Based on the FNL data of the United States NCEP and the analysis fields of Chinese GRAPES, the temperature data of Beijing sounder station are evaluated, including observation residuals, average deviations, standard deviations, probability density distributions, kurtosis coefficients, skewness coefficients, correlation coefficient and root mean square error. Then, the quality control of temperature data of sounding is conducted and the quality control effect is analyzed according to the evaluation results. The test results show that the sounding temperature is of good quality and the error is within ±1℃ basically. There is a slight difference between the evaluations of the FNL analysis and the GRAPES results. The threshold values of the suspicious value and error value selected at a single time can determined by itself adaptively according to the characteristics of residual distribution based on the analysis fields. The quality of the sounding temperature is improved and the data retain the original characteristics after quality control according to the evaluation. The quality control method can effectively eliminate the seasonal difference and the evaluation results of the two reference standards are basically consistent after quality control.