Durum wheat (Triticum durum) hyperspectral response to nitrogen fertilization across phenological stages under Mediterranean conditions

Together with maize and rice, wheat provides more than 50% of the world's caloric intake, demonstrating its importance for food security. To meet the growing global demand for food, wheat production will need to increase by more than 40% in the coming decades. Nitrogen (N) is a key nutrient for durum wheat as it significantly affects both vegetative growth and grain yield and quality. Since N fertilization represents one the main costs for wheat farmers, its management is critical to improving N use efficiency and reducing negative environmental impacts and production costs. Traditional methods for estimating plant N status are destructive, time-consuming and poorly representative of spatial and temporal field variability. Therefore, hyperspectral sensors provide a potential alternative to detect N stress, enabling targeted and timely agronomic actions. This study aims to assess the effects of different N fertilisation levels and the application of a soil endomycorrhizal inoculant on traditional biometric and eco-physiological variables in durum wheat. The objective is also to characterize wheat response to N fertilization levels through hyperspectral reflectance analysis, by detecting changes in canopy spectral signatures and exploring the relationships between measured variables and raw spectra. A field experiment was conducted during 2026 growing season at the Council for Agricultural Research and Economic Analysis (CREA) farm in Rutigliano (Bari, Southern Italy). A split-plot experimental design was established with N fertilization (0, 50, 100, and 150 kg ha⁻¹ N) as main plot factor and the application of the endomycorrhizal inoculant as subplot factor. Dry aboveground biomass (AGB), leaf area index (LAI), chlorophyll content index (CCI), relative water content (RWC) and stomatal conductance (gs) were collected at tillering, stem elongation, booting, milk ripening and dough ripening phenological stages, simultaneously with the acquisition of canopy spectral signatures using the ASD FieldSpec 4 Spectroradiometer (350–2500 nm). Yield response variables were also measured at harvest. N fertilization levels significantly affected AGB and CCI at all phenological stages, whereas LAI was significantly affected only from the booting stage onwards. Furthermore, the interaction effect between the two factors was statistically significant for CCI at milk and dough ripening. Across the raw spectrum, Pearson correlation coefficients between reflectance and AGB, LAI, and CCI exhibited negative values in the visible (Vis) region, shifting to positive values in the near infrared (NIR) via the red-edge transition region and returning negative in the shortwave infrared (SWIR) with peak absolute correlation coefficients of 0.75 reached at booting and milk ripening stages. Significant correlations were observed for gs and RWC only in restricted spectral ranges across phenological stages and with correlation coefficients remaining consistently low. When performed separately on Vis, NIR and SWIR spectral range, principal component analysis (PCA) showed that the first two PCs account for over 95% of the total variance in all three analyses. Observations segregated according to N treatments and across phenological stages, mainly along PC2 axis. The 720–760, 1180–1230 and 1300-1340 nm ranges exhibited the greatest weight on discriminating durum wheat status in response to N availability.