As the invasion of alien species intensifies, the native salt marsh vegetation, especially the ecosystem of Suaeda salsa (S. salsa), in Chinese coastal wetlands has been severely disrupted, significantly impeding its functionality within coastal wetland ecosystems. Chlorophyll content (Cab) is an important parameter for monitoring the growth and health status of vegetation. However, the remote sensing mechanism of Cab for different phenotypes of green and red vegetation influenced by betacyanin under different water and salt conditions is not clear. Therefore, to further clarify the remote sensing mechanism of S. salsa and improve the generality of Cab prediction for different phenotypes under different water and salt conditions, we designed a growth experiment under different groundwater levels and salt concentrations. This experiment was used to simulate the growth of S. salsa in different habitats in nature and investigate the relationship between their physicochemical parameters, spectra, and environmental stress. We propose using the specific absorption coefficient of betacyanin (Beta) to adjust PROSAIL-D to better fit the spectral characteristics of S. salsa. The relationship between Beta and Cab was utilized to further optimize the simulation accuracy of the PROSAIL-D model in different environments. Feature extraction algorithms, including differential enhancement of two-dimensional correlation spectroscopy (FD-2DCOS) and three-band models, were introduced to construct multiple spectral features for establishing a Cab prediction model for S. salsa. The research results showed that: (1) S. salsa had the best response of Cab and spectra to salinity under a higher groundwater level and showed a gradual reddening trend. (2) The spectral curves simulated by the adjusted PROSAIL-D model were more consistent with the spectral characteristics of S. salsa under different water-salt environments, and the simulation effect could be further optimized by using the relationship between Beta and Cab. (3) The FD-2DCOS analysis method we proposed could highlight the useful spectral information related to Cab better than the original 2DCOS algorithm. The final eight spectral features of (1/R701–1/R540)R540, Mean(D590–600), Mean(D660–670), Mean(D677–685), FDVI_NRG, FDVI_NIR, Skewness (D677–750) and Position(D677–750) had the best response to Cab of S. salsa, and had the least influence on LAI and soil background. (4) The particle swarm optimization random forest regression (PSO-RFR) model based on eight spectral features and adding 70% measured spectral data performed the best in predicting the actual Cab of S. salsa, with an RMSE of 2.326 μg·cm−2. Although further optimization and calibration of the adjusted PROSAIL-D model and Cab prediction model of S. salsa are needed in the future, this study extended its application to S. salsa, a special local salt marsh vegetation, and promoted the development of Cab monitoring of salt marsh vegetation in coastal wetlands.