Turbidites embedded in lacustrine sediment sequences are commonly used to reconstruct regional flood or earthquake histories. A critical step for this method to be successful is that turbidites and their trigger mechanisms are determined unambiguously. The latter is particularly challenging for prehistoric proglacial lake records in high‐seismicity settings where both earthquake‐generated and flood‐generated turbidites interrupt the background varved sedimentation. This calls for a new method to allow efficient and objective identification and classification of turbidites. This study examined turbidites in five long (9 to 17 m) sediment cores from Eklutna Lake, a proglacial lake in south‐central Alaska, using standard core logging and grain‐size data. A novel statistical approach is presented, in which varve‐thickness distributions were first analyzed to objectively identify the thickest turbidites and distinguish them from background sedimentation. For each turbidite, a selection of variables were then measured, including: basal grain‐size, thickness, magnetic susceptibility and spectrophotometric variables. Triggering mechanisms were discriminated by a combination of principal component analysis and clustering, and by calibration with historical events. Using this approach, a 2250 year long lake‐wide event stratigraphy was constructed, with 94 prehistoric events, including 24 earthquake and 70 flood events. Basal grain‐size and thickness variables turn out to be the most effective proxies for discrimination. This statistical approach is a powerful and new method to identify turbidites and their triggering mechanisms in long prehistoric sediment records. It opens up new prospects for palaeoseismological, palaeohydrological and palaeoclimate studies in proglacial lakes worldwide.