Application of machine learning in flood forecasting

Abstract

A flood is a costly natural disaster that imposes a considerable risk to many urban areas worldwide. Predicting the flood can help to alleviate the damage that it causes. In recent years, inspired by the success of Machine Learning (ML) in other fields, several papers have proposed ML-based algorithms for short-term and long-term flood prediction. In this study, we aim to give an overview of ongoing research in this area. We present several case studies from recent papers that employed machine learning for flood forecasting. Results of these studies have shown that ML models are powerful tools in flood prediction. We also briefly reviewed some of the commonly used machine learning models. However, the technical description of these methods is beyond the scope of this study and is not discussed in detail. Although ML models have shown promising results in flood forecasting, they do suffer from important limitations, and there is still room for their improvement. In this paper, we attempted to critically assess the shortcomings of existing algorithms and offer several suggestions for the future direction of this research field.