Atmospheric lidar has been widely used in horizontal scanning measurements for air pollution monitoring in recent years. The determination of the boundary value of the aerosol extinction coefficient (AEC) and the retrieval of the AEC profile are the key issues for quantitative atmospheric applications. In this work, a new method based on an improved Douglas-Peucker (DP) algorithm is proposed to find the linear region in the logarithmic lidar curve, from which the boundary value of the AEC is estimated. Combined with the classical Klett method, the AEC profile can be obtained in horizontal scanning measurements. The feasibility and performance of the improved DP algorithm has been validated through comparison studies of the AEC scanning map.
The experimental data evaluated in this work were obtained by a scanning Scheimpflug lidar (SLidar) system installed in Xianyang City, Shaanxi Province (Fig. 1), employing high power laser diodes as light sources and area image sensors as detectors. The elevation angle of the SLidar system during horizontal scanning measurements is about 3°. The scanning period is about 20 min with a rotation step of 2°. The DP algorithm has been proposed to automatically search linear regions in the logarithmic lidar signal. The DP algorithm decimates the original logarithmic lidar curve to a similar curve with fewer points through an iterative end-point fit algorithm. Besides, the variance of the distance between the logarithmic lidar signal and the corresponding line segment is used as the threshold to replace the farthest/maximum distance threshold in the classical DP algorithm, which can be used to obtain the linear region of the logarithmic lidar curve more accurately. The slope method is then employed for the determination of the AEC boundary value in the linear region, where the atmosphere is considered to be homogeneous. With the boundary value of the AEC as the input, the AEC profile can be obtained in horizontal scanning measurements according to the Klett method. Long enough linear regions (>3500 m) with an R2 correlation coefficient beyond 0.9999 are selected as reference signals, from which the AEC profile can be reliably obtained. By comparing with the AEC profile retrieved from the reference signal, the performances of the classical DP algorithm and the improved DP algorithm are evaluated.
It has been found out that the maximum distance threshold used in classical DP algorithm cannot be well adapted to different atmospheric conditions. On the other hand, the variance of the distance between the logarithmic lidar signal and the corresponding line segment, considering the deviations of all data points in the potential linear region, is more reliable for evaluating the linearity of the logarithmic lidar signal segment. According to comprehensive comparison studies, the variance threshold can be set to 1×10-4 to obtain the optimum result under various atmospheric conditions. The improved DP algorithm based on the variance threshold is then utilized to retrieve the boundary value of AEC from lidar signals. Finally, the AEC map can be obtained according to the Klett method (Fig. 14). It can be seen that the AEC retrieved by the classical DP algorithm (Fig. 7) is prone to sudden changes due to the influence of signal fluctuations. The improved DP algorithm can effectively avoid the influence of signal fluctuations, and the retrieved AEC result is more robust. To further verify the feasibility and accuracy of this method, statistical analysis on a one-week measurement is carried out. It can be seen from Fig. 15 that the AEC retrieved by the SLidar system and the proposed algorithm is in good agreement with the PM10 concentration reported by a nearby monitoring station with a correlation coefficient of 0.88. Figure 16 shows the relationship between PM10 concentration and aerosol extinction coefficient under different humidity conditions, implying that relative humidity also has a large impact on the relationship between AEC and the mass concentration of dry particles.
In this work, an improved DP algorithm based on a variance threshold is proposed for automatically searching linear region in the logarithmic lidar signal. The slope method is employed for the determination of the boundary value in a subinterval linear region, where the atmosphere is homogeneous. The AEC profile is then retrieved by the Klett method. The feasibility and performance of the classical and the improved DP algorithms are validated through detailed AEC comparison studies. It has been found that the retrieved AEC based on the improved DP algorithm is in good agreement with the PM10 concentration reported by a nearby air pollution monitoring station for a continuous measurement campaign. The promising results demonstrate that the improved DP algorithm can provide an effective approach for determining the boundary value in horizontal scanning lidar measurements.
