International Archives of the Photogrammetry, Remote Sensing and Spatial Information Sciences, Vol XXXV, Part B7. Istanbul 2004
characteristics of NDVI in the spring season. The NDVI values
over winter wheat showed the distinctive maximum in the first-
May and the sharp decrease after the time of maximum. Mixed
vegetation was a category comprised with rural settlements,
vegetable fields, bushes, miscellaneous grasses and small trees,
and its NDVI recorded the lower values than winter wheat
before May and the higher in June. The NDVI values over bare
land were the constantly lowest among the categories. The
category of forest showed the increasing tendency of NDVI in
the spring season and the level in June was higher than that of
mixed vegetation. From the practical point of view, the author
excluded forest and employed 3 categories for the further
examination, because forest area having a considerable acreage
was appeared only in a small part of the plain.
End-member values of NDVI of specific land use estimated by
the method mentioned above might not be fit each other among
counties. One of the causes could be the heterogeneous
distribution of atmospheric effect on the value of digital data.
Accuracy of geometric correction would be another cause
because the pixel value was sensitive to the components of land
use, which might be considerably modified by the slight gap of
location. Actually end-member values of winter wheat
calculated for 7 counties show similar temporal profile but
difference of the values in some cases (Figure 3). Therefore, the
author employed the averaged values to estimate winter wheat
sown area in this study.
07 r —Q—— Shunyi
a — Xushui
0.6 F ANE —1x—— Dacheng
—€— Wuyi
05r ^. ——3K——- Yangxin
——OÓ—— Gaotang
I 04 FX ——1 — Feixiang
e
Qu /]| NN (mir wi. mean
7 03}
02 +
01 FF
0 1 À 1 1 1 A 1 1 J
© OO. oc o — © © O 9
e — e e — N e + N
I I I | | | | | |
N . yous CN — CN : -—
aia ; x 2 e :
A. o A © > > = c
1743355255
Figure 3. Comparison of temporal change of end-member
values of winter wheat
A constraint in the first method exists in the condition that it
could require end-member values, which were obtained by the
combination of the high spatial resolution data. This condition
may not be satisfied for the wide area in the appropriate season
of sequential years. The second method had an intention to
mitigate this constraint. Figure 4 depicts the relation between 2-
temporal values of NDVI at mid-May and mid-June and the
probability density of land use in a pixel for the case of Shunyi
county, where a larger circle indicates the higher probability
density, e.g. the largest one is 95% followed by 85%, 75%, so
on. This figure evidently shows the feature that the position
would approach to the vertex of a triangle for the type of winter
wheat, bare land and forest according to the increase of
probability density.
140
fi Q winter wheat
S O mixed vegetation
> © bare land
9 0.25 F
I ® forest
T2902 k
©
e
=
2 015 +
2
> 01 F
5
x
©
= 0 i A 1 À A J
0 0.1 0.2 0.3 0.4 0.5 0.6
Maximum NDVI (May 11-20)
Figure 4. Relation between 2-temporal NDVI and the
probability density of land use
Figure 5 shows the values of winter wheat for 7 counties. It is
recognized that there is a common point, to which all the values
approach in accordance with the increase of probability density.
This could induce a schematic diagram as described in Figure 6.
© Shunyi
03r G Dacheng
S OFeixiang
v 025 F O Gaotang
= @Wuyi
£ 02 F > O Xushui
= Q Yangxin
= 0.15
t3
= 0.1
EE
2 °
'* 005 F
©
=
0 L À 1 d 1 J
0 0.1 0.2 0.3 0.4 0.5 0.6
Maximum NDVI (May 11-20)
Figure 5. Comparison of allocation by the probability density of
winter wheat sown area
NDVI (mid June)
Winter wheat (096)
dense X
O
Winter wheat (0%)
NDVI (mid May)
Figure 6. Schematic diagram of relation between winter wheat
percentage and 2-temporal NDVI
Internationc
ECC
Here, the a
line connec
winter whee
The author
percentage
the data on
the line of |
expressed |
parameters
and its coef
Figure 7 sh
area in and
7(a) repres
wheat sowr
33 by 33 p
linear unmi
of estimatic
and June 1-
1-10, and ^
at the tec
heterogenei
composite
application
describes t
percentage
However, |
estimated v
medium le
logistic fun
geometric
registration
Even thoug
of percenta
the distribu
spatial unit
percentage
county as a
linear unmi
obtained fr
method als
estimated v
to analyze
wheat in th
Because th
data for e
selected to
Figure 9 sl
fluctuation:
counties be
change wei
caused by |
and not b
estimated ı
characteris
