Assessment of nitrogen status in durum wheat: application of proximal hyperspectral and remote multispectral data

Nitrogen (N) management is a crucial agronomic practice in wheat cultivation, as N accounts for more than 40% of the mineral nutrients required for optimal wheat growth. However, due to its critical role, N is often overapplied by farmers to avoid deficiencies and ensure high yields, causing degradation of natural resources (e.g., soil and water), environmental pollution and economical losses. Applying N fertilizer at the right time and place and in the amount effectively required by wheat crop, could enhance nitrogen use efficiency (NUE), minimize potential N losses and support the aims of sustainable agriculture. Proximal and remote sensing technologies, based on spectral devices, could offer a valuable alternative to conventional, destructive, and time-intensive plant-based analyses. These technologies enable the acquisition of timely, large-scale, and continuous data on wheat growth and N status. This study assesses the potential of hyperspectral proximal sensing and multispectral remote sensing imagery to detect wheat N status, comparing data collected from ground-based and remote platforms with different spectral and spatial resolutions. The experiment was carried out in Rutigliano (Southern Italy) at the experimental farm of the Council for Agricultural Research and Economic Analysis (CREA). Durum wheat (Triticum durum) was sown over a 2500 m² area using a split-plot experimental design with three replicates. Four nitrogen rates (0, 50, 100, and 150 kg/ha) were applied as the main plot factor, while the application of a microgranular endomycorrhizal inoculum—comprising five Glomus species, plant growth-promoting rhizobacteria (PGPR), and beneficial fungi such as Trichoderma spp.—was used as the subplot factor. Twenty percent of the total N rate was applied at sowing, with the remaining fraction supplied at the end of tillering. At four key phenological stages, biometric and eco-physiological parameters were measured at 48 sampling points, including aboveground biomass (AGB), leaf area index (LAI), relative water content (RWC), gas exchange parameters, chlorophyll fluorescence, and chlorophyll concentration (µmol/m² of leaf surface). On the same dates, multispectral images were acquired using a Phantom 4 multispectral Unmanned Aerial Vehicle (UAV) at different flight altitudes and spatial resolutions. Additionally, hyperspectral reflectance data were collected using a ground-based spectroradiometer with a wavelength range of 350–2500 nm. Yield parameters were recorded at harvest. Multivariate statistical techniques were applied to analyse the complex spectral datasets, reduce dimensionality, and identify the most informative variables (e.g., specific wavelengths and vegetation indices) for discriminating levels of nutritional stress. The results demonstrate the potential of integrating proximal and remote sensing approaches for non-destructive, real-time monitoring of wheat nitrogen status. The combination of spectral data at varying spatial and spectral resolutions enables the identification of key indicators responsive to N variability. Furthermore, advanced statistical analyses facilitate the extraction of critical spectral features that distinguish plant nutritional conditions, paving the way for more precise input management and reduced environmental impact.