Statistical modeling of hydrometeorological events in poorly gauged coastal areas

Estimating extreme hydrometeorological events, such as storm surges or extreme precipitation , is crucial for effective flood risk management, particularly in poorly gauged or ungauged regions. As climate change intensifies, these events are expected to increase in frequency and severity, making reliable predictions even more vital for vulnerable areas. Traditional methods, such as asymptotic extreme value distributions, often face significant uncertainties when dealing with short observational records, which are common in many regions. This results in high uncertainties in extreme event prediction, thereby hindering effective preparedness and response strategies.

In this study we introduce an approach to hydrometeorological extremes that combines the Metastatistical Extreme Value Distribution (MEVD) and a flexible regionalization technique, aiming to overcome the limitations set by data scarcity in traditional at-site analysis methods. Unlike asymptotic methods, which uses only a small subset of the available observations, the MEVD method leverages the information contained in all observed events to infer the probability distribution of annual maxima. This approach is particularly beneficial when the data records are scarce, allowing for more accurate estimation of very rare events. Uncertainties can be further reduced by exploiting spatial information to compensate for the lack of information in time. The flexible regionalization approach proposed, unlike traditional regionalization methods, does not impose rigidly defined regions composed of statically homogeneous sites with predefined spatial boundaries. Rather it accounts for the observational information contained in the vicinity of the site where the estimation is being carried out by introducing a weight according to a similarity criterion. This feature allows for a seamless integration of data across varying spatial and temporal domains and a better representation of the continuous nature of hydrometeorological processes.

In this contribution the performance of the flexible MEVD-based regional approach is appliedcompared with that of state-of-the-art regionalization approaches.