Based on the data of microwave radiometer and radiosonde at Shapingba station in Chongqing from January 2012 to August 2014, the brightness temperature accuracy of microwave radiometer was verified by numerical simulation, and the variation characteristics of microwave radiometer retrieved products on sunny, cloudy and precipitation days were analyzed statistically. The results are as follows: (1) There was a high correlation between the brightness temperature at oxygen channel 53.85, 54.00 GHz and the temperature of radiosonde values on cloudy days, and there was also a high correlation between the simulated brightness temperature of MonoRTM model and the observed brightness temperature of microwave radiometer on sunny and cloudy days. (2) Under different weather conditions, the correlation between the temperature retrieved by microwave radiometer and radiosonde values was high. The temperature retrieved by microwave radiometer was obviously higher below 4.0 km on precipitation days, and the average absolute error below 3.8 km was less than 2 ℃ on sunny and cloudy days. The correlation between relative humidity retrieved by microwave radiometer and radiosonde values was worse than that of temperature at the same altitude. The average error of retrieved relative humidity was small at the height of 1.0-2.6 km on cloudy days, and it was also small and stable below 4.5 km on precipitation days. The average error of water vapor density retrieved by microwave radiometer was obviously larger below 4.0 km on precipitation days, while the average error on most altitude layers was smaller on cloudy days. (3) The average error of temperature retrieved by microwave radiometer below 4.2 km was large at 08:00 BST on precipitation days, but the average error of temperature and water vapor density retrieved by microwave radiometer was small on cloudy days. It showed that atmospheric profiles retrieved by microwave radiometer were available, and the inversion effect was better in stable atmospheric environment.