Lake Champlain is a natural freshwater lake that straddles the Canadian-US border in Eastern North America. In the spring of 2011, its water level was at a record high, causing massive floods. The severity of this flood was largely due to heavy rainfalls occurring in clusters over several days. For this reason, it has proven difficult to estimate the return period of this unprecedented event using standard extreme-value methods. A cluster of high precipitation can be defined as a streak of rainy days, at least one of which results in extreme rainfall. The purpose of this talk is to show how one can successfully estimate the return period of the spring precipitation that triggered the 2011 flood in the Richelieu River. To this end, a statistical model that takes into account clusters of extreme precipitation must be used. The model relies on a decomposition of clusters into polar coordinates.