Forecasting flood inundation in urban areas is challenging due to the lack of validation data. Recent developments have led to new genres of data sources, such as images and videos from smartphones and CCTV cameras. If the reference dimensions of objects, such as bridges or buildings, in images are known, the images can be used to estimate water levels using computer vision algorithms. Such algorithms employ deep learning and edge detection techniques to identify the water surface in an image, which can be used as additional validation data for forecasting inundation. In this study, a methodology is presented for flood inundation forecasting that integrates validation data generated with the assistance of computer vision. Six equifinal models are run simultaneously, one of which is selected for forecasting based on a goodness-of-fit (least error), estimated using the validation data. Collection and processing of images is done offline on a regular basis or following a flood event. The results show that the accuracy of inundation forecasting can be improved significantly using additional validation data.

The paper presents a new methodology for hydrodynamic-based flood forecast that focuses on scenario generation and database queries to select appropriate flood inundation maps in real-time. In operational flood forecasting, only discharges are forecasted at specific gauges using hydrological models. Hydrodynamic models, which are required to produce inundation maps, are computationally expensive, hence not feasible for real-time inundation forecasting. In this study, we have used a substantial number of pre-calculated inundation maps that are stored in a database and a methodology to extract the most likely maps in real-time. The method uses real-time discharge forecast at upstream gauge as an input and compares it with the pre-recorded scenarios. The results show satisfactory agreements between offline inundation maps that are retrieved from a pre-recorded database and online maps, which are hindcasted using historical events. Furthermore, this allows an efficient early warning system, thanks to the fast run-time of the proposed offline selection of inundation maps. The framework is validated in the city of Kulmbach in Germany.

The consideration of uncertainties in flood risk assessment has received increasing attention over the last 2 decades. However, the assessment is not reported in practice due to the lack of best practices and too wide uncertainty bounds. We present a method to constrain the model roughness based on measured water levels and reduce the uncertainty bounds of a two-dimensional hydrodynamic model. Results show that the maximum uncertainty in roughness generated an uncertainty bound in the water level of 1.26 m (90 % confidence interval) and by constraining roughness, the bounds can be reduced as much as 0.92 m.

In operational flood risk management, a single best model is used to assess the impact of flooding, which might misrepresent uncertainties in the modelling process. We have used quantified uncertainties in flood forecasting to generate flood hazard maps that were combined based on different exceedance probability scenarios. The purpose is to differentiate the impacts of flooding depending on the building use, enabling, therefore, more flexibility for stakeholders' variable risk perception profiles. The aim of the study is thus to develop a novel methodology that uses a multi-model combination of flood forecasting models to generate flood hazard maps with differentiated exceedance probability. These maps take into account uncertainties stemming from the rainfall–runoff generation process and could be used by decision makers for a variety of purposes in which the building use plays a significant role, e.g. flood impact assessment, spatial planning, early warning and emergency planning.

Operational real-time flood forecast is often done on the prediction of discharges at specific gauges using hydrological models. Hydrodynamic models, which can produce inundation maps, are computationally demanding and often cannot be used directly for that purpose. The FloodEvac framework has been developed in order to enable 2D flood inundations map to be forecasted at real-time. The framework is based on a database of pre-recorded synthetic events. In this paper, the framework is improved by generating a database based on rescaled historical river discharge events. This historical database includes a wider variety of runoff curves, including non-Gaussian and multi-peak shapes that better reflect the characteristics and the behavior of the natural streams. Hence, a hybrid approach is proposed by joining the historical and the existing synthetic database. The increased number of scenarios in the hybrid database allows reliable predictions, thus improving the robustness and applicability of real-time flood forecasts.