Weather Regime Diversity, Transitions, and Trends Using Hexagonal Self‐Organizing Maps

Abstract Persistent atmospheric circulation patterns, or weather regimes, strongly modulate surface weather and extremes, yet their internal diversity, transitions, and trends remain less understood for North America. We apply a self‐organizing map (SOM) framework to represent North American weather regimes using daily 500‐hPa geopotential height anomalies from 1940 to 2023. A new hexagonal SOM lattice was developed to minimize geometric artifacts, and a transition‐based metric was introduced to optimize the SOM to capture transitions among nodes. The resulting SOM identifies both “canonical” weather regimes, such as the Alaskan Ridge and Pacific Trough, and their intra‐regime diversity. Several regimes, including the North American‐Atlantic Ridge and Pacific Trough, are distributed across many SOM nodes, suggesting greater regime diversity and structural overlap with other regimes. In contrast, the Greenland High and Alaskan Ridge regimes exhibit less diversity. Node transitions generally follow preferred pathways, tending to move counterclockwise across the lattice, corresponding to shifts between large‐scale circulations with weaker, more meridional flow and stronger, zonally oriented flow. Trend analysis reveals increases in node frequency during spring and summer, particularly those dominated by the Alaskan and North American‐Atlantic Ridge regimes. This study expands our understanding of North American weather‐regime characteristics by demonstrating that straightforward, physically motivated modifications to established machine‐learning algorithms can enhance their utility to address questions about Earth systems.