Correction of ECMWF ensemble average precipitation forecast using two objective precipitation statistical methods

doi:10.11755/j.issn.1006-7639(2024)-02-0293

Abstract

Abstract:

Improving the accuracy of precipitation level forecast is helpful to optimize disaster warning and decision support. Based on the precipitation observation data in the time interval of 12 hours from January 2018 to January 2021 in Shandong Province and the ensemble prediction ensemble mean results of the European Centre for Medium-Range Weather Forecasting, the precipitation level forecast for 12 hours interval within 72 hours are statistically revised. Then, the effects of the original forecast of ECMWF ensemble mean precipitation forecast interpolation (EC_EPEM), the Model Output Statistics (MOS) prediction based on the EC_EPEM (EC_EPEM_MOS) and the Optimal Threat Score (OTS) prediction (EC_EPEM_OTS) are compared, and the improving effects of two statistical correction methods on precipitation level with the time interval of 12 hours prediction of the ECMWF ensemble forecast are discussed. The results indicate that the EC_EPEM_MOS has the best performance on the relatively smaller precipitation grades, while its correction effect is relatively poor for higher grades, even slightly lower than the EC_EPEM. The correction effect of the EC_EPEM_OTS is only lower than the EC_EPEM_MOS for 0.1 and 10.0 mm precipitation grades, and for the other grades it is optimal, especially for the larger grades, its correction effect is more obvious. The EC_EPEM_OTS has the best correction effect from 12 to 72 hours for both 50 mm and 100 mm precipitation grades, because the EC_EPEM_OTS increases the correction coefficient for a slightly larger grade, resulting in a low false report rate for large grades. At the same time, using a smaller correction coefficient for large precipitation also reduces the false report rate. The EC_EPEM_MOS is best in most parts of Shandong Province except for the mountains area in the middle parts for short prediction period and smaller precipitation, while the EC_EPEM_OTS is the best in the mountains area. For above medium grade, especially large precipitation, the EC_EPEM_OTS is the best in most areas of Shandong Province. The EC_EPEM_MOS correction prediction effectively reduces the problem of empty report of the EC_EPEM. The correction effect of the EC_EPEM_OTS is the best, and the rainfall area is closer to the observations in the processes of large-scale heavy rainfalls, and the overall distribution of precipitation is better grasped.

Key words: objective statistical method, precipitation forecast, optimal threat score correction, Shandong Province

摘要：

提升降水量级预报精度，有助于优化灾害预警与决策支持。选取2018年1月1日至2021年1月山东省逐12 h降水观测数据和欧洲中期天气预报中心（the European Centre for Medium-Range Weather Forecasting，ECMWF）的集合预报集合平均（Ensemble Prediction Ensemble Mean，EPEM）结果进行72 h内逐12 h降水量级预报统计订正，然后对比ECMWF集合平均降水预报插值的原始预报（EC_EPEM）、基于EC_EPEM的输出统计（Model Output Statistics，MOS）预报（EC_EPEM_MOS）、利用最优TS（Threat Score）评分订正（Optimal Threat Score，OTS）预报（EC_EPEM_OTS）的效果。结果表明：EC_EPEM_MOS在较小量级上表现最优，但在大量级上订正效果稍差，甚至略低于EC_EPEM；EC_EPEM_OTS仅在0.1、10 mm量级上低于EC_EPEM_MOS，其他量级均为最优，尤其在较大量级上订正效果更明显。在50、100 mm大量级上，EC_EPEM_OTS在12~72 h时效订正效果均最优，这是由于EC_EPEM_OTS在稍大量级上提高订正系数使得大量级降水漏报率减小，同时对大量级降水使用较小订正系数也减小了空报率。在较小量级降水中短期预报时效除了山东中部山区外EC_EPEM_MOS表现最佳，山区EC_EPEM_OTS最佳；中等以上量级、尤其较大量级降水，山东大部分地区EC_EPEM_OTS表现最佳。EC_EPEM_MOS订正预报有效地减小了EC_EPEM的空报问题。EC_EPEM_OTS的订正效果最佳，在大范围强降雨过程中与实况降雨分布更为接近，降水总体分布把握较好。

关键词: 客观统计方法, 降水预报, 最优TS评分订正, 山东

CLC Number:

P456.8

JIAO Yang, ZHENG Lina, ZHANG Yongjing, SU Yi. Correction of ECMWF ensemble average precipitation forecast using two objective precipitation statistical methods[J]. Journal of Arid Meteorology, 2024, 42(2): 293-304.

焦洋, 郑丽娜, 张永婧, 苏轶. 两种降水客观统计方法对ECMWF集合平均降水预报的订正研究[J]. 干旱气象, 2024, 42(2): 293-304.

Figures/Tables 10

Fig.1 Variation of accuracy of precipitation forecast in 12 hours interval with the change of running training period for prediction period of 12 to 72 hours for two correction algorithms （a） EC_EPEM_MOS，（b） EC_EPEM_OTS

Fig.2 The threat score （a）， empty report rate （b）， missing report rate （c） and forecast deviation （d） of 12 h cumulative precipitation prediction of EC_EPEM，EC_EPEM_MOS and EC_EPEM_OTS for 12 h prediction period during 2019-2020

Fig.3 The variation of threat score of prediction with the time interval of 12 h for precipitation with different grades of EC_EPEM，EC_EPEM_MOS and EC_EPEM_OTS for prediction period of 12 to 72 h during 2019-2020 （a） 0.1 mm，（b） 10 mm，（c） 25 mm，（d） 50 mm，（e） 100 mm

Tab.1 The threat scores of EC_EPEM， EC_EPEM_MOS and EC_EPEM_OTS for prediction period of 12 to 72 h with the time interval of 12 h from June to September during 2019-2020

降水量级/mm		预报时效/h
降水量级/mm		12	24	36	48	60	72
0.1~9.9	EC_EPEM	0.282	0.265	0.248	0.231	0.214	0.190
	EC_EPEM_MOS	0.303	0.285	0.268	0.246	0.229	0.198
	EC_EPEM_OTS	0.301	0.284	0.267	0.250	0.233	0.205
10.0~24.9	EC_EPEM	0.200	0.192	0.177	0.166	0.154	0.139
	EC_EPEM_MOS	0.205	0.193	0.183	0.175	0.167	0.159
	EC_EPEM_OTS	0.206	0.197	0.188	0.180	0.172	0.155
25.0~49.9	EC_EPEM	0.141	0.130	0.119	0.108	0.093	0.076
	EC_EPEM_MOS	0.132	0.119	0.105	0.092	0.080	0.058
	EC_EPEM_OTS	0.166	0.161	0.150	0.137	0.128	0.118
50.0~99.9	EC_EPEM	0.157	0.145	0.137	0.121	0.112	0.094
	EC_EPEM_MOS	0.140	0.127	0.113	0.100	0.087	0.066
	EC_EPEM_OTS	0.168	0.156	0.145	0.133	0.120	0.108

Fig.4 The spatial distribution of optimal products with high threat scores for 0.1 to 9.9 mm grades precipitation prediction with the time interval of 12 h for different prediction periods from June to September during 2019-2020 （a） 12 h，（b） 24 h，（c） 36 h，（d） 48 h，（e） 60 h，（f） 72 h

Fig.5 The spatial distribution of optimal products with high threat scores for 10.0 to 24.9 mm grades precipitation prediction with the time interval of 12 h for different prediction periods from June to September during 2019-2020 （a） 12 h，（b） 24 h，（c） 36 h，（d） 48 h，（e） 60 h，（f） 72 h

Fig.6 The spatial distribution of optimal products with high threat scores for 25.0 to 49.9 mm grades precipitation prediction with the time interval of 12 h for different prediction periods from June to September during 2019-2020 （a） 12 h，（b） 24 h，（c） 36 h，（d） 48 h，（e） 60 h，（f） 72 h

Fig.7 The spatial distribution of optimal products with high threat scores for 50.0 to 99.9 mm grades precipitation prediction with the time interval of 12 h for different prediction periods from June to September during 2019-2020 （a） 12 h，（b） 24 h，（c） 36 h，（d） 48 h，（e） 60 h，（f） 72 h

Fig.8 The spatial distribution of observations （a）， EC_EPEM （b）， EC_EPEM_MOS （c）， EC_EPEM_OTS （d） precipitation grades and optimal forecasting products （e） from 08：00 to 20：00 on July 22， 2020

Fig.9 The spatial distribution of observations （a）， EC_EPEM （b）， EC_EPEM_MOS （c）， EC_EPEM_OTS （d） precipitation grades and optimal forecasting products （e） from 20：00 on September 7 to 08：00 September 8， 2020

References 31

[1]	邓国, 龚建东, 邓莲堂, 等, 2010. 国家级区域集合预报系统研发和性能检验[J]. 应用气象学报, 21(5): 513-523.
[2]	顾震潮, 1958. 作为初值问题的高空大尺度天气数值预报与由地面温压场历史演变作预报的等值性[J]. 科学通报, 3(1): 19-20.
[3]	韩念霏, 杨璐, 陈明轩, 等, 2022. 京津冀站点风温湿要素的机器学习订正方法[J]. 应用气象学报, 33(4): 489-500.
[4]	李佰平, 智协飞, 2012. ECMWF模式地面气温预报的四种误差订正方法的比较研究[J]. 气象, 38(8): 897-902.
[5]	刘还珠, 赵声蓉, 陆志善, 等, 2004. 国家气象中心气象要素的客观预报—MOS系统[J]. 应用气象学报, 15(2): 181-191.
[6]	林春泽, 智协飞, 韩艳, 等, 2009. 基于TIGGE资料的地面气温多模式超级集合预报[J]. 应用气象学报, 20(6): 706-712.
[7]	麻巨慧, 朱跃建, 王盘兴, 等, 2011. NCEP、ECMWF及CMC全球集合预报业务系统发展综述[J]. 大气科学学报, 34(3): 370-380.
[8]	漆梁波, 曹晓岗, 夏立, 等, 2007. 上海区域要素客观预报方法效果检验[J]. 气象, 33(9): 9-18.
[9]	任宏利, 张培群, 李维京, 等, 2006. 基于多个参考态更新的动力相似预报方法及应用[J]. 物理学报, 55(8): 4 388-4 396.
[10]	盛春岩, 范苏丹, 荣艳敏, 等, 2020. 几种气温客观预报方法对比及最优集成预报研究[J]. 气象, 46(10): 1 351-1 361.
[11]	王雨, 闫之辉, 2007. 降水检验方案变化对降水检验评估效果的影响分析[J]. 气象, 33(12): 53-61.
[12]	吴启树, 韩美, 郭弘, 等, 2016. MOS温度预报中最优训练期方案[J]. 应用气象学报, 27(4): 426-434.
[13]	吴启树, 韩美, 刘铭, 等, 2017. 基于评分最优化的模式降水预报订正算法对比[J]. 应用气象学报, 28(3): 306-317.
[14]	徐姝, 熊明明, 王颖, 等, 2018. 改进的ECMWF集合预报融合产品在海河流域的检验与分析[J]. 气象与环境科学, 41(4): 41-46.
[15]	薛谌彬, 陈娴, 张瑛, 等, 2019. ECMWF高分辨率模式2 m温度预报误差订正方法研究[J]. 气象, 45(6): 831-842.
[16]	杨杰, 龚志强, 赵俊虎, 等, 2014. 基于模式误差分布特征的中国夏季旱涝预测可信度研究[J]. 物理学报, 63(14): 427-438.
[17]	杨璐, 王晓丽, 宋林烨, 等, 2023. 基于阵风系数模型的百米级阵风客观预报算法研究[J]. 气象学报, 81(1): 94-109.
[18]	赵华生, 黄小燕, 黄颖, 2018. ECMWF集合预报产品在广西暴雨预报中的释用[J]. 应用气象学报, 29(3): 344-353.
[19]	赵瑞霞, 代刊, 金荣花, 等, 2020. OTS、MOS和OMOS方法及其优化组合应用于72 h内逐3 h降水预报的试验分析研究[J]. 气象, 46(3): 420-428.
[20]	赵声蓉, 赵翠光, 邵明轩, 2009. 事件概率回归估计与降水等级预报[J]. 应用气象学报, 20(5): 521-529.
[21]	赵声蓉, 赵翠光, 赵瑞霞, 等, 2012. 我国精细化客观气象要素预报进展[J]. 气象科技进展, 2(5): 12-21.
[22]	郑志海, 任宏利, 黄建平, 2009. 基于季节气候可预报分量的相似误差订正方法和数值实验[J]. 物理学报, 58(10): 7 359-7 367.
[23]	智协飞, 陈雯, 2010. THORPEX国际科学研究新进展[J]. 大气科学学报, 33(4): 504-511.
[24]	智协飞, 季晓东, 张璟, 等, 2013. 基于TIGGE资料的地面气温和降水的多模式集成预报[J]. 大气科学学报, 36(3): 257-266.
[25]	智协飞, 赵忱, 2020. 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, 31(3): 303-314.
[26]	BARNETT T P, PREISENDORFER R W, 2010. Multi-field analog prediction of short-term climate fluctuations using a climate state vector[J]. Journal of the Atmospheric Sciences, 35(10): 1 771-1 787.
[27]	CARTER G M, DALLAVALLE J P, GLAHN H R, 1989. Statistical forecasts based on the National Meteorological Center’s numerical weather prediction system[J]. Weather and Forecasting, 4(3): 401-412.
[28]	GLAHN H R, LOWRY D A, 1972. The use of model output statistics (MOS) in objective weather forecasting[J]. Journal of Applied Meteorology and Climatology, 11(8): 1 203-1 211.
[29]	HAGEDORN R, BUIZZA R, HAMILL T M, et al, 2012. Comparing TIGGE multi-model forecasts with reforecast-calibrated ECMWF ensemble forecasts[J]. Quarterly Journal of the Royal Meteorological Society, 138(668): 1 814-1 827.
[30]	VISLOCKY R L, FRITSCH J M, 1995. Improved model output statistics forecasts through model consensus[J]. Bulletin of the American Meteorological Society, 76(7): 1 157-1 164.
[31]	World Meteorological Organization, 2012. Guidelines on ensemble prediction systems and forecasting[R]. WMO-No.1091, Geneva.