漂移克里金方法在雷达和雨量计联合估测降水中的应用

Merging radar and rain gauge data using Kriging with external drift (KED) for quantitative precipitation estimation.

文中介绍了一种新的融合雷达和雨量计数据开展定量估测降水研究的空间信息统计学方法—Kriging with external drift(KED)方法。该方法能很好地融合高精度、低时空分辨率的雨量计数据和低精度、高时空分辨率的雷达数据进行插值。通过变异函数描述降水场的空间结构信息,能够充分利用数据间的空间相关性,来改进估测精度和提高处理速度。利用其优良的数学特性,以期在定量估测降水业务研究上进行新的探索和尝试。选用湖南省有代表意义的3次降水过程资料,通过雷达直接估测降水(RAD)、变分校准(VAR)以及KED3种方法,分别与雨量计测量值进行对比分析,选用代表站进行交叉验证结果均表明:RAD的均方差、绝对误差、相对误差最大,VAR次之,而KED最小。KED估测的结果与雨量计测量降水最为接近,估测效果最好;3种方法与雨量计实测值计算一定范围的误差频率,KED估测值具有最小的均方差和最小的标准差,且误差分布相对集中在0值附近,斜度和峰度最佳,试验证明该方法不仅能提高降水估测精度,且优于其他方法,VAR均方差次之,RAD均方差效果相对较差。联合雷达、雨量计估测降水的实质是把雷达估测值与雨量计测量的结果相融合,以雨量计来校准雷达估测值,保留了雷达探测到降水的中、小尺度精细特征。校准后的雨量场数值接近雨量计测值,而且能够准确反映雷达测得的降水分布形式。

A new spatial information statistical method, Kriging with external drift （KED）, which merges radar and rain gauge data to make quantitative precipitation estimation, is introduced and analyzed. The essential of merging radar and rain gauge data to estimate rainfall is to calibrate the radar with rain gauge data and syncretize the result of the rain gauge records into the radar detection while keeping the meso and small scale features of radar data. High accuracy, low tempo spatial resolution rain gauge data is successfully combined with the low accuracy, high tempo-spatial resolution radar data using the NED interpolation. The covarianee function is used to reflect the spatial variance structure, and the spatial continuity of data is fully considered. As the spatial structure of KED predicted rainfall field is obtained by using variogram, the estimation precision can be improved and the processing speed can be accelerated by making full use of the spatial relation among the data. The KED method is expected to advance operational quantitative precipitation estimation. Three methods, namely the radar-based precipitation estimation （RAD）, the variation adjustment precipitation estimation （VAR）, and the NED estimation, are compared with each other and verified against the rain gauge data for three representative rainfall cases in Hunan Province. The results show that the mean-square deviation, absolute error and relative error from RAD are bigger than those from VAR, and those of KED are the smallest. The results from KED agree well with the rain gauge data. Error frequency calculations for the three methods and the rain gauge data show that KED has the smallest average error and standard deviation, and the error distribution for KED is located near 0. Moreover, the skewness and kurtosis of KED are the best, while those of VAR take the second place and RAD performs the worst. The magnitude of the KED calibrated precipitation field is close to the rain gauge record, meanwhile the precipitation distribution pattern detected by radar was well retained.