The relationships between QuickBird-derived SR[R, NIR] and chlorophyll
content at the landscape level were also significant with slightly lower R
2 (0.41)
(Figure 7.4b).
7.5 CONCLUSIONS AND DISCUSSION
With species percentage cover data, this study calculated Chl a + b content for
heterogeneous grasslands at the canopy and landscape levels. This canopy integration approach is simple and shown to be capable of estimating vegetation
chlorophyll content beyond leaf level. By relating the chlorophyll contents at leaf,
canopy, and landscape scales with remote sensing data, we found it is possible to
estimate chlorophyll contents at three levels using both ground and space remote
sensing data.
At the species level, 64% of the variation in Chl a + b was explained by labbased remote sensing–derived spectral index. Comparing with other similar studies,
we found that these R
2 values were generally lower. For example, Gitelson and
Merzlyak (1997) estimated chlorophyll content on higher plant leaves over a range
of species (i.e., Acer, Castanea, Vitis vinifera, and Fagus) using R750/R700
and found the R
2 values ranged from 90 to 98%. The relatively lower R
2
values found in this study could be explained by the fact that the leaf samples
were taken in the maximum growing season (July 1–15). In this season, vegetation
is uniformly green so that species have a similar level of chlorophyll content.
However, spectral data may vary due to differences in leaf attributes such as size
and shape and thickness. Consequently, those species with a similar level of
chlorophyll content may not have the same spectral indices values, leading to a
lower overall R
2 .
At the canopy level, 53% of the variation in chlorophyll content was explained
by field remote sensing–derived spectral index, while only 39% of the variation
was explained by satellite remote sensing. At the landscape level, 75% of the
variation in vegetation chlorophyll was explained by field remote sensing data, and
41% of the variation was explained by satellite remote sensing data. In comparison
with R
2 values at the canopy level, the lower R
2 values between satellite remote
sensing and chlorophyll data at canopy and landscape levels could be explained by
two reasons:
1. The different estimation capability between narrow-band index (calculated
from ground hyperspectral data) and broad-band index (calculated from
QuickBird) for quantifying chlorophyll. The narrow-band spectral indices
have long been demonstrated to be better predictors in comparison with
broad-band spectral indices for quantifying chlorophyll content.
2. Imperfect atmospheric and geometric corrections for the QuickBird image.
Although atmospheric effect was corrected, the spectral index derived from pixels
could still be affected by cloud/haze interference. As for the geometric correction, the
134
ESTIMATING GRASSLAND CHLOROPHYLL CONTENT
content at the landscape level were also significant with slightly lower R
2 (0.41)
(Figure 7.4b).
7.5 CONCLUSIONS AND DISCUSSION
With species percentage cover data, this study calculated Chl a + b content for
heterogeneous grasslands at the canopy and landscape levels. This canopy integration approach is simple and shown to be capable of estimating vegetation
chlorophyll content beyond leaf level. By relating the chlorophyll contents at leaf,
canopy, and landscape scales with remote sensing data, we found it is possible to
estimate chlorophyll contents at three levels using both ground and space remote
sensing data.
At the species level, 64% of the variation in Chl a + b was explained by labbased remote sensing–derived spectral index. Comparing with other similar studies,
we found that these R
2 values were generally lower. For example, Gitelson and
Merzlyak (1997) estimated chlorophyll content on higher plant leaves over a range
of species (i.e., Acer, Castanea, Vitis vinifera, and Fagus) using R750/R700
and found the R
2 values ranged from 90 to 98%. The relatively lower R
2
values found in this study could be explained by the fact that the leaf samples
were taken in the maximum growing season (July 1–15). In this season, vegetation
is uniformly green so that species have a similar level of chlorophyll content.
However, spectral data may vary due to differences in leaf attributes such as size
and shape and thickness. Consequently, those species with a similar level of
chlorophyll content may not have the same spectral indices values, leading to a
lower overall R
2 .
At the canopy level, 53% of the variation in chlorophyll content was explained
by field remote sensing–derived spectral index, while only 39% of the variation
was explained by satellite remote sensing. At the landscape level, 75% of the
variation in vegetation chlorophyll was explained by field remote sensing data, and
41% of the variation was explained by satellite remote sensing data. In comparison
with R
2 values at the canopy level, the lower R
2 values between satellite remote
sensing and chlorophyll data at canopy and landscape levels could be explained by
two reasons:
1. The different estimation capability between narrow-band index (calculated
from ground hyperspectral data) and broad-band index (calculated from
QuickBird) for quantifying chlorophyll. The narrow-band spectral indices
have long been demonstrated to be better predictors in comparison with
broad-band spectral indices for quantifying chlorophyll content.
2. Imperfect atmospheric and geometric corrections for the QuickBird image.
Although atmospheric effect was corrected, the spectral index derived from pixels
could still be affected by cloud/haze interference. As for the geometric correction, the
134
ESTIMATING GRASSLAND CHLOROPHYLL CONTENT