Effect of the dependence structure and time irreversibility of streamflow in flood inundation mapping with focus on the long-term persistence behaviour

The marginal structure of streamflow, with focus on the right tail behaviour, is considered as the main factor in flood risk assessment, while little is known on the effect of the temporal dependence structure and irreversibility of streamflow. Interestingly, the second-order dependence behaviour of streamflow is shown to highly deviate from a white noise behaviour (i.e., temporal independence), and to rather exhibit a Hurst-Kolmogorov (HK) behaviour, with strong autocorrelation and irreversibility at small scales and long-term persistence at large scales. The HK dynamics is known to be characterized by large uncertainty and variability, and therefore, it is expected to have a non-negligible impact on flood inundation mapping, especially in cases of successive storm events. Through benchmark experiments and real case scenarios, we investigate the influence of these effects in several output features of flood risk modelling such as flood depth, velocity, and duration, and we discuss possible consequences for insurance policies.