基于集合匹配尺度的邻域集合概率法及其在梅雨锋暴雨预报中的适用性

徐渊; 闵锦忠; 庄潇然; 陈云辉; 张璐; 田高山

doi:10.16032/j.issn.1004-4965.2024.029

基于集合匹配尺度的邻域集合概率法及其在梅雨锋暴雨预报中的适用性

doi: 10.16032/j.issn.1004-4965.2024.029

1.
江西省气象台，江西南昌 330096
2.
南京信息工程大学气象灾害教育部重点实验室/气象灾害预报预警与评估协同创新中心，江苏南京 210044
3.
江苏省气象台，江苏南京 210008
4.
民航西南空管局气象中心，四川成都 610202

基金项目:

国家自然科学基金项目 42192553

江西省气象局青年人才培养项目 “631”强降水定量客观预报、预警方法研究

江西省气象局面上项目 JX2023M24

详细信息

通讯作者:
闵锦忠，男，江苏省人，教授，主要从事中小尺度数值模拟、资料同化、集合预报等方面的研究。E-mail：minjz@nuist.edu.cn

中图分类号: P426.62
计量
- 文章访问数: 52
- HTML全文浏览量: 8
- PDF下载量: 10
- 被引次数: 0
出版历程
- 收稿日期: 2022-11-23
- 修回日期: 2024-02-20
- 网络出版日期: 2024-06-07
- 刊出日期: 2024-04-20

Neighborhood Ensemble Probability Method Based on Ensemble Agreement Scale and Its Application on Quantitative Probability Forecast of Meiyu Frontal Heavy Rainfall

1.
Jiangxi Meteorological Observatory, Nanchang 330096, China
2.
Key Laboratory, Meteorological Disaster of Ministry of Education (KLME)/Collaborative Innovation Center on Forecast and Evaluation of Meteorological Disasters (CIC-FEMD), Nanjing University of Information Science and Technology, Nanjing 210044, China
3.
Jiangsu Meteorological Observatory, Nanjing 210008, China
4.
Meteorological Center of CAAC Southwest Air Traffic Control Bureau, Chengdu 610202, China

摘要

摘要: 概率预报是由集合预报衍生、包含不确定性信息的客观产品，对业务决策服务有重要的参考价值。传统的邻域集合概率法中，邻域半径固定不变，不符合实际天气过程中牵涉甚广的尺度谱。为此引入基于集合匹配尺度的邻域集合概率法(Neighborhood Ensemble Probability based on Ensemble Agreement Scale，EAS_NEP)，并在中国南方典型的梅雨锋暴雨中开展准确性和预报技巧的定量检验评估，以期验证该方法在此类过程中的适用性，并促进其在实际业务中的推广使用。联合扰动初始场、侧边界和物理过程所得到的集合预报能较好地表征实际的预报不确定性，进一步在此基础上比较了格点概率法、不同半径的邻域集合概率法以及EAS_NEP的优劣。试验结果表明，EAS_NEP能根据集合成员间的一致性程度，自适应地调整邻域半径，其在集中型降水中所确定的邻域半径通常大于分散型降水。动态调整的邻域半径既避免了半径过大时的过度平滑与关键信息丢失，又消除了半径较小所带来的奇异点，其空间分布呈阶梯型，空间连续性更优。此外，BS(布莱尔评分)、FSS(分数技巧评分)和ROC曲线(相对作用特征曲线)等定量评估结果也体现出EAS_NEP相比传统方法正的预报技巧，尤其是在分散型降水和高阈值检验时优势更明显。以上结果表明，EAS_NEP在梅雨锋暴雨的预报中具有较好的应用前景，运用在业务中能有效提升概率预报质量。
- 概率预报 /
- 集合预报 /
- 集合匹配尺度 /
- 邻域集合概率 /
- 梅雨锋暴雨
Abstract: Probability forecast is an objective product derived from ensemble forecasts and contains information of uncertainty, which has important reference value for operational decision-making services. In the traditional neighborhood ensemble probability method (NEP), the neighborhood radius is always a constant and does not conform to the scale spectrum involved in the practical weather events. Therefore, the Neighborhood Ensemble Probability based on Ensemble Agreement Scale (EAS_NEP) was introduced, and comprehensive evaluations were conducted for two Meiyu frontal heavy rainfall events in southern China, in order to verify the applicability of this method in such events and promote its popularization in practical operation. The ensemble forecasts obtained by combining the initial condition, lateral boundary and physical process of perturbations can better represent the practical prediction uncertainty. Based on this, the Grid-to-grid Probability (GP), NEP with different radii and EAS_NEP were compared for accuracy, forecast skill and other aspects of performance. The results showed that EAS_NEP can adjust the neighborhood radius adaptively according to the agreement between ensemble members and its radii tend to be lager for concentrated precipitation than dispersed precipitation. These dynamically adjusted radii not only avoid excessive smoothing and key information loss when the neighborhood radius is too large, but also eliminate the singular points caused by small radius and possess stepped spatial distribution with better continuity. In addition, quantitative evaluation results such as BS, FSS and ROC curves also confirm that EAS_NEP has improved forecasting skills compared to traditional methods, especially in the case of dispersed precipitation and high threshold situation. The results suggest that EAS_NEP has a good application prospect in the Meiyu frontal heavy rainfall events and can effectively improve the quality of probability forecast.
- probability forecast /
- ensemble forecasts /
- ensemble agreement scale /
- neighborhood ensemble probability /
- Meiyu frontal heavy rainfall

HTML全文

图 1 三种理想情形下(a₁~a₃)，GP(b₁~b₃)，NEP(c₁~c₃)，EAS_NEP(d₁~d₃)所产生概率预报(单位：%)的空间分布^[11]

下载: 全尺寸图片幻灯片

图 2 2017年6月9日12时(a)、2021年7月3日18时(b)500 hPa位势高度(填色和蓝色等值线，单位：dagpm)、850 hPa相当位温(绿色等值线，单位：K)和风场(矢量，单位：m · s^-1)；2017年6月9日12时—10日12时(c)、2021年7月3日18时—4日18时(d)24 h累积降水量(单位：mm)

下载: 全尺寸图片幻灯片

图 3 模拟区域

下载: 全尺寸图片幻灯片

图 4 case1积分6 h(a₁)、12 h(b₁)、18 h(c₁)，case2积分6 h(a₂)、12 h(b₂)、18 h(c₂)逐小时降水量(单位：mm)的集合平均扰动-CTRL误差散点图

下载: 全尺寸图片幻灯片

图 5 EAS_NEP方案中case1和case2在研究区域的平均邻域半径随积分时间的变化

阈值为5 mm ·h^-1。

下载: 全尺寸图片幻灯片

图 6 case1 2017年6月9日12时—10日12时24 h累积降水的概率预报GP(a₁)，r = 1(b₁)，r = 4(c₁)，r = 8(d₁)，r = 16(e₁)，r = 24(f₁)，EAS_NEP(g₁)；case2 2021年7月3日18时—4日18时24 h累积降水的概率预报GP(a₂)，r = 1(b₂)，r = 4(c₂)，r = 8(d₂)，r = 16(e₂)，r = 24(f₂)，EAS_NEP(g₂)

阈值均为50 mm ·(24 h)^-1，相应的BS和FFS在图中标注。

下载: 全尺寸图片幻灯片

图 7 case1 GP(a₁)，r = 1(b₁)，r = 4(c₁)，r = 8(d₁)，r = 16(e₁)，r = 24(f₁)，EAS_NEP(g₁)；case2 GP(a₂)，r = 1(b₂)，r = 4(c₂)，r = 8(d₂)，r = 16(e₂)，r = 24(f₂)，EAS_NEP(g₂) 24 h累积降水量概率预报在各概率区间上的频率分布

阈值均为50 mm ·(24 h)^-1。

下载: 全尺寸图片幻灯片

图 8 case1中雨(a₁)，大雨(b₁)，暴雨(c₁)和case2中雨(a₂)，大雨(b₂)，暴雨(c₂)逐小时降水量概率预报的FSS

下载: 全尺寸图片幻灯片

图 9 case1 GP(a₁)，r = 1(b₁)，r = 4(c₁)，r = 8(d₁)，r = 16(e₁)，r = 24(f₁)，EAS_NEP(g₁)；case2 GP(a₂)，r = 1(b₂)，r = 4(c₂)，r = 8(d₂)，r = 16(e₂)，r = 24(f₂)，EAS_NEP(g₂) 24 h累积降水概率预报的ROC曲线

曲线下方相应的面积在图中标注。

下载: 全尺寸图片幻灯片

表 1 PPMP的物理过程及参数设置

集合成员序号	微物理方案	边界层方案	雨滴截断参数N_0r/m^-4	临界Richardson数Ri_c
1 (CTRL)	WSM6	MYJ	8×10⁶	0.5
2	WSM6	MYJ	8×10⁵	0.5
3	WSM6	MYJ	8×10⁷	0.5
4	WSM6	MYJ	8×10⁶	1
5	WSM6	MRF	8×10⁶	0.5
6	WSM6	MRF	8×10⁵	0.5
7	WSM6	MRF	8×10⁷	0.5
8	WSM6	MRF	8×10⁶	1
9	WSM5	MYJ	8×10⁶	0.5
10	WSM5	MYJ	8×10⁵	0.5
11	WSM5	MYJ	8×10⁷	0.5
12	WSM5	MYJ	8×10⁶	1
13	WSM5	MRF	8×10⁶	0.5
14	WSM5	MRF	8×10⁵	0.5
15	WSM5	MRF	8×10⁷	0.5
16	WSM5	MRF	8×10⁶	1

下载: 导出CSV

参考文献(29)

[1]	CUO L, PAGANO T C, WANG Q J. A review of quantitative precipitation forecasts and their use in short-to medium-range streamflow forecasting[J]. J Hydrometeorol, 2011, 12(5): 713-728.
[2]	汪瑛, 胡胜, 涂静, 等. 2016—2020年广东省定量降水预报检验评估[J]. 热带气象学报, 2021, 37(2): 154-165.
[3]	闵锦忠, 吴乃庚. 近二十年来暴雨和强对流可预报性研究进展[J]. 大气科学, 2020, 44(5): 1 039-1 056.
[4]	徐渊, 闵锦忠, 庄潇然. 考虑尺度敏感性的长江中下游暖区暴雨初始扰动方案比较分析[J]. 热带气象学报, 2023, 39(3): 386-401.
[5]	LEITH C E. Theoretical skill of Monte Carlo forecasts[J]. Mon Wea Rev, 1974, 102(6): 409-418.
[6]	CLEMEN R T, MURPHY A H. Objective and subjective precipitation probability forecasts: some methods for improving forecast quality [J]. Wea Forecasting, 1986, 1(3): 213-218.
[7]	THEIS S E, HENSE A, DAMRATH U. Probabilistic precipitation forecasts from a deterministic model: A pragmatic approach[J]. Meteor Appl, 2005, 12(3): 257-268.
[8]	FLORA M L, POTVIN C K, SKINNER P S, et al. Using machine learning to generate storm-scale probabilistic guidance of severe weather hazards in the Warn-on-Forecast system[J]. Mon Wea Rev, 2021, 149(5): 1 535-1 557.
[9]	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.
[10]	DEY S R A, ROBERTS N M, PLANT R S, et al. A new method for the characterization and verification of local spatial predictability for convective-scale ensembles[J]. Quart J Roy Meteor Soc, 2016, 142(698): 1 982-1 996.
[11]	BLAKE B T, CARLEY J R, ALCOTT T I, et al. An adaptive approach for the calculation of ensemble gridpoint probabilities[J]. Wea Forecasting, 2018, 33(4): 1 063-1 080.
[12]	ERICKSON M J, KASTMAN J S, ALBRIGHT B, et al. Verification results from the 2017 HMT-WPC flash flood and intense rainfall experiment[J]. J Appl Meteor Climatol, 2019, 58(12): 2 591-2 604.
[13]	罗亚丽, 孙继松, 李英, 等. 中国暴雨的科学与预报: 改革开放40年研究成果[J]. 气象学报, 2020, 78(3): 419-450.
[14]	WHAN K, SCHMEITS M. Comparing area probability forecasts of (extreme) local precipitation using parametric and machine learning statistical postprocessing methods[J]. Mon Wea Rev, 2018, 146(11): 3 651-3 673.
[15]	智协飞, 张珂珺, 田烨, 等. 基于神经网络和地理信息的华东及华南地区降水概率预报[J]. 大气科学学报, 2021, 44(3): 381-393.
[16]	倪允琪, 周秀骥. 中国长江中下游梅雨锋暴雨形成机理以及监测与预测理论和方法研究[J]. 气象学报, 2004, 62(5): 136-151.
[17]	LIU J Y, TAN Z M. Mesoscale predictability of mei-yu heavy rainfall[J]. Adv Atmos Sci, 2009, 26(3): 438-450.
[18]	宫宇, 罗亚丽. 梅雨锋前线状中尺度对流系统成熟阶段的空气垂直运动分析[J]. 热带气象学报, 2014, 30(4): 687-699.
[19]	梅疏影, 闵锦忠. 不同同化方案在一次梅雨锋暴雨预报中的应用研究[J]. 气象科学, 2018, 38(4): 432-441.
[20]	SCHWARTZ C S, KIAN J S, WEISS S J, et al. Toward improved convection-allowing ensembles: Model physics sensitivities and optimizing probabilistic guidance with small ensemble membership[J]. Wea Forecasting, 2010, 25(1): 263-280.
[21]	EBERT E E. Neighborhood verification: A strategy for rewarding close forecasts[J]. Wea Forecasting, 2009, 24(6): 1 498-1 510.
[22]	ZACHAROV P, REZACOVA D. Using the fractions skill score to assess the relationship between an ensemble QPF spread and skill[J]. Atmos Res, 2009, 94(4): 684-693.
[23]	WILKS D S. Statistical methods in the atmospheric sciences[M]. San Diego, USA: Academic Press, 2006: 294-298.
[24]	ZHANG Y J, ZHANG F Q, DAVIS C A, et al. Diurnal evolution and structure of long-lived mesoscale convective vortices along the Mei-yu front over the east China plains[J]. J Atmos Sci, 2018, 75(3): 1 005-1 025.
[25]	陈茂钦, 徐海明, 刘蕾, 等. WRF3.1微物理参数化方案对两例暴雨的集合预报试验及可预报性分析[J]. 气象科学, 2012, 32(3): 237- 245.
[26]	沈新勇, 马铮, 郭春燕, 等. 不同边界层参数化方案对一次梅雨锋暴雨过程湍流交换特征模拟的影响[J]. 热带气象学报, 2017, 33(6): 793-811.
[27]	ZHANG X B. Application of a convection-permitting ensemble prediction system to quantitative precipitation forecasts over southern China: Preliminary results during SCMREX[J]. Quart J Roy Meteor Soc, 2018, 144(717): 2 842-2 862.
[28]	ZHANG X B. Multiscale characteristics of different-source perturbations and their interactions for convection-permitting ensemble forecasting during SCMREX[J]. Mon Wea Rev, 2019, 147(1): 291-310.
[29]	张涵斌, 陈静, 智协飞, 等. GRAPES区域集合预报系统应用研究[J]. 气象, 2014, 40(9): 1 076-1 087.