Last updated 7 January 2026.
Weather regimes are recurrent, persistent and quasi-stationary circulation patterns which represent large-scale, low-frequency variability. The year-round North American regimes are here computed following the method defined in Lee et al. (2023), which uses a variance-normalisation technique to account for the seasonal cycle in the amplitude of the anomalies (following a similar method outlined in Grams et al. (2017) for the North Atlantic–European sector).
Here, the regimes are calculated using 00Z 500 hPa geopotential height from ERA5 at 1.0° resolution in the domain 180–30°W, 20–80°N from 1 January 1979 – 31 December 2025. The anomalies are computed with respect to a 60-day smoothed daily climatology, with the area-average trend in Z500 removed (6.0 m per decade). Prior to regime computation, the anomalies are filtered with a 10-day low-pass filter, then normalised by the area-average standard deviation for each calendar day (smoothed with a 60-day running mean), and finally transformed into a 12-dimensional principal component space. k-means clustering is then performed with k = 4. Each day is then assigned to one of the four clusters, plus a ‘no regime’ class, by minimum Euclidean distance to the centroid. ‘No Regime’ is here defined as a cluster at the origin in PC space (i.e., zero mean anomaly).
The dataset is openly available on Zenodo.
The four regimes are:
- Pacific Trough (PT), occurring on 25.5% of days
- Pacific Ridge (PR), occurring on 21.4% of days
- Greenland High (GH), occurring on 19.6% of days
- Alaskan Ridge (AKR), occurring on 19.5% of days
The remaining 14.0% of days are classified as No Regime, indicating the anomalies are closer to climatology than to any of the regime centroids.

A daily weather regime index (WRI), based on Michel and Rivière (2011), is then computed by projecting the normalised 500 hPa geopotential height anomalies for each day onto the composite of all days assigned to that regime, subtracting the mean projection, and dividing by the standard deviation over all days (see Lee and Messori 2024). This effectively shows the ‘amplitude’ of each pattern.
In the figures below, the assigned regime for each day (by minimum Euclidean distance in PC space) is shown with markers and thickened lines.
