|
Flood risk zoning in Nowshahr city using machine learning models
|
Hassan Kazemi  , Nabiullah Mansouri * , Seyed Ali Jozi |
| Department of Environmental Engineering, Faculty of Natural Resources and Environment, Islamic Azad University, Science and Research Branch, Tehran, Iran , Nmansouri@gmail.com |
|
|
Abstract: (1865 Views) |
The first step in developing the prevention and safety plan for cities is to identify the existing hazards and crises and to prioritize the risk factors according to historical records and scientific studies to explain the probability of danger. Flood is one of the most high-risk natural hazards that should be given particular attention in crisis management. The study aims to model and predict flood hazard according to the characteristics of Nowshahr port city in Mazandaran province using machine learning models. The criteria and indicators affecting flood hazard, including climate criteria (6-hour rainfall indicator), hydrological criteria (surface runoff height indicators, drainage density, and groundwater depth), topographic criteria (slope and altitude indicators), Soil science criteria (soil type indicator) and land use criteria (land use type indicator) were identified based on the literature review. The indicators were then converted into raster data in the ArcGIS environment and finally standardized by fuzzy method for using in models. K-nearest neighbor models (Fine KNN, Cubic KNN, Cosine KNN, Weighted KNN), decision tree (Fine tree, Medium tree, Coarse tree), ensemble learning (Subspace KNN)) and WLC were used in the present study for flood risk zoning. ROC method was used to validate the models and choose the best one based on assessing the accuracy and area below the diagram of the flood. The layers were prepared in the ArcGIS environment and transferred into MATLAB to perform the modelling process and define the coefficients. The results show that the ensemble learning model with an accuracy of 0.96 and the area below the graph of 0.90 is the best and the WLC model with the accuracy of 0.61 and the area below the graph of 0.62 is the weakest model of flood hazard prediction in Nowshahr. Therefore, the output result of the Ensemble (Subspace KNN) algorithm was chosen to present the final flood risk map.
|
|
| Keywords: zoning, flood risk, machine learning, ensemble learning. |
|
|
Full-Text [PDF 2014 kb]
(928 Downloads)
|
Type of Study: Research |
Subject:
Special
Received: 2021/09/11 | Accepted: 2022/03/7 | Published: 2022/03/7
|
|
|
|
|
|
|
| Send email to the article author |
|
|
|
| Add your comments about this article |
|
|
|
