Atmospheric CO₂ plays an important role in controlling climate change and its effect on the carbon cycle. However, detailed information on the dynamics of CO₂ vertical mixing remains lacking, which hinders the accurate understanding of certain key features of the carbon cycle. Differential absorption lidar (DIAL) is a promising technology for CO₂ detection due to its characteristics of high precision, high time resolution, and high spatial resolution. Ground-based CO₂-DIAL can provide the continuous observations of the vertical profile of CO₂ concentration, which can be highly significant to gaining deeper insights into the rectification effect of CO₂, the ratio of respiration photosynthesis, and the CO₂ dome in urban areas. A set of ground-based CO₂-DIAL systems were developed by our team and highly accurate long-term laboratory experiments were conducted. Nonetheless, the performance suffered from low signal-to-noise ratio (SNR) in field explorations because of decreasing aerosol concentrations with increasing altitude and surrounding interference according to the results of our experiments in Wuhan and Huainan. The concentration of atmospheric CO₂ is derived from the difference of signals between on-line and off-line wavelengths; thus, low SNR will cause the superimposition of the final inversion error. In such a situation, an efficient and accurate denoising algorithm is critical for a ground-based CO₂-DIAL system, particularly in field experiments. In this study, a method based on lifting wavelet transform (LWT) for CO₂-DIAL signal denoising was proposed. This method, which is an improvement of the traditional wavelet transform, can select different predictive and update functions according to the characteristics of lidar signals, thereby making it suitable for the signal denoising of CO₂-DIAL. Experiment analyses were conducted to evaluate the denoising effect of LWT. For comparison, ensemble empirical mode decomposition denoising was also performed on the same lidar signal. In addition, this study calculated the coefficient of variation (CV) at the same altitude among multiple original signals within 10 min and then performed the same calculation on the denoised signal. Finally, high-quality signal of ground-based CO₂-DIAL was obtained using the LWT denoising method. The differential absorption optical depths of the denoised signals obtained via LWT were calculated, and the profile distribution information of CO₂ concentration was acquired during field detection by using our developed CO₂-DIAL systems.
Keywords: CO2 monitoring; differential absorption lidar; high-quality signal acquisition; lifting wavelet transform.