广西台风直接经济损失预估模型研究

doi:10.11978/2024191

Abstract

Abstract:

China is among the countries most severely impacted by typhoon disasters globally, with Guangxi being one of the regions most affected in China. Guangxi experiences an average of five typhoons annually, with a maximum of up to nine. These typhoon disasters have inflicted significant economic losses on the region, severely hindering the high-quality socioeconomic development of Guangxi. Developing pre-assessment models for typhoon disaster losses provides a practical reference for government efforts in disaster prevention, mitigation, and post-disaster reconstruction, thereby supporting the region's accelerated socioeconomic development. In this study, 40 historical typhoon disasters that impacted Guangxi between 2001 and 2020, with relatively complete records and measurable economic losses, were analyzed. From three index layers—disaster-causing factors, disaster-bearing entities, and disaster prevention and mitigation capacities—12 typhoon impact factors were selected as input variables for a neural network model. To enhance data robustness, four types of neural networks, combined with cubic spline interpolation, were employed to construct an intelligent prediction model for typhoon disaster damage in Guangxi. By comparing the performance of these four neural networks on training and testing datasets, the most suitable model was selected. The practical application potential of this model was explored using Typhoons "Kalmaegi," "Mujigae," and "Chaba" as case studies, enabling dynamic prediction of typhoon-induced losses in Guangxi. The results indicate that the GA-BP (genetic algorithm-back propagation) neural network model exhibits the best performance. For the training set, the model achieved an R² of 0.9606 and an RMSE of 1.0669. For the testing set, it achieved an R² of 0.9073 and an RMSE of 1.1635. These results demonstrate that the predictions closely align with the actual typhoon damages, validating the effectiveness of the model. The study highlights the model's potential for dynamic real-world prediction of typhoon damage in Guangxi.

Key words: typhoon disasters, data augmentation, loss prediction, neural networks, Guangxi

CLC Number:

P429

HUANG Zhaoyong, XU Guilin, MO Zhiming. Research on a prediction model for direct economic losses caused by typhoons in Guangxi[J].Journal of Tropical Oceanography, 2025, 44(4): 56-66.

Figures/Tables 16

Fig. 1

Tab. 1

Tab. 2

Tab. 3

Tab. 4

Fig. 2

Tab. 5

Tab. 6

Fig. 3

Fig. 4

Fig. 5

Fig. 6

Tab. 7

Fig. 7

Fig. 8

Tab. 8

References 24

[1]	丁宏飞, 李演洪, 刘博, 等, 2016. 基于BP神经网络与SVM的快速路行程时间组合预测研究[J]. 计算机应用研究, 33(10): 2929-2932, 2936.
	DING HONGFEI, LI YANHONG, LIU BO, et al, 2016. Expressway’s travel time prediction based on combined BP neural network and support vector machine approach[J]. Application Research of Computers, 33(10): 2929-2932, 2936 (in Chinese with English abstract).
[2]	冯倩, 刘强, 2017. 基于SVM-BP神经网络的风暴潮灾害损失预评估[J]. 海洋环境科学, 36(4): 615-621.
	FENG QIAN, LIU QIANG, 2017. Pre-assessment for the loss caused by storm surge based on the SVM-BP neural network[J]. Marine Environmental Science, 36(4): 615-621 (in Chinese with English abstract).
[3]	韩鹏, 郭桂祯, 李鑫磊, 等, 2023. 基于地理探测器的福建省台风灾情影响因素分析[J]. 遥感技术与应用, 38(2): 487-495.
	HAN PENG, GUO GUIZHEN, LI XINLEI, et al, 2023. Research on influencing factors of typhoon disasters in Fujian Province based on geodetector[J]. Remote Sensing Technology and Application, 38(2): 487-495 (in Chinese with English abstract).
[4]	韩鹏, 郭桂祯, 孙宁, 等, 2022. 广东省台风灾害时空格局及影响因素研究[J]. 灾害学, 37(3): 112-117.
	HAN PENG, GUO GUIZHEN, SUN NING, et al, 2022. Spatiotemporal patterns and influencing factors of typhoon disasters in Guangdong Province, China[J]. Journal of Catastrophology, 37(3): 112-117 (in Chinese with English abstract).
[5]	李艳兰, 金龙, 史旭明, 等, 2021. 基于遗传—神经网络方法的广西台风灾害评估模型研究[J]. 气象与环境学报, 37(3): 139-144.
	LI YANLAN, JIN LONG, SHI XUMING, et al, 2021. Study on assessment model of typhoon disaster in Guangxi based on genetic-neural network method[J]. Journal of Meteorology and Environment, 37(3): 139-144 (in Chinese with English abstract).
[6]	李志娟, 2016. 台风灾害损失预测方法研究[D]. 广州: 华南理工大学.
	LI ZHIJUAN, 2016. The research of typhoon disaster loss prediction[D]. Guangzhou: South China University of Technology (in Chinese with English abstract).
[7]	林沛延, 林陪晖, 王俊, 等, 2023. 基于机器学习方法的浙江省台风灾害风险评估和动态风险预报[J]. 自然灾害学报, 32(4): 13-24.
	LIN PEIYAN, LIN PEIHUI, WANG JUN, et al, 2023. Typhoon disaster risk assessment and dynamic risk forecasts in Zhejiang Province based on machine learning methods[J]. Journal of Natural Disasters, 32(4): 13-24 (in Chinese with English abstract).
[8]	刘玉函, 唐晓春, 宋丽莉, 2003. 广东台风灾情评估探讨[J]. 热带地理, 23(2): 119-122.
	LIU YUHAN, TANG XIAOCHUN, SONG LILI, 2003. A discussion about the typhoon disaster evaluation of Guangdong[J]. Tropical Geography, 23(2): 119-122 (in Chinese with English abstract).
[9]	毛嘉铭, 刘光宇, 罗凯元, 2024. 结合数据增强及组合算法的短期光伏功率预测[J]. 电力系统及其自动化学报, 36(8): 133-141.
	MAO JIAMING, LIU GUANGYU, LUO KAIYUAN, 2024. Short-term photovoltaic power prediction using data augmentation and combined algorithms[J]. Proceedings of the CSU-EPSA, 36(8): 133-141 (in Chinese with English abstract).
[10]	牛海燕, 刘敏, 陆敏, 等, 2011. 中国沿海地区台风灾害损失评估研究[J]. 灾害学, 26(3): 61-64.
	NIU HAIYAN, LIU MIN, LU MIN, et al, 2011. Losses assessment of typhoon disaster in China coastal areas[J]. Journal of Catastrophology, 26(3): 61-64 (in Chinese with English abstract).
[11]	邱军, 孙力勇, 樊庆东, 2023. 一种基于三次样条插值的风电场出力预测评估方法[J]. 电气开关, 61(3): 16-18, 23.
	QIU JUN, SUN LIYONG, FAN QINGDONG, 2023. A method for predicting and evaluating wind farm output based on cubic spline interpolation[J]. Electric Switchgear, 61(3): 16-18, 23 (in Chinese with English abstract).
[12]	汪路, 卢莹, 赵海坤, 2023. 台风灾害时空特征分析与评估模型构建[J]. 灾害学, 38(4): 187-194.
	WANG LU, LU YING, ZHAO HAIKUN, 2023. Analysis of temporal and spatial characteristics of typhoon disasters and construction of evaluation model[J]. Journal of Catastrophology, 38(4): 187-194 (in Chinese with English abstract).
[13]	王阳, 陈薇伊, 马军山, 2022. 基于卷积神经网络的乳腺癌良恶性诊断[J]. 软件工程, 25(1): 6-9.
	WANG YANG, CHEN WEIYI, MA JUNSHAN, 2022. Diagnosis of benign and malignant breast cancer based on convolutional neural network[J]. Software Engineering, 25(1): 6-9 (in Chinese with English abstract).
[14]	温姗姗, 翟建青, THOMAS Fischer, 等, 2017. 1984—2014年影响中国热带气旋的经济损失标准化及其变化特征[J]. 热带气象学报, 33(4): 478-487.
	WEN SHANSHAN, ZHAI JIANQING, FISCHER T, et al, 2017. Variation of normalized economic losses from influential tropical cyclones in China for 1984—2014[J]. Journal of Tropical Meteorology, 33(4): 478-487 (in Chinese with English abstract).
[15]	徐金鹏, 郭新峰, 王瑞波, 等, 2023. 基于GAN数据增强的软件缺陷预测聚合模型[J]. 计算机科学, 50(12): 24-31.
	XU JINPENG, GUO XINFENG, WANG RUIBO, et al, 2023. Aggregation model for software defect prediction based on data enhancement by GAN[J]. Computer Science, 50(12): 24-31 (in Chinese with English abstract).
[16]	杨帆, 夏鸿崚, 2024. 基于数据集扩充的即时软件缺陷预测方法[J]. 南通大学学报(自然科学版), 23(1): 58-65.
	YANG FAN, XIA HONGLING, 2024. A just-in-time software defect prediction method based on data augmentation[J]. Journal of Nantong University (Natural Science Edition), 23(1): 58-65 (in Chinese with English abstract).
[17]	张恺, 杨隆浩, 高建清, 等, 2024. 基于累积置信规则库推理的台风灾害直接经济损失预测[J]. 灾害学, 39(1): 64-68, 74.
	ZHANG KAI, YANG LONGHAO, GAO JIANQING, et al, 2024. Typhoon disaster direct economic losses prediction based on cumulative belief rule-based system[J]. Journal of Catastrophology, 39(1): 64-68, 74 (in Chinese).
[18]	赵飞, 廖永丰, 张妮娜, 等, 2011. 登陆中国台风灾害损失预评估模型研究[J]. 灾害学, 26(2): 81-85.
	ZHAO FEI, LIAO YONGFENG, ZHANG NINA, et al, 2011. A pre-evaluation model for typhoon disasters in China[J]. Journal of Catastrophology, 26(2): 81-85 (in Chinese with English abstract).
[19]	GUO TENGJIAO, LI GUOSHENG, 2020. Study on methods to identify the impact factors of economic losses due to typhoon storm surge based on confirmatory factor analysis[J]. Natural Hazards, 100(2): 515-534.
[20]	LI JIACHUN, NIE BINGCHUAN, 2017. Storm surge prediction: present status and future challenges[J]. Procedia IUTAM, 25: 3-9.
[21]	RAJASEKARAN S, GAYATHRI S, LEE T L, 2008. Support vector regression methodology for storm surge predictions[J]. Ocean Engineering, 35(16): 1578-1587.
[22]	SUN HAI, WANG JIN, YE WENTAO, 2021. A data augmentation-based evaluation system for regional direct economic losses of storm surge disasters[J]. International Journal of Environmental Research and Public Health, 18(6): 2918.
[23]	UDDIN M J, LI YUBIN, SATTAR M A, et al, 2022. An improved cluster-wise typhoon rainfall forecasting model based on machine learning and deep learning models over the Northwestern Pacific Ocean[J]. Journal of Geophysical Research: Atmospheres, 127(14): e2022JD036603.
[24]	WANG SI, MU LIN, QI MENGNAN, et al, 2021. Quantitative risk assessment of storm surge using GIS techniques and open data: A case study of Daya Bay Zone, China[J]. Journal of Environmental Management, 289: 112514.

数据类型	数据名称	数据来源
基础数据	广西行政区划数据	中国科学院地理科学与资源研究所、国家测绘地理信息局标准地图服务官网
基础数据	广西高程数据	中国科学院地理科学与资源研究所、国家测绘地理信息局标准地图服务官网
社会经济数据	广西社会经济统计数据	《广西统计年鉴》
台风相关数据	广西台风灾情数据	《中国气象灾害年鉴》《中国海洋灾害公报》《中国民政统计年鉴》《南海区海洋灾害公报》《广西海洋灾害公报》
	广西风雨数据	国家气象信息中心
	广西水文数据	项目团队自研的近40年北部湾高分辨率海洋灾害数据集

指标层	指标因子
致灾因子危险性	最大风速/(m·s^-1) (X₁)
	最大降雨量/mm (X₂)
	中心气压/hPa (X₃)
	最大增水/cm (X₄)
	总水位/m (X₅)
	有效波高/m (X₆)
	平均波长/cm (X₇)
	平均波周期/s (X₈)
承灾体脆弱性	人口密度/每平方千米人数 (X₉)
	经济密度/(10⁸ yuan·km^-2) (X₁₀)
	城镇化率/% (X₁₁)
	农作物播种面积/1000hm² (X₁₂)
	全区电力消费量/10⁸ kWh (X₁₃)
防灾减灾能力	每千人拥有病床数 (X₁₄)
	每千人拥有卫生技术人员数 (X₁₅)
	卫生机构数量 (X₁₆)
	境内公路里程/km (X₁₇)
	境内铁路营业里程/km (X₁₈)

指标	重要性程度
最大风速(X₁)	32.48%
农作物播种面积(X₁₂)	21.07%
中心气压(X₃)	9.02%
总水位(X₅)	8.02%
人口密度(X₉)	7.95%
最大降雨量(X₂)	5.26%
最大增水(X₄)	4.32%
卫生机构数量(X₁₆)	4.04%
境内铁路营业里程(X₁₈)	2.20%
每千人拥有病床数量(X₁₄)	1.28%
每千人拥有卫生技术人员数量(X₁₅)	1.23%
城镇化率(X₁₁)	0.76%
平均波长(X₇)	0.48%
平均波周期(X₈)	0.47%
经济密度(X₁₀)	0.44%
有效波高(X₆)	0.41%
全区电力消费量(X₁₃)	0.35%
境内公路里程(X₁₇)	0.22%

综合层	模型输入	模型输出
致灾因子危险性	最大降雨量	直接经济损失/亿元
	中心气压
	最大风速
	最大增水
	总水位
承灾体脆弱性	人口密度
	城镇化率
	农作物播种面积
防灾减灾能力	每千人拥有病床数量
	每千人拥有卫生技术人员数量
	卫生机构数量
	境内铁路营业里程

参数类型	值
遗传代数	10
种群规模	5
优化变量边界	[-1, 1]
选择函数参数	0.09
交叉函数参数	2
变异函数参数	[2 gen 3]

Research on a prediction model for direct economic losses caused by typhoons in Guangxi

RichHTML

PDF (PC)

Like

Knowledge

Abstract

Cite this article

share this article

Figures/Tables 16

References 24

Related Articles 4

Metrics

Comments

Recommended 0

神经网络类型	训练集		测试集
神经网络类型	R²	RMSE	R²	RMSE
BP神经网络	0.9659	1.1185	-0.6589	20.9774
GA-BP神经网络	0.9458	2.3717	-0.2242	10.9302
卷积神经网络(CNN)	0.9839	1.0349	0.4827	14.9131
长短期记忆网络(LSTM)	0.9806	0.4110	-1.5828	8.8996

时刻	直接经济损失/亿元
时刻	海鸥	彩虹	暹芭
登陆前12小时	21.251	5.757	0.3108
登陆前8小时	23.116	6.867	0.6244
登陆前4小时	28.520	8.016	0.7127
灾后统计	27.930	9.190	0.7770

[1]	LI Shuzhao, WEI Che, SHEN Chen, SUN Guodong, YANG Yetao, LUO Jinhua, WANG Jiaolong. A machine learning-based method for predicting shallow subsurface CPT parameters in offshore wind farms [J]. Journal of Tropical Oceanography, 2025, 44(4): 14-24.
[2]	Xuyang CHEN, Baoliang LIU. Application of Real-time monitoring buoy in monitoring red tide [J]. Journal of Tropical Oceanography, 2018, 37(5): 20-24.
[3]	Chaoyu YANG, Danling TANG, Haibin YE. A study on retrieving chlorophyll concentration by using GF-4 data [J]. Journal of Tropical Oceanography, 2017, 36(5): 33-39.
[4]	HUANG Hu,DAI Zhi-jun,SHI Wei-yong,SHENG Kai. Deposition characteristics of beach profile in strong-tidal environment − A case study of Yintan, Guangxi during spring [J]. Journal of Tropical Oceanography, 2011, 30(4): 71-76.

神经网络类型	训练集		测试集
神经网络类型	R²	RMSE	R²	RMSE
BP神经网络	0.9572	1.5668	0.8843	2.7933
GA-BP神经网络	0.9606	1.0669	0.9073	1.1635
卷积神经网络(CNN)	0.9368	1.6186	0.8127	3.3191
长短期记忆网络(LSTM)	0.8672	3.5678	0.8053	4.3044