Changes in the rainfall event characteristics above the empirical global rainfall thresholds for landslide initiation at the pan-European level

Abstract

Rainfall-induced landslides are one of the most frequent natural disasters. Detection of changes in the characteristics of the rainfall events that can be regarded as triggering factor is therefore of primary importance for policy-makers, practitioners, civil protection units and others. This study evaluated changes in the frequency and intensity of rainfall events above selected empirical rainfall thresholds at the pan-European level. For areas classified as at least moderately susceptible to landslides according to the ELSUS v2 map (i.e. around 1.8 * 106 km2; around 37% of the area), changes were evaluated for the 1961–2018 period. The regional rainfall reanalysis data was used as input. Four different empirical rainfall thresholds were selected and changes in the frequency of rainfall events above the thresholds and their characteristics were analysed. The results showed that, on average, 14% more rainfall events above the threshold were recorded in the 1991–2018 period compared to 1961–1990 at the European level according to the selected empirical rainfall threshold. Moreover, several regions were found where frequency and intensity of rainfall events above the four selected global empirical rainfall thresholds were positive and statistically significant. These regions were for example Italy, Carpathian Mountains, British Isles and some other parts of Europe. Some of these areas are also regions where landslides susceptibility is high or even very high and where multiple landslides were recorded in the past. On the other hand, there are also regions where frequency and intensity of events above the threshold were negative and statistically significant. These regions were parts of Central Europe, Alps and Pyrenees. Also, in these regions landslides frequently occur. According to different empirical thresholds, no significant change in the frequency of rainfall events above the thresholds could be detected for around 50% of the investigated area.