Abstract: In order to estimate the retrieval ability of the one-dimensional variational (1DVAR) algorithm which was applied to obtaining temperature and humidity profiles from observations of the ground-based microwave radiometer, brightness temperature data observed by a multi-channel ground-based microwave radiometer at 00:00 and 12:00 UTC in Beijing of 2010 and 2011. First of all, the 432 cloudless samples have been obtained by the procuring cloudless sample method based on the simultaneous surface-based observing data, infrared brightness temperature (observed by the infrared sensor installed on the ground-based microwave radiometer) and radiosonde observations. Then, quality control over observed brightness temperature has been made according to the brightness temperature calculated by the radiative transfer model. After that, the simulated experiments using the radiosonde observations have been done, and we achieved that the accuracy of the retrieved temperature profiles are statistically better than this of background profiles below 3 km, and 1DVAR retrievals improved the background humidity profile from the ground surface up to 10 km. At last, the temperature and humidity profiles retrieved by the 1DVAR algorithm are compared with radiosonde observations. The result showed that the retrieved profiles achieved a root mean square error (RMSE) with respect to radiosonde observations less than 2.9 K for temperature profiles and less than 0.47 g/m3 for absolute humidity profiles below 10 km in height. Comparing with the neural network (NN) algorithm of the microwave radiometer, the retrieval results of the 1DVAR algorithm were closer to the real atmosphere.