WeatherBlend

Multi-model forecast blending for Bonehill Rocks, Dartmoor

Dry-window forecast

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

Bovey Tracey

2-hour dry window

Phase 3p — Gaussian copula MC over Phase 3o hourly P(wet)

Gaussian copula MC over the rich+orographic Phase 3o's hourly P(wet) marginals. Single empirical Σ per station, fit on train-split observed daytime wet/dry binary sequences (daytime-shape autocorrelation is lead-independent for truth, so per-lead Σs pool into one). Captures within-day wet/dry autocorrelation that an iid sampler misses. Bake-off Brier 0.1064 aggregate on the 9 (station, lead) Bonehill cells.

Target date (UTC) +24h +48h +72h Confidence Best start (UTC, P dry block)
100%99%100% confident dry (P=100% · run 5–9h) 13:00Z (99%)
100%100%100% confident dry (P=100% · run 9–9h) 14:00Z (99%)
—97%98% confident dry (P=97% · run 3–9h) 11:00Z (80%)
——100% confident dry (P=100% · run 9–9h) 13:00Z (99%)
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% · τ=26%)
99%98%99% 100% confident dry (P=99% · τ=26%)
—85%93% 50% confident dry (P=85% · τ=26%)
——99% 100% confident dry (P=99% · τ=24%)
Phase 3p — Gaussian copula MC over Phase 3o hourly P(wet)

Gaussian copula MC over the rich+orographic Phase 3o's hourly P(wet) marginals. Single empirical Σ per station, fit on train-split observed daytime wet/dry binary sequences (daytime-shape autocorrelation is lead-independent for truth, so per-lead Σs pool into one). Captures within-day wet/dry autocorrelation that an iid sampler misses. Bake-off Brier 0.1064 aggregate on the 9 (station, lead) Bonehill cells.

Target date (UTC) +24h +48h +72h Confidence Best start (UTC, P dry block)
100%98%100% confident dry (P=100% · run 5–9h) 14:00Z (98%)
100%100%100% confident dry (P=100% · run 9–9h) 13:00Z (99%)
—91%91% confident dry (P=91% · run 3–9h) 14:00Z (73%)
——100% confident dry (P=100% · run 9–9h) 12:00Z (98%)
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
93%94%85% 100% confident dry (P=93% · τ=25%)
99%98%97% 100% confident dry (P=99% · τ=25%)
—82%84% 33% confident dry (P=82% · τ=29%)
——97% 100% confident dry (P=97% · τ=40%)
Phase 3p — Gaussian copula MC over Phase 3o hourly P(wet)

Gaussian copula MC over the rich+orographic Phase 3o's hourly P(wet) marginals. Single empirical Σ per station, fit on train-split observed daytime wet/dry binary sequences (daytime-shape autocorrelation is lead-independent for truth, so per-lead Σs pool into one). Captures within-day wet/dry autocorrelation that an iid sampler misses. Bake-off Brier 0.1064 aggregate on the 9 (station, lead) Bonehill cells.

Target date (UTC) +24h +48h +72h Confidence Best start (UTC, P dry block)
99%96%99% confident dry (P=99% · run 5–9h) 13:00Z (98%)
100%100%100% confident dry (P=100% · run 9–9h) 12:00Z (98%)
—83%83% confident dry (P=83% · run 3–9h) 09:00Z (65%)
——100% confident dry (P=100% · run 9–9h) 11:00Z (98%)
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 3p — Gaussian copula MC over Phase 3o hourly P(wet)

Gaussian copula MC over the rich+orographic Phase 3o's hourly P(wet) marginals. Single empirical Σ per station, fit on train-split observed daytime wet/dry binary sequences (daytime-shape autocorrelation is lead-independent for truth, so per-lead Σs pool into one). Captures within-day wet/dry autocorrelation that an iid sampler misses. Bake-off Brier 0.1064 aggregate on the 9 (station, lead) Bonehill cells.

Target date (UTC) +24h +48h +72h Confidence Best start (UTC, P dry block)
92%88%96% confident dry (P=92% · run 5–9h) 12:00Z (92%)
99%99%99% confident dry (P=99% · run 9–9h) 11:00Z (98%)
—75%76% ambiguous (P=75% · run 3–9h) 08:00Z (61%)
——99% confident dry (P=99% · run 9–9h) 10:00Z (97%)
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%87%97% 50% confident dry (P=90% · τ=31%)
98%99%99% 83% confident dry (P=98% · τ=31%)
—17%63% 17% confident wet (P=17% · τ=23%)
——99% 83% confident dry (P=99% · τ=27%)
Phase 3p — Gaussian copula MC over Phase 3o hourly P(wet)

Gaussian copula MC over the rich+orographic Phase 3o's hourly P(wet) marginals. Single empirical Σ per station, fit on train-split observed daytime wet/dry binary sequences (daytime-shape autocorrelation is lead-independent for truth, so per-lead Σs pool into one). Captures within-day wet/dry autocorrelation that an iid sampler misses. Bake-off Brier 0.1064 aggregate on the 9 (station, lead) Bonehill cells.

Target date (UTC) +24h +48h +72h Confidence Best start (UTC, P dry block)
82%81%93% confident dry (P=82% · run 5–9h) 11:00Z (84%)
98%98%99% confident dry (P=98% · run 9–9h) 10:00Z (97%)
—65%68% ambiguous (P=65% · run 3–9h) 11:00Z (58%)
——98% confident dry (P=98% · run 9–9h) 09:00Z (97%)
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.