Although the optimization of a static Fabry–Perot interferometer (FPI)—used as a Doppler shift discriminator in wind lidar—has been proposed, it cannot be applied to the scanning FPI used in the high-spectral resolution lidar for temperature detection. After a comparison, the optimal scanning implementation is chosen and a new optimization scheme is proposed. The free spectral range (FSR) of the FPI is determined by the width of the Rayleigh spectrum. Then, for analytical purposes, the transmission of Rayleigh backscattering through an FPI is simplified to be a superposition of a Gaussian function and a constant background. The maximum likelihood estimation and the Cramer–Rao bound theory are used to derive an analytic expression of the temperature error. Thus, the effective reflectance of the FPI can be optimized. Finally, assuming known atmospheric temperature–pressure–density profiles, backscattering raw signals are simulated using the optimized parameters of the FPI and some other key system parameters of our existing lidar system. Comparisons between the assumed and retrieved temperature profiles revealed that error can be achieved in the altitude range of 15 to 40 km, even with the disturbance of aerosol contamination.