Abstract: The Three Gorges ground-based remote sensing vertical observation system, as the first pilot project of the "short board repair project" by the China Meteorological Administration, can monitor the fine structure and evolution of the atmospheric vertical profile in the Three Gorges Reservoir Region. The atmospheric temperature and humidity profiles are retrieved from the microwave radiometer in the Three Gorges ground-based remote sensing vertical observation system. Evaluating their retrieval bias and improving their accuracy is of great significance for precision monitoring of the vertical observation system in the Three Gorges Reservoir Region. By using microwave radiometer data, radiosonde data, automatic weather station data, and millimeter wave cloud gauge data of Three Gorges from November 2022 to June 2023, this study evaluates the bias of temperature and humidity at 83 height layers retrieved from microwave radiometer under sunny, cloudy, and rainy days. Then a linear regression correction method based on radiosonde data after the regrouping of the 83 height layers for temperature and humidity profile of the microwave radiometer was proposed. Finally, the applicability of the correction method by using the accuracy changes before and after the correction was evaluated using the observation from a rainy and snowy weather process. The results are as follows: (1) The temperature of the microwave radiometer has good consistency with radiosonde temperature under three weather conditions, and the bias of most height layers is less than 2 ℃. (2) On sunny days, the humidity of the microwave radiometer is higher than that of the radiosonde at all height layers. Above 6 km, the root mean square error (RMSE) of the humidity can reach 30% on cloudy and rainy days. (3) The correction method established after the regrouping of the 83 height layers can improve the temperature and humidity profile accuracy of the microwave radiometer. (4) The corrected temperature and humidity profiles of the microwave radiometer are closer to the reality in the case of a rain and snow event.