Satellite monitoring of crop traits for detecting nitrogen deficiency at critical nitrogen fertilization windows

How well satellite imagery tracks crop responses to nitrogen (N) supply and deficiency directly determines its practical value for in-season N management. Leaf chlorophyll content (LCC), leaf area index (LAI), and aboveground biomass (AGB) are key proxies of crop N nutritional status, and vegetation indices (VIs) are widely used to retrieve them. Yet how VIs respond across N rates and split-application windows, and whether they can detect early N deficiency, remain unclear. From 2021–2024, a winter wheat (Triticum aestivum L.) field experiment was conducted across eight fields with five N rates (0, 60, 120, 150, and 180 kg ha –1 ) and three fertilization windows: vegetation onset, early jointing, and shortly before heading. Twenty-nine VIs were derived from Sentinel-2 imagery. After the first fertilization window, VIs tracked crop traits poorly (|r| = 0.01–0.33) because N stress had not yet emerged. The second window was optimal for tracking LCC (|r| ≤ 0.64) and LAI (|r| ≤ 0.88); the third window was optimal for AGB (|r| ≤ 0.78). VIs were more sensitive to canopy structure (LAI, AGB) than to leaf chlorophyll. RECAI/TVI and CCCI detected N deficiency more reliably than traits predicted by random forest. However, none of the indices reliably separated N-insufficient from N-sufficient canopies, indicating that mild N stress leaves only yielded weak satellite-scale spectral signatures. Overall, our findings demonstrate the potential of satellite data to support crop N status monitoring during the latter two fertilization periods.