Apart from the estimation of magnitudes of precipitation, floods and storm surges, modelling of storm water flows in a densely populated urban area is required for designing coping strategies and making decisions. Incorporating surface runoff and conduit flow modelling capabilities has enabled the prediction of urban flood hazards. This study synthesises methodologies for simulating flood processes and evaluating flood hazards in urban environment. Existing models and their associated uncertainties are summarised, and state-of-the-art techniques to build up a numerical model for simulating urban floods and the applications to specific cases are illustrated. A schematic framework for urban flood hazard prediction is proposed, within which multi-source observation retrieval, physics-based modelling, parameter optimisation, uncertainty estimation, model-observation fusion, evaluation of compound effects of multiple factors and digital twin techniques are included. The major challenges and uncertainties in flood process modelling originate from input data, model structures, validation processes and compounding effects. Multidisciplinary techniques for estimating the input data and enhancing the efficiency and accuracy of the flood evaluation should be developed. Great efforts are needed in understanding the process-dependent indicators, coupled modelling and data-model assimilation. Determining the probability of compound floods and understanding the driving factors are also essential for evaluating flood risks.