The International Archives of the Photogrammetry, Remote Sensing and Spatial Information Sciences. Vol. XXXVII. Part B7. Beijing 2008
474
Figure 2(a)
Measured canopy chlorophyll content (g m' 2 )
Figure 2(b)
Figure 2(ab). Cross-validated prediction of grass variables in
Majella National Park, Italy, using narrow band NDVI. Left:
estimated LAI versus measured LAI; rights: canopy chlorophyll
content. The optimum wavebands are those reported in Table 2.
Figure 2 shows the relationships between the estimated and
measured LAI and canopy chlorophyll content using narrow
band NDVI. From the figure, it seems that saturation starts to
occur for canopy chlorophyll content greater than 2 (g nT 2 ) and
for LAI greater than 7(m 2 m' 2 ).
3.2 Red edge inflection point
The red edge inflection point (REIP) was calculated using two
methods. As can be observed from the results reported in Table
4, the relationships between measured and estimated grass
variables were not reliable using any of the methods. The R 2
and relative RMSE of the grass variables obtained from the
three methods were relatively similar.
Among the studied variables, estimation of canopy chlorophyll
content again yielded the highest R 2 values and the lowest
relative RMSE. Compared with regression models developed
using the optimum narrow band indices, the REIP methods
produced somewhat lower accuracies.
confirmed previous studies by researchers who suggested a
strong contribution by SWIR bands to the strength of
relationships between spectral reflectance and LAI (Cohen and
Goward, 2004; Darvishzadeh et al., 2008; Lee et al., 2004;
Nemani et al., 1993; Schlerf et al., 2005). Compared with the
narrow band NDVI, the narrow band SAVI2 gave somewhat
higher R2 and lower relative RMSE values for LAI. This result
is in agreement with that of Broge and Leblanc (2001), who
used simulated data and found SAVI2 to be the best vegetation
REIP method
Rev
RRMSE CV
Linear
LAI
0.49
0.39
interpolation
CCC
0.56
0.41
Linear
LAI
0.51
0.38
extrapolation
CCC
0.57
0.41
Table 4. Performance of red edge inflection point calculated
using different methods for predicting grass variables in Majella
National Park, Italy.
3.3 Partial least squares regression
The relationships between grass variables and reflectance
spectra were modeled using PLSR. Cross-validated results
using the entire reflectance spectra as inputs are shown in
Figure 3. The optimal number of PLSR factors preventing over
fitting was selected in two ways: (i) through visual inspection of
cross-validated RMSE versus the number of factors plots (not
shown), and (ii) by setting the condition that adding an extra
factor must reduce the RMSE (RMSE CV ) by >2%. The number
of factors in the final model were 4 for LAI and 5 for canopy
chlorophyll content models. Compared with other methods,
PLSR using entire reflectance spectra increased all R 2 values
(R 2 = 0.69, 0.74 for LAI and canopy chlorophyll content,
respectively) and decreased the Relative RMSE values
(RRMSE = 0.32, 0.34 for LAI and canopy chlorophyll content,
respectively).
4. DISCUSSION
The field experiment led to a large number of sample subplots
(191) with high variations in LAI. The canopy integrated
chlorophyll content (LAI x leaf chlorophyll content) strongly
reflects the variability of LAI and (to a lesser extent) leaf
chlorophyll content, expressed by the high inter-correlation
between LAI and canopy chlorophyll content (not shown).
Among the grass characteristics studied, canopy chlorophyll
content was most accurately estimated by nearly all of the
applied methods. The canopy chlorophyll content contains both
the structure and chlorophyll information of vegetation and can
be accurately estimated by canopy spectral reflectance.
The relationship between measured and estimated LAI was
better explained by multivariate calibration methods (PLSR)
than by univariate methods such as narrow band vegetation
indices and REIP. This is because a two-wavelength index
utilizes only a limited amount of the total spectral information
available in hyperspectral data (Lee et al., 2004).
The bands selected as the best combination of the vegetation
indices for LAI were found in the NIR to SWIR regions. This
index for LAI estimation. Moreover, the narrow band SAVI2
performed relatively well for canopy chlorophyll content. This
is due to the major influence of LAI in canopy chlorophyll
content and also to the fact that SAVI2 is relatively insensitive
to external factors such as soil background effects.
Although red edge has proved to respond more linearly to LAI
and chlorophyll when compared with the classical NDVI, which
often suffers from saturation problems (Danson and Plummer,