100%!

Finally, while there has been a large change to the behaviour of the GH in recent decades, we are not arguing that the trend will necessarily continue.

2025 saw a relatively low GH frequency for the second consecutive summer – but this has happened before prior to a spike. What will 2026 bring? 👀

We therefore encourage a more unified approach, analysing both initialised/seasonal models and uninitialised/climate models to help solve persistent modeling challenges across timescales.

Our study adds to a growing body of literature showing that many biases/missing trends in CMIP/climate models are present in seasonal models and must thus develop on short (subseasonal) timescales (e.g., @jbeverley.bsky.social doi.org/10.1038/s416...)

This trend is similar to known trends in Greenland blocking and its absence in CMIP (e.g., doi.org/10.1002/joc....). Here we show the key role of persistence, and that the missing trends are present in a state-of-the-art seasonal model – so even initialising the model does not solve the problem.

Yet we find that SEAS5 *can* produce extremely long-lived GH regimes, and summers with an extremely high frequency of GH days – even more so than observed. So, the issue is not an inability of the model to generate extremely persistent regimes, but the absence of the trend in persistence.

We then investigate whether a 10,000 member ensemble constructed from random sampling of @ecmwf.int SEAS5 hindcasts and forecasts over 1981–2024 can reproduce the observed trends.

Remarkably, the observed GH trends are not reproduced even in this very large ensemble!

The increased interannual variability means extremely GH-dominant summers (2012, 2016, 2019, 2023) have been juxtaposed with summers with relatively few GH days (2017, 2018, 2021, 2022, 2024).

Increased persistence leads to a greater increase in high-GH summers than a decrease in low-GH summers.

Using a year-round North American weather regime classification (doi.org/10.1175/JCLI...), we find a large increase in the summer frequency and interannual variability of the GH regime in recent decades, which we attribute to increased GH regime persistence through a first-order Markov model.

Indeed! We do tend to see an increased likelihood of storminess and wind extremes downstream following these events.

The poor old cold conveyor belt jet must wish it had a more headline-grabbing name…

Thanks! It's the projection of the full field each day onto each pattern mean. They don't sum to zero on each day – the regimes aren't orthogonal.

Yes indeed, thanks! (It was actually Christian emailing me about his preprint which motivated me to do this.)

Check out updated annual plots of the Weather Regime Index for 1979–2025 here simonleewx.com/north-americ...

Thanks Ray, I agree! We have a paper coming out imminently looking at summertime trends, but there is so much more to do – hopefully over the next few years.