Extreme flood estimation is a difficult problem, when dealing only with records of observed floods within the XXth century. Historical analysis allows to enlarge information on the major flood events which already occurred in the past. Some difficulties are related to its use, mainly the availability and exhaustiveness of historical information, the quality and accuracy of historical accounts, and the climatic variability and the morphological changes of the rivers during several centuries.This paper presents three "Historisk" projects, based on the collaboration between historians, archivists, hydraulicians and hydrologists for a better assessment of the flood risk. On the Isère river, a historical investigation allowed to rank the 1600-2000 floods into three classes, using criteria based on the flood damages. On the Guiers river, a hydraulic modelling provided a review of a historical discharge estimate available in a document from the XIXth century. On the Ardeche river, a detailed hydraulic modelling provided a discharge estimate of the main floods during the XIXth and XXth centuries, with an error analysis. A flood frequency analysis shows a good agreement between the historical approach and an hydrometeorological approach (Gradex model). Scientific exchanges during this Chinese-French conference could focus on the collaboration between the two countries on the extreme flood topic, as China has already investigated historical flood records of many basins.