AUTHOR=Szabó Andrea , Siphiwe Nxumalo Gift , Buday-Bódi Erika , Ademola Blessing , Tamás János , Nagy Attila 

TITLE=Evaluation of ground based spectral imaging for real time maize biomass monitoring

JOURNAL=Frontiers in Plant Science

VOLUME=Volume 16 - 2025

YEAR=2025

URL=https://www.frontiersin.org/journals/plant-science/articles/10.3389/fpls.2025.1566305

DOI=10.3389/fpls.2025.1566305

ISSN=1664-462X

ABSTRACT=Although point measurements of water management properties have become increasingly common, understanding the spatial heterogeneity of agricultural fields remains critical for advancing precision agriculture. Spectral analysis provides a non-destructive approach to evaluating plant biophysical properties, such as chlorophyll and carotenoids, which are critical for precision agriculture. This study addresses the challenge of precise plant trait prediction by integrating proximal sensing data with biomass observations to inform more effective water management strategies. This study predicts carotenoid and chlorophyll content from NDVI, and estimates dry and wet biomass from vegetation cover using multispectral Tetracam data. A key novel aspect of this study lies in the pioneering integration of proximal sensing with biomass information to improve the estimation of plant properties, offering practical applications for precision agriculture. The diagnostic results demonstrated varying model performances. The carotenoid prediction model, with a moderate R² (0.54), exhibited a slight overestimation, characterized by a Mean Bias Error (MBE) of 0.02 µg/g and a Normalized Root Mean Square Error (NRMSE) of 17%. Conversely, the chlorophyll prediction model showed improved accuracy, achieving an R² of 0.64, an MBE of 0.04 µg/g, and an NRMSE of 15.92%. Models predicting wet and dry biomass from vegetation cover yielded comparable performances, with R² values of 0.55 and 0.58, and low NRMSEs of 13.26% and 15.06%, respectively. These findings underscore the potential of combining proximal sensing and biomass data to enhance the prediction of plant properties, providing valuable insights for optimizing precision agriculture through machine learning.