Symposium on Remote Sensing for Resources Development and Environmental Management / Enschede / August 1986 
An evaluation of different green vegetation indices 
for wheat yield forecasting 
A.Giovacchini 
Aquater S.p.A., San Lorenzo in Campo, Pesaro, Italy 
ABSTRACT: In this study different spectral vegetation indices have been compared in order to select the more 
efficient index to forecast the final grain yield of the winter wheat. The study pointed out that at specific 
phenological stages, the maximum correlation between vegetation index values and the final grain yield can be 
obtained using indices belonging to different families. The maximum correlation value is yielded at the 
heading phase adopting vegetation indices based on the ratio concept. 
1 INTRODUCTION 
Crop production forecasting, if carried out with a 
sufficient reliability, assumes great economic and 
strategic interest. Remote Sensing technology could 
contribute significantly to increase forecasting 
reliability both through a better estimation of the 
cultivated surface and by directly or indirectly 
measuring the biophysical parameters influencing the 
crop yield. 
Generally the relationships between crop and 
remote sensing data have been analyzed using 
different types of vegetation indices. These 
spectral vegetation indices derive from the 
combination of one more spectral bands referring to 
the visible and near-infrared parts of the 
electromagnetic. 
Extensive scientific literature exists on the 
relationships between biological crop characteri 
stics and vegetation index values. Some authors 
relate spectral vegetation indices to wheat dry- 
matter accumulation (Tucker. G.J. et al. 1981; 
Aasc.J.K. et al. 1981). A strong correlation between 
spectral vegetation index values and final grain 
yield of wheat and corn crops, has been proven for 
specific phenological stages (Tucker G.J. et al. 
1980; Daughtry C.S.T. et al. 1984). Some authors 
have analyzed the possibility to estimate the leaf 
area index of crops by using different spectral 
vegetation indices (Goel. N.S. 1984; Badhwar C.D. 
1984). 
This study describes the relationships between the 
final grain yield of the winter wheat and some 
vegetation indices at several phenological stages. 
The objective of this analysis is two fold: 
- to verify if a unique vegetation index can be 
used to forecast the final grain yield of the winter 
wheat; 
- to select the most appropriate vegetation index 
in relation to the different phenological stages of 
the crop. 2 
2 MATERIAL AND METHODS 
The experiment was carried out taking into account 
30 fields of durum wheat (cultivar Creso). The crop 
fields did not belong to a specific experimental 
design, but were part of an observational study 
carried out on field crops located in the Cesano 
Valley (Central Italy). 
During the growing season of the crops, seven 
measurement surveys were carried out from March to 
July. By using an Exotech radiometre (mod 100), 
spectral data was collected in the four bands of the 
MSS Landsat sensor. Each wheat field had been 
previously stratified on the basis of the vegetation 
cover density by using color aerialphotographs at 
1:5.000 scale. Three radiometric measurements were 
carried out in each strata placing the instrument at 
3 m above the ground. In order to collect more 
representative spectral data, the radiometer was 
moved four times in each plot to obtain data 
referring to sample units of 2x2 square meters. The 
radiometer was calibrated on a standard panel before 
collection of the spectral data. At the same time, 
the following data referring to biophysical plant 
characteristics was collected: plant height, humid 
and dry biomass, % of weeds, surficial soil 
2 
moisture, n° of stalks per m ,• vegetation condition, 
% of green vegetation cover. In addition, at harvest 
the grain yield of each wheat field was assessed. 
3 RESULTS AND DISCUSSION 
For each measurement campaign spectral data was 
processed to calculate the vegetation indices 
reported on Tab. 1. For each vegetation index, the 
table reports the spectral bands used indicating the 
correspondent MSS Landsat bands. 
Both for each vegetation index and collection 
date, corresponding to a specific phenological 
stage, the coefficient of correlation between 
vegetation index values and final grain yield, were 
calculated. Tab. II shows these coefficients of 
correlation. 
As shown in Tab. II, all the vegetation indices 
yield the maximum correlation values at the heading 
stage. At this biophysical phase the indices based 
on the ratio concept yield the maximum correlation, 
as well as at the booting stage. 
At the first and second phenological stages 
considered in this study, the indices yielding the