Aims <p>The critical nitrogen concentration dilution curve is used to quantify the optimal nitrogen requirements of crops at different growth stages. However, studies using hyperspectral data for rapid and quantitative analysis of the nitrogen requirement of high-quality indica fragrant rice cultivars in South China are scarce.</p> Methods <p>Two high-quality indica fragrant rice varieties, Meixiangzhan. 2 and Nanjingxiangzhan, were grown in South China. Plant nitrogen concentration (PNC) and aboveground biomass (AGB) were measured, and canopy hyperspectral images were acquired at three critical growth stages. Critical nitrogen dilution curves were obtained, and the nitrogen nutrition index (NNI) and nitrogen requirement were calculated. Hyperspectral data were used to model and predict nitrogen requirements, enabling real-time, rapid, dynamic, and non-destructive estimation.</p> Results <p>The results showed that the critical nitrogen dilution curve was Nc = 1.96 W^–0.39 (R² = 0.85), Nc is the critical nitrogen concentration, and W is AGB. With a validation RMSE of 0.28 and an NRMSE of 24.57%, indicating high predictive accuracy. The nitrogen requirement during whole growth period ranged from –72.64 to 21.55&#xa0;kg/ha. Spectral estimation modeling indicated that the model based on the successive projections algorithm (SPA) for dimensionality reduction coupled with random forest (RF) performed best, achieving training and testing R² values of 0.93 and 0.84, with RMSEs of 0.09 and 0.24, respectively.</p> Conclusions <p>We demonstrate that the nitrogen requirement model developed using the SPA-RF algorithm exhibits superior accuracy and stability. The results provide a theoretical foundation and technical support for enabling quantitative precision fertilization of these rice varieties.</p>