atial
one
oral
that
user
get
nal
Vas
ose
ed.
the
ing
ing
)2)
tes
his
ind
ns
cal
no
er,
ce
Aigner, Edgar
2.3. Yield Data
Yield data were obtained directly from the farmers by questionnaires. The farmers are able to derive such
information on paddock scale during harvest from the amount of crop they sell, transport or store. This data
can be regarded as accurate and reliable.
Wheat, Barley and Canola are the main crops in the Eastern Wimmera. Due to the large size of the observation
targets (approx. 9 km?), several crop types can be found within one target and most of the times their
composition was only known to a certain extent. Therefore, areas to be included into the prediction model had
to show uniform behavior in space and time. This was tested by examining the variations in time and space
between smaller and larger subsets around the center pixel of the subset throughout the observation period.
3. MODEL DEVELOPMENT
3.1. Prediction Dates
The timing of the predictions is very important for the applicability of the prediction model as a management
tool. The farmers need reliable and timely forecasts to take management actions. The growing season of wheat
in the eastern Wimmera lasts from May to December; this means that predictions should be made available
throughout the growing season until the end of October, when there is the last opportunity to take economic
decisions (personal communication with the farmers). Predictions before September proved not to be reliable.
Thus, three prediction dates were tested: 10.September (DOY 253), just in time for taking additional
management actions, NDVI(max), at maximum green vegetation cover (usually between end of September
and Mid of October, ca. DOY 265 to 290) and 31.October (DOY 304), as the latest date for taking final
economic decisions (e.g. insurances).
3.2. Single Linear Regressions
Figure 2: Linear regressions of cumulated NDVI, GF, GSR and cumulated SDD for the growing
seasons 1995 to 1997 of wheat at the prediction date “NDVImax-
a) NDVI c) GSR
NDVI -Grain Yield for Wheat Growing Season Rainfall -Grain Yield for
at "NDVI max" (1 995-1 997) Wheat at "NDVI max" (1995-1 997)
45 400
40
r 350 +
= 3 oe e ^e T 2501
+ = T
2 20 + 3 = 200
15 + 6 150 T
101 100 +
5 - R? =0,6325 50 T
0 i : i i 0 4
0 1 2 3 4 5 0 1 2 3 4 5
Grain Yield (t/ha) Grain Yield (t/ha)
b) GF d) SDD
Comencement Grainfilling -Grain Yield SDD -Grain Yield for Wheat
for Wheat (1995-1997) at "NDVI max" (1995-1997)
250 200
150 t +
240 + 100 T e e t*,.,?*
e. 50 | ewe c?
+ a t + re ®
4 SOT + 18
á *«*
220 t +400 +
450 + +
210 + zi *£ -0217
200 0 1 2 3 4 5
0 1 2 3 4 5 Grain Yield (tha)
Grain Yield (tha)
The yield data’s correlation to four p
prediction date (NDVI), the date of the commencement o
duration of the grainfilling,
growing season rainfall (GSR) and the water stress index "stress degree days
arameters was examined: Cumulated daily NDVI from May, 1* to the
f the grainfilling stage (GF) as a measure for the
daily rainfall cumulated from May, 1* to the prediction date to indicate the
" (SDD), cumulated from July, 1*
to the prediction date. Figure 2 shows the linear regressions of those parameters with yield of wheat for 1995
through 1997. Cumulated NDVI (Fig. 2a), as an indicator for green vegetation growth, has a positive
correlation with crop yield. The offset of the regression line is due to the NDVI caused the by soil
background. The GF parameter (Fig. 2b) can be identified from the NDVI — MVC curvature (figure 1). The
International Archives of Photogrammetry and Remote Sensing. Vol. XXXIII, Part B7. Amsterdam 2000.
21
