CoxR

I'm surprised to have not seen the alaskan ridge mentioned much in this forum as a potential point of concern regarding the longevity of the block. The ensemble suites, particularly GEFS have intensified the signal for the Alaskan ridge over the past 36 hours, pushing colder air into Greenland and North America, strenghening the jet stream as a result and making it difficult to maintain the high latitude block. For me this is the biggest complication to the potential cold spell, it looks like some cold weather is likely from around the 15th, but the longevity will depend largely upon whether the Alaskan Ridge signal continues to strengthen or diminishes as we approach T0, if it diminishes then it will likely make for a much cleaner evolution. I believe the Alaskan Ridge/ North Pacific Ridge is what scuppered the potential cold spell in December 2021

A point on ensembles: Ensembles are a powerful tool most of the time for gauging the spread of uncertainty and the possible solutions going forward. However, in reference to the GEFS, there is a reason that the control and the operational (low resolution and high resolution) run on the same initial conditions are included in the ensembles, and that is to determine whether resolution of the model plays a key role in the outcome. In the past 4 or 5 days or so we've seen consistent disparity between the operational (high resolution) and the control (low resolution) runs and to me that suggests one thing - that the resolution of the model run has a key role. When both the control and the operational run are broadly in sync then resolution doesn't play as big a part and therefore you can rely more on the mean or the spread within the ensemble suite, and the operational run (particularly at longer lead times) does not hold as much weight. However, when the two (operational and control) are out of sync, then this starts to suggest that the resolution of the run is playing a key role and for me you can then place less trust in the ensemble suite, as all perturbations within it are run at a lower resolution. When the operational run is consistently diverging from the control run, for me, you should place far more emphasis on the operational run (depending on the model and the model ensemble suite) than you otherwise would do with respect to its ensemble suite, when the control and operational are both in sync with each other. Therefore, I think and have thought for a long time, that it isn't wise to post a control run when the operational isn't showing what you want and isn't worth highlighting the ensemble mean and suite as much when the control and operational runs are not in sync. I think it would be generally helpful for this forum if people were mindful of this going forward Original post: https://community.netweather.tv/topic/99706-model-output-discussion-into-2024/?do=findComment&comment=4990200

A point on ensembles: Ensembles are a powerful tool most of the time for gauging the spread of uncertainty and the possible solutions going forward. However, in reference to the GEFS, there is a reason that the control and the operational (low resolution and high resolution) run on the same initial conditions are included in the ensembles, and that is to determine whether resolution of the model plays a key role in the outcome. In the past 4 or 5 days or so we've seen consistent disparity between the operational (high resolution) and the control (low resolution) runs and to me that suggests one thing - that the resolution of the model run has a key role. When both the control and the operational run are broadly in sync then resolution doesn't play as big a part and therefore you can rely more on the mean or the spread within the ensemble suite, and the operational run (particularly at longer lead times) does not hold as much weight. However, when the two (operational and control) are out of sync, then this starts to suggest that the resolution of the run is playing a key role and for me you can then place less trust in the ensemble suite, as all perturbations within it are run at a lower resolution. When the operational run is consistently diverging from the control run, for me, you should place far more emphasis on the operational run (depending on the model and the model ensemble suite) than you otherwise would do with respect to its ensemble suite, when the control and operational are both in sync with each other. Therefore, I think and have thought for a long time, that it isn't wise to post a control run when the operational isn't showing what you want and isn't worth highlighting the ensemble mean and suite as much when the control and operational runs are not in sync. I think it would be generally helpful for this forum if people were mindful of this going forward

This is a big signal for SPV disruption at this lead time, and the signal has been getting consistently stronger...

Charts | Copernicus CLIMATE.COPERNICUS.EU

avert your eyes people, it's not a good one:

I think it's fair to say the GEM ensemble is pretty keen on an SSW... Original post: https://community.netweather.tv/topic/98196-model-output-discussion-new-year-and-beyond/?do=findComment&comment=4793564

I think it's fair to say the GEM ensemble is pretty keen on an SSW...

EC46 looks very good btw

I've been having a look at monthly England precipitation data from 1870 onwards to investigate which years are most analogous to 2022 and the results are very interesting. In terms of patterns of precipiation observed from January through September 2022, the top analogues are 1887 and 1967 with 8 out of the 9 months' precipitation anomalies having the same 'sign' (i.e. either negative or positive/ dry or warm) as Jan-Sep 2022. Other good matches are 1870, 1874, 1876,1885, 1893, 1929, 1933, 1935, 1961, 1976, 1995, 1996, 2000, 2005 and 2010 with 7 out of the 9 months in these years sharing the same precipitation anomaly signs. 14 out of the 17 winters in this set of analogues had below average or cold Central England Temperatures with a median temperature anomaly of -0.85 degrees celsius. Decembers were particularly cold in this set of analogues with a median CET anomaly of -1.55 degC and Februaries were also cold with a median CET anomaly of -0.36 degC, whereas Januaries came out average (-0.05 degC). Plotting all years in the analogue set that are within NCEP's 1948-Sep2022 data timeframe, the average sea level pressure anomalies in December look like this: And for Jan: For Feb: For these winters as a whole: Assuming precipitation will turn out above average this month, I expanded the analysis through October and the top analogue was once again 1967 with 9 out of the 10 months sharing the same precipitation anomaly signs that have been observed from January to October 2022. 1870, 1874, 1885, 1887, 1929, 1933, 1935, 1961, 1976, 2000 and 2005 all had 8 out of their first 10 months sharing the same precipitation anomaly signs. For this analogue set, 11 out of the 12 winters were below average or cold with a median CET anomaly of -0.89 deg C. Again - Decembers were very cold on average with a median CET anomaly of -1.41 degC. 10 of the 12 Decembers were below average or cold, with 5 having CET anomalies below -2 degC, 2 of which were below -3 degC. Februaries were colder on average in this analogue set compared to the Jan-Sep analogues, with a median anomaly of -1.08 deg C and 6 of the 12 having CET anomalies below -1.5 degC. Januaries remained a similarly mixed bag with a median CET anomaly of -0.04 degC. Again, plotting sea level pressure anomalies for years in this analogue set within the 1948-2022 timeframe, you get these charts for the corresponding winter months. December: For Jan: For Feb: For winter as a whole: I do think that this analogue set's winter SLP anomalies in the northern hemisphere bear quite a strong resemblance to the most recent seasonal output from JMA for the upcoming winter period, with the JMA's pattern generally shifted west relative to the anologue set and some discrepancies over Siberia. Here are the two side-by-side (sorry that they're not centered at the same longitude): Comparing average September SSTAs of the Jan-through September data analogues to the SSTAs observed at the mid-point of September this year: It's worth noting that the analogue SSTAs are biased negative because 4 of the 5 are earlier than the mid-point of the 1991-2020 reference period and that the observed SSTAs from last month are biased positive due to 2022 being a lot later than the mid-point of the reference period. Generally tropical SSTAs are very similar with the negative IOD and La Nina imprint present in both. I would say that north Atlantic SSTAs are fairly similar, with the warm Atlantic blob present last month centered further west relative to the analogue years, which could possibly partly explain the generally westward pattern shift of JMA's MSLP output for DJF relative to DJF in the analogue years but that's probably just conjecture. On the other hand, the north Pacific looks quite different, you would have to say. This analysis was generally just out of curiosity and I certainly wouldn't want to make a forecast purely based on observed monthly precipitation prior to the target month, but it does beg the question of to what extent predictability of future weather patterns can be gained from looking at patterns of precipitation. Whilst attempting to build my statistical model of CET and England Precipitation I have found statisically significant associations between precipitation in certain months and CET several months down the line. I would otherwise not have investigated these links were it not for several discussions I've seen on here (and YouTube) hypothesising a link between warm Septembers and warm winters and also dry Septembers and warm winters. The former has no connection statistically speaking with winter CET, but the latter does: there was a 99.6% confidence that a wet September was associated with colder Decembers if I recall correctly, but there was no statistically significant association between September and any other winter month. I have to say, I went into that analysis very sceptical about there being any connection, and was intending to dispel the myth but needless to say I was proved wrong and I will now be incorporating patterns of precipitation into my model, so well done to whoever first suggested that link

no, DWD is going for a stronger vortex as is Meteo-France. As I said, you have to compare the thick blue line to the thick orange line, not the black line. If the blue line (the ensemble mean forecast) is above the orange line (the model climate mean) then the model is suggesting a stronger vortex, even if it is below the observed climatology. This is reflected in it's forecasted MSLP anomalies through the winter, where as a consequence of a stronger-than-average polar vortex, pressure anomalies over the Arctic are generally negative: And here too for Meteo-France:

Signals for polar vortex strength are actually very mixed for early winter with a general convergence towards a stronger vortex in late winter. When analysing the zonal mean U10 charts, you compare the ensemble mean forecast (thick blue line) to the model climate mean (thick orange line), not to the observed climatological U10, as each model has it's own biases either towards a strong vortex e.g. CMCC or a weak vortex e.g. DWD. Looking at the forecasts, JMA and UKMO are going for a weaker-than-average PV in early winter before strengthening in late winter: ECMWF is going for an average to slightly weaker-than-average PV in early winter before strengthening to a stronger-than-average PV: CMCC, Meteo-France and DWD are going for a stronger-than-average PV throughout the winter: Here are the U10 verification stats of these models, provided by World Climate Service on twitter: Looking at these verification stats, models that have historically performed best at predicting December U10 (Meteo-France, DWD, CMCC and ECMWF) are forecasting an average to stronger-than-average PV in December and the model that has previously performed best for January (CMCC) is forecasting a stronger-than-average PV. So based on this, you'd have to say that a strong PV is more likely than a weak PV in early winter. However, the sample size of forecasts used in this verification analysis is small and models are updated regularly, so U10 forecasts for the likes of JMA and UKMO may well have improved over time. And looking through the U10 forecasts from October 2021, JMA, which is forecasting the weakest PV for the upcoming winter, probably performed best for winter 2021/22.

I'd take some comfort in the fact that the GloSea October update in 2010 forecasted these temperature anomalies for NDJ 2010/11: And the October update in 2009 forecasted these temperarture anomalies for DJF 2009/10:

A statistical model produced by Giacomo Masato - which performed very well for the summer period - has been very consistent since July and similar to the ECMWF but perhaps suggesting a scenario more favourable for snow. This was the 500hPa anomaly chart for NDJ init. in July: From August for NDJ: From September for NDJ: Released today for NDJ: I'm currently developing a statistical model for Central England Temperature and UK precipitation myself for lead times of up to 11 months. My model takes into account the Nino 3,4 and 3.4 SSTAs as well as northern hemisphere extratropical SSTAs , the SOI, the QBO, the NAO, the AO, the MJO, the Atlantic Tripole, the PDO, the NPO, the IOD, the AMO, Eurasian snow cover, North American snow cover, Arctic Sea Ice extent, patterns of precipitation and patterns of surface temperature. I have been keeping a close eye on December and I have to say, my model has been very strongly hinting at a cold December for a while. I'm still a fair way off the final product and it may not be ready before winter, but nevertheless I will keep this forum apprised when I do start getting results. The model will essentially be predicting global-warming-detrended Central England Temperature and so will give more of a hint as to whether the month will likely have relatively cold or warm synoptics (or neither), but I will attempt to translate predicted values of detrended CET into expected values when accounting for the warming trend within the CET dataset.