基于数据窗口标准差的边界层高度反演方法——以上海市为例

大气边界层具备一个重要特性，即在边界层顶部，气溶胶浓度在垂直分布上存在显著突变。利用该特性，采用主动遥感装置经过窗口平滑去除噪点后数据，提出基于梯度法改进的大气边界层高度反演方法——窗口标准偏差法。基于激光云高仪后向散射廓线数据，利用该方法反演边界层高度，在边界层高度下大气气溶胶混合均匀时，反演结果较为理想。在此基础上，将窗口标准偏差法反演结果与逐步曲线拟合法反演结果进行比较。结果表明：两种方法具有较好的相关性，相关系数为0.94。两种方法产生差异的原因在于窗口标准偏差法不考虑夹卷层厚度而逐步曲线拟合法考虑夹卷层厚度。窗口标准偏差法可以降低高空背景光噪音对反演结果的影响。通过该方法反演出的边界层高度，在时间序列上具有连续性强的特点，反演结果更有利于研究大气边界层高度的时间变化趋势。

Boundary layer height inversion method based on data window standard deviation: a case study in Shanghai

The atmospheric boundary layer has an important feature, that is, at the top of the boundary layer, there is a significant abrupt change in the vertical distribution of aerosol concentration.Using this feature, the active remote sensing device is used to remove the noise data through window smoothing, and an improved atmospheric boundary layer height inversion method based on gradient method-the window standard deviation method was proposed.Based on the backscattering profile data of the laser ceilometer, this method was used to retrieve the height of the boundary layer.In the case when the atmospheric aerosol is uniformly mixed at the boundary layer height, the inversion result is ideal.On this basis, comparing the inversion results of the window standard deviation method with the inversion results of the stepwise curve fitting method, it is found that the two methods have a good correlation, with a correlation coefficient of 0.94.The reason for the difference between the two methods is that the window standard deviation method does not consider the thickness of the entrainment layer while the stepwise curve fitting method considers the thickness of the entrainment layer.The window standard deviation method can reduce the influence of high-altitude background light noise on the inversion results.The height of the boundary layer retrieved by this method has the characteristics of strong continuity in time series, and the inversion results are more conducive to the study of the temporal change trend of the height of the atmospheric boundary layer.