Efficient nitrogen (N) management is critical for maximising yield while minimising environmental impacts in oilseed rape production. While Unmanned Aerial Vehicle (UAV)-based monitoring of N status has advanced rapidly, translating estimated N status into actionable fertilization strategies remains limited. This study proposes a multistage N topdressing recommendation framework for winter oilseed rape that integrates UAV multispectral data with prior agronomic knowledge using machine learning algorithms. The framework accurately estimated the nitrogen nutrition index (NNI), with the random forest model performing best ( r² =0.73 and RMSE = 0.11) for the validation dataset. By integrating estimated NNI, critical NNI thresholds, and optimal N uptake levels, dynamic, stage-specific N fertilizer topdressing rates were computed. A field experiment with varying basal N fertilizer rates was conducted to validate the framework, with UAV-guided topdressing performed after each monitoring event. Compared with the local conventional fertilization practice, the UAV-guided treatment with 90 kg N/ha basal fertilizer rates significantly improved yield by 20.2 % and N use efficiency by 80.1 %. This study bridges the gap between remote sensing-based diagnostics and in-field N fertilization, offering a feasible data-driven approach for real-time N management to enhance productivity and sustainability in oilseed rape cultivation. • Developed a UAV-based multistage N topdressing recommendation framework for rapeseed. • Integrated UAV spectral data with prior N uptake knowledge using machine learning. • Accurately estimated nitrogen nutrition index from UAV imagery. • Combined with 90 kg N/ha basal application, the framework achieved the highest yield.
Zhou et al. (Fri,) studied this question.