Deep learning-based response mechanism of nitrogen utilization, yield, and quality in mulched drip irrigation rice to nitrogen application rate

Nitrogen management is critical for coordinating productivity, grain quality, and environmental sustainability in film-mulched drip-irrigated rice systems. A 3a field experiment was conducted in Ningxia, Northwest China, to investigate the nonlinear responses of soil nitrogen dynamics, nitrogen use efficiency (NUE), yield formation, and grain quality to different nitrogen application rates under film-mulched drip irrigation (FMDI). Based on the experimental dataset, deep learning models including Deep Neural Network (DNN), Multilayer Perceptron (MLP), Convolutional Neural Network (CNN), and Transformer were employed to characterize complex agronomic response patterns under varying nitrogen conditions. The results showed that increasing nitrogen application significantly enhanced total nitrogen accumulation in deeper soil layers, suggesting potential downward redistribution of nitrogen under excessive nitrogen input. Agronomic traits including plant height, dry matter accumulation, filled-grain ratio, and 1000-grain weight generally increased with nitrogen application. NUE indicators improved under moderate nitrogen conditions but declined substantially when nitrogen application exceeded 180 kg ha −1 , indicating diminishing nitrogen utilization efficiency under excessive input. Grain quality responses exhibited clear trade-off characteristics: moderate nitrogen application improved protein accumulation and milling quality, whereas excessive nitrogen increased chalkiness and negatively affected eating-quality-related traits. Model comparison and robustness analyses further demonstrated that deep learning models outperformed conventional linear regression approaches in capturing nonlinear coupling relationships among soil nitrogen, NUE, yield components, and grain quality variables. Comprehensive evaluation suggested that 180 kg ha −1 represented the most balanced nitrogen application level for coordinating productivity, grain quality, nitrogen utilization, and environmental sustainability under the arid climatic and alkaline soil conditions of the Ningxia irrigation region. These findings provide scientific support for optimized nitrogen management and coordinated water-fertilizer regulation in FMDI rice production systems.