精细化网格的降水预报解析方法对比

Comparison of Parsing Methods for Fine Grid Precipitation Forecasting

以欧洲中期天气预报中心(ECMWF)精细化网格降水预报资料、自动气象站观测资料、中国气象科学数据信息网提供的三源融合降水产品为基础,研究最邻近点法、反距离权重法、最小二乘法和双线性插值法在格点降水向站点以及格点降水插值到更细网格中的优缺点和适用性。在模式预报相对准确精细的基础上,得出以下结论:1)最邻近点法在格点向站点解析方面预报技巧评分表现最好,其次为反距离权重法;2)将降水场变换到更高分辨率的格点降水,4种解析结果和原始场均存在一定的差异,其中最小二乘法的空间形态失真最为严重;3)反距离权重法在较粗网格向更细网格变换中的平均TS评分表现最好,而最邻近点法的ETS评分最高。

Based on ECMWF fine grid point precipitation forecast data and the three source fusion precipitation products by China Meteorological Science Data Information Network, we analyzed and compared the disadvantages and applicability of data conversion from grid points to station and grid point to more high resolution grid using different methods, including nearest neighbor method, inverse distance weighting, least square method and bilinear interpolation method. On the basis of the relative accuracy and fineness of the model forecast, the results show that: 1) The nearest neighbor method had the highest forecast skills on the data conversion from grid points to station, the second is the inverse distance weighting method;2) On the precipitation field parsing to more high resolution grid, four different methods and the original fields have some differences, but the least square method produces the most serious space pattern distortion;3) Average TS score of the inverse distance weighting method is the best one on parsing rough network grid to fine network grid, while the nearest neighbor method has a better ETS score.