The initial profile, which is calculated by the eigenvector regression algorithm in the groundbased hyperspectral remote sensing, has a significant impact on the accuracy of the physical retrieval. Based on the eigenvector regression algorithm, the similarities and differences between the retrievals of temperature profiles and water vapor profiles are analyzed using the radiance data observed by AERI and coincident radiosonde profiles. The optimal number of principal components is analyzed when retrieving the temperature and water vapor profiles. Considering both the accuracy and the information contained in the eigenvectors, the optimal numbers of principal components are both set to 7. In order to improve the accuracy of remote sensing, the surface temperature, humidity and pressure are introduced as the influence factors. The experiment results show that the introduction of surface pressure has a better performance than the other two single meteorological elements and the assemble of factors composed of all three meteorological elements, especially for the accuracy and stability of temperature and water vapor profiles in the middle and lower parts of the boundary layer. With the decrease of altitude, the RMSE of temperature profiles decreases to a maximum of 1.5 K, and the RMSE of temperature profiles decreases to a maximum of 0.42 g/kg. At the same time, the impact of logarithmic water vapor mixing ratio on the retrieval of water vapor profiles is analyzed. The result shows that the introduction of the logarithmic profile has little effect on the accuracy of the retrieval. However, the accuracy of the water vapor profile gains more than 12% when converting mixing ratio to relative humidity.