Detection of past extreme precipitation events and connection to recorded impacts: a multi-data and multi-method assessment over the Central-Eastern Alps

Extreme hydro-meteorological events are increasingly observed in southern Europe and especially in the European Alps, where they threaten ecological and socio-economic systems. To detect such events and analyse the changes in their occurrence, a proper definition of an extreme event is needed. Statistically, we define extremes from the tails of the probability distributions. However, these events are not necessarily extreme in terms of impact, and impact-related thresholds may vary spatially and temporally, i.e., single absolute thresholds do not necessarily reflect the extremes at all locations, in all time periods and all seasons. Moreover, the availability of harmonized and consistent datasets is crucial for investigating extremes in a transnational context. In this study, we focus on the identification and characterisation of extreme hydro-meteorological events affecting a transboundary Alpine region between Austria and Italy from 2003 to 2021 based on different definitions of extreme events considering spatiotemporal aspects and multiple datasets. Daily accumulated precipitation is used as the main proxy parameter to describe the potential for severe consequences, as it as it is the most broadly available quantity across different datasets compared to e.g., sub-daily precipitation sums. Moreover, its role as a triggering factor for various hazards (e.g., landslides, debris flows, pluvial and fluvial floods) is widely recognised. We analyse three different statistical methods for the detection of extreme events: (i) the identification of the highest daily precipitation amounts on a regional scale, (ii) the detection of daily precipitation values of high intensity on a local scale and (iii) the identification of exceptional daily precipitation records not in absolute terms but with respect to average conditions associated to a specific period of the year. All detection algorithms are applied to four gridded precipitation datasets, including both observation and reanalysis products, with different technical specifications. Subsequently, identified events for each method-dataset combination are blended with existing records of gravitational mass movements and fluvial floods in the Austrian-Italian border region to analyse the suitability of each combination to detect actual occurred impacts. First results indicate that most detected precipitation extremes relate to actual observed impacts (e.g., 74% for regional scale identification with reanalysis data). However, different method-dataset combinations have different strengths and weaknesses, which reflect inherent characteristics of the dataset and/or of the statistical method employed. Furthermore, some combinations show lower performance in detecting impactful events, because the dataset and method applied conflict with each other (e.g., a coarse-resolution dataset not resolving local-scale features conflicts with a statistical method searching for locally high intensities). The findings could contribute to better inform civil protection authorities about risks related to extreme hydrometeorological events, possibly affected by climate change.