High-impact climate events such as floods are highly destructive natural hazards causing widespread impacts on socio-ecosystems. However, processes leading to such events are still poorly understood, which limiting reliable prediction. This study takes advantage of centennial-long discharge series (1923–2010) and meteorological reanalysis (ERA-20C) to study processes generating the high-magnitude flood events (i.e. above the percentile 99.9) of the upper Rhône River (NW European Alps). A particular focus is paid to the role of precipitation on the flood generation to explore in what extent such events could be explained by only atmospheric variables. A flood typology is thus established using a hierarchical clustering analysis and three variables: long (8-day) and short (2-day) precipitation accumulations as well as an index characterizing the amplitude of the discharge increase during the 7 days prior to the flood day. The typology result in four classes, of which two are directly linked to precipitation. One results from heavy precipitation over two days (similar to <q>short-rain floods</q> in the literature) and the other one from a combination of short and long intense precipitation sequences (similar to <q>long-rain floods</q>). The two other types of floods cannot be explained by precipitation only, most probably involving ice and snow melting. The four events of highest magnitude (> 20 year return period) are of various types but are all triggered by heavy precipitation during the days preceding the floods. The role of the precipitation accumulations progressively decreases when considering floods of weaker magnitude, suggesting a higher diversity of processes involved in the generation of e.g. annual flooding. Our results highlight the needs to better understand the atmospheric processes leading to heavy precipitation accumulation since this would allow a better understanding of past and future trends of extreme flood events.