The subseasonal atmospheric variability in the extratropics is a key research topic within atmospheric science and of broad socioeconomic relevance (White et al., 2017). A far-reaching paradigm is that a small number of states, termed "weather regimes," can describe this variability. These are defined as recurrent or quasi-stationary states of the large-scale circulation (Hannachi et al., 2017). Weather regimes have implications for extended-range weather forecasting and in the understanding of climate variability (Merryfield et al., 2020).The concept of weather regimes was first introduced by weather forecasters in the late 1940s (Levick, 1949). A corresponding theory was developed by Charney and DeVore (1979), who hypothesized that the largescale circulation transitions between multiple equilibria states, based on a spectral model. The multiple equilibria viewpoint has been later authenticated for the barotropic case (Legras & Ghil, 1985), and in two-layer models (Yoden, 1983), yet was criticized using models with higher spectral resolution (Cehelsky & Tung, 1987). Faranda et al. (2016) argued for linking blocked regimes to unstable fixed points rather than stable equilibria. Nowadays, there is little doubt that weather regimes emerge from different statistical classifications applied to long archives (Hannachi et al., 2017). However, it is a source of debate whether these regimes represent metastable states of the atmosphere (Majda et al., 2006), or are merely useful statistical categorizations lacking physical grounding (Fereday, 2017).A variety of procedures has been used to classify weather regimes (Hannachi et al., 2017). They typically seek to define a low-dimensional phase-space reflecting the key aspects of the atmosphere's variability. A common choice is to compute the Empirical Orthogonal Functions (EOF) of the data. Clustering algorithms