华南暴雨区域集合预报中不同同化方案的影响试验研究

基于全球集合预报系统(GEFS)资料，利用WRF中尺度模式及GEFS动力降尺度获取区域集合预报初值场，通过对同化后的分析场进行模式积分实现华南前汛期区域集合预报。对2019年6月10日的一次华南前汛期暴雨过程进行不同同化方案的试验:混合同化(Hybrid)、三维变分(3Dvar)、集合卡尔曼滤波(EnKF)和对比试验(Ctrl)四组试验的对比分析，探讨具有不同背景误差协方差矩阵的同化方案对区域集合预报集合扰动和集合离散随时间演变特征的影响，评估不同试验的降水模拟效果。(1) Hybrid对模式初始场有较好的改善作用，而3DVar和EnKF对初始场的改善作用不明显。(2) 对风场、温度场和湿度场，在前期预报中Hybrid的预报误差小于3DVar和EnKF，在中后期的预报中，3DVar和EnKF的预报误差得到改善，且好于Hybrid。同样，集合扰动能量，Hybrid和Ctrl在前期预报发展好于3DVar和EnKF，而在中后期的预报3DVar和EnKF好于Hybrid和Ctrl。(3) 从24 h累积降水评分中，整体上同化试验好于Ctrl，3DVar和EnKF好于Hybrid，且3DVar对大中雨级别的降水评分较好，而EnKF对暴雨以上级别的降水评分较好。(4) 对于集合统计检验分析，同化试验的AUC值都大于Ctrl的AUC值，24 h累积降水量阈值在10~100 mm的AUC值，3DVar最好；而125 mm阈值的AUC值，EnKF最好。 

EXPERIMENTAL STUDY ON INFLUENCE OF DIFFERENT ASSIMILATION SCHEMES ON ENSEMBLE FORECAST OF TORRENTIAL RAIN IN SOUTH CHINA

Based on the Global Ensemble Forecast System data, the present study uses the WRF model and the GEFS dynamic downscaling method to obtain the regional ensemble forecast initial states. Moreover, the assimilated analysis field is integrated to achieve the reginal ensemble forecast for a precipitation event during annually first rainy season in South China. Four tests, namely Hybrid, 3DVar, EnKF, and Ctrl, are carried out for a torrential rain process on June 10, 2019 in South China. We also explore the evolution characteristics of the ensemble disturbance and ensemble spread for the assimilation schemes with difference background error covariance matrices and evaluate the precipitation simulation performance of different tests. The results show that: (1) Hybrid can improve the initial field of the model, while 3DVar and EnKF failed. (2) For wind, temperature, and relative humidity, the forecast error of Hybrid in the early forecast is less than that of 3DVar and EnKF. In the middle and late forecast, the forecast error of 3DVar and EnKF is reduced and is smaller than that of Hybrid. As for the ensemble disturbance energy, Hybrid and Ctrl are better than 3DVar and EnKF in the early forecast, and 3DVar and EnKF are better than Hybrid and Ctrl in the middle and late forecasts. (3) According to the 24-hour cumulative precipitation scores, the assimilation test is better than Ctrl, 3DVar and EnKF are better than Hybrid; 3DVar scores the best in heavy and moderate rainfall and EnKF scores the best in torrential rain and above. (4) For 24-hour cumulative precipitation ensemble statistical analysis, the area under the curve value of the assimilation test is greater than that of Ctrl; 3DVar performs the best in the 10mm~100mm cumulative precipitation threshold and EnKF performs the best in the 125mm cumulative precipitation threshold.