The International Archives of the Photogrammetry, Remote Sensing and Spatial Information Sciences. Vol. XXXVII. Part B7. Beijing 2008
Correlation analysis between chlorophyll content and
spectral reflectance
Towards exploring the bands range which could have high
significant correlation with chlorophyll content, correlation
analysis was carried out in 400-900nm spectral region. Linear
and nonlinear correlation analysis was performed respectively
to locate the optimal spectral regions (Figure 1 and Figure 2).
Figure 1. Variation with wavelength (400-900nm) in linear
correlation coefficient (R) between reflectance and chlorophyll
content
0.8
0.7
0.6
0. 5
R 2 0.4
0.3
0.2
0. 1
0
400 500 600 700 800 900
wavelength (nra)
Figure 2. Variation with wavelength in the coefficient of
determination (R 2 ) obtained when regressing reflectance against
chlorophyll content, using an exponential model
The reflectance in 400-500nm region showed the lowest
correlation both linear and nonlinear with chlorophyll content. It
has been previously shown (e.g., Gitelson et al., 1996a) that the
red wavelength reflectance near 670nm is only sensitive to
chlorophyll a levels below 0.003mg/cm 2 . Such low levels of
chlorophyll can only be found in highly stressed or senescing
leaves, such as yellow autumn leaves of deciduous species (Datt,
1998). Though the nonlinear correlation between reflectance
near 670-680nm and chlorophyll content is higher, due to the
strong absorption by chlorophyll, the reflectance at 680nm
remained nearly constant with increasing chlorophyll content.
So 400-500nm and 670-680nm spectral regions are not the
optimal band to estimate chlorophyll content. And the
reflectance near 705nm showed highest correlation both linear
and nonlinear with mixed leaves chlorophyll content.
Correlation analysis between chlorophyll content and the
first derivative of reflectance
The first derivative of the reflectance curves can provide
information about rate of change along the curve and at what
wavelength these changes occur. Such insight is not apparent
with simple reflectance curves. Chlorophyll variations can
induce changes in the first derivatives of the reflectance spectra.
The first derivatives of reflectance spectra can be defined as:
DR: =
Pi+\ Pi-\
A/t
(1)
where p i+l and p j _ l are the reflectance at wavelength i +1
and i -1 respectively, AX is the wavelength difference between
i +1 and i-1, so AX is equal to 2 here. According to this
equation, the first derivatives of the reflectance of fresh leaves
in LOPEX93 database can be calculated.
Figure 3. Variations with wavelength in the correlation
coefficient obtained when regressing chlorophyll content
against 5R, using a linear model
Figure3 illustrated the variations with wavelength in the
coefficient for the relationship between the amplitude of the
first derivative, 5R, and chlorophyll content. The plot was
constructed using a linear regression model to derive the R
value. The plot showed a good degree correlation in several
bands within the 400-800nm region. However, the highest
correlation between 8R and chlorophyll content was near
530-545nm and 725-735nm.
wavelength (run)
Figure 4. Variations with wavelength in the coefficient of
determination (R 2 ) obtained when regressing 8R against
chlorophyll content, using an exponential model
Figure 4 illustrated the variations with wavelength in the
coefficient of determination for the relationship between the
amplitude of the first derivative, 8R, and chlorophyll content.
The plot was constructed using an exponential regression model
to derive the R 2 value, the best fit model for most wavelengths,
and shows a good degree correlation in several bands within the
1392