Performance of RGB-Based Vegetation Indices for Temporal Monitoring of Maize Grown for Silage

Remote sensing based on RGB imagery acquired by remotely piloted aircraft provides an accessible approach for monitoring crop development; however, the performance of vegetation indices and their ability to represent final biomass may vary throughout the crop cycle. This study evaluated the performance of RGB-based vegetation indices for temporal monitoring of maize grown for silage and investigated the relationship between spectral information and crop biomass. The study was conducted in a 6.0 ha field at the Federal Institute of Southeast Minas Gerais, Rio Pomba Campus, Brazil. Thirty-three georeferenced sampling points were established, and aerial surveys were performed throughout the crop cycle using an RGB camera mounted on a remotely piloted aircraft. Five vegetation indices were evaluated: visible normalized difference vegetation index (visible NDVI), visible atmospherically resistant index (VARI), green leaf index (GLI), combined index (COM), and leaf area index (LAI). Green and dry matter yields were determined at harvest. VARI, visible NDVI, and GLI showed consistent temporal responses to crop development, whereas COM and LAI were less suitable for describing crop dynamics. Although RGB information acquired closer to harvest showed stronger associations with green matter yield than earlier imagery, its predictive ability remained limited. Increasing model complexity did not consistently improve prediction. These results demonstrate that RGB-based vegetation indices are useful for temporal monitoring of maize grown for silage, but spectral information in the visible domain alone is insufficient to adequately predict final biomass variability.