The characterization of Extraordinary Extreme Events (EEEs) for the assessment of design rainfall depths with high return periods

Abstract

The occurrence of rainfall Extraordinary Extreme Events (EEEs) in Mediterranean areas causes serious concerns to the engineers involved in the design of flood and landslide risk mitigation plans as well as of strategic hydraulic engineering structures, such as dams. These extraordinary maxima are characterized by very low frequencies and spatial extent scales that are smaller than those of ordinary maxima, and are usually identified as outliers by classical regional frequency analysis. Extreme Value mixture models, such as the Two-Component Extreme Value distribution, have been introduced in regional frequency analysis in order to overcome this problem. Nevertheless, the rainfall maxima series available for carrying out these regional analyses present coarse spatial spacing and small temporal extent, when compared with EEEs spatial structures and frequencies. Thus, regional statistical analyses with mixture models tend to overestimate EEEs return periods as well. This study presents a new operational statistical approach to properly retrieve the EEEs frequency from the available database and thus to avoid dramatic underestimations of the rainfall depth at very high return periods. The proposed approach implies the analysis of the EEEs at a given support scale in order to assess (a) the percentage of EEEs among the annual maxima and (b) the conditional distribution of the annual maxima given the occurrence of an EEE. The descriptive properties of the proposed procedure have been tested in Italy, by comparing the performances of the proposed procedure in predicting rainfall depths at very high return periods (i.e., larger than 100 years) with those provided by the TCEV-based regional analysis currently adopted as national reference approach. K E Y W O R D Sdaily annual maxima, design rainfall depths, extraordinary extremes, rainfall extremes, rainfalltriggered hazards, regional frequency analysis

Comparison of ERA5-Land and UERRA MESCAN-SURFEX Reanalysis Data with Spatially Interpolated Weather Observations for the Regional Assessment of Reference Evapotranspiration

“…Indeed, the access to observed weather data can be even more difficult for areas extending across regional or national boundaries. In Italy, for instance, weather monitoring networks are managed by regional and national services, without a common data sharing policy and data distribution platform [77]. Pan European gridded agrometeorological maps released to support the EU Common Agricultural Policy, are derived by interpolating observed weather data collected in the different Member States, with uneven distributions, not consistent with what would be required based on the topographic complexity (e.g., Agri4Cast Resources Portal, Gridded Agro-Meteorological Data in Europe [78]).…”

The role of citizen science to assess the spatiotemporal pattern of rainfall events in urban areas: a case study in the city of Genoa, Italy

“…The inclusion of observations from "citizen science" measuring stations shows how in this case the precipitation field shows a higher spatial variability with values of cumulative precipitation on the event extremely variable within 1-2 km. This demonstrates how both the meteorological radar, which has a spatial resolution usually of the order of km, and the authoritative ground measurement networks, which, in the best cases, in Italy, have densities of around 0.03 stations per km 2 (Pelosi et al , 2020), are not suitable for adequately describing the precipitation field.…”

Regional Assessment of Sub-Hourly Annual Rainfall Maxima

“…When higher-order parameters (i.e., the shape and scale parameters) of the parent distribution of the annual rainfall maxima are assumed to be constant with the duration [15,19], it is even possible to decouple the dependence on the return period from the dependence on the duration, thus allowing more robust estimates of the higher-order parameters by pooling observations at different durations, achieving an extension of the sample size. This regional procedure delivers reliable predictions at high return periods [20] beyond the available sample size, which is typically not larger than 50 years at sub-daily durations and 18-20 years at sub-hourly durations. Under this hypothesis, the annual rainfall maxima for a given duration d and given return period T can be computed as follows:…”