ıptions
AS the
ie. grid
| map
I type,
series
se this
vals is
lutions
y (e.g.
spatial
n (e.g.
ery 16
al and
in the
ensing
erlying
ver the
hnique
remote
> 2.0. a
ng the
VI's as
tool.
| based
erature
a state
etween
tes can
he link
bsolute
Seville,
sat-TM
1992.
rofiles.
on the
hermal
km at
not be
in the
| value
tion of
es, this
canopy
g crop
raction
ccurate
'S have
lues in
nalised
DH
where IR and R are the reflectance in the infrared and red range of
the electromagnetic spectrum respectively.
To correct for soil background reflectance Clevers (1988) defined
the so-called Weighted Difference Vegetation Index (WDVI):
WDVI=IR-(IRs/Rs)R (2)
where IR, and R, is the infrared and red reflectance of the bare
soil respectively.
This WDVI is used for estimating the Leaf Area Index LAI
according to the inverse of an exponential function:
LAI=1/a*In(1-WDVI/WDVIe) (3)
with a and WDVI” as two empirical parameters. Bouman et al.
(1992) confirmed the exponential relationship between LAI and
WDVI. They empirically found consistent parameters for various
years, locations and growing conditions for some main agricul-
tural crops, like sugar beets, potatoes, wheat, barley and oats.
The advantage of WDVI is its higher saturation ceiling for
biomass than NDVI has and therefore the WDVI has a greater
sensitivity to changes in biomass. NDVI is widely used because it
is less sensitive to differences in view angle and atmospheric
conditions (which are considerable in NOAA/AVHRR images).
WDVI and NDVI profiles are made from NOAA/AVHRR
composites, Landsat- TM and SPOT images.
4.2 Comparison of satellite derived data and CGMS-
simulations
4.2.1 Regional average profiles using high resolution satellite
data
Based on the combined use of the high-resolution (HR) derived
crop classification and the CORINE land cover database,
concerning irrigated and non-irrigated land, crop specific
vegetation index profiles are composed of the 7 HR satellite
images for winter wheat and sunflower for both irrigated fields
and non-irrigated fields. A baseline is subtracted from the VI
values to correct for the different radiometric characteristics of the
two sensors used, and for differences in atmospheric conditions
between the different observation dates.
The mean curves for the whole study area are displayed in Figure
1. The simulation results using the standard data sets for this area
are also plotted in this figure.
For winter wheat the decrease in simulated LAI due to ripening of
the crop is clearly shown. It is also in agreement with the decrease
Las
a
o
NDY!
o ‘ re mae ws dL ++ — ud a 0
IDILI 20.f*b 10-Apr 30-May 19-Ju!
LAI
in NDVI as derived from the available satellite data. Both
simulated LAI and observed NDVI reveal a difference in
vegetation development between irrigated land (potential growth)
and non-irrigated land (water-limited growth).The available
satellite images allow to monitor the final part of the winter wheat
growing season only. For sunflower the entire growing season is
observed. Simulated crop growth agrees well with the observed
NDVI-profiles. For sunflower the differences in vegetation
development between irrigated land and non-irrigated land are
also clearly visible. WOFOST simulations indicate insufficient
water availability for rainfed sunflower crops for the first days in
May. This corresponds to the satellite observed differences in
NDVI profiles. In the next section field specific situations are
discussed.
4.2.2 Field specific profiles of irrigated and non-irrigated
sunflower
In this study, calibration of WOFOST is based on field specific
profiles. Sunflower has been selected as an example to indicate
the use of remote sensing in calibration of crop simulation,
because the available HR images cover the whole growing cycle
for sunflower. For selected fields individual NDVI and WDVI
profiles have been made. Figure 2 and Figure 3 show the field-
average profiles of irrigated and non-irrigated sunflower. The
profiles indicate differences between irrigated and non-irrigated
land in total biomass amount and in the date of maximum LAI.
Also the earlier saturation of NDVI compared to WDVI can be
seen for irrigated situations in these figures.
4.2.3 Regional average profiles using low resolution satellite
data
In addition to the analysis of HR satellite data also vegetation
index profiles as derived from NOAA/AVHRR satellite data have
been analysed. The increase in NDVI in the beginning of the
growing season is negligible, while WDVI clearly shows a
stronger relation with crop growth. On the other hand WDVI is
more sensitive to atmospheric disturbances as the dips in the
WDVI-profile are much larger comparing to the NDVI-profile
(Nieuwenhuis et al. 1996).
For the test-site in Spain relatively small scale agriculture is
found. NOAA/AVHRR is not well-suited to obtain crop specific
information. In addition to the poor resolution of 1 km, the
geometrical distortion of the images of the Seville test-site caused
severe problems.
$
NDVI
3-sep
Figure 1 Simulated LAI (bold and dashed lines for irrigated and non-irrigated plots respectively) and observed NDVI (Wl; irrigated
plots, 0: non-irrigated plots) for winter wheat (lett) and suntlower (right) using high resolution satellite data for Seville 1992
Intemational Archives of Photogrammetry and Remote Sensing. Vol. XXXII, Part 7, Budapest, 1998 145
