Machine Learning for Flood forecasting

This project aims to develop and validate the feasibility of a machine learning (ML) approach to model and quantify overland flow volume during floods. The study will be conducted using existing knowledge of traditionally used hydrodynamic (HD) modelling and remote sensing (RS) combined with recently emerged ML technology. In recent years with technological advancement, ML techniques are considered as a future alternative to traditional flood inundation modelling in addition to many other aspects of water resource assessment.  ML approaches are specifically useful to reduce computational time and multiple scenario modelling. However, these emerging ML approaches need to be adapted and tested rigorously to evaluate if they are useful for case specific (such as overland flow estimation) application. This project will validate the approach and guide further work such that subsequent work can lead to production of a robust ML tool. The tool will be able to assess different aspects of seasonal floods, such as flood susceptibility mapping, inundation depth and overland flow volume, discharge-inundation area relationship. This will contribute in the area of digital water to address key issues of quantifying overland flow and identifying missing water (if any) in a river basin.