Published in Scientific Papers. Series A. Agronomy, Vol. LXVI, Issue 1
Written by Oleg OSINNII, Oleksandr AVERCHEV, Sergiy LAVRENKO, Pavlo LYKHOVYD
Rice is one of the major food crops with a growing demand on the global market. The need for water-saving and environmentally friendly technologies presses current agricultural science to look for alternative ways of rice irrigation. The most prospective one is drip irrigation. Yield prediction is also of great importance for sustainable agriculture. The goal of the study was to create a pilot model for drip-irrigated rice yield prediction in the conditions of the South of Ukraine using spatial normalized difference vegetation index. The index values were taken from OneSoil AI platform for the drip-irrigated rice cultivated in 2016-2017 within the framework of cultivation technology studies. The highest index value was recorded in the stage of “tillering-heading” and applied for the regression and neural network-based models. It was established that the performance of various regression models was quite similar in fitting quality and accuracy, while neural network-based one provided significantly higher precision. It is reasonable to simulate drip-irrigated rice yield with a good accuracy (MAPE<15%) using simple linear regression model. Further improvement of predictions is expected through the increase of the sample size.
[Read full article] [Citation]
