Spatial flood susceptibility assessment using boosting and bagging in machine learning techniques
مریم جهانبانی 1 , hossein aghamohammadi 2 , Mohammad Hassan vahidnia 3 , Zahra Azizi 4
1 - دانشکده منابع طبیعی و محیط زیست، دانشگاه آزاد اسلامی، واحد علوم و تحقیقات، تهران، ایران.
2 - Assistant Professor, Department of Remote Sensing and GIS, Faculty of Natural Resources and Environment, Science and Research Branch, Islamic Azad University, Tehran, Iran
3 - Assistant Professor, Department of Remote Sensing and GIS, Faculty of Natural Resources and Environment, Science and Research Branch, Islamic Azad University, Tehran, Iran
4 - Assistant Professor, Department of Remote Sensing and GIS, Faculty of Natural Resources and Environment, Science and Research Branch, Islamic Azad University, Tehran, Iran.
Keywords: flood, spatial information system, Ensemble machine learning models, Adaptive Boosting algorithm, Bagging algorithm.,
Abstract :
Every year flooding causes countries billions of dollars’ worth of damage that threatens the livelihood of individuals. As a result, it poses significant socio-economic threats to populations worldwide. Therefore, it should be controlled and restrained. In this regard, machine learning algorithms, along with geographic information systems, are primary tools that are effective in flood control modeling and analysis. The purpose of this research is to identify a part of flood-sensitive regions across the Heraz catchment area in Mazandaran province using ensemble methods in machine learning algorithms. The research process is as follows: first, the data of flood points were prepared. Next, 70% of approximately 240 sample positions were used for modeling and map preparation. The remaining 30%, which were randomly selected, were used to validate the produced maps. Then, the effective factors, including slope angle, slope direction, topography, soil type, land cover, distance from the river, annual rainfall, normalized difference vegetation index, index of sediment transmittance, index of topographic wetness, and index of stream density have been used to weight the impact of each factor using machine learning algorithms. Based on the results of this study, the system performance characteristic curve (ROC) was drawn, and the area under the curve (AUC) was calculated to validate the flood-prone area map. Findings demonstrated that the Adaptive Boosting model is more accurate than the Bagging model in preparing a flood sensitivity map. Predictive susceptibility mapping plays a pivotal role in enabling urban planners and managers to mitigate and safeguard proactively against the adverse consequences of flooding. Flood management authorities in the Ministry of Energy can employ the proposed ensemble model to assist disaster management and mitigate hazards in future studies. .