IAPRS & SIS, Vol.34, Part 7, "Resource and Environmental Monitoring", Hyderabad, India,2002
provides an estimate of pre-anthesis crop period. A scatter plot
of D with crop emergence date is given in Fig 3.
5.5
Yield 2 12.129 * Gmax = 2.7239 .
R? - 0.8338
5.0 4
Wheat Yield (t/ha)
A
e
2.5 T r T
0.45 0.50 0.55 0.60 0.65
Profile Predicted Peak NDVI
Fig 1. Relationship between wheat yield (t/ha) and profile
predicted peak NDVI (Gx) for the combined dataset
Table 2: Details of Regression Analysis
(Yield 7 Intercept * Slope * G4)
s Dataset | #N | Slope* T r s.e.e.
1 TE 18 13.783 | -3.583 |.0.869 | 0.206
(1.338)
2 Punjab 16 Cn 3.109 | 0.716 | 0.251
3 | Combined 34 12.129 | -2.724 | 0.834 | 0.232
Har+Pun (0.957)
4 Subset 15 13.968 | -3.721 | 0.899 | 0.197
Haryana (1.301)
5 Subset 11 12.503. .|--3:077- | 0.682 | 0.303
Punjab (2.869)
6 Subset 26 12.568 | -2.996 | 0.834 | 0.247
Combined (1.143)
N : No. of districts; s.e.e. : Standard error of estimation
* Standard error in slope is given in parenthesis.
A decrease in this period (D), with delay in spectral emergence
with r^70.939 was noticed. The range of D is quite similar to
pre-anthesis period range observed by Hundal and Kaur (1997).
The decrease in D with delayed crop emergence is analogous to
experimental observation of decrease in pre-anthesis period
with delay in sowing observed by Saini er a/. (1986). Thus, T,
and T,. obtained from crop spectral profile contain
information about crop phenology.
In the present study, radiance conversion from digital number
and built-up area subtraction method were applied to normalize
the multi-date and multi-sensor data; which may not be an
adequate radiometric normalization for such type of data set. In
future such multi-date data set would be normalized for
multiplicative and additive non-target signals
6. CONCLUSIONS
Using IRS WiFS data for 2000-2001 rabi season a non-linear
model (Badhwar model) based district-wise wheat spectral
profiles were generated with multi-date NDVI values
normalised by built-up area subtraction. The fittings of spectral
profiles were found to be statistically significant as tested by
366
Chi-square test. The spectral profile parameters T,, & and B
were found to be statistically significant. The other spectral
profile parameters like- Tax, Gmax, and © were also estimated
and linear regression analyses were made to related these
parameters (including T,, o and ß) with district-wise wheat
yields. To test the overall significance of the regressions and the
significance of the values of the regression coefficients
statistical analysis was performed. It was found that district-
wise wheat yields were highly correlated with profile predicted
peak NDVI (Ga). The rmse of yield prediction for an
independent set of eight districts was found to be 0.181 t/ha at
the mean observed yield of 4.546 t/ha (i.e. < 4%).
5.1
491 y=0.9822x v
R? 2 0.8647 .°
Predicted Yield (t/ha)
o 90 RB
~ © - w un ~
e
®
w
un
3.5 3.7 3.9 4.1 4.3 4.5 4.7 4.9 5.1
Obserevd Yield (tha)
Fig 2. Validation of yield model at an independent set of 8
districts.
The derived pre-anthesis period D from time of crop emergence
(Ty) to time of peak NDVI (T,4,) decreased with delay in
spectral emergence as observed in field study. The study will be
extended to additional seasons with more data set for predicting
wheat yield.
ACKNOWLEDGEMENTS
The authors are grateful to Dr. R. R. Navalgund, former Deputy
Director (RESA), Space Applications Centre (Presently,
Director NRSA). A great deal of encouragement was provided
by Shri J. S. Parihar, Group Director, Agricultural Resources
Group, SAC in carrying out this study.
©
a
1
©
e
1
70 4
65 4
60
y = -0.7358x + 322.03
55 R? = 0.9385
Pre-anthesis Period (Tmax-To)
50
320 325 330 335 340 345 350 355 360
Crop Emergence Day From Jan 1, 2000
Fig 3. Spectral profile parameters used for crop phenology.
\
REFERENCES
Badhwar, G. D., 1980. Crop emergence date determination
from spectral data. PE&RS, 46(3) pp.369-377.
Badhwar, G. D., Carnes, J. G. and Austin, W. W., 1982. Use of
Landsat-derived temporal profiles for corn-soybean feature
