Precipitation-driven typology of storms in the Alps

Numerous advances in precipitation science hinge on our ability to accurately categorize storms into physically meaningful classes, particularly to differentiate between convective and non-convective phenomena. Nonetheless, achieving such classifications remains a challenge for the research community. Here, we propose a precipitation-driven typology of storms in the Alps developed through a straightforward methodology for unsupervised classification. From a vast sub-hourly precipitation dataset, we extracted over 790,000 independent storm time series. To categorize these, we employed a resampling-based partitioning algorithm, optimal in clustering big data. Four storm features (i.e., the maximum intensity, total volume, total duration, and coefficient of variation) drove our typology on an algorithmic basis. The algorithm revealed five dominant storm classes, which we termed as “convective”, “stratiform”, “short stratiform”, “intermittent minor” and “minor” based on a physically-informed examination of their features. Three other features (i.e., the month of the storm initiation, solar time at the first occurrence of the maximum intensity, and lightning count) were used for an independent validation of the classes, together with investigations on the extent to which each class was clustered in space. Consistency with anticipated physical patterns suggests the potential utility of our proposed typology across various modelling applications. These include class-specific stochastic simulation of storms, class-informed bias adjustment of climate model projections or the development of multi-class extreme value analyses. Detailed investigations of its climatological traits revealed, among others, higher convective activity in recent years and specific Alpine regions. We provide the historical occurrences of the proposed storm typology as an open dataset.