Urban flood forecasting using a hybrid modeling approach based on a deep learning technique

Abstract Climate change is contributing to the increasing frequency and severity of flooding worldwide. Therefore, forecasting preparing for floods while considering extreme climate conditions are essential decision-makers prevent manage disasters. Although recent studies have demonstrated potential long short-term memory (LSTM) models rainfall-related runoff, there remains room improvement due lack observational data. In this study, we developed a flood model based on hybrid modeling approach that combined rainfall-runoff deep learning model. Furthermore, proposed method time using several representative rainfall variables. The study focused urban river basins, amounts, duration, distribution create virtual scenarios. Additionally, simulated results were used as input data forecast under other conditions. prediction achieved high accuracy with correlation coefficient >0.9 Nash[ndash]Sutcliffe efficiency >0.8. These indicated would enable reasonable occurrences their timing only forecasted information.