Direct detection Doppler wind lidar based on Fabry-Perot interferometer has been demonstrated for its capability of atmospheric wind detection ranging from troposphere to mesosphere with high temporal and spatial resolution. However, bias would emerge while retrieving the radial velocity during the observation and an external wind reference is needed to eliminate this bias. Reasons for this bias are theoretically analyzed and results show that the ambient temperatures of Fabry-Perot interferometer (FPI) and seeder laser are main effects. Then this effect of temperature is learned by experiment. By precise control of the ambient temperature of FPI and seeder laser separately, laser transmission through calibrated FPI, which depends upon the temperature, is studied. Using the instrumental function of the FPI, the frequency bias dependence on temperature is obtained. The experimental results show that the temperature coefficient of frequency bias for seeder laser is 1650 MHz/K while the one for FPI is 799 MHz/K, which means the precision of ambient temperature should be theoretically better than 0.004 K for seeder laser and 0.007 K for FPI in 355 nm Doppler wind lidar system.