004
The
cate
n of
onse
| for
'rted
thod
ising
d in
nade
' the
data
rmal
veen
Ithus
soll
ative
21 by
dard
cel,
area.
with
ever,
rated
soil
ctrum
with
iat are
ire 2);
meters
nsitive
pheric
te tool
> most
ophyll
inuous
"'ASI-2
| width
cessing
International Archives of the Photogrammetry, Remote Sensing and Spatial Information Sciences, Vol XXXV, Part B7. Istanbul 2004
techniques were then applied to the airborne CASI-2 imagery
acquired for the Houghton-le-Spring test site during 2002 and
2003.
4.1 Red edge position
The red edge is defined as the ramp in green vegetation
reflectance between the red and near infrared wavelengths,
between 670 nm and 780 nm (Figure 3). It is related to the
concentration of chlorophyll in the vegetation canopy (Miller et
al, 1990) and may be characterised by the position of
maximum slope or Red Edge Position (REP). The slope and
position of the red edge change under stress conditions, with the
REP moving towards shorter wavelengths in response to
senescence (Rock et al, 1988) The REP is therefore an
appropriate tool for investigating the effect of subsidence on
vegetation.
reflectance (%)
507
blue green red near infra red
45:
40 1 A
" /
30: /
2
un
1
| red edge
0 | TTE T T T T 7
450 550 650 750 850 950
wavelength (nm)
Figure 3. Typical vegetation reflectance spectrum in the
CASI-2 spectral range. The red edge is highlighted.
A method was required to estimate the REP from the discrete
bands that bracket the red edge in the CASI-2 vegetation
bandset. Several methods are available for estimating the REP
from discontinuous data, each of which has it advocates and
detractors. Two methods were assessed; Guyot and Baret
(1988) method and Lagrangian interpolation (Dawson and
Curran, 1998). The former method, as implemented by Clevers
et al. (2001) is computationally simple but approximated the red
edge to a straight line and is therefore less sensitive to shifts in
the REP. The Lagrangian method involves calculation of the
first derivative and the fitting of a polynomial function in order
to estimate the REP. Although it is computationally more
intensive it theoretically provides a more accurate estimate of
the REP.
Accurate measurement of the REP can be made from the ASD
spectra by derivative analysis and the position compared to that
calculated using the Guyot and Baret, and Lagrangian methods
applied to the simulated CASI-2 data set. Clevers et al. (2001)
noted the possibility of artefacts in the calculated REP where
more than one maximum is present in the first derivative. This
phenomena has been noted in ASD spectra from the test site
and requires further investigation.
717
The REP has indicated species changes in areas affected by
subsidence and shifts in REP have also been noted within
species in affected areas.
4.2 Chlorophyll absorption feature
The chlorophyll absorption feature is found between the red
peak and the red edge of the vegetation reflectance spectrum.
Increases in canopy chlorophyll result in a minor deepening and
major broadening of this feature (Banninger, 1991). Continuum
removal, which is widely used in geological spectral studies, is
a normalisation technique that emphasises the wavelength and
depth of individual absorption features (Clark and Roush,
1984). The chlorophyll absorption feature is described by the
reflectance values of ten CASI-2 bands between 450 nm and
750 nm.
reflectance (%)
50 -
45 17
40 - y
35 A
E
| 7 /
30 4 continuum // /
2s 7 /
/
20. /
15 /
10 - of / chlorophyll
/ / absorption
dient, T
31 nian cu, feature
oL : : ien
450 550 650 750 850
wavelength (nm)
normalised
reflectance
10 normalised continuum
. he pass
/
/
/
/
/
\ FWHM |
=—->
depth
05 ept
/
\
\
X
a | |
| position of minimum
1
550 750
wavelength (nm)
Figure 4. Continuum removal of the chlorophyll absorption
feature and parameters.
The continuum is a convex hull connecting the maxima at 550
nm and 750 nm, defined by CASI-2 band positions (Figure 4).
Continuum removed values for each band are obtained by
dividing the reflectance value for each band by the convex hull
value, resulting in values between zero and one. The discrete
band positions of the CASI-2 dictate that the position of the
