ETKF协方差膨胀方案对WRFDA混合同化及预报的影响

Impact of Different Covariance Inflation Schemes of ETKF on WRFDA Hybrid Assimilation and Forecast

为讨论集合变换卡尔曼滤波不同协方差膨胀方案对集合-变分混合同化及预报的影响,开展了中国中东部区域一次连续大范围降水过程的连续10天的循环同化和预报试验。结果表明:4种不同的协方差膨胀方案相对于无协方差膨胀的方案,均有效地提高了混合同化和预报的效果。将同化时次之前所有膨胀系数平均值作为新膨胀系数的方案,同化和预报的效果均是最差的;其他3种协方差膨胀方案效果较为接近略有区别:对于风场,将预报误差协方差投影到集合子空间的方案和采用平均新息协方差信息的方案表现较好;对于温度场、湿度场和降水预报,采用平均新息协方差信息的方案和采用了同化时次前两次集合预报比率的方案较好。

To study the impact of different ETKF（Ensemble Transform Kalman Filter）covariance inflation schemes on Hybrid data assimilation and forecast,five experiments are conducted over most China area from 10 to 20 July 2011. These five experiments include four experiments with different covariance inflation schemes and one experiment without any covariance inflation schemes. The results show that：All the experiments with covariance inflation perform better than experiment without covariance inflation. The inflation sheme which averages the inflation factors performs worst;The performance of the other three experiments with covariance inflation shows similar results,but differences still exsit：For wind,the scheme which uses projection factor in ensemble subspace and the scheme which uses averaged innovation value show the better results than the other two. For temperature,humidity and preciptation,the sheme which uses the ratio of the spread between previous and current cycle and the scheme uses averaged innovation value perform the best.