================================================================================
FORECASTSEASON.COM | THE SEASONAL WIRE
Generated: 2026-10-16T18:52:14Z  |  Initialisation: 1 October 2026  |  Route JSONs on disk: 8
Period: November 2026 -> March 2027 (5 whole months; lead 1 is the run's own month and is part elapsed)
Systems: SEAS5 (51 members, reforecast 1993-2016), C3S multi-system (10 systems, reforecast 1993-2016), NMME (4 models, reforecast 1991-2016)
Source: C3S seasonal-monthly-single-levels and IRI/NMME, each system verified
        against its OWN reforecast and ERA5 monthly means at 1 degree.
Format: machine-readable. Sections marked ##; tables are pipe-delimited with a
        header row that names the unit of every column. Probabilities are
        percent and 33.3 is climatology; anomalies are degrees Celsius unless
        the column says otherwise. ASCII only.
Terms:  p_above_pct = chance the month lands in the upper third of that system's
        own trend-adjusted climatology. beats_clim_pct = share of land in the box
        where that system's reforecast scored better than issuing climatology.
        rpss_land = ranked probability skill score over land; 0 is climatology.
================================================================================

## WHAT THIS IS
Every outlook forecastseason.com publishes, in one file, for broadcast desks,
models and agents. Nothing here is invented and nothing is computed by this
file: each number is lifted from the route that published it, and the
ENDPOINTS section at the bottom says which route and where its full JSON
lives. Where a route is absent from this build its section says so rather
than being quietly missing.

## THE HEADLINE
C3S multi-system puts S. America temperature at 74 percent likely above normal for November 2026 - the strongest tilt in this run.
Its own reforecast beat climatology over 96 percent of that box; the three
systems differ by 11 points over the United States in February 2027.

## BROADCAST LINES
1. C3S multi-system puts S. America temperature at 74 percent likely above normal for November 2026 - the strongest tilt in this run.
2. Its own reforecast beat climatology over 96 percent of that box; the three systems differ by 11 points over the United States in February 2027.
3. C3S multi-system takes 9 of 10 whole-month scorings on reforecast RPSS over land
4. Skill decays with lead: over the United States, 54-61 percent of the box beat climatology at November 2026, 46-89 percent by March 2027

## SEASON CLOCK
Autumn in the northern hemisphere, 26 percent elapsed
Runs to the December solstice at 2026-12-21T20:50:13Z
This is astronomy, not a forecast: Meeus, Astronomical Algorithms ch 27.

## NEXT RELEASES
date_utc|centre|cadence|days_away
2026-11-05T12:00Z|ECMWF SEAS5|5th of the month, 12 UTC|20
2026-11-05T12:00Z|Copernicus C3S multi-system|5th of the month, 12 UTC|20
2026-11-08T12:00Z|NMME|around the 8th|23
2026-11-19T20:00Z|NOAA CPC outlooks|third Thursday|35

## THE OUTLOOK BY REGION
One row per region, field and target month. Every system on the same six
boxes, the same land mask and the same averaging code, so the columns are
directly comparable. part_elapsed=1 marks the initialisation's own month.
region|field|month|lead|part_elapsed|seas5_p_above_pct|c3s_p_above_pct|nmme_p_above_pct|seas5_beats_clim_pct|c3s_beats_clim_pct|nmme_beats_clim_pct
United States|t2m|October 2026|0|1|44.2|43.9|40.5|85|98|96
United States|t2m|November 2026|1|0|25.9|35.7|32.2|54|60|61
United States|t2m|December 2026|2|0|37.4|37.6|35.0|41|56|56
United States|t2m|January 2027|3|0|32.3|32.7|37.1|47|72|42
United States|t2m|February 2027|4|0|24.3|27.1|35.8|53|90|93
United States|t2m|March 2027|5|0|33.5|28.1|31.5|46|89|64
United States|precip|October 2026|0|1|25.0|39.0|33.8|78|89|76
United States|precip|November 2026|1|0|45.5|44.6|43.9|41|60|52
United States|precip|December 2026|2|0|40.5|44.1|41.0|47|61|61
United States|precip|January 2027|3|0|41.4|44.8|43.5|37|67|52
United States|precip|February 2027|4|0|47.6|43.8|45.0|42|61|62
United States|precip|March 2027|5|0|48.0|41.7|44.1|43|56|45
Europe|t2m|October 2026|0|1|34.7|36.2|38.0|77|81|82
Europe|t2m|November 2026|1|0|36.3|38.5|40.1|38|70|68
Europe|t2m|December 2026|2|0|48.6|43.8|31.6|46|74|42
Europe|t2m|January 2027|3|0|46.6|44.6|45.7|29|44|19
Europe|t2m|February 2027|4|0|45.5|42.9|39.8|34|82|28
Europe|t2m|March 2027|5|0|38.8|36.9|41.4|47|68|62
Europe|precip|October 2026|0|1|37.3|38.8|36.1|64|86|67
Europe|precip|November 2026|1|0|36.8|38.0|34.0|29|47|35
Europe|precip|December 2026|2|0|35.7|38.7|37.0|26|40|29
Europe|precip|January 2027|3|0|40.3|39.4|38.3|32|64|43
Europe|precip|February 2027|4|0|46.1|42.3|38.2|27|59|38
Europe|precip|March 2027|5|0|36.8|40.8|39.8|38|46|42
Asia|t2m|October 2026|0|1|41.4|47.9|38.9|94|99|95
Asia|t2m|November 2026|1|0|40.5|47.4|38.0|42|66|53
Asia|t2m|December 2026|2|0|39.5|44.4|37.4|51|74|48
Asia|t2m|January 2027|3|0|47.5|47.7|40.8|40|65|52
Asia|t2m|February 2027|4|0|50.5|52.0|44.2|51|83|56
Asia|t2m|March 2027|5|0|39.7|48.7|46.4|52|85|70
Asia|precip|October 2026|0|1|41.0|40.1|37.0|71|78|72
Asia|precip|November 2026|1|0|39.1|42.9|36.8|26|43|39
Asia|precip|December 2026|2|0|41.4|42.2|38.4|30|50|41
Asia|precip|January 2027|3|0|41.8|41.8|40.1|29|38|27
Asia|precip|February 2027|4|0|42.7|41.8|39.6|30|51|33
Asia|precip|March 2027|5|0|40.8|41.9|39.6|28|47|36
Africa|t2m|October 2026|0|1|46.7|59.7|46.2|91|99|98
Africa|t2m|November 2026|1|0|59.0|58.3|50.1|56|81|82
Africa|t2m|December 2026|2|0|63.0|62.7|61.0|65|81|71
Africa|t2m|January 2027|3|0|73.5|69.6|61.3|52|72|67
Africa|t2m|February 2027|4|0|74.8|75.2|62.8|73|91|88
Africa|t2m|March 2027|5|0|71.7|74.3|64.4|67|74|79
Africa|precip|October 2026|0|1|52.9|45.5|37.8|51|59|50
Africa|precip|November 2026|1|0|34.7|40.4|39.6|31|40|33
Africa|precip|December 2026|2|0|30.7|34.8|36.3|30|34|30
Africa|precip|January 2027|3|0|33.0|33.0|37.1|18|32|23
Africa|precip|February 2027|4|0|32.6|33.9|33.6|18|27|26
Africa|precip|March 2027|5|0|30.2|33.0|37.4|20|24|25
S. America|t2m|October 2026|0|1|67.2|68.7|59.6|95|100|99
S. America|t2m|November 2026|1|0|78.4|74.0|66.4|89|96|94
S. America|t2m|December 2026|2|0|79.2|77.7|74.8|68|85|77
S. America|t2m|January 2027|3|0|82.3|80.6|78.6|80|92|92
S. America|t2m|February 2027|4|0|82.8|82.0|78.9|86|94|92
S. America|t2m|March 2027|5|0|83.2|81.4|78.3|78|88|78
S. America|precip|October 2026|0|1|31.0|30.1|30.5|82|89|80
S. America|precip|November 2026|1|0|26.7|27.7|32.4|68|74|62
S. America|precip|December 2026|2|0|28.0|26.4|24.7|59|79|55
S. America|precip|January 2027|3|0|30.8|27.4|25.7|46|67|59
S. America|precip|February 2027|4|0|31.1|28.8|26.8|55|69|51
S. America|precip|March 2027|5|0|33.7|31.7|32.2|36|57|39
Oceania|t2m|October 2026|0|1|22.1|42.6|37.2|98|100|99
Oceania|t2m|November 2026|1|0|39.8|42.8|40.8|89|95|89
Oceania|t2m|December 2026|2|0|45.9|47.6|51.3|79|96|91
Oceania|t2m|January 2027|3|0|57.5|53.8|57.1|44|54|37
Oceania|t2m|February 2027|4|0|59.0|55.9|58.8|76|85|86
Oceania|t2m|March 2027|5|0|57.0|57.0|58.5|77|88|71
Oceania|precip|October 2026|0|1|46.2|38.7|43.9|71|81|83
Oceania|precip|November 2026|1|0|34.4|39.4|33.8|57|68|65
Oceania|precip|December 2026|2|0|37.3|38.2|32.4|56|74|63
Oceania|precip|January 2027|3|0|31.4|33.8|32.4|38|56|36
Oceania|precip|February 2027|4|0|31.7|31.6|30.4|57|70|47
Oceania|precip|March 2027|5|0|37.2|36.3|35.5|51|78|59

## VERIFICATION
Reforecast skill, land-masked, from three separately written implementations.
C3S multi-system scores highest at 9 of 10 whole-month scorings. Note the
reforecast periods differ: SEAS5 1993-2016; C3S multi-system 1993-2016; NMME
1991-2016 - so a comparison down a column is exact and one across is not.
system|field|month|lead|part_elapsed|rpss_land|beats_clim_pct
SEAS5|t2m|October 2026|0|1|+0.2480|91
C3S multi-system|t2m|October 2026|0|1|+0.2682|97
NMME|t2m|October 2026|0|1|+0.2556|96
SEAS5|t2m|November 2026|1|0|+0.0560|61
C3S multi-system|t2m|November 2026|1|0|+0.0944|78
NMME|t2m|November 2026|1|0|+0.0958|72
SEAS5|t2m|December 2026|2|0|+0.0420|60
C3S multi-system|t2m|December 2026|2|0|+0.0762|78
NMME|t2m|December 2026|2|0|+0.0541|66
SEAS5|t2m|January 2027|3|0|+0.0320|51
C3S multi-system|t2m|January 2027|3|0|+0.0816|71
NMME|t2m|January 2027|3|0|+0.0588|60
SEAS5|t2m|February 2027|4|0|+0.0740|64
C3S multi-system|t2m|February 2027|4|0|+0.1172|87
NMME|t2m|February 2027|4|0|+0.1034|77
SEAS5|t2m|March 2027|5|0|+0.0640|61
C3S multi-system|t2m|March 2027|5|0|+0.1019|81
NMME|t2m|March 2027|5|0|+0.0830|73
SEAS5|precip|October 2026|0|1|+0.1480|67
C3S multi-system|precip|October 2026|0|1|+0.1542|76
NMME|precip|October 2026|0|1|+0.1139|67
SEAS5|precip|November 2026|1|0|+0.0040|39
C3S multi-system|precip|November 2026|1|0|+0.0416|52
NMME|precip|November 2026|1|0|+0.0172|45
SEAS5|precip|December 2026|2|0|+0.0060|39
C3S multi-system|precip|December 2026|2|0|+0.0394|53
NMME|precip|December 2026|2|0|+0.0198|46
SEAS5|precip|January 2027|3|0|-0.0140|31
C3S multi-system|precip|January 2027|3|0|+0.0288|48
NMME|precip|January 2027|3|0|+0.0033|36
SEAS5|precip|February 2027|4|0|-0.0070|34
C3S multi-system|precip|February 2027|4|0|+0.0253|50
NMME|precip|February 2027|4|0|+0.0065|40
SEAS5|precip|March 2027|5|0|-0.0170|33
C3S multi-system|precip|March 2027|5|0|+0.0122|48
NMME|precip|March 2027|5|0|-0.0076|37

POOLING: C3S multi-system pools 10 systems and its consensus beats its own best member at 10 of 10 scorings (rainfall 5 of 5)
POOLING: NMME pools 5 models and its consensus beats its own best member at 6 of 10 scorings (rainfall 2 of 5)
DISAGREEMENTS: over 60 region-months where the two multi-models differ, C3S multi-system has the better reforecast in 50

## THE ANALOG BENCH
This site's own forecast, and the only one here that is not somebody else's
model. It is on a rolling three-month axis and reports an anomaly in degrees,
NOT a tercile probability, so it does not belong in the table above and is
not put there. signal_c is the anomaly with the warming trend removed.
Analogues drawn from the ocean state through July-August-September 2026.
region|season|field|anom_c|signal_c|skilful_pct
United States|OND 2026|t2m|+1.42|+0.99|31
Europe|OND 2026|t2m|+0.89|+0.28|45
Asia|OND 2026|t2m|-0.08|-0.62|52
Africa|OND 2026|t2m|+1.14|+0.59|76
S. America|OND 2026|t2m|+0.94|+0.60|81
Oceania|OND 2026|t2m|+1.05|+0.86|92
United States|NDJ 2027|t2m|+1.83|+1.32|32
Europe|NDJ 2027|t2m|+0.28|-0.46|15
Asia|NDJ 2027|t2m|+0.02|-0.49|38
Africa|NDJ 2027|t2m|+0.94|+0.38|57
S. America|NDJ 2027|t2m|+1.00|+0.66|89
Oceania|NDJ 2027|t2m|+1.26|+1.08|76
United States|DJF 2027|t2m|+2.12|+1.65|27
Europe|DJF 2027|t2m|+0.29|-0.58|4
Asia|DJF 2027|t2m|+1.21|+0.76|30
Africa|DJF 2027|t2m|+1.26|+0.70|48
S. America|DJF 2027|t2m|+1.17|+0.83|85
Oceania|DJF 2027|t2m|+1.09|+0.82|75
United States|JFM 2027|t2m|+1.94|+1.29|27
Europe|JFM 2027|t2m|+0.79|-0.18|2
Asia|JFM 2027|t2m|+1.68|+1.16|37
Africa|JFM 2027|t2m|+1.32|+0.75|49
S. America|JFM 2027|t2m|+1.34|+0.98|78
Oceania|JFM 2027|t2m|+1.28|+1.05|70
United States|FMA 2027|t2m|+1.45|+0.76|42
Europe|FMA 2027|t2m|+1.70|+0.84|6
Asia|FMA 2027|t2m|+1.75|+1.12|32
Africa|FMA 2027|t2m|+1.56|+0.99|54
S. America|FMA 2027|t2m|+1.41|+1.03|81
Oceania|FMA 2027|t2m|+1.14|+0.97|80
United States|OND 2026|tp|+4.67|+3.37|25
Europe|OND 2026|tp|+8.39|+4.79|24
Asia|OND 2026|tp|+3.53|+2.45|32
Africa|OND 2026|tp|-5.11|-2.69|46
S. America|OND 2026|tp|-10.20|-8.56|74
Oceania|OND 2026|tp|-5.92|-8.99|75
United States|NDJ 2027|tp|+6.05|+5.98|30
Europe|NDJ 2027|tp|+5.20|+1.68|22
Asia|NDJ 2027|tp|+3.58|+2.85|29
Africa|NDJ 2027|tp|-6.01|-4.11|46
S. America|NDJ 2027|tp|-9.77|-8.91|51
Oceania|NDJ 2027|tp|+2.01|-1.98|44
United States|DJF 2027|tp|+6.46|+6.38|33
Europe|DJF 2027|tp|+5.78|+2.95|17
Asia|DJF 2027|tp|+1.55|+1.26|25
Africa|DJF 2027|tp|-6.67|-5.27|40
S. America|DJF 2027|tp|-16.48|-15.36|35
Oceania|DJF 2027|tp|+0.75|-0.30|45
United States|JFM 2027|tp|+4.32|+4.64|41
Europe|JFM 2027|tp|+9.16|+6.69|28
Asia|JFM 2027|tp|+2.61|+2.39|20
Africa|JFM 2027|tp|-1.93|-0.73|43
S. America|JFM 2027|tp|-19.88|-18.48|26
Oceania|JFM 2027|tp|+7.81|+3.21|46
United States|FMA 2027|tp|+9.17|+9.10|49
Europe|FMA 2027|tp|+9.42|+7.69|18
Asia|FMA 2027|tp|+4.61|+4.15|25
Africa|FMA 2027|tp|+2.01|+2.92|42
S. America|FMA 2027|tp|-13.57|-12.58|30
Oceania|FMA 2027|tp|+8.48|+4.91|52

## CITY BRIEFINGS
40 cities, the rolling season each is closest to, and whether a forecast for
that gridcell has ever verified. useful=0 means the reforecast for this cell
did NOT beat climatology and the numbers on that row should not be acted on.
An EMPTY rainfall cell is not zero: it means the climatology there is too dry
for a tercile to carry meaning and the route declined to publish one.
city|country|season|t2m_anom_c|p_warm_pct|t2m_useful|precip_anom_pct|p_wet_pct|precip_useful
Rome|Italy|NDJ 2027|+1.0|52|1|+9|38|1
London|United Kingdom|NDJ 2027|+0.7|50|0|+4|42|0
Paris|France|NDJ 2027|+1.1|52|1|+4|36|0
Madrid|Spain|NDJ 2027|+0.7|57|1|+12|45|0
Berlin|Germany|NDJ 2027|+0.7|43|0|+0|34|0
Moscow|Russia|NDJ 2027|+0.7|42|0|+6|44|1
Athens|Greece|NDJ 2027|+0.5|52|0|-7|28|0
Reykjavik|Iceland|NDJ 2027|-0.5|22|1|-5|26|0
Bergen|Norway|NDJ 2027|+0.0|38|1|+1|30|0
Cairo|Egypt|NDJ 2027|+0.5|48|0|||0
Lagos|Nigeria|NDJ 2027|+0.7|93|1|+0|31|0
Nairobi|Kenya|NDJ 2027|+0.3|54|1|+18|48|0
Cape Town|South Africa|NDJ 2027|+0.2|42|0|+3|42|0
Kinshasa|DR Congo|NDJ 2027|+0.9|100|1|+23|56|0
Marrakesh|Morocco|NDJ 2027|+0.9|57|0|+4|37|1
Tokyo|Japan|NDJ 2027|+1.0|70|0|+32|58|0
Singapore|Singapore|NDJ 2027|+0.8|99|1|+42|58|0
Mumbai|India|NDJ 2027|+1.8|88|1|||0
Delhi|India|NDJ 2027|+0.2|46|0|||0
Shanghai|China|NDJ 2027|+0.5|48|1|+22|57|1
Bangkok|Thailand|NDJ 2027|+2.1|91|1|||0
Dubai|United Arab Emirates|NDJ 2027|+0.4|50|1|||0
Jakarta|Indonesia|NDJ 2027|+1.3|100|1|-11|33|0
Manila|Philippines|NDJ 2027|+0.8|95|1|-115|3|1
Cherrapunji|India|NDJ 2027|+0.5|59|1|||0
New York|United States|NDJ 2027|+0.5|37|0|+40|65|1
Seattle|United States|NDJ 2027|+1.1|59|1|-9|28|0
Miami|United States|NDJ 2027|-0.6|24|1|+55|80|0
Mexico City|Mexico|NDJ 2027|+0.5|61|0|-7|27|1
Vancouver|Canada|NDJ 2027|+1.4|62|1|+14|40|0
Chicago|United States|NDJ 2027|+0.0|29|1|+8|40|1
Sao Paulo|Brazil|NDJ 2027|+1.3|76|1|-2|38|1
Buenos Aires|Argentina|NDJ 2027|-0.5|16|1|+20|42|1
Lima|Peru|NDJ 2027|+2.8|100|1|+61|97|1
Bogota|Colombia|NDJ 2027|+1.8|100|1|-53|8|1
Manaus|Brazil|NDJ 2027|+1.8|100|1|-57|11|0
Sydney|Australia|NDJ 2027|+0.3|41|1|-6|30|1
Melbourne|Australia|NDJ 2027|-0.5|24|0|+0|36|1
Auckland|New Zealand|NDJ 2027|-0.4|13|1|+5|34|0
Alice Springs|Australia|NDJ 2027|+0.2|37|1|+7|42|1

## WEATHER REGIMES
How SEAS5 assembles each month across the seven Euro-Atlantic weather regimes
(Grams et al. 2017, fitted on ERA5). fc_pct is the share of the 51 members'
days in that regime; clim_pct is the model's own 1993-2016 share; tilt_pp is
their difference in points. rpss is the leave-one-out tercile skill of that
monthly frequency over the 24 reforecast years - at or below zero the tilt on
that row has no record of verifying and should not be acted on.
month|part_elapsed|regime|fc_pct|clim_pct|tilt_pp|rpss|acc
2026-10|1|AT|8|14|-6|-0.11|+0.31
2026-10|1|EuBL|22|15|+6|+0.21|+0.43
2026-10|1|ZO|32|13|+19|+0.21|+0.46
2026-10|1|ScTr|12|16|-4|+0.45|+0.79
2026-10|1|ScBL|15|19|-4|+0.26|+0.60
2026-10|1|AR|9|11|-2|+0.19|+0.41
2026-10|1|GL|2|12|-10|+0.21|+0.35
2026-11|0|AT|16|15|+1|+0.14|+0.30
2026-11|0|EuBL|7|14|-6|-0.07|-0.33
2026-11|0|ZO|15|16|-1|+0.03|-0.21
2026-11|0|ScTr|10|14|-4|+0.04|-0.11
2026-11|0|ScBL|18|14|+4|-0.07|-0.18
2026-11|0|AR|15|13|+1|+0.18|+0.33
2026-11|0|GL|19|14|+6|+0.04|-0.16
2026-12|0|AT|17|16|+1|+0.07|-0.12
2026-12|0|EuBL|8|13|-5|-0.01|+0.35
2026-12|0|ZO|18|18|-0|+0.08|+0.07
2026-12|0|ScTr|11|14|-3|+0.01|-0.30
2026-12|0|ScBL|14|13|+1|-0.17|-0.21
2026-12|0|AR|18|12|+6|-0.02|-0.24
2026-12|0|GL|14|14|-0|+0.00|-0.07
2027-01|0|AT|15|16|-1|+0.15|+0.48
2027-01|0|EuBL|11|13|-3|+0.02|+0.10
2027-01|0|ZO|9|20|-11|+0.06|+0.23
2027-01|0|ScTr|12|12|+0|+0.15|+0.10
2027-01|0|ScBL|14|11|+2|+0.01|-0.01
2027-01|0|AR|17|13|+4|-0.06|-0.00
2027-01|0|GL|24|15|+8|+0.06|-0.01
2027-02|0|AT|12|16|-4|+0.10|+0.50
2027-02|0|EuBL|10|13|-4|+0.04|-0.12
2027-02|0|ZO|7|20|-12|+0.01|-0.07
2027-02|0|ScTr|14|11|+3|+0.05|+0.03
2027-02|0|ScBL|8|11|-3|+0.04|-0.34
2027-02|0|AR|19|12|+6|+0.13|+0.06
2027-02|0|GL|31|17|+14|+0.18|+0.25
2027-03|0|AT|14|17|-3|-0.20|-0.52
2027-03|0|EuBL|7|13|-6|-0.09|-0.28
2027-03|0|ZO|12|17|-5|+0.16|+0.38
2027-03|0|ScTr|12|12|-0|+0.21|+0.63
2027-03|0|ScBL|14|12|+3|+0.10|+0.06
2027-03|0|AR|21|14|+7|+0.06|+0.13
2027-03|0|GL|20|15|+5|+0.21|+0.34
2027-04|0|AT|18|18|+1|+0.03|+0.35
2027-04|0|EuBL|15|14|+0|-0.06|-0.07
2027-04|0|ZO|14|16|-2|+0.04|+0.01
2027-04|0|ScTr|10|12|-2|+0.02|-0.07
2027-04|0|ScBL|17|12|+5|+0.03|+0.06
2027-04|0|AR|17|16|+2|-0.01|-0.08
2027-04|0|GL|10|13|-3|+0.04|-0.03

## POLAR VORTEX
Probability of a major sudden stratospheric warming - the Charlton-Polvani
(2007) criterion: zonal-mean zonal wind at 10 hPa and 60N turns easterly
between 1 Nov and 31 Mar, events 20 westerly days apart, and the wind must
return westerly for 10 days before 30 Apr or it is the vortex's annual spring
breakdown rather than a sudden warming. p_any is the share of members with at
least one. READ IT AGAINST model_clim and observed_clim, not against zero: a
warming happens in about half of all winters, so a bare probability near 50
percent carries almost no information. window is this initialisation's own
reach - it cannot see past it, and the observed rate beside it is counted
over the same window.
field|value|note
p_any|94%|51 members
p_any_confirmed|94%|excluding events too late in the run to prove recovery
model_clim|76%|this model's own rate over the reforecast winters
observed_clim|50%|ERA5, same window, 24 winters
window|2026-10-02 to 2027-05-04|forecast reach
provisional|0|member-forecasts flagged

month|p_pct
2026-12|20
2027-01|53
2027-02|29
2027-03|16

Skill over the reforecast winters. bss at or below zero means the probability
was no better than issuing the climatological rate every year; roc 0.50 is no
discrimination. acc_u_djf is the anomaly correlation of DJF-mean vortex
strength, which is a different and easier question than any individual
warming. THE _conf COLUMNS ARE THE ONES TO SCORE ON. The plain columns count
every wind reversal inside the window, and a later init_month reaches further
into spring, where a reversal can be the vortex's final warming rather than a
sudden one - which raises obs_rate and flatters a model that forecasts a
warming almost every year. The _conf columns drop those from both sides.
init_month|winters|obs_rate|model_rate|bss|roc|obs_rate_conf|model_rate_conf|bss_conf|roc_conf|acc_u_djf|bias_u_djf_ms
08|24|0.46|0.69|-0.151|0.48|0.29|0.49|-0.104|0.51|+0.20|-3.3
09|24|0.71|0.85|+0.007|0.59|0.50|0.72|-0.144|0.46|+0.23|-4.5
10|24|0.50|0.76|-0.265|0.45|0.50|0.76|-0.265|0.45|+0.11|-4.4

## WHAT NOT TO TRUST
Lead 1 is the initialisation's own month and is already part elapsed when
this run is issued. It is included above with part_elapsed=1 and is excluded
from every headline on this site.

A tercile probability near 33.3 is a model saying nothing. Most of these are
nearer 33 than not, and that is the honest state of monthly seasonal
forecasting rather than a fault in the run.

Seasonal rainfall is far weaker than seasonal temperature everywhere, and
over much of the world a rainfall tercile is not worth issuing at all. The
beats_clim_pct column is what says where.

Skill decays fast with lead. Over the United States, temperature beat
climatology over 54-61 percent of the box at November 2026 and 46-89 percent
by March 2027. Treat the far leads as context, not guidance.

The systems do not share a reforecast period - SEAS5 1993-2016; C3S
multi-system 1993-2016; NMME 1991-2016 - so their tercile boundaries come
from slightly different climates. Comparisons within one system are exact;
comparisons across them are close but not matched.

Two forecasts on this site disagree, and comparing their published headline
numbers gives the wrong answer: put on identical footing at
/assets/compare/compare.json the ranking reverses. Do not rank systems by
numbers taken from different routes without checking the footing first.

Do not rank the vortex initialisation months by roc or bss. A later
init_month runs a longer window into spring, where a wind reversal can be the
vortex's final warming rather than a sudden one and the window is too short
to apply the 30 April recovery test. That raises obs_rate and flatters a
model that forecasts a warming almost every year. Init month 09 scores above
0.55 on roc and below it on roc_conf - the gain is the window, not the
forecast. Score on the _conf columns.

## ENDPOINTS
path|route|present|holds
/assets/feed.json|08|1|This feed: headline, broadcast lines, the regional outlook, and this index
/wire.txt|08|1|The whole outlook as plain text, section-marked and pipe-delimited
/assets/seas5/latest.json|01|1|Route 01 pointer: newest initialisation and every cycle published
/assets/seas5/<cycle>/catalog.json|01|?|Route 01 catalog: 16 fields, 7 leads, 5 seasons, the 51-member layers, and the colour LUT every raster is encoded with
/assets/probs/probs.json|10|1|Route 10: SEAS5 tercile probabilities, raw and trend-adjusted, with RPSS and ACC by lead
/assets/multi/multi.json|03|1|Route 03: ten C3S systems, the consensus, and each system's own reforecast score
/assets/nmme/nmme.json|02|1|Route 02: four NMME models, the consensus, and each model's own reforecast score
/assets/analog/analog.json|09|1|Route 09: this site's own analog forecast, its analog years and its cross-validated skill
/assets/compare/compare.json|11|1|Route 11: the matched verification of routes 09 and 10 over the same years
/assets/cities/cities.json|07|1|Route 07: 40 city briefings, each with the skill of its own gridcell attached
/assets/regimes/regimes.json|12|1|Route 12: the seven Euro-Atlantic weather regimes, SEAS5's monthly frequencies against its own climatology, with ACC and RPSS
/assets/vortex/vortex_skill.json|13|1|Route 13: Brier skill, ROC area and DJF-U60 anomaly correlation for the stratospheric vortex, per initialisation month over 24 reforecast winters
/assets/dailyfc/latest.json|16|1|Route 16 pointer: newest daily MEMBER initialisation. 51 members on a daily axis - percentiles, spread and absolute-threshold chances; no reforecast behind it, so no anomalies, no terciles and no skill score
/assets/dailyfc/<cycle>/catalog.json|16|?|Route 16 catalog: 8 fields x 215 daily frames x (mean, percentiles, spread, chance layers), each with its own legend, units and colour LUT, plus the per-place plume index
/assets/sfs/sfs.json|15|1|Route 15: NOAA SFS Beta, TWELVE monthly leads from 31 members against a 1991-2025 reforecast - tercile probabilities raw and trend-adjusted, RPSS and ACC at every lead, the same on each WHOLE three-month season the start covers (DJF/MAM/JJA/SON, in skill_season and terciles_season), UNSCORED snowfall probabilities in `snow` and `snow_season` (ERA5 here has no snowfall, so they carry no RPSS and no ACC), and the Nino-3.4 plume from the model's own ocean (experiment beta1)
/assets/nino/nino.json|03|1|Route 03 ENSO: Nino 3.4, 3, 4, 1+2, relative Nino 3.4 and the IOD from the ten Copernicus systems - every member, each system's mean and ACC/RMSE vs ERA5, the one-vote consensus, and the strongest months in the ERA5 record
/assets/sfs/enso.json|15|1|Route 15: Nino 3.4, 3, 4, 1+2, relative Nino 3.4 and the IOD from the SFS ocean - beta1 and beta2, 31 members x 12 leads, each lead's ACC/RMSE vs ERA5 beside persistence and the ensemble's spread
/assets/sfs/beta2/sfs.json|15|1|Route 15: the same document for SFS experiment beta2, a different model on the same 31x12 live and 11x35 reforecast design; published only for starts whose calendar month beta2 has a reforecast for

================================================================================
End of wire. Regenerated whenever the site is rebuilt; the routes behind it refresh monthly on the 9th.
================================================================================