Mike Poole

ECM clusters T120-T168: Support for the easterly on this timeframe growing, with clusters 2 (contains the op) and 3 - both take the low well south under the block. Cluster 1 more inclined to hold it out west, but this is the solution on the decline, I think. T192-T240: I think this timeframe is still work in progress! The UK rather in a no man’s land that could go one of a number of ways. More runs needed…

Rain All Night If each run takes 1 minute, 1000 member suites are easily possible…

minus10 Of course, we don’t know how long the AI model took to train itself to get to the point where it can knock out runs in 1 minute! And let’s not forget it requires as input the self-consistent T0 dataset calculated from the input data - that is a non-trivial calculation in itself which would take a lot more than 1 minute, it kind of gets that ‘for free’ as the main ECM models need it too.

ECM showing that there is potential in the upcoming setup. Close but no cigar at T216, but the potential to land something more significant is there before that. While I’m on, here’s the translated blurb for the AIFS model on Meteociel, for those interested: ‘These maps present data from the experimental European CEP/ECMWF AIFS model available in Open-Data. The maps are updated randomly according to the runs but generally one to two hours after the classic deterministic model is released. The resolution of the Open-Data model is approximately 0.25/0.28° (or approximately 25km/30km). There are fewer fields available than for the classic model. All fields are available up to 360h for runs of 00Z, 06Z, 12Z and 18Z. These data are provided on an experimental basis for testing forecasts made by artificial intelligence/machine learning, the calculation of the model by the AI is done in one minute (!). The model was trained with ERA5 reanalyses and classic ECMWF HRES runs.’ The model takes 1 minute to run - wow!

Good spot! It will be much easier to bring this into the discussion more if it is in the same format and with the same plotting variables as all the others!

ECM op well supported in the clusters T192-T240: It’s cluster 2 with 18 members. Fine margins with that high pressure to the north between the easterly and cluster 3 which has a southwesterly. Cluster 1 looks a bit of a mess. I still think the easterly may have some legs, it has shown for a while now, and like others have said, I think is has the potential to upgrade as the models factor in the details of the strat warming over the next week or so.

Met4Cast I completely agree. This winter was a major test for the new version of the 46 which, as well as running every day, also switched from being run as an extension to the EPS, to being run separately from T0 on a low resolution 36km horizontal grid. Which (I think) has actually contributed to its output essentially being reduced to noise long before the timeframe it is supposed to provide guidance on! ECMWF must surely have a rethink on that one!

Met4Cast February is a fail of epic proportions really. Whether your long range forecasting method of choice is the seasonal models, the ECM 46, or the GSDM, or ENSO, or the MJO, or the strat, (or tea leaves ), all of them at some stage (or indeed many stages) pointed to a decent spell of cold in February. The worst offender was the ECM 46 which persisted with it well beyond the point where everything else had given up the ghost.

Good post. However, the clusters are not a mean representation in the graphics. This, for me, is exactly the thing that sets them apart from other representations for the week 2 range. The graphics are from the representative member, not the mean. ECM website explains: ‘The most representative member of each cluster scenario is the member of the cluster which has the minimum difference from the RMS of all the ENS members in that cluster.’ Because it is an actual calculated solution and not averaged, it is therefore as physically possible as any op run.

ECM clusters again flirting with the easterly in the T192-T240 timeframe: Both clusters 1 and 2 take the low to the SW on a very southerly track, allowing flow from the east or southeast across the UK, or part of it. Together these are 46 of the 51 members. Still fluctuating suite to suite, to be honest, as yesterdays 12z had more mild solutions. You’d feel that with the strat situation as it is, the trend would against the Atlantic option, though. And as I said yesterday, if we do get the continental flow, it may require some patience before any significant cold air heads our way.

IDO The mean won’t verify. It can’t. It’s the average of many physically possible runs, but that doesn’t yield a pressure pattern that is physically possible itself. In this circumstance, it will be either one of the cold evolutions or one of the mild evolutions that have been showing on the op runs and the ensembles, and we just don’t know yet. I do think that if it went the way of the easterly, it would need to properly persist for quite a while to develop a suitable cold pool, with the reversal in the strat, that might not be as unlikely as it sounds.

I think a stronger showing for the easterly on the ECM clusters T192-T240: Both clusters 2 and 3 are now orientating the high more favourably, with the low initially SW taking a more southerly track, setting up an easterly flow in between. Cluster 1 has the low too strong and the high poorly orientated and a mild flow for the UK.

nick sussex Yes, the ECM is out of the same stable as yesterday’s GEM 12z. Would need some luck to land that - the thing is though, the winter cold-synoptics shield is switched off now, so at least we are in a fair game. Sadly, I never feel we are in a fair game when such things show themselves on the model output in January.

ECM clusters tonight, from T192 right through: It is cluster 2 (on both timeframes) that continues to provide interest for me, as this one could potentially draw some very cold air from the east/north east. The Atlantic low looks to stay camped west, allowing for the UK feed to be from a continental source. Cluster 1 also has blocking highs to the north, but never oriented such as to bring air from the east, and the UK is in mild air. Could easily go one way or the other here. The zonal wind reversal looks to be really quite severe now: The mean is bottoming out at about -18m/s. Expect the models to firm up on things a bit once the reversal happens about next Monday, the main effect of that to be seen in the trop may still be beyond the 2 week range, but something is suggesting quite a quick response in what we are seeing already - interesting times.

GEM 12z is up for a full on March cold spell by day 10: Incoming! I’ve been watching this possibility for a couple of days now, it’s been there on the ECM clusters, not favourite, but it is in the mix, and now on this GEM op run.

Met4Cast Means at that range possibly hiding some quite different options within a theme of a block to the north somewhere. Extended ECM clusters T264+ show this quite nicely: For the UK, all depends on where the block is and how it is angled. Cluster 1 (17 members) looks potentially quite cold. Cluster 2 (16 members) looks quite settled. The other two clusters have weaker high pressure with the UK remaining unsettled, cluster 4 in particular being mild.

GFS 12z going with the easterly that was suggested on the clusters this morning from day 11. T264: T312, and the cold air finally beginning to make itself felt.

Just perhaps worth watching a possible easterly setup around the 10 day timeframe. Clusters at T192-T240: Cluster 2 (25 members) showing more organised heights to the north, this was there in the extended clusters yesterday with just 11 members, so maybe an increasing signal. Some work to do to get from there to a proper easterly flow, the low to the SW would need to go under, or keep well away, and for any cold interest at this time of year, flow from anywhere south of east would not cut it either. But this would be post wind reversal in the strat, so worth watching as the models sort themselves out after that.

ECM clusters tonight have some interesting setups, T192-T240: Strong (red border) signal for blocking around Scandi there, clusters 1,5 and 6 the potential to bring winds from the east, but probably south of east on clusters 1 and 5 so its unlikely to be particularly cold. Clusters 2,3,4 have UK in a southwesterly. T264+: A variety of patterns there, but some high latitude blocking in most, with clusters 3 and 4 showing the coldest evolutions. Cluster 6 on its own with high pressure into the UK in week 2. Zonal winds chart showing a strong reversal consistently now, could strengthen an existing signal for blocked patterns into mid-March, interestingly the mean does make it back into positive territory into April:

Just looking at the uncertainty on the cold plunge next week, T144. The ECM clusters have 5 scenarios: Looking at the representative members at T144, the 580 hPa temperatures are as follows (in order for clusters 1-5): Clusters 1 to 4 have low pressure over the UK. Cluster 5, although cold might well be dry. Cluster 2 looks like it will be cold enough for snow in many places, cluster 3 also locally pretty cold, cluster 1 marginal, cluster 4 very marginal.

MJB For me, unfortunately, the evidence seems to increasingly suggest that we can no longer get the right set up, during the months that matter, or at least it is now extremely unlikely. It is not a timing issue, March (and April) are different and winds from the north or east have always been a little more likely in spring than winter. It is the almost total absence of any snowy setups in the winter months during the last decade (since Jan 2013 actually) that is worrying.

GFS, GEM and UKMO all showing a trough dropping over the UK with cold air from the NW. T144: Certainly some interest in the north for snow, especially with elevation.

ECM clusters T192-T240: Clusters 2 and 3 there doing their best to pull in an easterly under a Scandi high. I think appearances might be a little deceptive, as I’m sceptical that there’s any cold there, so a toothless easterly is on the cards. A northwesterly as cluster 5 might actually bring colder air. T264+: Longer term the Scandi high option in cluster 1 with 22 members is the preferred evolution. Cluster 2 moves to an Atlantic ridge. Not sure whether these evolutions are related to events in the strat, zonal winds on the 46 dayer trend to a slightly stronger reversal than yesterday, but seems a little soon to equate that with the Scandi high signal.

GEM manages to bring in some rather surprisingly cold air country wide along with that low at T240:

joggs Yes, it is certainly a possibility that has done something, it did spew a whole load of water vapour into the strat. However, it is not clear cut, because that would as I understand it make the strat colder and the vortex stronger, and we’ve had a weak vortex all winter, just one that would not deliver for UK coldies.