基于融合TC-WREM模型的热带气旋大风半径估算研究

周必高; 鲁小琴; 吴贤笃; 仇欣; 谢海华; 朱忠勇; 郑建琴

doi:10.16032/j.issn.1004-4965.2024.065

基于融合TC-WREM模型的热带气旋大风半径估算研究

doi: 10.16032/j.issn.1004-4965.2024.065

周必高^{1, 2},
鲁小琴^3, ,,
吴贤笃¹,
仇欣⁴,
谢海华^{1, 2},
朱忠勇^{1, 2},
郑建琴^{1, 2}

1.
温州市气象局，浙江温州 325027
2.
温州市台风监测预报技术重点实验室，浙江温州 325027
3.
中国气象局上海台风研究所，上海 200030
4.
南京大学大气科学学院，江苏南京 210023

基金项目:

温州市科技局基础公益科研项目 S2023012

南京大学中尺度灾害性天气教育部重点实验室基金 LMSWE-2201

上海市2021年度“科技创新行动计划”自然科学基金项目 21ZR1477300

详细信息

通讯作者:
鲁小琴，女，上海市人，研究员，主要从事台风结构客观分析技术研究及台风多源数据融合应用。E-mail：luxq@typhoon.org.cn

中图分类号: P435
计量
- 文章访问数: 17
- HTML全文浏览量: 6
- PDF下载量: 1
- 被引次数: 0
出版历程
- 收稿日期: 2023-06-21
- 修回日期: 2024-08-18
- 网络出版日期: 2025-01-08
- 刊出日期: 2024-10-20

Research on the Estimation of Tropical Cyclone Gale Radius Based on a Fusion TC-WREM Model

ZHOU Bigao^{1, 2},
LU Xiaoqin^{3
, ,},
WU Xiandu¹,
QIU Xin⁴,
XIE Haihua^{1, 2},
ZHU Zhongyong^{1, 2},
ZHENG Jianqin^{1, 2}

1.
Wenzhou Meteorological Bureau, Wenzhou, Zhejiang 325027, China
2.
Wenzhou Key Laboratory of Typhoon Monitoring and Forecasting Technology, Wenzhou, Zhejiang 325027, China
3.
Shanghai Typhoon Institute, China Meteorological Administration, Shanghai 200030, China
4.
School of Atmospheric Sciences, Nanjing University, Nanjing 210023, China

摘要

摘要: 利用2001—2020年美国联合台风警报中心(JTWC)热带气旋(Tropical Cyclone，TC)最佳资料数据集和静止气象卫星云图，建立了基于多层感知器神经网络模型(Multi-Layer Perceptron，MLP)和卷积神经网络(Convolutional Neural Network，CNN)融合的TC大风半径估算模型(TC Wind Radii Estimation Model，TCWREM)。该模型利用MLP和CNN分别对TC属性数据和卫星云图中与TC大风半径相关联的核心特征进行预提取，最终通过融合TC-WREM模型开展大风半径估算。融合的TC-WREM模型能实现对TC属性数据和卫星云图底层特征的深度客观挖掘，较单独的MLP和CNN模型的估算误差降低7%~24%。以TC近地面8级大风半径(R8)估算为例，针对2021年台风“烟花”的独立样本估算检验显示分象限R8估算平均绝对误差(Mean Absolute Error，MAE)分别为39、33、40和51 km，均值为41 km，误差中位值约40 km，优于业务估算精度(为大风半径的25%~40%)及西北太平洋和大西洋同类研究估算结果。由于融合TC-WREM模型的输入为易获取的TC属性数据和静止气象卫星云图，因此该模型易于在业务中进行推广，从而可改善国内TC大风半径估算模型缺乏的现状。
- 热带气旋 /
- 大风半径估算 /
- 卷积神经网络模型 /
- 多层感知器神经网络模型 /
- 融合TC-WREM模型 /
- 西北太平洋
Abstract: In this paper, a fusion TC Wind Radii Estimation Model (TC-WREM) that combines a MultiLayer Perceptron net (MLP) and a Convolutional Neural Network (CNN) is established by using the Tropical Cyclone (TC) best track data set and the static satellite cloud images. This model utilizes MLP and CNN to pre-extract the core features associated with TC wind radius from TC attribute data and satellite cloud images and ultimately performs gale wind radius estimation. The fused TC-WREM model in this study can achieve deep and objective mining of TC attribute data and underlying features of satellite cloud images, whose estimation error is reduced by about 7%-24% compared to individual MLP and CNN models. Taking the estimation of 17.2 m ·s^-1 wind radius of TC In-fa in 2021 as an example, the fused TCWREM model has higher estimation accuracy than the independent MLP and CNN model. Independent sample testing shows that the mean absolute estimation error in 4 quadrants is 39, 33, 40, and 51 km, respectively, with an average of 41 km, respectively, which is superior to that of other similar research. The fused TC-WREM model is advantageous due to its utilization of easily obtainable TC attribute information and geostationary meteorological satellite cloud images as inputs. This makes it suitable for operational use and addresses the current lack of domestic TC gale radius estimation models.
- tropical cyclone /
- wind radius estimation /
- Convolutional Neural Network /
- multi-layer perceptron net /
- fusion TC-WREM model /
- western North Pacific

HTML全文

图 1 不同时次TC卫星云图增强显示

a. 2002年7月3日06 UTC；b. 2012年8月25日12 UTC；c. 2018年9月16日06 UTC。

下载: 全尺寸图片幻灯片

图 2 融合TC-WREM模型架构图

下载: 全尺寸图片幻灯片

图 3 融合TC-WREM模型训练流程图

下载: 全尺寸图片幻灯片

图 4 2021年7月18日12时—26日00时“烟花”逐样本R8估算和观测值的对应散点图

下载: 全尺寸图片幻灯片

图 5 7个台风R8估算偏差箱图

下载: 全尺寸图片幻灯片

图 6 0505台风“海棠”R8估算MAE演变

下载: 全尺寸图片幻灯片

图 7 影响华东地区的6个TC不同象限估算AEA箱图

下载: 全尺寸图片幻灯片

表 1 TC匹配记录属性数据

序号	标识(时间-名称)	月份	东经/°E	北纬/°N	最大风速/(m·s^-1)	最低气压/hPa	NE/km	SE/km	SW/km	NW/km
1	2001072200-	7	150.3	25.0	20.56	994	111	111	111	111
2	2001072206-	7	149.6	25.2	23.13	991	120	120	120	120
3	2001072212-	7	149.0	25.4	25.7	987	111	93	93	111
……	……	……	……	……	……	……	……	……	……	……
7 587	2020111418-VAMCO	11	107.9	17.1	43.69	975	259	157	176	278
7 588	2020111500-VAMCO	11	107.6	7.1	38.55	983	204	130	102	213
7 589	2020111512-VAMCO	11	105.7	18.4	20.56	996	65	37	46	83

下载: 导出CSV

表 2 不同模型对R8各象限的估算性能

估算模型	MAPE/%				AEA/%
估算模型	NE	SE	SW	NW	NE	SE	SW	NW
CNN	0.308	0.246	0.279	0.294	69.22	75.43	72.14	71.62
MLP	0.222	0.185	0.226	0.285	77.83	81.50	77.32	71.50
TC-WREM	0.211	0.167	0.193	0.283	79.17	84.24	81.72	71.71

下载: 导出CSV

参考文献(18)

[1]	陈联寿, 丁一汇. 西太平洋台风概论[M]. 北京: 科学出版社, 1979.
[2]	COCKS S B, GRAY W M. Variability of the outer wind profiles of western North Pacific typhoons: Classifications and techniques for analysis and forecasting[J]. Mon Wea Rev, 2002, 130(8): 1 989-2 005.
[3]	GUO X, TAN Z M. Tropical cyclone fullness: a new concept for interpreting storm intensity[J]. Geophysi Res Lett, 2017, 44(9): 4 324-4 331.
[4]	ZHUO J Y, TAN Z M. Physics-augmented deep learning to improve tropical cyclone intensity and size estimation from satellite imagery[J]. Mon Wea Rev, 2021, 149(7): 2 097-2 113.
[5]	DEMUTH J L, DEMARIA M, KNAFF J A, et al. Evaluation of Advanced Microwave Sounding Unit tropical-cyclone intensity and size estimation algorithms[J]. J Appl Meteor Climatol, 2004, 43(2): 282-296.
[6]	DEMUTH J L, DEMARIALM, et al. Improvement of Advanced Microwave Sounding Unit tropical cyclone intensity and size estimation algorithms[J]. J Appl Meteor Climatol, 2006, 45(11): 1 573-1 581.
[7]	DVORAK. Tropical cyclone intensity analysis and forecasting from satellite imagery[J]. AMS Journals, 1975, 103(5): 420-430
[8]	DVORAK V F. Tropical cyclone intensity analysis using satellite data. National Oceanic and Atmospheric Administrator[R]. NOAA Technical Report NESDIS 11, September 1984. NOAA--S/T 84-232. cover page4. Includes, bibliographical references (pages 46-47).
[9]	KNAFF J A, DEMARIA M, MOLENAR D A, et al. An automated, objective, multiple-satellite-platform tropical cyclone surface wind analysis[J]. J Appl Meteor Climatol, 2011, 50(10): 2 149-2 166.
[10]	KNAFF J A, LONGMOE S P, MOLENAR D A. An objective satellite-based tropical cyclone size climatology[J]. J Climate, 2014, 27(1): 455-476.
[11]	KNAFF J A, SAMPSON C R. After a decade are Atlantic tropical cyclone gale force wind radii forecasts now skillful? [J]. Wea And Forecasting, 2015, 30(3): 702-709.
[12]	KNAFF J A, SLOCUM C J, MUSGRAVE K D, et al. Using routinely available information to estimate tropical cyclone wind structure [J]. Mon Wea Rev, 2016, 144(4): 1 233-1 247.
[13]	KOSSIN J P, KNAFF J A, BERGER H I, et al. Estimating hurricane wind structure in the absence of aircraft reconnaissance[J]. Wea Forecasting, 2007, 22(1): 89-101.
[14]	LU X Q, YU H, YANG X M, et al. Estimating tropical cyclone size in the Northwestern Pacific from geostationary satellite infrared images [J]. Remote Sensing, 2017, 9(7): 728.
[15]	LU X Q, WONG W K, YU H, et al. Tropical cyclone size identification over the Western North Pacific using support vector machine and general regression neural network[J]. J Meteor Soc Japan SerII, 2022, 100(6): 927-941.
[16]	ZEILER M D, FERGUS R. Visualizing and understanding convolutional networks[C]//Fleet D, Pajdla T, Schiele B, et al. Computer VisionECCV 2014, Cham: Springer.
[17]	MATSUYU M, MORI K, MITARI Y, et al. Subject independent facial expression recognition with robust face detection using a convolutional neural network [J]. Neural Networks, 2003, 16(5-6): 555-559.
[18]	SAMPSON C R, FLLKADA E M, KNAFF J A, et al. Tropical cyclone gale wind radii estimates for the Western North Pacific[J]. Wea Forecasting, 2017, 32(3): 1 029-1 040.