基于K-means聚类与集成学习算法的小流域山洪灾害易发性评估

doi:10.3969/j.issn.0255-8297.2024.03.002

摘要/Abstract

摘要： 为了更好地分析空间异质性对山洪灾害易发性评估的影响，建立了基于K-means聚类与集成学习算法的小流域山洪灾害易发性评估模型。首先，选取中国江西省12 338个小流域为研究区，对各时段不同频率降雨量指标进行K-means聚类。其次，以误差平方和与平均轮廓系数为聚类效果评价指标，将小流域分为2个类内聚集、类外分散的子集。最后，针对不同子集，从几何特征、环境特征以及降水特征3个方面选取平均坡度、形心高程、形状系数、最长汇流路径比降、地形湿度指数、归一化植被指数、距离河流最近距离、降雨量、洪峰模数以及汇流时间10个山洪影响因素，应用自适应增强算法与极致梯度提升算法进行山洪灾害易发性评估。研究发现，降水是导致山洪灾害的重要因素，江西省高降水区域山洪灾害易发程度普遍高于低降水区，同时省内高风险区分布较为分散，主要分布在东北区域与西北边缘区域。对聚类后两类相似小流域分别进行山洪易发性评估，接受者操作特征曲线下面积值均在0.90以上，精度较聚类前有所提高。聚类策略作为易发性评估模型的前驱过程，可以有效解决小流域异质性问题。

关键词: 空间异质性, K-means聚类, 集成学习, 自适应增强, 极致梯度提升, 山洪灾害

Abstract: In this paper, a model based on K-means clustering and ensemble learning approaches is developed to properly analyze the impact of spatial heterogeneity on the assessment of flash flood disaster susceptibility. Firstly, 12 338 catchments in Jiangxi Province, China, are selected as the study area, where the K-means clustering is performed on different frequency rainfall indicators for each period. Secondly, using the error sum of squares and mean contour coefficients as the clustering evaluation index, the small catchment datasets are divided into two subsets. Finally, for different subsets, ten flash flood influencing factors such as average slope, normalized difference vegetation index and rainfall are selected from geometric characteristics, environmental characteristics, and precipitation characteristics. The adaptive boosting (AdaBoost) and eXtreme gradient boosting (XGBoost) models are applied to evaluate the susceptibility of flash floods. It is found that precipitation is an important factor in flash floods disaster, and flash floods are more likely to occur in high precipitation areas in Jiangxi Province. Meanwhile, the distribution of high-risk areas is dispersed, mainly in the northeastern region and the northwestern edge. The area under the receiver operating characteristic curve (AUC) values of similar catchments could increase to 0.90 or above after clustering. The clustering model effectively addresses the heterogeneity of catchments as a precursor process for susceptibility assessment.

Key words: spatial heterogeneity, K-means clustering, ensemble learning, adaptive boosting (AdaBoost), eXtreme gradient boosting (XGBoost), flash floods disaster

中图分类号:

P208

管筝, 印涌强, 张晓祥, 陈跃红. 基于K-means聚类与集成学习算法的小流域山洪灾害易发性评估[J]. 应用科学学报, 2024, 42(3): 388-404.

GUAN Zheng, YIN Yongqiang, ZHANG Xiaoxiang, CHEN Yuehong. Estimating Flash Flood Disaster Susceptibility Based on K-means Clustering and Ensemble Learning Approaches[J]. Journal of Applied Sciences, 2024, 42(3): 388-404.

参考文献

[1] Reichenbach P, Rossi M, Malamud B D, et al. A review of statistically-based landslide susceptibility models [J]. Earth Science Reviews, 2018, 180: 60-91.
[2] Rahmati O, Haghizadeh A, Stefanidis S. Assessing the accuracy of GIS-based analytical hierarchy process for watershed prioritization; Gorganrood River basin, Iran [J]. Water Resources Management, 2016, 30(3): 1131-1150.
[3] Youssef A M, Pradhan B, Sefry S A. Flash flood susceptibility assessment in Jeddah City (Kingdom of Saudi Arabia) using bivariate and multivariate statistical models [J]. Environmental Earth Sciences, 2015, 75(1): 12.
[4] Zhong M, Zeng T, Jiang T, et al. A Copula-based multivariate probability analysis for flash flood risk under the compound effect of soil moisture and rainfall [J]. Water Resources Management, 2021, 35(1): 83-98.
[5] Khosravi K, Pham B T, Chapi K, et al. A comparative assessment of decision trees algorithms for flash flood susceptibility modeling at Haraz watershed, northern Iran [J]. Science of the Total Environment, 2018, 627: 744-755.
[6] Costache R. Flash-flood potential assessment in the upper and middle sector of Prahova River catchment (Romania): a comparative approach between four hybrid models [J]. Science of the Total Environment, 2019, 659: 1115-1134.
[7] Hosseini F S, Choubin B, Mosavi A, et al. Flash-flood hazard assessment using ensembles and Bayesian-based machine learning models: application of the simulated annealing feature selection method [J]. Science of the Total Environment, 2020, 711: 135161.
[8] Arabameri A, Saha S, Chen W, et al. Flash flood susceptibility modelling using functional tree and hybrid ensemble techniques [J]. Journal of Hydrology, 2020, 587: 125007.
[9] Bui D T, Tsangaratos P, Ngo P T T, et al. Flash flood susceptibility modeling using an optimized fuzzy rule based feature selection technique and tree based ensemble methods [J]. Science of the Total Environment, 2019, 668: 1038-1054.
[10] Ma M H, Zhao G, He B S, et al. XGBoost-based method for flash flood risk assessment [J]. Journal of Hydrology, 2021, 598: 126382.
[11] Chen W, Li Y, Xue W F, et al. Modeling flood susceptibility using data-driven approaches of Naïve Bayes tree, alternating decision tree, and random forest methods [J]. Science of the Total Environment, 2020, 701: 134979.
[12] Dodangeh E, Choubin B, Eigdir A N, et al. Integrated machine learning methods with resampling algorithms for flood susceptibility prediction [J]. Science of the Total Environment, 2020, 705: 135983.
[13] Bandara K, Bergmeir C, Smyl S. Forecasting across time series databases using recurrent neural networks on groups of similar series: a clustering approach [J]. Expert Systems with Applications, 2020, 140: 112896.
[14] Lin K R, Chen H Y, Xu C Y, et al. Assessment of flash flood risk based on improved analytic hierarchy process method and integrated maximum likelihood clustering algorithm [J]. Journal of Hydrology, 2020, 584: 124696.
[15] Jiang S J, Zheng Y, Wang C, et al. Uncovering flooding mechanisms across the contiguous United States through interpretive deep learning on representative catchments [J]. Water Resources Research, 2022, 58(1): e2021WR030185.
[16] Zhai X Y, Zhang Y Y, Zhang Y Q, et al. Simulating flash flood hydrographs and behavior metrics across China: implications for flash flood management [J]. Science of the Total Environment, 2021, 763: 142977.
[17] 张帆, 张永勇, 陈俊旭, 等. 多种机器学习模型对不同洪水类型特征指标模拟效果评估[J]. 地理科学进展, 2022, 41(7): 1239-1250. Zhang F, Zhang Y Y, Chen J X, et al. Performance of multiple machine learning model simulation of process characteristic indicators of different flood types [J]. Progress in Geography, 2022, 41(7): 1239-1250. (in Chinese)
[18] Xu H S, Ma C, Lian J J, et al. Urban flooding risk assessment based on an integrated Kmeans cluster algorithm and improved entropy weight method in the region of Haikou, China [J]. Journal of Hydrology, 2018, 563: 975-986.
[19] 樊建勇, 单九生, 管珉, 等. 江西省小流域山洪灾害临界雨量计算分析[J]. 气象, 2012, 38(9): 1110- 1114. Fan J Y, Shan J S, Guan M, et al. Research on analysis and calculation method of critical precipitation of mountain torrents in Jiangxi Province [J]. Meteorological Monthly, 2012, 38(9): 1110-1114. (in Chinese)
[20] 张若婧, 陈跃红, 张晓祥, 等. 基于参数最优地理探测器的江西省山洪灾害时空格局与驱动力研究[J]. 地理与地理信息科学, 2021, 37(4): 72-80. Zhang R J, Chen Y H, Zhang X X, et al. Spatial-temporal pattern and driving factors of flash flood disasters in Jiangxi Province analyzed by optimal parameters-based geographical detector [J]. Geography and Geo-Information Science, 2021, 37(4): 72-80. (in Chinese)
[21] 郑彦辰, 李建柱, 荣佑同, 等. 降雨时空分布量化及其在洪水过程分类中的应用[J]. 水利学报, 2022, 53(5): 560-573. Zheng Y C, Li J Z, Rong Y T, et al. Quantification of rainfall spatial and temporal distribution characteristics on the flood hydrograph and its application in flood type classification [J]. Journal of Hydraulic Engineering, 2022, 53(5): 560-573. (in Chinese)
[22] Tien B D, Pradhan B, Nampak H, et al. Hybrid artificial intelligence approach based on neural fuzzy inference model and metaheuristic optimization for flood susceptibility modeling in a high-frequency tropical cyclone area using GIS [J]. Journal of Hydrology, 2016, 540: 317-330.
[23] Liu Y S, Yang Z S, Huang Y H, et al. Spatiotemporal evolution and driving factors of China's flash flood disasters since 1949[J]. Science China Earth Sciences, 2018, 61(12): 1804-1817.
[24] Ragettli S, Zhou J, Wang H, et al. Modeling flash floods in ungauged mountain catchments of China: a decision tree learning approach for parameter regionalization [J]. Journal of Hydrology, 2017, 555: 330-346.
[25] 李青, 王雅莉, 李海辰, 等. 基于洪峰模数的山洪灾害雨量预警指标研究[J]. 地球信息科学学报, 2017, 19(12): 1643-1652. Li Q, Wang Y L, Li H C, et al. Rainfall threshold for flash flood early warning based on flood peak modulus [J]. Journal of Geo-information Science, 19(12): 1643-1652. (in Chinese)
[26] Khosravi K, Nohani E, Maroufinia E, et al. A GIS-based flood susceptibility assessment and its mapping in Iran: a comparison between frequency ratio and weights-of-evidence bivariate statistical models with multi-criteria decision-making technique [J]. Natural Hazards, 2016, 83(2): 947-987.
[27] 郭良, 丁留谦, 孙东亚, 等. 中国山洪灾害防御关键技术[J]. 水利学报, 2018, 49(9): 1123-1136. Guo L, Ding L Q, Sun D Y, et al. Key techniques of flash flood disaster prevention in China [J]. Journal of Hydraulic Engineering, 2018, 49(9): 1123-1136. (in Chinese)
[28] Roy P, Chandra P S, Chakrabortty R, et al. Threats of climate and land use change on future flood susceptibility [J]. Journal of Cleaner Production, 2020, 272: 122757.
[29] 周志华. 机器学习[M]. 北京: 清华大学出版社, 2016.
[30] Rousseeuw P J. Silhouettes: a graphical aid to the interpretation and validation of cluster analysis [J]. Journal of Computational and Applied Mathematics, 1987, 20(1): 53-65.
[31] Freund Y, Schapire R E. A decision-theoretic generalization of on-line learning and an application to boosting [J]. Journal of Computer and System Sciences, 1997, 55(1): 119-139.
[32] 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, 2016: 785-794.
[33] Khosravi K, Shahabi H, Pham B T, et al. A comparative assessment of flood susceptibility modeling using multi-criteria decision-making analysis and machine learning methods [J]. Journal of Hydrology, 2019, 573: 311-323.
[34] 吴广建, 章剑林, 袁丁. 基于K-means的手肘法自动获取K值方法研究[J]. 软件, 2019, 40(5): 167-170. Wu G J, Zhang J L, Yuan D. Automatically obtaining K value based on K-means elbow method [J]. Computer Engineering & Software, 2019, 40(5): 167-170. (in Chinese)
[35] Yao J, Zhang X X, Luo W C, et al. Applications of stacking/blending ensemble learning approaches for evaluating flash flood susceptibility [J]. International Journal of Applied Earth Observation and Geoinformation, 2022, 112: 102932.
[36] Xiong J N, Li J, Cheng W M, et al. A GIS-based support vector machine model for flash flood vulnerability assessment and mapping in China [J]. ISPRS International Journal of GeoInformation, 2019, 8(7): 297.