UAV-Based Hyperspectral Monitoring of Foliar Nitrogen in Spinach (Spinacia oleracea L.): Comparing Flight-Specific, Window-Based, and Time-Integrated Modeling Approaches

Spinach (Spinacia oleracea L.) is an economically important leafy vegetable whose nitrogen management remains comparatively understudied with remote sensing tools relative to major row crops. This study evaluated the potential of unmanned aerial vehicle (UAV)-based hyperspectral imagery to estimate foliar nitrogen content (%N) in field-grown spinach at a commercial plot across two growing seasons and four nitrogen fertilization treatments. Nine spectral indices were related to destructively measured %N at four phenological stages of the Biologische Bundesanstalt, Bundessortenamt und CHemische Industrie (BBCH) scale (BBCH 16, 30, 45, and 49) under three modeling scenarios, ranging from flight-specific regression to a continuous model incorporating days after sowing (DAS) as a proxy for crop development, with model performance assessed out-of-sample by repeated cross-validation. Foliar nitrogen content differed significantly among treatments from BBCH 30 onward and only marginally at the earliest stage, with the two highest doses converging into a statistically indistinguishable class. The Vogelmann red-edge index 3 (VOGREI3) yielded the most accurate estimates, peaking at BBCH 45 (cross-validated coefficient of determination, R2 = 0.845) before declining significantly at BBCH 49, a loss consistent with spectral saturation of the closed canopy. A single continuous model incorporating DAS covered the BBCH 45 and BBCH 49 flights with a relative error of 12.4%, against 9.6% and 14.9% for the corresponding flight-specific models, and required no stage-specific recalibration within the calibration seasons. In an unseen season, plot ranking was preserved, and a few reference samples corrected the bias. UAV hyperspectral imagery combined with development-stage-aware modeling therefore shows potential for non-destructive nitrogen monitoring in spinach once locally calibrated, although transferability across further seasons, sites and cultivars remains to be established.