Nowcasting Floods in Detailed Scales Considering the Uncertainties Associated with impact-based Practical Applications

Impact-based nowcasting systems at detailed scales, to the street level, have become essential in flood risk management. This is achieved by focusing on predicting the impacts of flood events rather than merely forecasting weather conditions. This approach leverages advancements in 2D hydrodynamic modelling, high-performance computing (HPC), and detailed rainfall forecasting to improve the precision of early warning systems. However, its real-world implementation is hindered by challenges such as the coarse temporal resolution of weather forecasts and inherent modelling uncertainties. This study investigates the uncertainties and challenges associated with impact-based nowcasting systems, using the Mandra town (Greece) as a case study. We demonstrate the feasibility of applying a comprehensive framework that integrates 2D hydrodynamic modelling, HPC, and temporally disaggregated rainfall forecasting. Our findings show that the Alternating Block Method (ABM) effectively captures storm dynamics, mitigating significant underestimations that arise from coarser forecast inputs. Additionally, we assess various flood impact indices to manage modelling uncertainties. Our results highlight that similarities exist in the flood indices when storms are mild with short return periods. However, discrepancies between indices increase with storms of longer return periods, underscoring the critical need for careful index selection. This research provides new insights into enhancing flood nowcasting accuracy and effectiveness, particularly in small to medium-sized catchments. Moreover, it offers evidence that the scientific community along with the stakeholders such as Civil Protection, local governments, and others should focus orient their efforts on more reliable flood indices, as the discrepancies between the methodologies investigated increase with the severity of the events.