Aigner, Edgar 
data compared to 1997’s data shows both, a great difference in values and in yield. This might be due to non- 
linearity effects, but non-linear regression could not be proved statistically for the whole database by testing 
diverse kind of fits. Taking 1995's data out of the database and modeling its yield using regression coefficients 
derived from the remaining data—records, hence, must lead to an overestimation. The same is valid for the 
modeling of 1997's yield, while 1996 yield is slightly underestimated. This fact shows, that this model needs 
testing with additional data from a longer observation period. For the prediction date “NDVI, the residual 
analysis shows no systematical over— or underestimation throughout 1995 to 1997. Again, the reason for this is 
the fact, that “NDVI, is not a fix prediction date, like the others. All that also underlines the importance of 
the “GF” parameter, as it is a time variant measure as well. This is confirmed by the fact that the error being 
made at the prediction date 10.Sep. without “GF”is much higher than the one made at ‘31.Oct.”. 
5. CONCLUSION 
It was the aim of this study to test the applicability of the NOAA-AVHRR in combination with 
meteorological for crop yield estimation in the Eastern Wimmera, South-Eastern Australia. 
It was shown, that the parameters NDVI, GSR, GF and SDD, correlate with grain yield of wheat, cereals 
(wheat and barley) and canola. Three, for the farmers important prediction dates were examined, the 10^ of 
September, the date of the maximum NDVI value in the NDVI-MVC curvature and the 31* of October. Yield 
estimates using the linear regressions of the single parameters are not accurate enough for the Farmers' 
requirements. Therefore the parameters were included into multiple linear regressions as independent variables 
to predict grain yield. This significantly increased correlations, and yield estimations at the prediction date 
“NDVI max could be made with an accuracy in the order of the farmers' estimates, or better. It is obvious that 
this model needs further examination. The inclusion of additional data and a longer observation period, where 
more diverse agrometeorological conditions are taken into account, will probably improve the results. Also, 
the quality of the data used certainly can be improved by an enhanced processing setup. Tests with 1998's 
data, however, made another major shortcoming of this model apparent. 1998 was affected by frosts late in the 
growing season end of October and beginning of November (1), that is after the prediction dates. This results in 
an overestimation of grain yield by 1 to 2 t/ha, as none of the effects caused by the frost days is reflected in the 
independent variables. Bad weather conditions or diseases occurring after the predictions, therefore, are not 
taken into account by this model. Due to the large subsets used, the environment of the observed paddocks has 
to be uniform. If this is not the case, for example due to small water—bodies, no reliable predictions are 
possible. If the remote sensing data can be derived from smaller subsets, also the accuracy of the method 
described might increase. Further investigation is necessary, before the described model can become an useful 
farming tool. If the results are at least confirmed, and an operational setup can be implemented, the method 
has the potential to benefit the farming community in South Eastern Australia. The relatively simple and easily 
accessible data used, might allow the model to be applied to other regions in the world with minor 
modifications. 
6. REFERENCES 
Becker, F., Li, Z.L. (1990): "Towards a local split window method over land surfaces" 
In: International Journal of Remote Sensing, Vol. 11, No. 3, pp. 369—393. 
Cabezón, M.P., Taylor, J.C. (1994): "Yield Forecast Model for Wheat and Barley in Andalucia" 
In: Proceedings of the Yield Forecasting Seminar, Villefranche 24—27 October, Eurostat-JRC-DGVI-FAO, 
pp. 433-442. 
Demircan, A (1995) *Die Nutzung fernerkundlich bestimmter Pflanzenparameter zur flüchenhaften 
Modellierung von Ertragsbildung und Verdunstung" 
Münchener Geographische Abhandlungen, Reihe B, Band 20, Institut für Geographie der Universität 
München, Kommisionsverlag: GEOBUCH- Verlag, München. 
Dilley, A.C., Elsum, C.C. (1994): "Improved AVHRR Data Navigation Using Automated Land Feature 
Recognition to correct a Satellite Orbital Model" 
CSIRO Division of Atmospheric Research Technical Paper No. 34. 
Dilley, A.C., Edwards, M. (1998): "The automatic processing of ASDA format NOAA HRPT data at CSIRO 
DAR" 
CSIRO Division of Atmospheric Research Internal Paper No. 6. 
  
24 
International Archives of Photogrammetry and Remote Sensing. Vol. XXXIII, Part B7. Amsterdam 2000. 
bem CMM emi 
P. 
— 
bend CS 
laud FUN bed