North American wintertime weather regimes

These are computed following the method defined in Lee et al. (2019) and Lee et al. (2022) [although several other studies have computed similar regimes], updated to use daily 00Z ERA5 1.5° data in December-March from 1 December 1979 through 31 March 2022.

To obtain the regimes, empirical orthogonal function (EOF) analysis is performed on the daily linearly-detrended 500 hPa geopotential height anomalies in the region 20-80°N 180-30°W, retaining the leading 12 EOFs to filter small-scale variability. k-means clustering is then performed, with k set to 4. Each day is assigned to a cluster by the minimum Euclidean distance to the centroid in the 12-dimensional principal component (PC) space (although it turns out only the leading 3 EOFs are important). The four regimes are:

  • Arctic High (ArH; alternatively Greenland High), occurring on ~20% of DJFM days.
  • Arctic Low (ArL; alternatively Pacific Ridge), occurring on ~24% of DJFM days.
  • Alaskan Ridge (AKR), occurring on ~25% of DJFM days.
  • Pacific Trough (PT), occurring on ~31% of DJFM days.
North American weather regimes: Composite-mean 500 hPa geopotential height anomalies for all days assigned to each of the four North American wintertime weather regimes. The % of DJFM days assigned to each regime is also shown.

A daily weather regime index, based on Michel and Rivière (2011), is then computed by projecting the 500 hPa geopotential height anomalies for each day onto the composite of all days assigned to that regime (truncated to the first 12 PCs), and dividing by the standard deviation. This effectively shows the ‘strength’ of each pattern (but remember that the patterns are not orthogonal; a large positive index for one regime will inevitably mean a negative projection for another).

In the figures here, those days with an index exceeding 1 standard deviation are shown with markers. The top sub-panel denotes the (binary) regime assignment by minimum Euclidean distance, but the index shows that often the story is more complex. The method here does not include a ‘no regime’ assignment, but one could interpret days where the index is less than 1 as ‘no regime’.

DJFM weather regimes 1979/80 – 2021/22

2022 2021 2020 2019 2018 2017 2016 2015 2014 2013 2012 2011 2010 2009 2008 2007 2006 2005 2004 2003 2002 2001 2000 1999 1998 1997 1996 1995 1994 1993 1992 1991 1990 1989 1988 1987 1986 1985 1984 1983 1982 1981 1980

Proportion of DJFM days assigned to each regime per year