Large-Scale Meteorological Patterns Associated with Extreme Precipitation in the US Northeast

Both K-means clustering and Self-Organizing Maps are used to identify large-scale circulation patterns associated with extreme precipitation in the US Northeast.  The process is examined here, as well as the resulting patterns. Tropopause height is used as the typing variable, as it provides insights into three-dimensional synoptic-scale circulation, jet streak locations, Rossby wave-breaking and PV streamers, as well information regarding low-level diabatic heating. The K-means algorithm is applied on extreme precipitation days only, which are defined as the top 1% of daily precipitation (1979-2008) at 35 Northeast stations. Two objective techniques for identifying the optimum number of clusters are applied -- both techniques recommend a 3-pattern solution, and a secondary 6-pattern solution. Separate seasonal typing supports the 6-pattern solution. In contrast, SOM analysis is performed for all days 1979-2008, utilizing a 5x6 rectangular nodal grid. The SOM pattern-space reveals several patterns that are more likely to support extreme precipitation; these particular patterns are similar to those identified using K-means typing. While the K-means technique provides a succinct solution with a small set of patterns associated with extreme precipitation days, the SOM technique allows us to evaluate the persistence (duration) of extreme precipitation patterns days, the preferred transition to other patterns, and a measure of how the specific patterns differ on extreme precipitation days versus non-extreme precipitation days. The patterns themselves include two version of tropopause ridges, two flavors of cold-season East-Central US troughs, and cold-season and warm-season Eastern US troughs.