) at 30
ground plots. To minimize any discrepancies due to the
phenological variation of leaf development, the field
measurements were conducted as close to the date of satellite
data acquisition. Although the May 8" of satellite data
acquisition is slightly earlier than the field measurement (late
June to early July), we believe that it did not cause any serious
problem since the leaf development in 2003 started very early
and the canopy condition between May and June was not much
different.
Figure 2. Distribution of 30 forest stands of LAI measurements
within the study area of the Kyongan Watershed.
For the study, we obtained Landsat-7 ETM+ data acquired on
May 8, 2003. ETM+ images were geo-referenced,
