High-precision temperature retrieval with Pure Rotational Raman Lidar via a hybrid denoising approach

Pure Rotational Raman Lidar (PRRL) is a powerful optical tool for atmospheric temperature profiling, but its echoes are three to four orders of magnitude weaker than Mie scattering and easily contaminated by background light and detector noise. As a result, long-range signals are often buried in noise, limiting detection range and retrieval accuracy in environmental optics applications. To address this challenge, we propose a VMD-DE-SSA hybrid denoising approach tailored to the characteristics of PRRL signals, in which Differential Evolution (DE) adaptively optimizes the parameters of Variational Mode Decomposition (VMD), followed by Singular Spectrum Analysis (SSA) for secondary denoising. Simulation studies demonstrate that the proposed approach improves the signal-to-noise ratio (SNR) of PRRL by over 25% while reducing signal RMSE and MAE by over 69.2% and 74.8% compared with conventional denoising techniques. When applied to 23 days of valid PRRL measurements, our approach significantly improved temperature inversion, achieving an average daily RMSE of less than 0.57 K and MAE below 0.73 K. Our findings demonstrate that the proposed VMD-DE-SSA technique is a robust tool for advancing PRRL capabilities, enabling more precise and reliable temperature measurements for remote sensing applications.