The drift phenomenon exists in the process of radisonde meteorological observation, and the characteristics of drifting data are of great significance for evaluating the accuracy of ground-based microwave radiometer retrieval products. By researching the drift characteristics of radisonde data at the 0 to 10 km height range of Jinghe National Basic Station in Xi’an from 31 October 2017, to 31 December 2023, on the basis of dividing the station area and the drifting area, this study discusses the influence of drift on temperature evaluation of ground-based microwave radiometers under different weather conditions. The results show that both the Root Mean Squared Error (RMSE) and Correlation Coefficient (Corr) of the drift area are larger than those of the station area, and the Corr is smaller than that of the station area, exhibiting distinct interval characteristics. For clear-sky conditions, the RMSE decreases while the Corr increases in the middle and lower troposphere, and the maximum regional difference occurs at 4.25 km. For cloudy conditions, both the RMSE and Corr increase, and the maximum regional difference in the correlation coefficient occurs at 5.25 km. As altitude increases, for clear-sky conditions, the RMSE and Corr tend to approach those of the station area, while for cloudy conditions, only the Corr approaches and overlaps around 8 km. Due to the combined influence of sounding drift and weather conditions, there are differences in the evaluation indexes of different weather in the same region. The RMSE of clear-sky weather is larger than that of cloudy weather, and the Corr is smaller. The RMSE of the drifting region is close, and the Corr of clear sky is small. Under long-distance sounding drift conditions, wind effects lead to changes in kinetic and thermal energy, particularly within the 7 km altitude range in the troposphere, where wind speed differences under different weather conditions are significant. These factors are the main reasons for the large deviations in temperature evaluation metrics of ground-based microwave radiometers. The study provides a reference for using sounding data to evaluate the data accuracy and product application of new atmospheric detection equipment.