TY - JOUR JF - E.E.R.Journal JO - E.E.R. VL - 11 IS - 1 PY - 2021 Y1 - 2021/6/01 TI - Risk analysis of urban flood in Bandar Abbas using Machine Learning model and Analytic Hierarchy Process TT - تحلیل ریسک و آسیب‌پذیری سیلاب شهری بندرعباس با استفاده از مدل‌های ماشین بردار پشتیبان و بیشینه بی‌نظمی N2 - Extended abstract 1- Introduction Floods are one of the natural events that cause human casualties and damage to buildings, facilities, gardens, fields, and natural resources every year. Urbanization disturbs the balance of slopes through indirect intrusion within watersheds, kills vegetation, soil compaction, and changes in the profile of waterways, increases the severity of floods, and increases the amount of sediment generated. At the foot of the mountain, which includes the city's physical fabric expansion area, the natural drainage pattern disrupts and increases the risk of urban flooding. On the one hand, because of its geographical position and the heavy rainfall regime over a short period of time, and on the other hand, because of its significant growth and development, especially during the last decade, and because of its location, the town of Bandar Abbas faces flood problems. On the other hand, flood risk zoning has not been considered so far in order to be used in the planning and management of flood protection and control in Bandar Abbas, and not much work has been done in this area in the form of research and even studies. Flood risk zoning is therefore important in order to forecast the degree of flood damage under various circumstances and the economic and social basis for flood control and containment systems. Risk modeling and flood vulnerability mapping will play an important role in future decision-making, flood management, and land management in the area of the study in some cases. 2- Methodology In general, the first step in the implementation of research in watershed management, environmental, natural resources, agriculture, etc. projects is the preparation of the data used in that project. The data needed to investigate the hazard, vulnerability, and risk of urban floods in Bandar Abbas in the first stage are: 1- Height 2- Land slope 3- Distance from water table 4- Water transport capacity of canal 5- Distance from river 6- Distance from collection network of surface runoff was obtained from Iran Water Resources Management Organization and city demographic data Bandar Abbas was prepared by the Statistics Center of Iran. A map of four key factors, including building quality, urban density, population density and, socio-economic status, has been prepared to examine the vulnerability. Two models of Maximum Entropy (MaxEnt) and Support Vector Machine (SVM) were used to investigate the risk, vulnerability, and risk of flooding and urban flooding in Bandar Abbas. Then, the Area under Curve (AUC) obtained from the curve ROC was used in order to test the model's efficiency. After estimating flood hazard, a hierarchical analysis model was used in order to estimate flood vulnerability in this report. Finally, the map of flood risk was obtained based on hazard and vulnerability maps. 3- Results Based on the Maximum Irregularity Model, the results obtained from the flood risk prediction map showed that the southern parts of Bandar Abbas had a greater likelihood of flooding. It is also likely that parts of Bandar Abbas city center will be flooded. Bandar Abbas western and eastern areas are less likely to be flooded. Support has also shown that the southern, eastern, and southwestern regions are listed as likely to undergo urban flooding in order to help control urban floods. Using the SVM, the flood prediction map shows that the southern, eastern, and southwestern areas are more likely to flood; however, the northern and northwestern parts of Bandar Abbas are less likely to flood. The AUC was used in order to prepare the models. In the two phases of training and validation, the accuracy of the model suggests the highest irregularity. The Maximum Entropy Model, based on these curves, was 99.7% accuracy in the training phase and 94.2% accuracy in the validation phase. Therefore, in both the training and validation phases, the MaxEnt had excellent performance (area under the curve about 90%). The findings of hierarchical analysis have shown that the most important effective criterion for vulnerability is population density. The standard of construction, urban density and socio-economic status were ranked second, and fourth, respectively. Finally, on the basis of the risk map review, it can be claimed that there is a higher degree of risk in the southern parts of Bandar Abbas and parts of Bandar Abbas city center. 4- Discussion & Conclusions The results of the model accuracy evaluation show that SVM has accuracy in flood probability spatial prediction and in identifying areas vulnerable to flooding. It can also be seen that the SVM had better performance than the model of the support vector machine. Since most urban areas such as Bandar Abbas lack reliable hydraulic and hydrological knowledge, a new approach to flood management and urban flooding may be the use of machine learning models and historical urban flooding events. Machine learning models are highly capable of spatial analysis of flood events and urban flooding, as well as extracting the relationships between environmental variables and flood events, based on the tests conducted. The flood risk map revealed in this analysis that the central and southern sections of Bandar Abbas are more susceptible to flood. Prioritization and expert studies have shown that among the factors influencing vulnerability, the population density factor is the most significant. The map of vulnerability based on different factors also showed that there is a greater degree of vulnerability in the central and coastal areas. Comprehensive flood risk analysis has shown that there is a high risk of flooding in the southern and central parts of Bandar Abbas and these areas have a high priority for urban runoff and flood control. SP - 36 EP - 57 AU - ahmadi, yusef AU - bazrafshan, ommolbanin AU - salajeghe, ali AU - holisaz, arashk AU - Azare, Ali AD - Natural Resources Department, Faculty of Agriculture & Natural Resources, University of Hormozgan, BandarAbbas, Iran KW - Flood Vulnerability KW - Spatial Prediction KW - Support vector machines KW - Maxent model UR - http://magazine.hormozgan.ac.ir/article-1-603-en.html DO - 10.52547/jeer.11.1.36 ER -