A Monte Carlo simulation study of the factors influencing the performance of flood early warning systems

Abstract

In recent decades, flood early warning systems (FEWSs) have been widely used as complementary non-structural mitigation measures in order to improve the population resilience to floods. FEWS research focusses mainly on flood forecasting techniques or social aspects of warning response, and end-to-end modelling frameworks that represent the entire FEWS forecast-decisionresponse/impact chain are rarely developed. A generic Monte Carlo simulation framework has been developed that represents an end-to-end FEWS in a versatile way, allowing factors influencing FEWS performance to be explored which cannot be analysed easily based on limited realworld data. The framework has been applied to a simulated generic fluvial case, where factors influencing FEWS performance in terms of reliability and economic effectiveness are explored. A new reliability performance measure based on inundation maps has been proposed. The framework has also been used to explore factors controlling the performance of a simulated FEWS representing an urban polder in Nanjing, China, with performance metrics based on waterlogging and pumping costs. For the generic fluvial case, the main results show that: i) the correlation between forecasts and observed values controls reliability; ii) probabilistic forecasts based on optimising a probabilistic threshold are robust to forecast biases in the mean and variance, iii) a FEWS based on uncertain forecasts is characterised by an optimal lead time that represents a balance between an adequate time to act in response and a reasonably good forecast; iv) the performance of the proactive action is the most important factor influencing the economic effectiveness of a FEWS. For the simulated flood-prone polder system case study, the results show that probabilistic forecasts of storm rainfall and runoff volume can considerably enhance the waterlogging and pumping metrics. The results of this research can be used to improve the performance of fluvial FEWSs, and to design FEWSs for polder systems.