Dataset accompanying “Investigating the limits of spectroscopy for the estimation of foliar N and P in apple”: Hyperspectral reflectance, foliar nutrient concentrations and associated metadata

This dataset was generated to support research investigating the use of hyperspectral reflectance for the estimation of foliar nitrogen (N) and phosphorus (P) concentrations in apple (<i>Malus domestica</i>) trees. This article and the dataset it describes accompany an original research article submitted to <i>Computers and Electronics in Agriculture</i> entitled "Investigating the limits of spectroscopy for the estimation of foliar N and P in apple" [1]. Data were collected from a controlled potted experiment involving 150 'Golden Delicious' apple trees grown under varying nutrient supply regimes, including full nutrient supply, nitrogen- and phosphorus-deficient treatments, and trees infected with '<i>Candidatus</i> Phytoplasma mali'. The experiment was conducted over the 2023 growing season at the Laimburg Research Centre in South Tyrol, Italy. All data and the accompanying code for its analysis is freely available in the associated GitHub repository [2]. Spectral data were collected using the Spectral Evolution SR-3500 field spectroradiometer with an attached leaf clip, producing high-resolution hyperspectral reflectance profiles (350-2500 nm) from the adaxial surface of fully expanded leaves. A total of 1189 leaf spectra were recorded and were matched to chemically analysed leaf samples. Corresponding foliar N and P concentrations (and others) were determined through laboratory analysis using the Dumas combustion method for nitrogen and ICP-OES following acid digestion for phosphorus. The dataset includes metadata detailing tree treatments, sampling dates, infection status, and shoot growth metrics. Additionally, R scripts used for data processing, spectral pre-treatment (including multiplicative scatter correction and Savitzky-Golay derivatives), feature selection (VIP and mRMR), and model development are provided. The dataset is suitable for reuse in the development and benchmarking of spectral models for nutrient estimation, especially in the context of field-based or remote sensing applications in horticulture. Its wide range of foliar nutrient values, inclusion of multiple physiological stresses, and detailed documentation make it a valuable resource for researchers working in precision agriculture, plant phenotyping, chemometrics, and hyperspectral data analysis.