利用一种自动识别算法移除天气雷达反射率因子中的亮带

该文介绍了一种自动识别和移除雷达反射率因子资料中亮带的算法, 并对该算法进行了初步测试。该算法利用的是插值到直角坐标系中的雷达反射率因子资料, 其配置和运行也相对简单, 但却对移除亮带比较有效。首先, 设定一套雷达反射率因子垂直廓线的理想模板, 这些理想的模板能够在最大程度上反映不同亮带存在区域的雷达实际反射率因子的垂直廓线特征。然后, 在水平方向每个点上, 进行理想模板和实际反射率因子垂直廓线在垂直和水平两个方向上的拟合和差异计算, 来自动识别雷达反射率因子中存在的连续亮带区域。最后, 利用亮带之上和亮带之下的反射率因子值对亮带中的反射率因子值进行插值纠正, 就可以移除亮带。利用位于天津塘沽的我国新一代天气雷达 (CINRAD/SA) 的反射率因子资料, 通过个例分析和准业务运行试验, 均表明这个简单算法可以识别和移除绝大多数影响雷达定量降水估计的反射率因子亮带区域, 但是实际雷暴区域的反射率因子特征受到该算法的影响比较小。计算分析还表明, 在京津地区的初夏, 上述亮带区域一般容易出现在2.5 km左右的高度处。

An Automatic Identification Algorithm for the Removal of Bright Band from Reflectivity of CINRAD/SA

An automatic identification algorithm for the removal of bright band from radar reflectivity data is introduced and tested by case analysis and pre-operational run. The algorithm has good results of removal of bright band based on the radar reflectivity data which have been transformed into a regular 3-dimensional Cartesian grid while configuration and running setup are relative simple. Firstly, the algorithm performs the recognition of vertical bright band profile. The algorithm considers each point independently, and attempts to find the evidence of a bright band by using a pattern recognition approach. Based on the typical reflectivity of the vertical profile attributes in the bright band deducted from statistics, a set of ideal templates of radar reflectivity vertical profiles is established, which can resemble the actual radar reflectivity characteristics in different bright band areas at the utmost. For each horizontal grid the algorithm attempts to couple the templates to the measured vertical profile of radar reflectivity. The couple function is obtained by computing the slope and the correlation of the least squares regression between the template and the measured profile. Secondly, the algorithm finds continuous regions of bright band. Regions of bright band may be further recognized by the horizontal uniformity of the signature. The algorithm attempts to use the goodness-of-fit and height-of-best-fit field computed above to gain further clues of a bright band. A grid kernel-based approach is used, in which a kernel is composed of 5×5 grids in the field. For the grids in the kernel, the algorithm counts the number of difference between contiguous goodness-of-fit values exceeding a given threshold or the difference in the height-of-best-fit values for contiguous points exceeding a given threshold. A high interested value is assigned to points with low numbers, since this indicates uniformity. So bright band areas are identified as those fields with high interested values. Finally, the algorithm removes the bright band signature. It removes the continuous regions of bright band by correcting the reflectivity within the bright band, which employs linear or non-linear interpolation methodology based on the measured reflectivity values in the profile above the band and the measured values below the band. Some thunderstorm cases analysis and pre-operational running tests suggest that the algorithm is able to successfully identify and remove most bright band regions, which influences quantitative precipitation estimation, based on CINRAD/SA radar reflectivity located at Tanggu of Tianjin. However, the algorithm has less effect on actual storms reflectivity data. Results of the analysis also demonstrate that the bright band regions usually appear at 2.5 km over Beijing and Tianjin in early summer. Because the algorithm is relatively easy to configure and run, it is appropriate to use in real-time or operational mode. Of course the algorithm has limitations and needs to be continuously improved for actual operational applications.