The accuracies of microwave radiometer （MWR） measurements obtained at Baoshan Station of Shanghai from July 2018 to July 2019 are evaluated by comparing brightness temperature of MWR against those calculated from radiosonde soundings at the same site with radiative transfer model. Beyond that, the performances of the MWR calibration techniques and the effects of radome replacement are estimated. The results show that the observed brightness temperature from MWR agrees well with the simulated ones from radiosonde soundings in clear-sky conditions. The correlation coefficients between the two datasets are over 0.96 in all channels, with root mean square errors being 0.15-2.68 K. The performances of the V-band channels are better than those of K-band channels. Moreover, the features of brightness temperature bias vary with channels, including random deviations, systematic biases and biases with significant seasonal variations. The absolute calibration with liquid Nitrogen （LN2 calibration） could significantly reduce the systematic bias in most of K-band channels. But the brightness temperature from V-band channels do not change obviously after calibration. By replacing the radome periodically, brightness temperature biases in rainy conditions might be reduced significantly. And the recovery time of brightness temperature in rainy conditions might also be shortened. The results also indicate that the radome made of new material used in this study is more efficient than the original one in reducing the negative impacts of precipitation in MWR accuracy. It works for about 4 months.