Title

結合隨機森林與自組織映射於淹水災害分區評估

Translated Titles

Assessment of flood hazard zoning using an integrated approach of random forest and self-organizing map

DOI

10.6342/NTU201902265

Authors

黃韻如

Key Words

淹水災害分區模式 ; 淹水潛勢圖 ; 隨機森林 ; 自適應增強 ; 自組織映射 ; Flood hazard zoning model ; Flood susceptibility map ; Random forest ; Adaptive boosting ; Self-organizing map

PublicationName

臺灣大學土木工程學研究所學位論文

Volume or Term/Year and Month of Publication

2019年

Academic Degree Category

碩士

Advisor

林國峰

Content Language

繁體中文

Chinese Abstract

台灣位於西太平洋的颱風路徑要衝,每年平均受到三至四場颱風侵襲,帶來豐沛的雨量。又因台灣地形陡峭,上游逕流短時間內即抵達下游地區,容易造成淹水災害。而近年來,受到氣候變遷的影響,台灣颱風侵襲數量減少,但強颱比例上升,且極端降雨日數增加,這樣的變化,使降雨誘發淹水災害的機會上升。淹水不僅造成經濟損失、交通道路阻斷,也會威脅到人民的性命安全。因此,本研究期望建立淹水災害分區模式,映射淹水災害分區圖,提供作為防災參考依據,以達到減災的目的。 本研究提出以機器學習結合聚類型演算法建立淹水災害分區模式,其中包含兩個部分:第一部分,使用機器學習法中的隨機森林 (random forest, RF) 以及自適應增強 (adaptive boosting, AdaBoost) 分別建立淹水潛勢模式產生潛勢值;第二部分,將潛勢值匯入自組織映射 (self-organizing map, SOM) 完成淹水災害分區模式。同時,本研究考慮兩種不同的SOM輸入項:(1)僅使用自身網格潛勢值,(2)使用自身及周圍網格潛勢值。除此之外,亦以傳統用於災害分區的自然斷點法 (natural breaks, NB) 建立淹水災害分區模式,與本研究提出之模式相互比較。 本研究根據歷史淹水事件與區域,選擇台灣宜蘭縣平地地區共十個鄉鎮作為研究區域。蒐集2004年到2015年的期間造成淹水的8場颱風與1場豪雨事件資料,以及12個淹水影響因子,包含高程、坡度、坡向、曲率、平面曲率、剖面曲率、地形濕度指數、逕流強度指數、與河道距離、與排水渠距離、土地使用與最大小時累積雨量。 結果顯示,本研究提出使用自身及周圍網格的潛勢值建立之模式能夠改善模式對於淹水風險的劃分,且提出的模式優於傳統模式。其中又以RF結合使用自身及周圍網格作為輸入項的SOM所建立之淹水災害分區模式表現最好。最終,本研究探討淹水影響因子對於淹水風險等級的關係,其中以土地使用、與河道距離、高程、坡度與降雨對於淹水風險劃分的影響最劇。未來可依據本研究所發展之淹水災害分區模式,協助相關管理機關擬定適當的防災策略。

English Abstract

Taiwan is located on the main track of western Pacific typhoons, and approximately three to four typhoons hit Taiwan per year. The heavy rainfall frequently results in severe flooding in downstream areas. Moreover, it is accepted that the number of typhoons which attack Taiwan will decrease, the proportion of strong typhoons and days of extreme rainfall will increase considerably in recent years due to climate change. This changes have increased the chances of rainfall-induced flooding. Flood cause hung losses of human life, property, and devastations to environment. Therefore, the flood hazard zoning model is an important tool for flood prevention and can efficiently mitigate the disasters. In this study, a flood hazard zoning model based on machine learning and cluster algorithm is proposed. The model contains two steps. First, two kinds of machine learning, random forest (RF) and adaptive boosting (AdaBoost), are employed to construct flood susceptibility models to yield flood susceptibility values, respectively. Second, the flood susceptibility values are then input to self-organizing map (SOM) to obtain the flood hazard zones. In addition, two different inputs for SOM are considered: (i) only the flood susceptibility value of the grid itself is used as input, and (ii) flood susceptibility values of the self and surrounding grids are used as inputs. After construct all flood hazard zoning models, decide the proposed models which using the better inputs for SOM in the RF based models and AdaBoost based models. For comparison with the proposed models, the traditional models based on the natural break (NB) are also constructed. Ten villages at Yilan County in northeastern Taiwan are selected as the study area. All the flood-related factors and the historical flooding data are based on GIS technology. Flood events from 2004 to 2015 are collected. Moreover, twelve flood-related factors are used in this study, namely elevation, slope, aspect, curvature, plan curvature, profile curvature, topographic wetness index, stream power index, distance to river, distance to drainage, land use and maximum hour rainfall. The results confirm that the proposed models with the flood susceptibility values of the self and surrounding grids does improve the assessment performance. The proposed models are better performed than the traditional models. Furthermore, the proposed model, based on the RF integrated with the SOM using the flood susceptibility values of the self and surrounding grids as input, yields the most reasonable flood hazard zoning results. Finally, regarding the relation of factors and flood hazard zoning, the land use, distance to drainage, elevation, slope and the maximum 1-h rainfall have great influence on flood hazard zoning. The proposed flood hazard zoning model is expected to be useful to support the formulation of adequate disaster mitigation strategies.

Topic Category 工學院 > 土木工程學研究所
工程學 > 土木與建築工程
Reference
  1. 1. Abella, E. A. C., Van Westen, C. J., 2008. Qualitative landslide susceptibility assessment by multicriteria analysis: a case study from San Antonio del Sur, Guantánamo, Cuba. Geomorphology 94(3-4), 453-466.
  2. 2. Alfieri, L., Salamon, P., Bianchi, A., Neal, J., Bates, P., Feyen, L., 2014. Advances in pan‐European flood hazard mapping. Hydrological processes 28(13), 4067-4077.
  3. 3. Al-Abadi, A. M., 2018. Mapping flood susceptibility in an arid region of southern Iraq using ensemble machine learning classifiers: a comparative study. Arabian Journal of Geosciences 11(9), 218.
  4. 4. Bui, D. T., Tuan, T. A., Klempe, H., Pradhan, B., Revhaug, I., 2016. Spatial prediction models for shallow landslide hazards: a comparative assessment of the efficacy of support vector machines, artificial neural networks, kernel logistic regression, and logistic model tree. Landslides 13(2), 361-378.
  5. 5. Breiman, L., Friedman, J.H., Olsen, R.A., Stone, C.J., 1984. Classification and regression trees. Wadsworth and Brooks, New York, USA.
  6. 6. Breiman, L., 1996b. Out-of-bag estimation. Technical Report, Statistics Department, University of California
  7. 7. Breiman, L., 2001. Random Forests. Machine Learning 45, 5-32.
  8. 8. Chen, W., Xie, X., Wang, J., Pradhan, B., Hong, H., Bui, D. T., Duan, Z., Ma, J., 2017. A comparative study of logistic model tree, random forest, and classification and regression tree models for spatial prediction of landslide susceptibility. Catena 151, 147-160.
  9. 9. Chang, L. C., Shen, H. Y., Chang, F. J., 2014. Regional flood inundation nowcast using hybrid SOM and dynamic neural networks. Journal of hydrology 519, 476-489.
  10. 10. Chung, C. J. F., Fabbri, A. G., 2003. Validation of spatial prediction models for landslide hazard mapping. Natural Hazards 30(3), 451-472.
  11. 11. Das, S., 2018. Geographic information system and AHP-based flood hazard zonation of Vaitarna basin, Maharashtra, India. Arabian Journal of Geosciences 11(19), 576.
  12. 12. Freund, Y., Schapire, R. E., 1997. A decision-theoretic generalization of on-line learning and an application to boosting. Journal of computer and system sciences 55(1), 119-139.
  13. 13. Gnecco, G., Morisi, R., Roth, G., Sanguineti, M., Taramasso, A. C., 2017. Supervised and semi-supervised classifiers for the detection of flood-prone areas. Soft Computing 21(13), 3673-3685.
  14. 14. Hallegatte, S., Green, C., Nicholls, R. J., Corfee-Morlot, J., 2013. Future flood losses in major coastal cities. Nature climate change 3(9), 802.
  15. 15. Hong, H., Liu, J., Bui, D. T., Pradhan, B., Acharya, T. D., Pham, B. T., Zhu, A., Chen. W., Ahmad, B. B., 2018. Landslide susceptibility mapping using J48 Decision Tree with AdaBoost, Bagging and Rotation Forest ensembles in the Guangchang area (China). Catena 163, 399-413.
  16. 16. Kohonen, T., 1982. Self-organized formation of topologically correct feature maps. Biological cybernetics 43(1), 59-69.
  17. 17. Khosravi, K., Nohani, E., Maroufinia, E., Pourghasemi, H. R., 2016. 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. Natural Hazards 83(2), 947-987.
  18. 18. Kanani-Sadat, Y., Arabsheibani, R., Karimipour, F., Nasseri, M., 2019. A new approach to flood susceptibility assessment in data-scarce and ungauged regions based on GIS-based hybrid multi criteria decision-making method. Journal of Hydrology 572, 17-31.
  19. 19. Lin, G. F., Chen, L. H., 2006. Identification of homogeneous regions for regional frequency analysis using the self-organizing map. Journal of Hydrology 324(1-4), 1-9.
  20. 20. Lin, G. F., Lin, H. Y., Chou, Y. C., 2013. Development of a real-time regional-inundation forecasting model for the inundation warning system. Journal of Hydroinformatics, 15(4), 1391-1407.
  21. 21. Lin, G. F., Chang, M. J., Huang, Y. C., Ho, J. Y., 2017. Assessment of susceptibility to rainfall-induced landslides using improved self-organizing linear output map, support vector machine, and logistic regression. Engineering geology 224, 62-74.
  22. 22. Lian, J., Xu, H., Xu, K., Ma, C., 2017. Optimal management of the flooding risk caused by the joint occurrence of extreme rainfall and high tide level in a coastal city. Natural Hazards 89(1), 183-200.
  23. 23. Liu, K., Wang, M., Cao, Y., Zhu, W., Yang, G., 2018. Susceptibility of existing and planned Chinese railway system subjected to rainfall-induced multi-hazards. Transportation Research Part A: Policy and Practice 117, 214-226.
  24. 24. Jhong, B. C., Wang, J. H., Lin, G. F., 2016. Improving the Long Lead-Time Inundation Forecasts Using Effective Typhoon Characteristics. Water resources management 30(12), 4247-4271.
  25. 25. Jhong, B. C., Wang, J. H., Lin, G. F., 2017. An integrated two-stage support vector machine approach to forecast inundation maps during typhoons. Journal of hydrology 547, 236-252.
  26. 26. Micheletti, N., Foresti, L., Robert, S., Leuenberger, M., Pedrazzini, A., Jaboyedoff, M., Kanevski, M., 2014. Machine learning feature selection methods for landslide susceptibility mapping. Mathematical Geosciences 46(1), 33-57.
  27. 27. Mondal, S., Mandal, S., 2018. Landslide susceptibility mapping of Darjeeling Himalaya, India using index of entropy (IOE) model. Applied Geomatics 1-18.
  28. 28. Nandi, A., Mandal, A., Wilson, M., Smith, D., 2016. Flood hazard mapping in Jamaica using principal component analysis and logistic regression. Environmental Earth Sciences 75(6), 465.
  29. 29. Pham, B. T., Bui, D. T., Prakash, I., Dholakia, M. B., 2016. Rotation forest fuzzy rule-based classifier ensemble for spatial prediction of landslides using GIS. Natural Hazards 83(1), 97-127.
  30. 30. Quinlan JR, 1993 C4.5: programs for machine learning. Morgan Kaufmann, San Mateo, CA, USA
  31. 31. Rojas, O., Mardones, M., Rojas, C., Martínez, C., Flores, L.,2017. Urban growth and flood disasters in the coastal river basin of south-central Chile (1943–2011). Sustainability 9(2), 195.
  32. 32. Siddique, M. S., Schwarz, J., 2015. Elaboration of multi-hazard zoning and qualitative risk maps of Pakistan. Earthquake Spectra 31(3), 1371-1395.
  33. 33. Sameen, M. I., Pradhan, B., Lee, S., 2018. Self-learning random forests model for mapping groundwater yield in data-scarce areas. Natural Resources Research 1-19.
  34. 34. Trigila, A., Iadanza, C., Esposito, C., Scarascia-Mugnozza, G., 2015. Comparison of Logistic Regression and Random Forests techniques for shallow landslide susceptibility assessment in Giampilieri (NE Sicily, Italy). Geomorphology 249, 119-136.
  35. 35. Vojtek, M., Vojteková, J., 2019. Flood Susceptibility Mapping on a National Scale in Slovakia Using the Analytical Hierarchy Process. Water 11(2), 364.
  36. 36. Wilby, R. L., Beven, K. J., Reynard, N. S., 2008. Climate change and fluvial flood risk in the UK: More of the same? Hydrological Processes: An International Journal 22(14), 2511-2523.
  37. 37. Wang, Z., Lai, C., Chen, X., Yang, B., Zhao, S., Bai, X., 2015. Flood hazard risk assessment model based on random forest. Journal of Hydrology 527, 1130-1141.
  38. 38. Wu, T., Li, H. C., Wei, S. P., Chen, W. B., Chen, Y. M., Su, Y. F., Liu, J. J., Shih, H. J., 2016. A comprehensive disaster impact assessment of extreme rainfall events under climate change: a case study in Zheng-wen river basin, Taiwan. Environmental Earth Sciences 75(7), 597.
  39. 39. Xu, H., Ma, C., Lian, J., Xu, K., Chaima, E., 2018. Urban flooding risk assessment based on an integrated k-means cluster algorithm and improved entropy weight method in the region of Haikou, China. Journal of hydrology 563, 975-986.
  40. 40. Yilmaz, I., 2010. Comparison of landslide susceptibility mapping methodologies for Koyulhisar, Turkey: conditional probability, logistic regression, artificial neural networks, and support vector machine. Environmental Earth Sciences 61(4), 821-836.
  41. 41. Zheng, F., Westra, S., Leonard, M., Sisson, S. A., 2014. Modeling dependence between extreme rainfall and storm surge to estimate coastal flooding risk. Water Resources Research 50(3), 2050-2071.
  42. 42. 周佳,陳維婷,羅敏輝,李明安,許晃雄,洪志誠,鄒治華,盧孟明,洪致文,陳正達,鄭兆尊,2017。臺灣氣候變遷科學報告2017–物理現象與機制。