Short-term quantitative precipitation forecast experiments based on blending of nowcasting with numerical weather prediction.
Short-term quantitative precipitation forecast experiments based on blending of nowcasting with numerical weather prediction.
复制标题
作者：
Yeung Linus
为了克服中尺度数值预报模式（NWP）在对流尺度上的短时定量降水预报（QPF）和雷达实时降水预报技术的不足，提出了一种融合两种预报源的QPF方案。对数值预报模式QPF应用相位校正技术，通过与雷达反射率和自动雨量计资料的定量降水估计（QPE）相比较，校正它们的系统定时或定位误差。将强度校准技术应用于相位校正的模型QPF以调整其强度，使得校准的强度分布可以匹配从QPE计算的实际分布。第三，混合的QPF是作为基于雷达的临近预报和相位校正的强度调整模型QPF之间的加权平均值获得的，其中根据参数化的双曲正切函数在不同的前置时间分配权重。这些参数可以根据一个指数动态地变化，该指数可以量化相对于模式QPF的外推临近预报的技能，并且发现这种指数对于不同的降水类型是不同的。一般来说，在几个小时内给予临近预报较高的权重。在大约3小时或更长的时间，在2011年和2012年的5次典型暴雨过程中，采用该混合方案对京津地区进行了模拟，结果表明：QPF检验结果表明，在0-6的预报范围内，混合QPF的预报性能得到了显著的改善，总体预报水平普遍高于雷达临近预报和模式QPF单独预报，因此，混合QPF预报方案在实际应用中具有很好的应用前景。
In order to overcome the deficiency of the quantitative precipitation forecast(QPF) by a mesoscale numerical weather prediction(NWP) model for the very short range at convective scales and the vanishingly low skill of radar-based rainfall now-castings beyond the first couple of hours,a QPF scheme to blend these two sources has been developed.It consists of the following three major steps.Firstly,a phase correction technique is applied to the NWP model QPF to correct their systematic timing or location errors by comparing to the corresponding quantitative precipitation estimation(QPE) derived from the radar reflectivity and automatic raingauge data.Secondly,an intensity calibration technique is applied to the phase-corrected model QPF to adjust its intensity so that the calibrated intensity distribution can match the actual distribution calculated from QPE. Thirdly,the blended QPF is obtained as a weighted average between the radar-based nowcast and the phase-corrected intensity-adjusted model QPF with the weights at the different lead times assigned according to a parameterized hyperbolic tangent function. Such parameters can vary dynamically according to an index that could quantify the skill of extrapolative nowcast relative to the model QPF and,such an index is found to be different for the different precipitation types.In general,higher weights are given to nowcast in the couple of hours.At about 3 hours and beyond,higher weights are given to model QPF as the now-cast skill drops.This blending scheme was applied to five typical heavy rain cases in 2011 and two in 2012 over the Beijing-Tianjin -Hebei region with about eighty tests done.The QPF verification results indicated that the performance of the blended QPF was significantly improved in the 0-6 forecast range with the overall forecast skill generally higher than either radar-based nowcast or model QPF taken individually.Given such encouraging results,the present blending scheme is expected to be promising as an effective tool when deployed for operation.