Mathematical modeling of oilseed yield in the South of Ukraine according to aerospace monitoring data

Abstract

Goal. To develop mathematical models of the yield of the main oilseeds grown in the South of Ukraine, using data from aerospace monitoring of their crops. To evaluate the reliability and accuracy of models, determine the limiting factors and the possibilities of their scientific and practical use. Methods. The study was performed using data from the average monthly value of the normalized differential vegetation index (NDVI) for 2021–2022 for oilseeds (winter rape and sunflower) located in the South of Ukraine (in Kherson, Odesa, Mykolaiv, Zaporizhzhia, Dnipropetrovsk, and Kirovograd regions). Satellite monitoring data was obtained on the GIMMS Global Agricultural Monitoring System platform, NASA. Images of Terra MODIS NDVI in combination with GDA Crop Specific Masks were used. Mathematical processing of the results and the construction of models were carried out using linear regression algorithms, and the reliability and accuracy of the models were evaluated using F- and P-statistics. Building models and graphic work were done in Python 3.13 environment. Results. Regression models were developed of winter rape and sunflower yields by NDVI value. It was established that they had high adequacy (coefficient of determination for winter rape — 0.61, sunflower — 0.70), and accuracy (average absolute error for winter rape — 0.13 t / ha, or 5.6%, for sunflower — 0.16, or 7.8%). However, it should be noted that, despite the high accuracy, the models had limited practical value, since the detailed statistical evaluation confirmed the insufficient level of mathematical reliability of the model for winter rape (P-statistics for the model of winter rape — 0.13, sunflower — 0.007 at the minimum required level P < 0.05). Conclusions. According to the parameters of the seasonal dynamics of the value of satellite NDVI, mathematical models of the yield of the main oilseeds grown in the South of Ukraine were developed. They had high adequacy and accuracy, but the model for winter rape was of purely theoretical importance due to the insufficient level of reliability of predictions.