Abstract: Betacyanin, as a natural non-photosynthetic pigment, helps to elucidate the physiological responses and resistance of plants to different environmental stress factors. In order to improve the accuracy of betacyanin estimation in plant leaves, the red Suaeda salsa, which is rich in located in betacyanin and located in the southern restoration area of Tiaozini mudflat wetland in Dongtai, Jiangsu, was selected as a research subject. Based on measured spectral data and biochemical parameters, the PROSPECT-D leaf radiative transfer model was optimized and calibrated; using simulated leaf reflectance spectra as the data source, three training set selection methods were used in conjunction with three inversion algorithms: Partial Least Squares Regression (PLSR), Particle Swarm Optimization-Random Forest (PSO-RF), and Support Vector Machine (SVM) to construct remote sensing inversion models for betacyanin content in the leaves of the coastal wetland Suaeda salsa. The results show that the simulated leaf reflectance spectra of the optimized model fit well with the measured spectra (R²=0.996, RMSE=0.011); statistical regression PLSR and machine learning algorithms SVM, PSO-RF all achieved high-precision inversion of betacyanin content in Suaeda salsa leaves, among which the PLSR model had the highest accuracy (R²=0.86, RMSE=1.60, RPD=2.14); it provides technical support for further implementation of fine remote sensing quantitative monitoring of Suaeda salsa pigment content on a large scale.