WeatherBlend

Multi-model forecast blending for Sennen, Cornwall

Dry-window forecast

P(at least one N-hour dry block in 09:00–18:00 local). 3b champion plus Monte-Carlo challengers.

Lands End

2-hour dry window

Phase 3q — Gaussian copula MC over Phase 3c hourly P(wet)

Gaussian copula MC over Phase 3c's hourly P(wet) marginals — the 3p engine bound to the rich (non-orographic) 3c instead of 3o, for coastal locations without a 3o model. Single empirical Σ per station from train-split observed daytime wet/dry sequences. Captures within-day wet/dry autocorrelation an iid sampler misses; scored against the 3b champion at Sennen.

Target date (UTC) +24h +48h +72h Confidence Best start (UTC, P dry block)
100%100%100% confident dry (P=100% · run 5–9h) 13:00Z (99%)
100%100%100% confident dry (P=100% · run 8–9h) 12:00Z (99%)
—97%96% confident dry (P=97% · run 2–9h) 15:00Z (84%)
——100% confident dry (P=100% · run 5–9h) 09:00Z (96%)
Monte-Carlo P(an 2-hour dry block runs from each start hour) — one line per forecast horizon, soonest target day shown. The peak is the best time to set off; the height is how likely it stays dry. Each hour is its own probability, so the lines need not sum to the daily "any dry window" figure.

3-hour dry window

Phase 3b — lean (53 features)

53-feature LightGBM. Champion.

Target date (UTC) +24h +48h +72h Model agreement Confidence
99%98%99% 100% confident dry (P=99%)
99%98%98% 100% confident dry (P=99%)
—93%94% 50% confident dry (P=93%)
——98% 100% confident dry (P=98%)
Phase 3q — Gaussian copula MC over Phase 3c hourly P(wet)

Gaussian copula MC over Phase 3c's hourly P(wet) marginals — the 3p engine bound to the rich (non-orographic) 3c instead of 3o, for coastal locations without a 3o model. Single empirical Σ per station from train-split observed daytime wet/dry sequences. Captures within-day wet/dry autocorrelation an iid sampler misses; scored against the 3b champion at Sennen.

Target date (UTC) +24h +48h +72h Confidence Best start (UTC, P dry block)
100%99%99% confident dry (P=100% · run 5–9h) 13:00Z (98%)
100%99%100% confident dry (P=100% · run 8–9h) 11:00Z (98%)
—89%89% confident dry (P=89% · run 2–9h) 14:00Z (75%)
——100% confident dry (P=100% · run 5–9h) 08:00Z (94%)
Monte-Carlo P(an 3-hour dry block runs from each start hour) — one line per forecast horizon, soonest target day shown. The peak is the best time to set off; the height is how likely it stays dry. Each hour is its own probability, so the lines need not sum to the daily "any dry window" figure.

4-hour dry window

Phase 3b — lean (53 features)

53-feature LightGBM. Champion.

Target date (UTC) +24h +48h +72h Model agreement Confidence
98%98%99% 100% confident dry (P=98%)
98%95%99% 100% confident dry (P=98%)
—90%90% 33% confident dry (P=90%)
——99% 83% confident dry (P=99%)
Phase 3q — Gaussian copula MC over Phase 3c hourly P(wet)

Gaussian copula MC over Phase 3c's hourly P(wet) marginals — the 3p engine bound to the rich (non-orographic) 3c instead of 3o, for coastal locations without a 3o model. Single empirical Σ per station from train-split observed daytime wet/dry sequences. Captures within-day wet/dry autocorrelation an iid sampler misses; scored against the 3b champion at Sennen.

Target date (UTC) +24h +48h +72h Confidence Best start (UTC, P dry block)
99%98%98% confident dry (P=99% · run 5–9h) 13:00Z (97%)
99%98%99% confident dry (P=99% · run 8–9h) 09:00Z (97%)
—80%78% confident dry (P=80% · run 2–9h) 13:00Z (69%)
——98% confident dry (P=98% · run 5–9h) 08:00Z (91%)
Monte-Carlo P(an 4-hour dry block runs from each start hour) — one line per forecast horizon, soonest target day shown. The peak is the best time to set off; the height is how likely it stays dry. Each hour is its own probability, so the lines need not sum to the daily "any dry window" figure.

5-hour dry window

Phase 3q — Gaussian copula MC over Phase 3c hourly P(wet)

Gaussian copula MC over Phase 3c's hourly P(wet) marginals — the 3p engine bound to the rich (non-orographic) 3c instead of 3o, for coastal locations without a 3o model. Single empirical Σ per station from train-split observed daytime wet/dry sequences. Captures within-day wet/dry autocorrelation an iid sampler misses; scored against the 3b champion at Sennen.

Target date (UTC) +24h +48h +72h Confidence Best start (UTC, P dry block)
91%86%94% confident dry (P=91% · run 5–9h) 12:00Z (90%)
99%96%98% confident dry (P=99% · run 8–9h) 09:00Z (97%)
—60%66% ambiguous (P=60% · run 2–9h) 12:00Z (54%)
——94% confident dry (P=94% · run 5–9h) 08:00Z (89%)
Monte-Carlo P(an 5-hour dry block runs from each start hour) — one line per forecast horizon, soonest target day shown. The peak is the best time to set off; the height is how likely it stays dry. Each hour is its own probability, so the lines need not sum to the daily "any dry window" figure.

6-hour dry window

Phase 3b — lean (53 features)

53-feature LightGBM. Champion.

Target date (UTC) +24h +48h +72h Model agreement Confidence
90%94%94% 83% confident dry (P=90%)
93%79%90% 100% confident dry (P=93%)
—58%54% 0% ambiguous (P=58%)
——84% 67% confident dry (P=84%)
Phase 3q — Gaussian copula MC over Phase 3c hourly P(wet)

Gaussian copula MC over Phase 3c's hourly P(wet) marginals — the 3p engine bound to the rich (non-orographic) 3c instead of 3o, for coastal locations without a 3o model. Single empirical Σ per station from train-split observed daytime wet/dry sequences. Captures within-day wet/dry autocorrelation an iid sampler misses; scored against the 3b champion at Sennen.

Target date (UTC) +24h +48h +72h Confidence Best start (UTC, P dry block)
85%80%90% confident dry (P=85% · run 5–9h) 11:00Z (85%)
98%88%96% confident dry (P=98% · run 8–9h) 08:00Z (96%)
—50%53% ambiguous (P=50% · run 2–9h) 11:00Z (46%)
——88% confident dry (P=88% · run 5–9h) 08:00Z (85%)
Monte-Carlo P(an 6-hour dry block runs from each start hour) — one line per forecast horizon, soonest target day shown. The peak is the best time to set off; the height is how likely it stays dry. Each hour is its own probability, so the lines need not sum to the daily "any dry window" figure.

Dry hour = all four 15-min readings ≤ 0.1 mm. Search bounded to 09:00–18:00 local.