Weather Station Status



This website uses WeeWX for weather conditions reporting.
Latest weather observation: 15/09/2026 02:33.

Component Status Age
h:m:s
Latest update time as of
15/09/2026 02:33
Live WeeWX observationsCurrent0:00:2915/09/2026 02:33
Met Office forecastCurrent0:28:4715/09/2026 02:05
Trearddur Bay tide dataCurrent1:33:0115/09/2026 01:00
Vestaboard weather updateNOT Current4:03:46> 2:00:00
14/09/2026 22:30
Automated ntfy monitoringCurrent0:03:4715/09/2026 02:30
Web configuration backupCurrent0:18:4215/09/2026 02:15
WeeWX configuration backupCurrent0:03:4615/09/2026 02:30
Local wind forecast modelCurrent0:00:3015/09/2026 02:33
External forecast inputsCurrent0:03:0715/09/2026 02:30

Experimental Ravenspoint model

Four-hour wind forecast accuracy

Confidence Established

These are live results from forecasts published on the landing page, checked against subsequent Ravenspoint observations. The system now favours whichever input has recently performed best. Lower error figures are better.

ForecastAssessed7-day wind error30-day wind errorTrendGust errorDirection errorMet Office wind errorSailing decisionsRange coverage
+1 hour 367 1.6 kt 1.6 kt Collecting 2.9 kt 13° 1.7 kt 89% exact; safety score 94% (17 unsafe suitable calls) 50%: 51% of 367
80%: 83% of 367
+2 hours 367 1.8 kt 1.8 kt Collecting 3.4 kt 21° 1.7 kt 84% exact; safety score 92% (19 unsafe suitable calls) 50%: 52% of 367
80%: 84% of 367
+3 hours 367 1.9 kt 1.9 kt Collecting 3.5 kt 21° 1.7 kt 80% exact; safety score 92% (16 unsafe suitable calls) 50%: 55% of 367
80%: 82% of 367
+4 hours 367 1.9 kt 1.9 kt Collecting 3.5 kt 22° 1.7 kt 80% exact; safety score 94% (7 unsafe suitable calls) 50%: 55% of 367
80%: 85% of 367

2200 forecast outcomes assessed in total. Unsafe “suitable” calls carry five times the penalty of an ordinary classification error. Forecast ranges target 80% coverage.

Model safeguard: candidate 20260914T144751Z-cc471ebc was held back. candidate must improve composite error by at least 2%: 4.45 versus required 4.43

Forecast data quality

Healthy Latest observation age: 7 minutes · Excluded implausible rows: 11 · Units: metric-normalised-v2.

Weather regime: Showery / unstable.

Safety guard active Local model remains within the live safety limit. Comparisons: 2200; local unsafe suitable calls: 196; persistence: 196.

Upstream model-feature archive: 36.1 days; RAF Valley 1729 records; Met Office land 719 records. Collecting safely

Met Office revision archive: 2375 snapshots, 466 changed forecast versions and 121520 timestamped forecast rows retained.

Current upstream changes: RAF Valley steady -1.0 kt; Met Office land stations steady +1.4 kt. Upstream changes remain shadow inputs until the 60-day readiness gate is met.

Current forecast decision audit

HorizonSelected sourceWeightsExcludedReasonChange outlook
+1hlocallocal 85%, persistence 15%corrected, historicalDifferential, metOfficelocal had the lowest recency-weighted error; sources more than 8% worse were excluded.Steady; direction shift 0%
+2hcorrectedcorrected 85%, metOffice 15%local, historicalDifferential, persistencecorrected had the lowest recency-weighted error; sources more than 8% worse were excluded.Steady; direction shift 100%
+3hmetOfficemetOffice 85%, corrected 15%local, persistencemetOffice had the lowest recency-weighted error; sources more than 8% worse were excluded.Falling; direction shift 100%
+4hmetOfficemetOffice 85%, corrected 15%local, persistencemetOffice had the lowest recency-weighted error; sources more than 8% worse were excluded.Rapidly decreasing; direction shift 100%
+5hmetOfficemetOffice 100%local, corrected, persistencemetOffice had the lowest recency-weighted error; sources more than 8% worse were excluded.Rapidly decreasing; direction shift 100%
+6hmetOfficemetOffice 100%local, corrected, persistencemetOffice had the lowest recency-weighted error; sources more than 8% worse were excluded.Falling; direction shift 100%

Direct gust and change models

+1h: gust classifiers 18 kt or more, 20 kt or more, 25 kt or more, 30 kt or more passed; wind-change classifier held back; direction-change classifier held back.
+2h: gust classifiers 18 kt or more, 20 kt or more, 25 kt or more, 30 kt or more passed; wind-change classifier held back; direction-change classifier held back.
+3h: gust classifiers 18 kt or more, 20 kt or more, 25 kt or more, 30 kt or more passed; wind-change classifier held back; direction-change classifier held back.
+4h: gust classifiers 18 kt or more, 20 kt or more, 25 kt or more, 30 kt or more passed; wind-change classifier held back; direction-change classifier held back.
+5h: gust classifiers 18 kt or more, 20 kt or more, 25 kt or more, 30 kt or more passed; wind-change classifier held back; direction-change classifier held back.
+6h: gust classifiers 18 kt or more, 20 kt or more, 25 kt or more, 30 kt or more passed; wind-change classifier held back; direction-change classifier held back.

Accuracy by observed wind direction (30 days)

DirectionAssessedWind errorGust errorDirection error80% error allowanceRange coverage
N302.5 kt4.9 kt77°±2.9 kt50%: 40% of 30
80%: 100% of 30
S13851.6 kt3.5 kt17°±2.5 kt50%: 56% of 1385
80%: 87% of 1385
SW4032.0 kt3.2 kt28°±3.0 kt50%: 48% of 403
80%: 87% of 403
W3222.0 kt2.4 kt18°±3.0 kt50%: 60% of 322
80%: 84% of 322
NW605.0 kt4.7 kt31°±6.9 kt50%: 10% of 60
80%: 32% of 60

Direction-calibration safeguards

+1h SW sector: Held back — held back: held-out error 2.38 versus 1.75 without correction.
+2h SW sector: Applied — held-out error improved from 2.26 to 2.13.
+3h SW sector: Applied — held-out error improved from 2.61 to 2.16.
+4h SW sector: Applied — held-out error improved from 2.54 to 2.18.
+5h SW sector: Applied — held-out error improved from 2.43 to 1.99.
+6h SW sector: Applied — held-out error improved from 2.74 to 2.14.

Nearby-station lead analysis

Positive lead means the nearby station changes before Ravenspoint.

StationBest leadCorrelationSamplesRepeatable
RAF Valley0 min0.871388Yes
Met Office Isle of Anglesey30 min0.87694Yes
Met Office Gwynedd0 min0.77664Yes
Met Office Conwy0 min0.68692Yes

Recent model candidates

20260914T144751Z-cc471ebcHeld back. candidate must improve composite error by at least 2%: 4.45 versus required 4.43
20260913T144751Z-558c2147Held back. candidate must improve composite error by at least 2%: 4.45 versus required 4.43
20260912T144711Z-45cd143dHeld back. candidate must improve composite error by at least 2%: 4.46 versus required 4.43
20260911T144711Z-5f6cad72Held back. candidate must improve composite error by at least 2%: 4.47 versus required 4.43
20260910T144711Z-c932f610Held back. candidate must improve composite error by at least 2%: 4.47 versus required 4.43
20260909T144711Z-ea3572a0Held back. candidate must improve composite error by at least 2%: 4.48 versus required 4.43
20260908T143238Z-6fff5a84Held back. candidate must improve composite error by at least 2%: 4.48 versus required 4.43
20260907T143211Z-a0c367f7Held back. candidate must improve composite error by at least 2%: 4.48 versus required 4.43
20260906T141732Z-6596af76Held back. candidate must improve composite error by at least 2%: 4.48 versus required 4.43
20260905T141711Z-e55fdeaaHeld back. candidate must improve composite error by at least 2%: 4.49 versus required 4.43

Met Office warning feed

Current Source: NSWWS API · Active warnings: 0 · Last checked: 3 minutes ago.

Three-hour trend indicator accuracy

This is a separate rolling seven-day check of the “increasing, steady or falling” message shown with live wind. It is not the six-hour forecasting model above.

Forecasts assessed110
Fully correct51 (46%)
Trend direction correct54%
Average wind-strength error4.0 kt
18-knot threshold correct87%

Weekly accuracy history

Weekly wind-trend forecast accuracy Blue columns show fully correct predictions as a percentage. The red line shows average wind-strength error in knots. 0% 25% 50% 75% 100% 2026-W32: 0% correct from 2 forecasts; 8.8 kt average error 32 2026-W33: 50% correct from 132 forecasts; 4.0 kt average error 33 2026-W34: 38% correct from 168 forecasts; 4.6 kt average error 34 2026-W35: 39% correct from 164 forecasts; 4.1 kt average error 35 2026-W36: 37% correct from 168 forecasts; 4.3 kt average error 36 2026-W37: 44% correct from 99 forecasts; 4.2 kt average error 37 2026-W38: 52% correct from 27 forecasts; 2.9 kt average error 38 8.8 kt 0 kt Fully correct Average error (kt)