CORRECTION METHOD BASED ON NEIGHBORHOOD OPTIMAL PROBABILITY FOR HOURLY PRECIPITATION FORECAST FROM GRAPES RAPID UPDATING CYCLE ASSIMILATION AND FORECASTING SYSTEM
-
摘要: 针对高分辨率数值天气预报的时空不确定性, 利用邻域最优概率方法对华南区域GRAPES快速更新循环同化预报系统的24 h预报进行逐时降水订正和检验评估。结果表明: (1)邻域法能改善模式降水预报的空间不确定性, 最优邻域半径随降水等级增加而减小, 强降水的最优邻域半径约为60 km; (2)通过引入时间滞后因子, 可进一步改善模式不同时间起报的不确定性, 结合Brier评分确定了时间滞后窗为4 h; (3)提出基于邻域最优概率阈值的降雨进行分级订正方法, 有效提升了降水客观预报能力, 晴雨预报较模式全部为正技巧, TS评分达到0.89以上, 总体提升幅度约5.3%;强降水预报同样均为正技巧, TS评分呈先降后升趋势, 在12 h时效前后预报效果最优, 进一步提升了GRAPES快速更新循环同化预报系统的业务预报水平。Abstract: To tackle the spatial-temporal uncertainty of high-resolution numerical models, in this paper, the neighborhood optimal probability method is developed to improve the hourly precipitation forecast in the24 h forecasts from the GRAPES rapid updating cycle assimilation and forecasting system (GRAPES_GZ_R) for southern China. Resultsshow that: (1) The spatial neighborhood probability can help reduce the spatial uncertainty of precipitation forecast from the GRAPES_GZ_R. The optimal neighborhood radius decreases with the increase of precipitation grade, and it is about 60 km for heavy rainfall prediction. (2) The time-lagged factor can help further reduce the uncertainty of forecasting initialized at different time, and the time-lagged window is determined to be 4 h by using the Brier score. (3) The precipitation classification correction method based on the optimal probability threshold of neighborhood is proposed, which effectively improves the objective precipitation forecast. Compared with previous forecasts, the present clear-rainy forecast shows all positive skills with the TS score above 0.89 and the overall improvement being about 5.3%. The heavy rain forecast also shows all positive skills; corresponding TS score decreases first and then increases, and finally achieves the best effect at about 12 h, which further improves the operational application capacity of the GRAPES_GZ_R.
-
Key words:
- neighborhood method /
- time-lagged /
- optimal probability /
- precipitation forecasts
-
图 7 2020年5月22日03时降水实况(a), GRAPES_GZ_R模式不同时次起报的降水(b~f)，领域概率法2020年5月21日14时起报的强降水概率（g）和小时降水量（h)
-
[1] 童文雪, 李刚, 孙娟珍. 针对对流降水预报的BJ-RUC系统1小时更新循环方案研究[J]. 热带气象学报, 2018, 34(2): 177-187. [2] 陈子通, 黄燕燕, 万齐林, 等. 快速更新循环同化预报系统的汛期试验与分析[J]. 热带气象学报, 2010, 26(1): 49-54. [3] 陈子通, 徐道生, 戴光丰, 等. 热带高分辨率模式(TRAMS-V3.0)技术方案及其系统预报性能[J]. 热带气象学报, 2020, 36(4): 444-454. [4] CHAKRABORTY A. The skill of ECMWF medium-range forecasts during the year of tropical convection 2008[J]. Mon Wea Rev, 2008, 138(10): 3 787-3 805. [5] 李俊, 杜钧, 陈超君. "频率匹配法"在集合降水预报中的应用研究[J]. 气象, 2015, 41(6): 674-684. [6] 代刊, 朱跃建, 毕宝贵. 集合模式定量降水预报的统计后处理技术研究综述[J]. 气象学报, 2018, 76(4): 493-510. [7] 智协飞, 吕游. 基于频率匹配法的中国降水多模式预报订正研究[J]. 大气科学学报, 2019, 42(6): 814-823. [8] 智协飞, 赵忱. 基于集合成员订正的强降水多模式集成预报[J]. 应用气象学报, 2020, 31(3): 303-314. [9] THEIS S E, HENSE A, DAMRATH U. Probabilistic precipitation forecasts from a deterministic model: A pragmatic approach[J]. Meteorological Applications, 2005, 12(3): 257-268. [10] 潘留杰, 张宏芳, 陈小婷, 等. 基于邻域法的高分辨率模式降水的预报能力分析[J]. 热带气象学报, 2015, 31(5): 632-642. [11] 马申佳, 陈超辉, 智协飞, 等. 基于时空不确定性的对流尺度集合预报效果评估检验[J]. 气象学报, 2018, 76(4): 578-589. [12] 刘雪晴, 陈静, 陈法敬, 等. 降水邻域集合概率方法尺度敏感性试验[J]. 大气科学, 2020, 44(2): 282-296. [13] 牛若芸. 基于最优概率的中期延伸期过程累计降水量分级订正预报[J]. 热带气象学报, 2019, 35(6): 721-731. [14] 伍红雨, 邹燕, 刘尉. 广东区域性暴雨过程的定量化评估及气候特征[J]. 应用气象学报, 2019, 30(2): 233-244. [15] EBERT E E. Neighborhood verification: A strategy for rewarding close forecasts[J]. Weat Forecasting, 2009, 24(6): 1 498-1 510. -
下载:
点击查看大图
图(8)
计量
- 文章访问数: 289
- HTML全文浏览量: 49
- PDF下载量: 30
- 被引次数: 0
粤公网安备 4401069904700003号