The surface polarized reflectance is able to fully reflect the physical characteristics of surface, such as vegetation classification, plant biomass estimation, leaf angle distribution and surface water content. The bidirectional polarization distribution function is a useful tool for quantitatively describing the surface polarized reflectance. The multi-angle polarized remote sensing reveals a significant advantage in capturing the radiation and polarization information. The polarization and directionality of the earth reflectance (POLDER) instrument is the only sensor which has provided a long-term trend of polarized measurements. Its combination of multi-spectral, polarization and multi-angle observations has considerable capability for retrieving ocean, land, cloud and aerosol properties. Until now, data obtained from POLDER has been widely used to study various surface bidirectional polarized reflectance models, especially Nadal model. However, parameters of Nadal model reveal a low accuracy in China region. In this study, the parameters of Nadal model suitable for China region are obtained and analyzed based on the POLDER-2 polarized reflectance data. Based on the modified parameters of Nadal model, polarized reflectance under different surface types is further analyzed. Our results show that the polarized reflectance retrieved from modified parameters of Nadal model reveals better correlation with the POLDER-2 products than the polarized reflectance from Nadal official parameters under different surface types. The polarization properties of three typical surfaces (forest, grassland and desert) are further investigated and reveal that 1) different surface polarized reflectances decrease with the increase of the scattering angle, and the polarized reflectance of the same object decreases as the normalized difference vegetation index increases; 2) significant discrepancies exist between the polarized reflectances of different surfaces, the polarized reflectance of forest is the lowest in the three surface types, then that of grass is the second lowest, and desert reveal the largest value (about twice that of forest), 3) the discrepancies of polarized reflectance between different surfaces have an increasing trend as satellite view zenith angle increases. This study will provide a priori knowledge for the detection of surface polarization properties and aerosol parameters based on multi-angle polarization remote sensing data, and also establish a good foundation for the quantitative applications of GF-5 satellite multi-angle polarization imager to be launched soon in China.