What’s that coming over the hill, is it a weak vortex?

Only a few years ago, ECMWF’s then-twice-weekly 51-member extended-range forecasts were not publicly available — something that is almost hard to comprehend nowadays, as we have daily, 101-member ensemble forecasts available for free on the ECMWF website. The ensemble size is spectacular, and increases forecast reliability. But I’m more interested in what we gain from daily initialisations.

Moving to daily initialisations means we can better understand the sensitivity of the extended-range forecast to the initial conditions. We should no longer have scenarios where the Thursday forecast said one thing, only for the Monday forecast to jump to something else. A spectacular ‘jump’ occurred between the forecasts issued on 29 January and 1 February 2018, when forecasts that had grown more confident in a strong vortex suddenly flipped to an increased risk of a major SSW (which eventually happened on 12 February). My first PhD paper looked at that forecast evolution.

In theory, then, a forecast which predicts the same thing in the extended-range when fed a variety of initial conditions over several days or weeks could be on to something. It suggests the model is being attracted to a specific evolution regardless of the initial state (and indeed, regardless of the actual evolution of the weather over that time). Since that’s beyond the medium-range Lorenzian deterministic predictability limit, such behaviour suggests a potential window of opportunity for subseasonal predictability — i.e., an occasion when there is some constraint on the atmospheric evolution 2-6 weeks ahead, in what is otherwise often a predictability desert (not a predictability dessert, which would be tasty).

That brings us to the reason for this blog: since at least 24 October, ECMWF’s daily 101-member extended-range ensemble forecast has been predicting a weaker-than-normal stratospheric polar vortex to develop by the first week of December. This isn’t a small signal either — the latest forecast, initialised on 2 November (Fig. 1), shows an ensemble mean that’s ~10 m/s weaker than the model’s own climatology by the end of the run. [The model’s own climatology accounts for any mean-state biases in the vortex strength.]

The weak vortex is more than 4 weeks into the forecast, yet has persisted through at least 10 days of initial conditions (and 10 days of real-world weather evolution). This seems very striking to me. Sufficient that I’m writing this blog instead of working on a manuscript!

The weak vortex does not appear to be accompanied by a large number of ensemble members showing a major sudden stratospheric warming (SSW; i.e., easterly zonal winds at 10 hPa 60°N). A weak vortex isn’t the same as an SSW, and we wouldn’t expect the ensemble mean to show an SSW until the medium range anyway. That said, SSWs in early December are rare — and so perhaps the relatively small number of members that reverse to easterlies is in fact anomalously large (I don’t have the statistics… ECMWF, can you make an easterly wind probability plot like the C3S ones?).

So, this looks like a signal worth noting. But why is it there? Much research has found that blocking in the Ural Mountains region [e.g., Kolstad and Charlton-Perez 2011; Peings 2019, White et al. 2019] can enhance upward-propagating planetary wave activity and serve as an SSW precursor. When planetary waves break in the stratosphere, they exert a westward drag on the zonal winds and warm the stratosphere. The stratospheric vortex strength is on average effectively an integrator of the amount of wave activity over the preceding month to six weeks, so this forecast suggests prolonged increased upward wave activity.

Although I haven’t got the data to explicitly make the link, the same forecasts have also been showing remarkably persistent Ural blocking (Fig. 2) that has also been robust to differences in initial conditions. At least in the 2 November run, there is an ensemble-mean anomalous ridge over the Urals from week 2 to the end of the run! There is also generally an anomalous trough over eastern Siberia/North Pacific. Both of these constructively interfere with the climatological mean stationary wavenumber-1 pattern (cf. Fig. 2 here with Fig. 3 in White et al. 2019).

So, it seems that the model’s weak vortex forecast is consistent with its tropospheric forecast, and that the troposphere keeps evolving the same way, run-to-run. But why?

The North Pacific trough is consistent with a deepened Aleutian Low, which is a typical response to the tropospheric Rossby wave train induced by El Niño. That’s a key reason why the vortex is — on average — weaker during El Niño winters. Climate models also generally indicate that El Niño winters see the highest SSW frequency, but observations don’t necessarily line up with this, which is probably just an effect of the small and noisy observational sample size.

But why is the Ural blocking there and so persistent? I’m less sure. One thing that struck me looking at the Z500 anomalies is the presence of persistent troughing near the British Isles and western Europe that is present alongside the Ural blocking. The regime forecasts seem confident in NAO+, even in the ensemble mean. Perhaps this persistent cyclonic flow and associated wave breaking is helping to build the block downstream? One could also ask why there is a persistent cyclonic anomaly in the Atlantic, too. I am running out of steam at this point, but I think this goes to show the multi-scale nature of the problem and how remarkable it is that all these components have persisted run-to-run.

Now, it’s important to also note that the stratosphere is not a slave to the troposphere. The same tropospheric patterns do not always induce the same stratospheric response, while not every extreme stratospheric event has an extreme tropospheric precursor. The state of the stratosphere matters. In this case, it’s interesting that the vortex becomes extremely strong for the time of year during the medium-range. In fact, it’s likely to set new date-records, and might even qualify as a “strong vortex event” (definitions differ, but broadly when the zonal winds exceed ~40 m/s). You can see that in Fig. 1, where the ECMWF forecast exceeds the 90th percentile (thin red line) of the model climate. After this, the winds begin to weaken, and this evolution has also been persistent run-to-run. Meanwhile, Fig. 3 shows the latest 10-day forecast from NASA’s GEOS model (orange) alongside the MERRA-2 climatology for the whole season, with the GEOS forecast exceeding the date-record maxima.

I mention all this because wave propagation and breaking in the stratosphere can be enhanced when potential vorticity gradients are steep, and steep PV gradients can manifest as strong zonal winds. This is usually something we think about more in the second half of the winter, when the vortex edge has been sharpened by wave activity through the winter (a bit like peeling an onion), but it could tie in here, especially since these vortex conditions are more typical in midwinter.

Now, I’m not a forecaster, and I’ve certainly seen enough subseasonal forecasts ‘evaporate’ to know that a big and persistent signal doesn’t mean you should bet your house on it. Verification statistics for subseasonal forecasts are low, and often my interest is more centred around understanding why models say what they say, rather than whether the forecast will verify. But if this case does verify… then, given the lead time at which it first appeared, it would be a remarkable case worth investigating further.

In the shorter term, I’d be on the lookout for the Scandinavia-Greenland pattern. This is a transient (synoptic-scale) pattern that is characterised by cyclones tracking up the east coast of Greenland with an accompanying ridge building over Scandinavia. It is usually associated with anticyclonic wave breaking, and often subsequent development of the Ural high, as well as enhanced upward wave propagation itself. Perhaps I’m biased because I wrote two papers about it (in 2019 and 2020). But it’s been appearing in some medium-range output, such as that shown in Fig. 4 which went on to develop a Ural high and saw a significant amount of upward wave activity inducing a weakened vortex by the end of the run. GFS fantasy land, yes, but it’s the sort of processes that the ECMWF subseasonal run suggests we’d be looking for.

So, potentially interesting times ahead. Or maybe this blog will age terribly and serve as a reminder of the challenges of extended-range forecasting… either way, if you’ve read this far, thank you! It’s always a pleasure to share things like this online, and even more so when people find the time and interest to engage with it. Cheers!