基于CMA-MESO的分级短时强降水概率预报方法研究

doi:10.11755/j.issn.1006-7639(2022)-04-0700

摘要/Abstract

摘要：

随着预报服务需求不断增长和预报内容日趋精细化，仅针对20 mm·h^-1以上的短时强降水预报已不能完全满足业务需要，开展不同雨强等级的短时强降水预报方法研究显得十分必要。利用2016—2019年6—8月中国南方9省1市的国家及区域气象站共51 355站次短时强降水样本，将雨强R分为4个等级：20≤R<30 mm·h^-1、30≤R<50 mm·h^-1、50≤R<80 mm·h^-1及R≥80 mm·h^-1(分别对应I、Ⅱ、Ⅲ、IV级)。将各级样本与同时段CMA-MESO(China Meteorological Administration mesoscale model)数值预报模式初始场进行时空匹配，提取22个相关物理量建立数据集并进行百分位值统计;利用XGBoost(extreme gradient boosting)机器学习方法对物理量进行重要性排序以确定权重系数;应用连续概率预报方法，选用升、降半岭函数作为隶属函数，建立不同等级短时强降水概率预报模型。运用该模型在2020年汛期进行实时业务预报，并对湖北省2020年6—8月15次大暴雨过程0~36 h预报时效的逐小时不同等级短时强降水概率预报产品进行检验，结果表明：I级概率预报产品60%阈值的TS评分(0.145)最好，对应命中率为55.7%;Ⅱ级概率预报产品65%阈值的TS评分(0.083)最好，对应命中率为39.1%;Ⅲ级概率预报产品70%阈值的TS评分(0.03)最好，对应命中率为21.7%;IV级概率预报产品80%阈值的TS评分(0.005)最好，对应命中率为5.8%。对不同等级雨强个例对比检验表明，各级概率预报产品对CMA-MESO模式在同时次不同等级短时强降水预报上均有较好的订正作用。对3次强降水过程逐小时预报检验表明，I级概率预报产品命中率为40%~80%，空报率为50%~90%，预报时效达36 h，普遍优于同时次CMA-MESO降水量预报。本研究对不同等级短时强降水分型建模并在实际预报中有较好的参考性，能够对CMA-MESO的降水预报起到订正作用。

关键词: CMA-MESO, 分级降水, 短时强降水, 概率预报

Abstract:

With continuous growth of forecast service demand and increasingly refined forecast content, the forecast of short-term heavy precipitation above 20 mm·h^-1 can not meet the forecast service demand fully. It is very necessary to carry out research on forecast methods about short-term heavy precipitation with different rainfall intensity. The 51 355 samples of short-term heavy rainfall from national and regional meteorological stations in nine provinces and one city in southern China from June to August during 2016-2019 were divided into four rainfall grades according to their rainfall intensity (R), namely, I: 20≤R<30 mm·h^-1, II: 30≤R<50 mm·h^-1, III: 50≤R<80 mm·h^-1, and IV: R≥80 mm·h^-1. The samples of all rainfall grades were spatiotemporal matched with the initial field of CMA-MESO (China Meteorological Administration mesoscale model) in the same period, and the percentile statistics were applied to 22 physical quantities extracted from these samples. The XGBoost (extreme gradient boosting) machine learning method was used to rank importance of those 22 physical quantities to determine their weight coefficients. Based on the continuous probability prediction method, the ascending and descending half ridge functions were selected as the membership function, the probability prediction models of short-term heavy precipitation with different rainfall grades were established. The real-time operational prediction was carried out in flood season of 2020 using these prediction medels, and the hourly probability prediction products of short-term heavy precipitation with different rainfall grades for 0-36 h prediction time during 15 heavy rainstorm precesses in Hubei Province from June to August 2020 were tested. The results show that for the grade I probability prediction products, the TS score (0.145) using 60% as threshold works best, with a corresponding hit rate of 55.7%; for the grade II probability prediction products, the TS score (0.083) using 65% as threshold works best, with a corresponding hit rate of 39.1%; for the grade III probability prediction products, the TS score (0.03) using 70% as threshold works best, with a corresponding hit rate of 21.7%; for the grade IV probability prediction products, the TS score (0.005) using 80% as threshold works best, with a corresponding hit rate of 5.8%.The results also suggest that probability prediction products help to correct the CMA-MESO model in predicting short-term heavy precipitation with different rainfall grades at the same time. The hourly prediction test of three heavy precipitation processes shows the hit rate of 40%-80%, the false rate of 50%-90%, and 36 h prediction time for the grade I probability forecast products, which are generally better than CMA-MESO precipitation forecast at the same time. A model was established to forecast short-term heavy precipitation with different grades in this study, and it outperforms existing numerical models and can be a good reference for meteorologists to forecast short-term heavy precipitation and correct precipitation forecast biases in CMA-MESO.

Key words: CMA-MESO, graded precipitation, short-term heavy rain, probability forecast

中图分类号:

钟敏, 肖安, 许冠宇. 基于CMA-MESO的分级短时强降水概率预报方法研究[J]. 干旱气象, 2022, 40(4): 700-709.

ZHONG Min, XIAO An, XU Guanyu. Study on probability forecast method about graded short-term heavy rain based on CMA-MESO[J]. Journal of Arid Meteorology, 2022, 40(4): 700-709.

图/表 9

图1 17个物理量(a)和5个物理量(b)对雨强的重要性得分

Fig.1 The importance score of each physical quantity to rainfall intensity for 17 (a) and 5 (b) physical quantities

表1 不同等级短时强降水样本水汽及层结特征物理量百分位统计

Tab.1 Percentile statistics of physical quantities about water vapor and stratification characteristic during short-term heavy precipitation samples with different grades

物理量	百分位	I级	Ⅱ级	Ⅲ级	IV级
T_d700/℃	75%	11.31	11.29	11.26	10.91
	50%	10.20	10.15	9.99	9.72
	25%	8.83	8.65	8.50	7.44
T_d850/℃	75%	19.73	19.82	19.90	19.91
	50%	18.90	19.00	19.11	18.99
	25%	18.06	18.19	18.33	18.23
T_d925/℃	75%	22.95	23.05	23.05	23.12
	50%	22.02	22.15	22.31	22.21
	25%	21.08	21.26	21.49	21.32
RH₅₀₀/%	75%	93.62	92.57	91.27	93.05
	50%	83.00	80.42	79.00	79.51
	25%	67.71	65.38	65.45	65.00
RH₇₀₀/%	75%	95.00	94.50	93.91	93.07
	50%	89.00	88.00	87.67	84.22
	25%	80.53	79.31	79.67	71.96
RH₈₅₀/%	75%	97.47	97.23	97.41	95.92
	50%	93.54	93.00	93.00	90.13
	25%	87.94	87.18	87.08	84.60
RH₉₂₅/%	75%	95.74	95.31	95.00	93.56
	50%	90.00	89.36	89.71	88.00
	25%	82.38	82.00	81.58	80.68

表2 不同等级短时强降水样本不稳定条件物理量百分位统计

Tab.2 Percentile statistics of physical quantities about unstable condition during short-term heavy precipitation samples with different grades

物理量	百分位	I级	Ⅱ级	Ⅲ级	IV级
BLI/℃	75%	-1.01	-1.15	-1.30	-1.66
	50%	-2.19	-2.63	-2.53	-2.75
	25%	-3.49	-3.65	-3.67	-3.90
CAPE/ (J·kg^-1)	75%	1764	1824	1939	2008
	50%	893	1047	1181	1082
	25%	248	358	487	401
ΔT_850-500/℃	75%	23.19	23.37	23.41	24.51
	50%	22.15	22.30	22.42	23.52
	25%	21.32	21.47	21.61	22.83
K指数/℃	75%	38.27	38.30	39.08	39.09
	50%	37.42	37.48	38.23	38.00
	25%	36.12	36.15	37.00	36.00

表3 不同等级短时强降水样本动力条件物理量百分位统计

Tab.3 Percentile statistics of physical quantities about dynamic condition during short-term heavy precipitation samples with different grades

物理量	百分位	I级	Ⅱ级	Ⅲ级	IV级
Div₈₅₀/(10^-5s^-1)	75%	0.95	0.35	-0.43	-0.90
	50%	-1.17	-1.51	-2.05	-2.11
	25%	-3.71	-3.88	-4.14	-4.32
Div₉₂₅/(10^-5s^-1)	75%	0.72	0.08	-0.92	-1.67
	50%	-1.79	-2.22	-3.29	-4.09
	25%	-5.17	-5.58	-6.36	-7.50
Divq₈₅₀/ (10^-8g·cm^-2·hPa^-1·s^-1)	75%	3.78	0.12	-4.03	-6.89
	50%	-11.70	-16.90	-21.51	-26.94
	25%	-49.37	-54.03	-54.76	-64.70
Divq₉₂₅/ (10^-8g·cm^-2·hPa^-1·s^-1)	75%	2.20	-1.70	-6.95	-9.55
	50%	-16.59	-25.19	-39.51	-42.59
	25%	-68.42	-83.07	-95.62	-108.12
V₇₀₀/(m·s^-1)	75%	6.58	6.30	6.51	6.09
	50%	4.19	3.94	3.64	3.24
	25%	2.01	1.98	1.81	1.56
V₈₅₀/(m·s^-1)	75%	5.94	5.87	5.39	6.07
	50%	3.20	3.01	2.83	2.83
	25%	0.91	0.86	0.69	0.48
w₇₀₀/(10^-2m·s^-1)	75%	6.76	6.35	5.65	5.11
	50%	1.25	1.08	0.84	0.92
	25%	-1.54	-1.45	-1.25	-0.55

表4 不同等级短时强降水样本LCL和H0百分位统计

Tab.4 Percentile statistics of physical quantities of LCL and H0 during short-term heavy precipitation samples with different grades

物理量	百分位	I级	Ⅱ级	Ⅲ级	IV级
LCL/hPa	75%	982.24	983.82	983.86	983.41
	50%	963.51	967.76	970.37	967.15
	25%	940.18	948.97	952.82	945.92
H₀/m	75%	5656.11	5656.68	5656.74	5650.09
	50%	5515.68	5515.68	5520.86	5522.82
	25%	5349.48	5347.85	5344.15	5314.91

图2 不同等级短时强降水样本925 hPa露点温度(a)及BLI(b)概率密度分布

Fig.2 The probability density distribution of 925 hPa dew point temperature (a) and benefit lifting index (b) during short-term heavy precipitation samples with different grades

表5 不同等级短时强降水概率预报产品的不同概率阈值预报检验

Tab.5 The forecast test of different probability thresholds of probability prediction products about short-term heavy precipitation with different grades

概率阈值/%	I级			Ⅱ级			Ⅲ级			IV级
概率阈值/%	HR/%	FR/%	TS	HR/%	FR/%	TS	HR/%	FR/%	TS	HR/%	FR/%	TS
60	55.7	83.6	0.145	54.6	91.4	0.080	51.9	97.4	0.026	50.4	99.8	0.002
65	40.7	81.8	0.144	39.1	90.3	0.083	35.0	97.1	0.027	34.1	99.7	0.002
70	25.8	79.8	0.128	24.7	88.9	0.083	21.7	96.7	0.030	22.4	99.7	0.003
75	13.0	78.0	0.089	12.4	87.5	0.066	10.8	96.1	0.029	12.3	99.6	0.004
80	5.1	76.8	0.043	5.1	85.8	0.039	4.8	95.1	0.025	5.8	99.4	0.005

图3 2020年7月5日(a、b)、7日(c、d)及6月27日(e、f)、8月8日(g、h)08：00起报的16 h时效I级(a)、20 h时效II级(c)、10 h时效III级(e)、14 h时效IV级(g)短时强降水概率预报产品(彩色填色区，单位：％)与CMA-MESO模式预报的对应时次1 h降水量(b， d， f， h，彩色填色区，单位：mm) (左图从上到下黑色圆点分别为20 mm·h-1及以上、30 mm·h-1及以上、50 mm·h-1及以上、80 mm·h-1及以上短时强降水实况)

Fig.3 The probability prediction products (color shaded areas， Unit： %) about short-term heavy precipitation with grade I for 16 h prediction time (a)， with grade II for 20 h prediction time (c)， with grade III for 10 h prediction time (e)， with grade IV for 14 h prediction time (g) forecasted from 08：00 BST on July 5 (a， b) and 7 (c， d)， June 27 (e， f) and August 8 (g， h) in 2020 and 1 h precipitation forecasted by the CMA-MESO model at corresponding time (b， d， f， h， color shaded areas， Unit： mm) (The black dots in the left figures from top to bottom are observed precipitation with rainfall intensity of 20 mm·h-1 and above， 30 mm·h-1 and above， 50 mm·h-1 and above， 80 mm·h-1 and above， respectively)

图4 I级概率预报产品20 mm·h-1及以上短时强降水与CMA-MESO模式预报的对应时次10 mm·h-1及以上短时强降水的命中率(a)、空报率(b)及TS评分(c)

Fig.4 The hit rate (a)， false alarm rate (b) and TS score (c) of the grade I probability product about short-term heavy precipitation with rainfall intensity of 20 mm·h-1 and above and corresponding time short-term heavy precipitation with rainfall intensity of 10 mm·h-1 and above predicted by the CMA-MESO model

参考文献 35

[1]	俞小鼎. 短时强降水临近预报的思路和方法[J]. 暴雨灾害, 2013, 32(3):202-209.
[2]	DOSWELL Ⅲ C A. Severe convective storms—an overview[J]. Meteorological Monographs, 2001, 28(50):1-26.
[3]	孙继松. 短时强降水和暴雨的区别与联系[J]. 暴雨灾害, 2017, 36(6):498-506.
[4]	樊李苗, 俞小鼎. 中国短时强对流天气的若干环境参数特征分析[J]. 高原气象, 2013, 32(1):156-165.
[5]	黄艳, 俞小鼎, 陈天宇, 等. 南疆短时强降水概念模型及环境参数分析[J]. 气象, 2018, 44(8):1033-1041.
[6]	庄晓翠, 赵江伟, 李健丽, 等. 新疆阿勒泰地区短时强降水流型及环境参数特征[J]. 高原气象, 2018, 37(3):675-685.
[7]	陈元昭, 俞小鼎, 陈训来. 珠江三角洲地区重大短时强降水的基本流型与环境参量特征[J]. 气象, 2016, 42(2):144-155.
[8]	周晋红, 赵彩萍, 董春卿. 太原汛期短时强降水环流分型及环境参量分析[J]. 干旱气象, 2019, 37(3):392-399.
[9]	王成福, 谢蕊, 吉哲君, 等. 甘南高原大到暴雨天气过程分型及特征[J]. 干旱气象, 2019, 37(1):97-108.
[10]	程鹏, 罗汉, 陈佩璇, 等. 庆阳黄土高原短时强降水统计特征及预警指标[J]. 干旱气象, 2020, 38(2):319-328.
[11]	田付友, 郑永光, 张涛, 等. 我国中东部不同级别短时强降水天气的环境物理量分布特征[J]. 暴雨灾害, 2017, 36(6):518-526.
[12]	张武龙, 康岚, 杨康权, 等. 四川盆地不同强度短时强降水物理量特征对比分析[J]. 气象, 2021, 47(4):439-449.
[13]	STENSRUD D J, CORTINAS J, BROOKS H E. Discriminating between tornadic and nontornadic thunderstorms using mesoscale model output[J]. Weather and Forecasting, 1997, 12(3):613-632.
[14]	HART R E, FORBES G S, GRUMM R H. The use of hourly model-generated soundings to forecast mesoscale phenomena. Part Ι: initial assessment in forecasting warm-season phenomena[J]. Weather and Forecasting, 1998, 13(4): 1165-1185.
[15]	LEE B D, WILHELMSON R B. The numerical simulation of nonsupercell tornadogenesis. Part Ш: parameter tests investigating the role of CAPE, vortex sheet strength and boundary layer vertical shear[J]. Journal of the Atmospheric Sciences, 1999, 57(4): 2246-2261.
[16]	何立富, 周庆亮, 谌芸, 等. 国家级强对流潜势预报业务进展与检验评估[J]. 气象, 2011, 37(7):777-784.
[17]	沈澄, 孙燕, 魏晓奕, 等. 基于物理量参数的江苏短时强降水预报模型的研究[J]. 气象, 42(5):557-566.
[18]	雷蕾, 孙继松, 王国荣, 等. 基于中尺度数值模式快速循环系统的强对流天气分类概率预报试验[J]. 气象学报, 2012, 70(4):752-765.
[19]	曾明剑, 张备, 吴海英, 等. 基于接近度概念的强对流天气预报方法研究[J]. 高原气象, 2015, 34(5):1357-1368.
[20]	陈秋萍, 冯晋勤, 李白良, 等. 福建强天气短时潜势预报方法研究[J]. 气象, 2010, 36(2):28-32.
[21]	斯琴, 王佳津, 荀学义, 等. 基于T639对流参数的内蒙古强对流天气潜势预报方法初探[J]. 干旱气象, 2016, 34(5):906-911.
[22]	裴尕桑, 靳双龙, 曾晓青, 等. 青海一次强降水过程的物理量预报应用试验[J]. 干旱气象, 2018, 36(3):507-515.
[23]	BENEVIDES P, CATALAO J, NICO G. Neural network approach to forecast hourly intense rainfall using GNSS precipitable water vapor and meteorological sensors[J]. Remote Sensing, 2019, 11, 966. DOI: 10.3390/rs11080966.
[24]	LIU Y, ZHAO Q Z, YAO W Q, et al. Short-term rainfall forecast model based on the improved BP-NN algorithm[J]. Scientific Reports, 2019,9, 19751. DOI:10.1038/s41598-019-56452-5.
[25]	XIE H S, WU L D, XIE W, et al. Improving ECMWF short-term intensive rainfall forecasts using generative adversarial nets and deep belief networks[J]. Atmospheric Research, 2021, 249, 105281.DOI:10.1016/j.atmosres.2020.105281.
[26]	于翡, 黄丽萍, 邓莲堂. GRAPES-MESO模式不同空间分辨率对中国夏季降水预报的影响分析[J]. 大气科学, 2018, 42(5):1146-1156.
[27]	黄丽萍, 陈德辉, 邓莲堂, 等. GRAPES_Meso V4.0主要技术改进和预报效果检验[J]. 应用气象学报, 2017, 28(1):25-37.
[28]	孙继松, 戴建华, 何立富, 等. 强对流天气预报的基本原理与技术方法[M]. 北京: 气象出版社, 2014:145-158.
[29]	章国材. 强对流天气分析与预报[M]. 北京: 气象出版社, 2011:74-109.
[30]	吴翠红, 王晓玲, 等. 湖北省中尺度暴雨诊断分析研究[M]. 北京: 气象出版社, 2013.
[31]	CHEN T Q, GUESTRIN C. XGBoost: a scalable tree boosting system[C]// Proceedings of the 22nd ACM SIGKDD International Conference on Knowledge Discovery and Data Mining. California: ACM, 2016.
[32]	田付友, 郑永光, 张小玲, 等. 2017年5月7日广州极端强降水对流系统结构、触发和维持机制[J]. 气象, 2018, 44(4):469-484.
[33]	俞小鼎, 姚秀萍, 熊廷南, 等. 多普勒天气雷达原理与业务应用[M]. 北京: 气象出版社, 2006:170-172.
[34]	王季方, 卢正鼎. 模糊控制中隶属度函数的确定方法[J]. 河南科学, 2000, 18(4):348-351.
[35]	ZADEH L A. Fuzzy sets as a basis for a theory of possibility[J]. Fuzzy Sets and Systems, 1978, 1(1): 3-28.