The Impact of Water Availability on the Discriminative Status of Nitrogen (N) in Sugar Beet and Celery Using Hyperspectral Imaging Methods

Abstract

A pot experiment was conducted to determine the impact of water availability on the discriminatory status of nitrogen (N) in plants using hyperspectral imaging. Nitrogen deficiency causes a significant decrease in chlorophyll concentration in plant leaves regardless of water availability. Five different classification algorithms were used to discriminate between nitrogen concentrations in plants at different levels of water availability. Several statistical parameters, including kappa and overall classification accuracy for calibration and prediction, were used to determine the efficiency and accuracy of the models. The Random Forest model had the highest overall accuracy of over 81% for sugar beet and over 78% for celery. Additionally, characteristic electromagnetic wavelengths were identified in which reflectance correlated with nitrogen and water content in plants could be recorded. It was also noted that the spectral resolution between the N and High Water (HW)/Low Water (LW) treatments was lower in the short-wave infrared (SWIR) region than in the visible and near-infrared (VNIR) region.