Spatiotemporal clustering of streamflow extremes and relevance for flood insurance

Recent research has revealed the significance of Hurst-Kolmogorov dynamics, which is characterized by strong correlation and high uncertainty in large scales, in flood risk assessment as for example in inundated flood duration. However, classic risk estimation for flood insurance practices is formulated under the assumption of temporal independence of extreme flood events, which is unlikely to be tenable in real-world hydrometeorological processes exhibiting long range dependence. Additionally, insurable flood losses are considered as ideally independent and non-catastrophic in financial terms due to the widely spread perception of limited risk regarding catastrophically large flood losses.

As the accurate risk assessment is a fundamental part of flood insurance and reinsurance practices, this study investigates the effects of lack of fulfillment of these assumptions, paving the way for a deeper understanding of the underlying clustering mechanisms of streamflow extremes. For this purpose, a spatiotemporal analysis of the daily flow series from the US-CAMELS dataset is applied, comprising the impacts of clustering mechanisms on return intervals, duration and severity of the over-threshold events which are treated as proxies for collective risk. Moreover, stochastic approaches are developed and an exploratory analysis is introduced regarding the stochastic aspects of the correlation between the properties of the extreme events and the actual claims records of the FEMA National Flood Insurance Program which are recently published.

Furthermore, regarding precipitation mechanisms, the presence of persistence in annual rainfall is expected to induce clustering in rainfall extremes, which should be manifested by clustering of floods. Therefore, in the framework of a case study, it is investigated whether the collective risk estimated using the former as a proxy, i.e. the magnitude of the rainfall peak over threshold events in a year, is correlated with the actual compensations given.

Eventually, the current insurance practices and actual compensations given in the agriculture domain in Greece are reviewed, while inspecting the underlying stochastic assumptions and evaluating changes in the estimated risk in the case that these assumptions are not valid.