快速动态分级法进行雷达定量估测降水方法研究

在使用雷达资料进行降水估测(QPE)时，对动态分级Z-I关系算法进行大幅度优化后，发现当时间精度较高时，使用CTF2判别函数所得QPE结果误差较大。提出了用同时满足偏差为0且方差最小作为最优解判别式的快速动态分级法(FDC)。通过留出法进行交叉验证，结果表明，快速动态分级法得到的结果整体相关性好，能较好地表现出降水的时空分布特征，反演降水的平均误差为-0.13 mm/h，计算速度快，具有较高的业务应用价值。 

AN APPROACH FOR RADAR QUANTITATIVE PRECIPITATION ESTIMATE BASED ON FAST DYNAMIC CATEGORICAL METHOD

The outcome of Radar Quantitative Precipitation Estimate (QPE) obtained by CTF2 discriminant deviates much in optimizing dynamic categorical Z-I relationship with high time accuracy. The Fast Dynamic Categorical (FDC) algorithm based on dynamic categorical Z-I relationship is proposed to satisfy both the 0 bias and the minimum variance. It shows that the correlation coefficients of predicted and actual rainfall resulted from this method are higher and the spatial and temporal distribution of precipitation is better embodied by hold-out and cross-validation, compared with the fast dynamic categorical method. With fast calculating speed and average error of inversion precipitation of -0.13mm h-1, this approach is highly valuable in precipitation forecast operation.