Inferring extreme river floods from everyday hydrologic dynamics

The idea that extreme river floods are intrinsically different from smaller, more frequent floods underlain the estimation of possible flood magnitudes for decades, and was further advocated in recent years. At the same time, newly developed approaches proved the possibility of predicting the occurrence of extreme floods based on the features of more ordinary runoff events, in a way refuting the initial claim.

Here we give an overview of these approaches that enable inferring extreme floods from everyday hydrologic dynamics, by focusing on recent results by the authors. The methods, which are rooted in the Physically-based (PhEV) and Metastatistical (MEV) Extreme Value distributions, account for the role of antecedent catchment conditions (considered stochastically) and runoff generation processes in shaping the flood hazard.

We show that the possible occurrence of extreme floods and the emergence of heavy-tailed flood distributions and flood divides (i.e., marked increments of the magnitude of rarer floods) are predicted by metrics of everyday discharge dynamics. We present how knowledge of runoff generation processes can be used in the MEV framework to predict extraordinarily large river floods. We finally show that combining the MEV and PhEV frameworks allows for obtaining reliable estimates of rare floods with no need of a careful preliminary choice of the distribution of ordinary events with a particular tail, currently a critical step of the MEV approach.