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.

Bellever Dartmoor

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)
90%89%93% confident dry (P=90% · run 1–9h) 15:00Z (91%)
100%100%99% confident dry (P=100% · run 5–9h) 09:00Z (99%)
—100%100% confident dry (P=100% · run 9–9h) 15:00Z (99%)
——100% confident dry (P=100% · run 9–9h) 15:00Z (100%)
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
72%39%50% 50% confident dry (P=72% · τ=44%)
99%100%96% 83% confident dry (P=99% · τ=44%)
—100%100% 100% confident dry (P=100% · τ=36%)
——100% 100% confident dry (P=100% · τ=45%)
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)
80%80%86% confident dry (P=80% · run 1–9h) 14:00Z (86%)
99%99%97% confident dry (P=99% · run 5–9h) 08:00Z (98%)
—100%100% confident dry (P=100% · run 9–9h) 14:00Z (98%)
——100% confident dry (P=100% · run 9–9h) 12:00Z (99%)
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
27%24%22% 50% confident wet (P=27% · τ=29%)
100%93%96% 83% confident dry (P=100% · τ=29%)
—99%100% 100% confident dry (P=99% · τ=42%)
——100% 100% confident dry (P=100% · τ=43%)
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)
69%69%77% ambiguous (P=69% · run 1–9h) 13:00Z (81%)
98%97%94% confident dry (P=98% · run 5–9h) 08:00Z (97%)
—100%100% confident dry (P=100% · run 9–9h) 13:00Z (98%)
——100% confident dry (P=100% · run 9–9h) 11:00Z (99%)
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)
61%58%67% ambiguous (P=61% · run 1–9h) 12:00Z (73%)
94%93%88% confident dry (P=94% · run 5–9h) 08:00Z (96%)
—99%99% confident dry (P=99% · run 9–9h) 12:00Z (97%)
——99% confident dry (P=99% · run 9–9h) 10:00Z (99%)
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
36%7%8% 33% confident wet (P=36% · τ=51%)
95%88%71% 67% confident dry (P=95% · τ=51%)
—100%100% 100% confident dry (P=100% · τ=52%)
——79% 83% confident dry (P=79% · τ=56%)
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)
49%49%56% ambiguous (P=49% · run 1–9h) 11:00Z (64%)
88%88%83% confident dry (P=88% · run 5–9h) 08:00Z (94%)
—98%98% confident dry (P=98% · run 9–9h) 11:00Z (97%)
——98% confident dry (P=98% · run 9–9h) 09:00Z (99%)
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.