Predição da produtividade da soja e delimitação de zonas de manejo utilizando índices de vegetação

Vegetation indices enable the estimation of soybean yield, contributing to crop monitoring and decision-making in precision agriculture systems. The objective of this study was to evaluate the spatial and temporal variability of NDVI and NDRE vegetation indices across different soybean growing seasons, aiming at grain yield estimation, management zone delineation, and the development of specific predictive models. The study was conducted in a 122.65 ha commercial field located in Ponta Porã, Mato Grosso do Sul, Brazil, during the 2019/2020 to 2024/2025 growing seasons. Yield data obtained from a combine harvester yield monitor and Sentinel-2 satellite imagery were used to calculate NDVI and NDRE. The indices were correlated with yield using simple and multiple linear regression models, which were evaluated based on the coefficient of determination (R²), root mean square error (RMSE), and normalized RMSE (nRMSE). Both indices showed significant correlations with yield; however, NDRE exhibited superior predictive performance. Multiple linear regression models improved soybean yield prediction accuracy by integrating vegetation indices. NDRE showed a strong relationship with yield across management zones, with better performance in Zone 2 (R² = 0.76; nRMSE = 7.7%) than in Zone 1 (R² = 0.67; nRMSE = 7.8%), indicating high predictive accuracy in both zones. The results demonstrate the potential of NDRE for soybean yield estimation, management zone delineation, and the development of predictive models applicable to precision agriculture.