Purpose. Early yield prediction is an important task of modern agriculture, providing great opportunities for better crop management and enhance the advantages of implementation of the systems of precision agriculture. Winter wheat is the major cereal crop in Ukraine. In order to forecast winter wheat (Triticum aestivum L.) grain yields prior to harvesting in the systems of precision agriculture, we developed prediction models on the basis of remotely sensed normalized difference vegetation index values at the stages of the crop tillering (stage 5) and heading (stage 10.1).

Methods. The model of grain yield prediction has been developed on the basis of regression analysis of the field yield data of the crop, obtained during 2017-2018 at the research fields of the Institute of Irrigated Agriculture of NAAS, in connection to the spatial vegetation index values in corresponding stages of the crop growth. Polynomial regression analysis was implemented in order to determine the link between the yields and vegetation index values at the two stages of the crop development. Statistical analyses were performed at p > 0.05.

Results. The results of the study revealed the possibility of early (up to 60-70 days in advance in case of use the index values at the tillering stage) winter wheat grain yield prediction by linking the values of normalized difference vegetation index of the crop to its productivity. Approximation of the developed polynomial regression models proved that their accuracy is enough to provide reliable yielding forecasts: the mean absolute percentage error of the models is 7.76-8.53%, R2 values for the prediction is 0.9331-0.9454.

Conclusions. The developed polynomial regression models allow obtaining early grain yield prediction using spatial normalized difference vegetation index values. The models are easy to use and will be especially practical in the systems of precision agriculture.