A robust classification of tropospheric wind profiles from ERA reanalyses (1940–2024)

Vertical wind structure plays a central role in tropospheric dynamics, yet its variability is rarely characterized beyond surface-level fields. Summarizing this variability over multi-decadal timescales requires reducing the dimensionality of wind profiles while preserving their dynamical content. Here, we develop an objective classification of vertical wind regimes using the ERA reanalysis (1940–2024), to identify a compact set of representative tropospheric structures and quantify their temporal evolution.

We first derive spatio-temporal averaged wind profiles from the reference regions defined within the IPCC framework. Although these regions are based on surface climate characteristics, the resulting regional wind profiles provide a baseline against which we compare new wind profile classifications from vertical climate variability.

Dominant modes of variability are extracted using empirical orthogonal functions applied to multi-level wind profiles. Clustering (k-means and hierarchical approaches) is then performed in the reduced phase space to identify dynamical regimes, with robustness assessed through bootstrap resampling and multiple validation metrics. We show that a limited number of regimes capture most of the tropospheric wind variance over the 84-year period, each characterized by distinct vertical shear and directional signatures. The length of the record allows us to examine persistence, transition probabilities, and modulation across seasonal to multi-decadal variability.

Overall, this framework provides a physically interpretable compression of vertical wind variability over a uniquely long ERA dataset, offering new diagnostic tools for atmospheric dynamics and a potentially valuable input for transport, dispersion, and predictability studies.