Leaf Nutrient Retrieval Using Hyperspectral Sensing and Machine Learning

Over recent years, hyperspectral sensing has become recognized as a cost-effective, zero-waste, and rapid technology for nutrient monitoring by measuring leaf spectral reflectance across a high number of narrow adjacent bands. This study investigates the potential of multi-/hyperspectral sensing, combined with machine learning techniques, to predict leaf nutrient concentrations using both leaf-level and satellite data. Α selection of linear and non-linear regression models is here investigated exploiting a publicly available leaf-level dataset, with a focus on nitrogen estimation as case study. This research is part of the DEMETRA project, which aims to advance nutrient estimation techniques and support open-data efforts.