K-Means Cluster Analysis Identification of Idealized Relative Anomaly Patterns in Annual Total Precipitation across New England States’ NCDC Climate Divisions Encompassing the 1895–2018 Period of Record

The New England states, defined as Connecticut, Rhode Island, Massachusetts, New Hampshire, Vermont, and Maine are composed of 15 climatic sub-regions, part of a nationwide arrangement of 344 Climate Divisions organized state-by-state by the NOAA National Centers for Environmental Information (NCEI), formerly the National Climatic Data Center (NCDC). Based on area-averaging techniques, single-valued month-to-month precipitation statistics have been compiled, division-by-division, since 1895, and several years ago, using new, improved areal averaging techniques, the entire division-by-division statistics were redone, by year. While the geographical extent of the New England states is not great, there are contrasts in topography and proximity to the Atlantic Ocean, which could conceivably induce contrasts in annual precipitation relative anomaly character across the divisions, and to explore this possibility, the nature of idealized annual precipitation anomaly modes are resolved using K-Means Clustering Analysis, the K-Means treatment further enhanced by an optimizing data mining training/testing capability - the V-Fold Cross Validation Algorithm. Period of record is 1895 thru 2018, some 124 seasons. Previous studies along these lines investigated variations in seasonal (July-June) total precipitation modes across the seven California Climate Divisions (Fisk, 2015) and also for 12 near-coastal Climate Divisions for California, Oregon, and Washington (Fisk, 2016), in each case, a “drilling-down” undertaken to illuminate pattern configuration contrasts relative to El Nino, Neutral, and La Nina ENSO episodes. It is known that the North Atlantic Oscillation (“NAO”) can influence New England area winter precipitation significantly (presumably November-March), but since the analysis here covers a complete individual calendar year period (January-December), and the NAO influence crosses the latter and beginning portions of two successive calendar years, respectively, the NAO signal is not likely to be clearly expressed in the clustering results.

The New England area analysis produced six clusters of annual precipitation variability across the 15 climate divisions; these are presented in both graphical and tabular form, along with supporting explanations.