基于激光雷达观测的大气边界层自动识别局部最优点算法

摘要

The atmospheric aerosols have significant influence on human health,the environment and the climate system.Theatmospheric boundary layer (ABL) reflects processes of the near-surface atmosphere and con centration of pollutants.Ground-based laser radar can monitor the vertical dis tribution of atmospheric aerosols stably and continuously.It provides dynamic i nformation for timing observations of the ABL and environmental forecasting,if aerosols can be monitored and evaluated using lidar technology.There is a gap i n the study of ABL observations during the presence of a residual layer and aero sol intrusion,as well as deficiencies in the accuracy and poor computational ef ficiency of the gradient method.This paper combines the physical meaning of the latter method with characteristics of a lidar timing chart and local optimum mo del,which based on space-time proximity.Then a polarization-Mie scattering l idar system is used to observe the vertical distribution of aerosols over time a t Taihu observation site,which is in a newly developed area of the city of Wuxi,Jiangsu Province,China.Observation and analysis is carried out for two cases in terms of pollution at the end of 2012.Then corresponding estimation model w as built with gradient method and local optimum model based on range-corrected signals.In the case of steady weather and mixed pollution,results of the gradi ent method and local optimum model were very similar.However,the gradient meth od has more error in the case of pollution intrusion with the residual layer.Th e local optimum model based on the space-time proximity theory considers vertic al eigenvalues and horizontal correlations,thereby greatly reducing the effects of low clouds,signal interference,weak signals,bi-layered aerosols,and res idual layer condition.Compared with the gradient method,the local optimum mode l had a smaller O(n) and greater stability in computer automatic identification.ABL identification in the case with the residual layer and aerosol intrusion wa s solved with use of lidar technology and the local optimum model.The accuracy and computational efficiency problems of the gradient method were resolved using automatic operation.%大气边界层高度是影响近地面大气物理运动的主要因素,同时也是影响地面污染物浓度的一个重要因素.地基激光雷达可以对大气气溶胶的垂直分布进行连续稳定的监测,应用激光雷达技术对大气边界层进行连续观测可以为环境监测与预报提供指导性的动态信息.针对存在残留层以及外来污染物输入情况时边界层高度变化检测的可靠性及计算效率问题,结合梯度法的物理意义与激光雷达时序图的图形图像学特征,提出了一种基于时空邻近度的边界层局部最优点识别算法.以江苏省无锡市新区偏振米散射激光雷达太湖观测站点的气溶胶垂直观测数据为例,通过对2012年底两次污染事件进行观测分析,分别使用梯度法和局部最优点法进行大气边界层高度的自动识别.实验结果表明,在静稳状态和污染混入后的情况下,梯度法与局部最优点识别法的结果较为接近,但梯度法在处理污染混入状态以及存在残留层的情况下误判率较高.基于时空邻近度的局部最优点算法通过对垂直特征值以及水平相关性的控制,有效地消除了在弱信号、噪声信号、低云以及存在残留层和外来污染等情况下导致的计算机误判现象,在减小算法时间复杂度的同时在计算机自动识别结果具有更高的稳定性,弥补了梯度法在自动化运行中的识别精度与计算效率的不足.