September 2022 C.E.T. and EWP forecast contests

[Roger J Smith]

CET _ As the 14.6 is from 14.58 rounded, and today's mins appear to be around 5 on average, with the limited warming so far, there probably is a good chance of ending up at 14.4. A daily mean of 10.7 or below is likely enough to produce that result (assuming they don't change any numbers already posted).

EWP _ We were probably close to 78 mm going into the rainfall event and will likely end on about 91-92.

[The PIT]

Sunny Sheffield at 14.6C +0.5C above normal. Rainfall 47.3mm 71.9% of the monthly average.

[Roger J Smith]

16 minutes ago, summer blizzard said:

When was the last colder second half.

Assuming an average of 12.0 this year (16th-30th) then the answer would be 2012 (11.3). In 2015 it was 12.3. 

Going back further, only these years were colder in the last half century: 1972 (11.1), 1974 (10.2), 1979 (11.9), 1986 (11.4), 1990 (11.6), 1993 (11.2). 

The following were just above where this year seems to be headed: 1973 (12.1), 1988 (12.2), 1994 (12.3), 2015 (12.3).

The warmest were 2006 (16.5), 1985 (15.4), 2011 (15.1), 2014 (15.0), 1998 (14.8), 1989 and 2021 (14.7), 1999, 2016 and 2019 (14.6).

[richie3846]

5 hours ago, Roger J Smith said:

Assuming an average of 12.0 this year (16th-30th) then the answer would be 2012 (11.3). In 2015 it was 12.3. 

Going back further, only these years were colder in the last half century: 1972 (11.1), 1974 (10.2), 1979 (11.9), 1986 (11.4), 1990 (11.6), 1993 (11.2). 

The following were just above where this year seems to be headed: 1973 (12.1), 1988 (12.2), 1994 (12.3), 2015 (12.3).

The warmest were 2006 (16.5), 1985 (15.4), 2011 (15.1), 2014 (15.0), 1998 (14.8), 1989 and 2021 (14.7), 1999, 2016 and 2019 (14.6).

I felt a shudder when I read '2012 11.3c'. That was the summer/autumn without mercy, I never want to experience that again in my life thanks! It was absolutely and utterly gross.

[damianslaw]

A very contrasting month, one of those switcharound ones, where a very mild first half is cancelled out by a colder second half to produce a near average finish, and a mean that doesn't reflect the month at all.

[Stationary Front]

Looks like it’s confirmed at 14.4 shame as I had 14.5 for my guess!

[ItNeverSnowsInWindsor]

Wow! Some drop that!

Wouldn’t mind a similar drop during Xmas!   

[Don]

1 hour ago, Stationary Front said:

Looks like it’s confirmed at 14.4 shame as I had 14.5 for my guess!

Yep after a very warm first week, an impressive drop and comfortably under 15C!  As for my guess, one of my better this year at 0.4C out from 14.8C

[Frigid]

Wow I actually guessed it right this time . That drop was huge, second half of September was notably cool compared to the first. 

[damianslaw]

Whilst comfortably over the 61-90 average which is fast becoming very out of date to compare mean against, why officially still use it I don't know, Sept 2022 was bang on the 01-20 average, and only 0.4 degrees above the 81-10 average, which perhaps should become the mean we measure against. Therefore a very near average month.

Anyone know why 61-90 still used? When did 41-70 stop being used in 1990? or 2000? Either way its more than 30 years since 1990.

Edited October 1, 2022 by damianslaw

[The PIT]

Sunny Sheffield finished on 14.5C +0.5C so average basically. Rainfall finished on 62.3mm 94.7% of the monthly average.  Pretty average month then despite the warm wet start followed by a dry spell of 15 days where only 2.1mm of rain fall.

[Don]

2 hours ago, Frigid said:

Wow I actually guessed it right this time . That drop was huge, second half of September was notably cool compared to the first. 

The closest I've got to being correct this year was back in February when I was 0.3C out.  At least this time my guess was too high which is what I was hoping for!!

[LetItSnow!]

1 hour ago, damianslaw said:

Whilst comfortably over the 61-90 average which is fast becoming very out of date to compare mean against, why officially still use it I don't know, Sept 2022 was bang on the 01-20 average, and only 0.4 degrees above the 81-10 average, which perhaps should become the mean we measure against. Therefore a very near average month.

Anyone know why 61-90 still used? When did 41-70 stop being used in 1990? or 2000? Either way its more than 30 years since 1990.

I think 1961-1990 is still used as for them remind of the change in temperature - that period was the last with temperature largely pre-heated levels. Many years and months in recent times have only had relatively small anomalies compared to 1991-2020, so some people may assume the difference is not so big and so it's not a big deal. Like, for example, the year 2021 barely had an anomaly in temperature in the C.E.T. with 1991-2020, but around +0.7C against 1961-1990, and some areas of the country having a slightly cooler than average year.

Having said that, I think September 2022 is the first month since June 2022 to have a 1961-1990 anomaly below +1.0C (+0.8C). The last month coming out cooler than 1961-1990 being May 2021 (-0.9).

[ItNeverSnowsInWindsor]

1 hour ago, Don said:

The closest I've got to being correct this year was back in February when I was 0.3C out.  At least this time my guess was too high which is what I was hoping for!!

I’d love to guess a winter month below average and it coming in even colder

[Don]

2 minutes ago, prolongedSnowLover said:

I’d love to guess a winter month below average and it coming in even colder

Happened to me in January and December 2010!!

[Walsall Wood Snow]

Wow I managed to guess correct for once. Anyway at 14.4c September 2022 shares the same mean CET with September's:

1702

1724

1727

1736

1747

1843

1854

1869

1926

1936

1945

1970

1981

2002

 

[DR(S)NO]

On 27/08/2022 at 11:39, DR(S)NO said:

Cool and wet towards the end of the month for me.

14.5c and 85mm. Thank you please.

Happy with my predictions this month for a change

[Roger J Smith]

As to the use of 1961-1990, it is definitely an outlier among national weather services, Ireland and Canada use 1981-2010 in their data tables, and the USA already adjusted from 1981-2010 to 1991-2020 almost a year ago. So here's my usual pre-official listing of the CET order of finish based entirely on the table of entries, to be confirmed by J10 later. Numbers in brackets are order of entry. EWP forecasts are also shown and top "best combined" now based on the scoring order from the EWP tracker estimate of 89.7 mm, which may change a bit on 5th when final values are shown in tables. I will just edit in any changes in the best combined feature of this post and edit this wording. The CET rank order ignores effect of late penalties in the scoring table, but for "best combined" each late entry is dropped one rank per late day (as the EWP rank already includes a similar rank change). 

Rank __ CET _ EWP _ Forecaster (order of entry) ____ best combined top results

_01 __ 14.4 _ -- -- _ Walsall Wood Snow (32) _____

_02 __ 14.4 _110.0_  Frigid (33) __________________ (2, 28) = 30 _8th best combined

_03 __ 14.5 _110.0_ Neil N ( 3 )  __________________ (3, 30) = 33 _t10th best combined

_04 __ 14.5 _ 85.0 _ DR(S)NO ( 21 ) _______________ (4, 7.) = 11 __2nd best combined

_05 __ 14.5 _ 90.0 _ Stationary Front ( 31 ) _______ (5, 2) =7 best combined

_06 __ 14.5 _ 57.0 _ stargazer ( 36 ) ______________ 

_07 __ 14.5 _ 87.0 _ Mr Maunder ( 43 ) ___________ (7, 5) =12 __ 3rd best combined

_08 __ 14.3 _ -- -- __ Summer Sun ( 44 ) __________

_09 __ 14.3 _115.0_ davehsug ( 48 )  _____________ (9,36) = 45 __ 18th best combined

_10 __ 14.6 _ 67.0 _ Reef ( 27 ) ___________________ (10,33) = 43 _t15th best combined

_11 __ 14.6 _ -- -- __ J.P.  ( 38 ) ____________________

_12 __ 14.2 _ 76.0 _ freeze (41) ___________________ (12,20) = 32 _ 9th best combined

_13 __ 14.2 _ 85.0 _ Federico ( 42 ) _______________ (13, 9) = 22 _ 4th best combined

_(10) _ 14.2 _ 76.0 _ average 1991-2020 __________

_14 __ 14.2 _ 76.0 _ daniel* ( L1-6 ) _______________ (15, 21) = 36 _ 13th best combined

_15 __ 14.7 _ 80.0 _ summer blizzard ( 7 ) ________ (14, 15) = 29 _ 7th best combined

_16 __ 14.1 _ 82.5 _ Weather26 ( 8 ) ______________ (16, 12) = 28 _ 6th best combined

_17 __ 14.1 _ 79.0 _ SLEETY ( 11 ) _________________ (17, 16) = 33 _ t10th best combined

_18 __ 14.7 _ 66.0 _ Godber 1 ( 20 ) _______________

_19 __ 14.1 _ 63.0 _ virtualsphere ( 22 ) ___________

_20 __ 14.7 _ 61.5 _ prolongedSnowLover ( 46 ) ___

_21 __ 14.1 _ -- -- _ Mark Smithy ( L1-2 ) ___________

_22 __ 14.7 _ 90.0 _ J10 ( L1-5 ) ____________________ (23, 3) = 26 _ 5th best combined

_23 __ 14.8 _118.0_ I Rem Atl 252 ( 45 ) ___________

_24 __ 14.8 _ 77.0 _ Don ( 51 ) _____________________ (24, 19) = 43 _t15th best combined 

_25 __ 14.8 _ 69.0 _ Midlands Ice Age ( L1-3 ) _____

_26 __ 14.0 _ -- -- _ Kentish Man ( L1-4 ) ___________

_(26) _ 14.0 _ 77.2 _ average 1981-2010 ___________

_27 __ 14.9 _ 86.0 _ SteveB ( 9 ) ___________________ (27, 6) = 33 _t10th best combined

_28 __ 14.9 _ 72.7 _ Roger J Smith ( 18 ) ___________

_29 __ 14.9 _ 99.0 _ weather-history ( 35 ) _________ (29,14) = 43 _t15th best combined

_30 __ 14.9 _ -- -- __ Mark Bayley ( 30 ) _____________

_31 __ 14.9 _ -- -- __ damianslaw ( 49 ) _____________

_(27) _ 14.9 _ 76.5 _ consensus ____________________

_32 __ 15.0 _ 78.0 _ Bobd29 ( 2 ) ___________________ 17th EWP

_33 __ 15.0 _ 85.0 _ summer8906 (25) _____________ (33, 8.) = 41 _t14th best combined

_34 __ 15.0 _ 68.0 _ sunny_vale (29) ________________

_35 __ 15.0 _ 60.1 _ Jonboy ( 39 ) ___________________ 

_36 __ 15.0 _ 75.0 _ Mulzy (40) _____________________

_37 __ 15.1 _ 62.0 _ dancerwithwings ( 5 ) _________

_38 __ 13.7 _ 75.0 _ snowray ( 6 ) __________________

_39 __ 15.1 _ 63.0 _ Wold Topper ( 24 ) ____________ 

_40 __ 15.1 _ 83.0 _ February1978 ( 52 ) ___________ 11th EWP

_41 __ 15.1 _109.0_ noname_weather ( L1-1 ) _____

_42 __ 15.2 _ 77.0 _ Emmett Garland ( 34 ) ________ 18th EWP

_43 __ 15.2 _ 81.0 _ Feb1991Blizzard ( 26 ) ________ 13th EWP

_44 __ 15.3 _ 40.2 _ Polar Gael ( 10 ) ______________ 

_45 __ 15.3 _ 65.0 _ seaside60 ( 47 ) _______________ 

_46 __ 15.4 _ 90.0 _ Weather Enthusiast91 ( 4 ) ___ 1st EWP 

_47 __ 15.5 _ -- -- __ Stretford End ( 12 ) ___________

_48 __ 15.5 _ 72.0 _ Summer18 ( 13 ) ______________

_49 __ 15.5 _ -- -- __ richie3846 ( 14 ) ______________

_50 __ 15.5 _ -- -- _ DiagonalRedLine ( 16 ) ________ 

_51 __ 15.5 _ 60.0 _ The PIT ( 23 ) _________________   

_52 __ 15.6 _ 61.0 _ rwtwm ( 17 ) __________________

_53 __ 15.8 _ 60.0 _ Ed Stone ( 19 ) ________________

_54 __ 15.8 _65.0 _ syed2878 ( 28 ) ________________

_55 __ 12.9 _ 95.2 _ Kirkcaldy Weather ( 37 ) ______ 10th EWP

_56 __ 15.9 _ 73.0 _ Tidal Wave ( 50 ) ______________

_57 __ 16.5 _ 88.0 _ shillitocettwo ( 15 ) ___________ 4th EWP

_58 __ 17.0 _ 40.0 _ Earthshine ( 1 ) _______________

(top 20 EWP are shown, best combined rank totals end at 45)

Edited October 3, 2022 by Roger J Smith

[Roger J Smith]

The EWP tracker has ended up at 89.7 mm. This is just a bit below my scoring estimate of 91.1, and close enough that it will only change a few scores by slight amounts if it survives the final edit on the 5th (when a table value is posted). Since that could be back up closer to 91 than this value, I will hold off making any changes to scoring material posted so far. Whatever value they hit on the fifth, I don't expect changes to be more than superficial to the scoring already shown. This does not change the first four scores at all, but breaks a tie for 5th in favour of the lower of the two values. Check back around the 5th to see what the final value is and any required edits to scoring tables. 

(later edit _ the tables with EWP info have been adjusted to fit 89.7 mm, will be adjusted again if necessary after 5th Oct)

Edited October 3, 2022 by Roger J Smith

[Frigid]

2 hours ago, Roger J Smith said:

The EWP tracker has ended up at 89.7 mm. This is just a bit below my scoring estimate of 91.1, and close enough that it will only change a few scores by slight amounts if it survives the final edit on the 5th (when a table value is posted). Since that could be back up closer to 91 than this value, I will hold off making any changes to scoring material posted so far. Whatever value they hit on the fifth, I don't expect changes to be more than superficial to the scoring already shown. This does not change the first four scores at all, but breaks a tie for 5th in favour of the lower of the two values. Check back around the 5th to see what the final value is and any required edits to scoring tables. 

Ah, if only I kept my original ewp forecast of 90mm. That with the 14.4c prediction would've been epic.

[J10]

Excel Sept 22 CET.xlsx
PDF Summary Sept 2022 Summary.pdf

Seasonal and Monthly

This month 2 players got it spot on Walsall Wood Snow and Frigid.
A further 7 were 0.1c out.

Overall

The same top 5 remains however there have been some changes in order.

Freeze now leads (from 2nd)
richie3846 2nd (from 1st)
Stationary Front 3rd (from 2nd)
February1978 4th (from 4th)
seaside 60 5th (from 3rd)

Sept 2022 Summary.pdf Sept 22 CET.xlsx

Edited October 2, 2022 by J10

[Roger J Smith]

Here is an update on how consensus and the two normals are doing. The contest was scored using legacy CET values Dec to Apr and v2.0 (all that now exists) in May to August. These tables show the actual values, then the errors, and the contest ranks of the entrants who have similar values for average error. This gives an approximate idea of how the "robots" are doing although their scores might have added up somewhat differently as the forecasters with those average errors ranked differently in specific months. 

 

(a) actual values

ROBOT ______ DEC _ JAN _ FEB _ MAR _ APR _ MAY _ JUN _JUL _AUG _ SEP

Consensus __ 4.8 __ 5.4 __ 5.6 __ 7.1 __ 8.4 _ 12.3 _ 15.1 _ 17.5 _17.4 _ 14.9

1981-2010 ___4.6 __ 4.4 __ 4.4 __ 6.6 __ 8.5 _ 11.7 _ 14.4 _ 16.6 _16.5 _ 14.0

1991-2020 ___5.0 __ 4.7 __ 4.9 __ 6.7 __ 8.9 _ 11.9 _ 14.6 _ 16.8 _16.6 _ 14.2

contest CET _ 6.4 __ 4.6 __ 6.8 __ 7.9 __ 9.1 _ 13.1 _ 14.9 _ 18.2 _18.7 _ 14.4

 

(b) absolute errors

ROBOT _____ DEC _ JAN _ FEB _ MAR _ APR _ MAY _ JUN _JUL _AUG _SEP __ average

Consensus __ 1.6 __ 0.8 __ 1.2 __ 0.8 __ 0.7 __ 0.8 __ 0.2 __ 0.7 __1.3 _ 0.5 __ 0.86

1981-2010 ___1.8 __ 0.2 __ 2.4 __ 1.3 __ 0.6 __ 1.4 __ 0.5 __ 1.6 __ 2.2 _ 0.4__ 1.24

1991-2020 ___1.4 __ 0.1 __ 1.9 __ 1.2 __ 0.2 __ 1.2 __ 0.3 __ 1.4 __ 2.1 _ 0.2 __ 1.00

SUMMARY: Consensus was warmer than both normals for all months except Dec (where it was between them) and Apr (when it was below both). The actual value was warmer than both normals in Dec, Feb, Mar, Apr, May, June, July, and August, so the contest consensus was correctly placed in Feb, Mar, May, (marginally so) June, July and August. In December and April, our consensus was too low relative to normal, and in January and June it was too high (not by much in June though). In September, the result was between the normals and our consensus with generally similar errors. The overall performance of consensus remains considerably better than the 1991-2020 normal and more considerably better than the cooler 1981-2010 normal value. The lowest average error would have been achieved by any strategy of making forecasts around 1.2 above 1991-2020 averages. The average error of those forecasts (1991-2020 + 1.2) would have been around 0.62 deg.  

 

(c) Contest ranks (now compared against all ranked forecasters incl those at 8/10)

optimal strategy (91-20)+1.2 (avg error 0.62) __ 4th (seaside60 0.56, freeze 0.58, seaside60 0.56)

consensus (avg error 0.86) ____________________ 16th (the three above and 12 others are ahead of 0.86)

1991-2020 (avg error 1.00) ____________________ 24th (another eight were 0.87 to 0.99)

1981-2010 (avg error 1.24) ____________________ 40th (another 16 were 1.01 to 1.23)

(Average error is well correlated with total score but not perfectly so, however it can be inferred that contest ranks overall for the three "robots" would be similar to the above.)

Obviously playing an optimal strategy is a good one, if you can guess in advance what that is; at the moment, any optimal strategy between adding 0.5 and 1.5 to 1991-2020 would put you pretty close to the lead held by the three mentioned above;  you could be well into the top ten, even just using 1991-2020 is good for 24th place in average error. Using 1981-2010 gets you no further into the table than lower middle ranks (51 are ranked based on at least eight of ten entries). Since average error is not the only component of CET contest scoring, it would take a more rigorous analysis to work out the points totals for these automatic type strategies but they are bound to be close in rank to the average error. 

An equally good strategy is to predict consensus plus 0.7, this results in an average error of just 0.58, which is equal to second lowest average error among forecasters. Anything in the range of consensus plus 0.5 to 1.0 is similar in outcome (slightly less accurate than 0.58).

[Weather-history]

EWP for September is 89.9mm, wettest September since 2019 and wettest month since February. 

 

[Roger J Smith]

The 89.9 mm EWP posted (eventually) will not change any rankings or scores, and will just make slight changes to average errors which I will edit into the table already posted back around the 29th of Sep or on page nine (as I'm seeing it, maybe it's not the same pagination for every account here). So those average error values will be changed later today, I hope. Since they were last calculated from 89.7 mm and we've had ten months, they will only change by .02 mm for those with ten forecasts entered, a little more for others. 

Already noted somewhere but there was one daily record CET value, a high minimum of 15.4 on the 4th. (in v 2.0). It was a relatively weak record as there are higher ones here and there to the end of the month. The lowest minimum of 3.7 on 17th was not quite a record low for the date, those start approaching frost levels in mid-September (1-2 C records are common).