Simon’s Stratosphere Watch #2

When I wrote the first SSW blog on the evening of Tuesday 22 February, I did not expect to be writing the next one so soon after with such increased uncertainty. But there are many aspects to the way the world evolved over the last week which I would not have expected! Thanks to everyone who read blog #1.

Sunday 27 February


In the medium-term, a minor split of the stratospheric vortex will occur. The specifics of this event are uncertain and rapidly-evolving. However, forecasts remain confident that the vortex will not be significantly affected and will continue into April in a generally stronger-than-normal state, albeit perhaps slightly weaker than earlier forecasts suggested, weakening with the seasonal cycle.


Medium-term evolution

During the next couple of days, a spell of downward-propagating wave activity/wave reflection will occur, associated with EP-flux divergence in the stratospheric jet and thus acceleration of the zonal-mean winds. The reflection event appears to arise due to vertical curvature in the zonal-mean wind profile and a split-jet structure in the upper stratosphere, which has previously been related to wave reflection (including during early 2020). This was in the forecast earlier in the week, but the uncertainty in what happens afterwards has grown. The evolution is now looking quite remarkable, as the wave activity swings around from being strongly divergent to strongly convergent. Per the 00Z GFS from 26 Feb, the divergence of EP-flux at 12Z on 27 Feb yields a U10-60 acceleration equivalent to +29 m/s/day (Fig. 1, left). But less than two days later (Fig. 1, right), at 00Z on 2 March, the U10-60 tendency due to the EP-flux convergence is -30 m/s/day! The result is a U forecast which looks like it came straight out of game of line rider.

Figure 1: EP-flux vectors (arrows), zonal mean zonal winds (fill), and zonal wind tendency due to divergence of EP-flux (contours). The tendency at 10 hPa 60°N is shown in the bottom right. (Left) 12Z 27 Feb 2022 during period of downward-propagating wave activity leading to zonal wind acceleration. (Right) 00Z 2 March 2022 during period of upward-propagating wave activity leading to zonal wind deceleration. Figures from Zac Lawrence /

The uncertainty in the zonal wind forecast grows during this upward burst of wave activity, which is at a remarkably short lead-time for noteworthy stratospheric uncertainty. Forecasts have been trending weaker as the transient strengthening/reflection event nears and forecasts converge on its specifics. I am making the assumption that the reflection/strengthening event is important for the subsequent weakening because analysis of ‘chiclet’ charts (Fig. 2) shows a co-appearance of the weakening signature with the prior brief strengthening (associated with the wave activity divergence). However, it is possible these are both related to one common medium-range event which appeared in the forecasts at the same time.

Figure 2: ‘Chiclet’ chart (forecast lead-time vs. valid time) for the ensemble mean U10-60 from the last 17 days of 00Z GEFS runs. The brief, sharp intensification event stands out for lead-times less than 2 weeks, with the subsequent weakening simultaneously appearing.

The wave activity convergence and zonal deceleration manifests as a minor split of the stratospheric vortex (Fig. 3), associated with an amplified wavenumber-2 disturbance. A smaller vortex is sheared off over eastern North America, with a much larger vortex located over Eurasia and concomitant European and Aleutian anticyclones. It is visually very appealing, but not a major SSW… the zonal winds at 60°N only drop to 7-8 m/s or so in most current forecasts. (At this point, do you laugh or cry that we can achieve a vortex split but not an SSW?) In a zonal-mean dynamics sense, the wave breaking does not sufficiently decelerate the jet, but it can be also thought of in terms of the very uneven split and the strength of the cyclonic circulation around the Eurasian vortex owing to its very cold interior.

Figure 3: A split of the 10 hPa stratospheric vortex, but not a major SSW. Still, it looks pretty cool. The North American child vortex is so small that my code doesn’t even assign it an L marker.

So, what’s going on here to drive this split? Since this uncertainty has developed quickly, and the details remains uncertain even on timescales of less than 10 days, an obvious candidate is uncertainty in the troposphere. Indeed, looking at tropospheric forecasts during this time shows a pattern which would be conducive to driving this sort of stratospheric evolution (Fig. 4): anomalous ridges over Scandinavia and Alaska and anomalous troughs over the northwestern North Pacific and northwestern North Atlantic is a classic pattern for constructive interference with stationary wavenumber-2 in the troposphere. Constructive interference between the weather patterns and the climatological-mean stationary wave pattern can amplify upward-propagating wave activity from the troposphere which can then converge in the stratosphere.

Figure 4: (top) Z500 anomaly forecast for 00Z 1 March from GFS 00Z 27 Feb []. (bottom) Z500 eddy heights (all waves) and decomposition into waves 1 and 2 for DJFM in ERA5.

This tropospheric pattern appeared in forecasts simultaneously with the weaker vortex forecasts. For example, the ECMWF forecast for the second North Atlantic Z500 EOF (which, when positive, indicates an anomalous ridge over Scandinavia) for the week beginning 28 February jumped from around 0 to around +1.5 (and outside the spread of the prior forecast) between the forecasts initialised on Monday 21 and Thursday 24 February. We wrote a paper in 2020 arguing that models do not properly represent variability in the Scandinavia-Greenland pattern (similar to the EOF2 used by ECMWF) and its impact on weakening the vortex. So is this a case of a poorly forecast troposphere driving enhanced vertically-propagating wave activity leading to a split of the stratospheric vortex — like early February 2018?

Life is never that simple. The stratospheric wave activity is predicted to become very high (Fig. 5; GFS and GEOS forecasts peak at >40 K m/s at 100 hPa on 3 March, at the top end of climatology for this time of year) but is not clearly related to anomalously strong tropospheric forcing — if anything, the tropospheric heat flux is rather low at that time.

Figure 5: GFS analyses and forecasts (from 00Z 26 Feb) of 40-80°N eddy heat flux anomalies. Note the >2 sigma event in the middle and lower-stratosphere in the coming days, which is not clearly related to an anomalous tropospheric source. Note also how the heat flux has been unusually low this year so far, leading to a strong vortex.

This is not unusual: anomalously strong tropospheric wave driving does not always induce an anomalously weak stratospheric vortex, and anomalously weak stratospheric vortex states are not always preceded by anomalously strong tropospheric wave driving. This effect is related to the ability of the stratosphere to focus wave propagation: not all stratospheric states are equally receptive to an upward-propagating wave packet (the easiest example is to imagine strong tropospheric wave driving during a period of stratospheric easterlies following a major SSW — the waves cannot propagate into the easterlies, so the stratosphere is in control and not a slave to the troposphere).

Fully understanding what happens at the start of March to induce such strong stratospheric wave activity without strong tropospheric wave activity would probably take a bit more than a blog post. One possibility is that the vertically coherent, ‘sharpened’ vortex which develops during the reflection event is a particularly good guide for any amount of wave activity. It’s also possible that the downward-propagating wave event, primarily confined to the Western Hemisphere, helps to amplify the European ridge in the lower stratosphere (which itself constructively interferes with the mean stratospheric wave field), such that even a small amount of upward-propagating wave activity from the troposphere is sufficient to ‘ping’ a wavenumber-2 disturbance. Some resonant behaviour is possible.

In any case, the complexities of this interaction between the spheres likely contributes to why it was not well-forecast. Although these are processes that models can struggle to represent well (see also an upcoming paper led by Zac Lawrence, which should appear on WCD soon), the systematic shift across multiple different forecast systems suggests to me that this is more a case of the occurrence of something inherently unlikely and unpredictable on longer timescales. Some people like to criticise weather forecast models given any instance like this, but it must be remembered that the most likely outcome does not always occur — some forecast shifts aren’t entirely the fault of the model.

Long-term forecast

Looking further ahead, I don’t currently see any reason to shift the extended-range forecast too far from what I discussed last time. The split event — presuming it evolves as is currently being modelled, which seems a reasonable assumption at these lead-times — is only very brief. It does not induce the strong easterlies and extended recovery period that occur during a major SSW. The large Eurasian vortex returns quickly to the pole, taking the stratospheric NAM strongly positive once more. Things then continue in largely the same vein as before this synoptic-scale event occurred: the vortex responds “’tis but a scratch!” and continues kicking.

Having said that, I do think the events of the next week take a bit of a chunk of intensity out of the vortex, and thus probably bring the final warming a bit closer (seen in Thursday’s ECMWF extended-range) and with that limit the potential for an exceptionally strong late-season vortex. To “forecast the forecast”, I anticipate longer-term model confidence will improve from the first few days of March onwards (less stratospheric spread in the Thursday ECMWF forecast versus that on Monday…?).

Leave a Reply

Fill in your details below or click an icon to log in: Logo

You are commenting using your account. Log Out /  Change )

Facebook photo

You are commenting using your Facebook account. Log Out /  Change )

Connecting to %s

%d bloggers like this: