The original data acquired by Raman lidar system is calibrated by ten equations. The temperature is inversed, and it is compared with standard temperature at the same moment and the same effective altitude qualitatively. Then 50 groups of data are selected as statistical samples. From the perspective of quantitative analysis (similarity evaluation, difference evaluation, and comprehensive evaluation), the characteristic statistics of the samples is carried out by weight coefficient, error diagnosis and standard processing. Quality evaluation parameters and analysis method of each level are filtered. The acquired evaluation standards of each level are all based on the sample library with lager amount, which can represent average behavior of the temperature detection of lidar better. The test of real-time detection samples beyond the sample library indicate that the comprehensive multilevel control technique can describe the advantages and disadvantages of temperature data detection of lidar effectively. Temperature detection effect of lidar at different weather conditions exists differences.