When detecting water vapour, the traditional tools have low spatio-temporal resolution. Lidar can improve the credibility of weather forecast. According to the theory of error analysis, the error sources of water vapour and their contributions are investigated based on the experimental data. Meanwhile, the computed relative errors are compared with that calculated relative errors from Raman lidar and radiosonde data. It turns out that the error sources on water vapour observed by Raman lidar includes the calibration constant, the transmission correction and the Raman scattering signals. The calibration constant error is about 4% and constant with height, and the major contribution to the total error under 1.5 km height. The transmission correction error increases slowly along with the height and less than 4% in the condition of clean air. The Raman scattering signals error is less than 20% under 3 km height, but it becomes the main source above 3 km height. The comparison shows that the relative error derived by Raman lidar agrees well with the calculated relative error. In summary, the analysis results above are helpful to promote the application of Raman lidar in weather forecast.