WeatherBlend
Multi-model forecast blending for Bonehill Rocks, Dartmoor
Rainfall skill
Eyeball first: P(wet) trajectories against a 0/1 wet-hour indicator from the same ≥ 0.1 mm threshold the blender was trained on. Dry-window predicted vs observed verdict follows. Stations sit on separate EA gauges, so flip between them via the sub-nav.
P(wet) vs observed wet-hour
P(wet) is the blender's probability that the next hour sees ≥ 0.1 mm. Truth is plotted as discrete dots at 0 (dry) or 1 (wet) using that same 0.1 mm threshold, so the only meaningful comparison at any hour is where P(wet) sits relative to the dot — a connecting line would imply non-existent intermediate states. Hours with fewer than 4 of 4 15-min readings are dropped to avoid flipping wet↔dry at the boundary.
Bellever Dartmoor
Phase 3a (lean)
Phase 3c (rich)
Three-way comparison — +24h lead
Dry window — predicted vs observed
One row per target UTC day × window length. The "observed" column is blank for dates beyond the last full rainfall day.
3-hour dry window
| Target date (UTC) | +24h | +48h | +72h | Observed |
|---|---|---|---|---|
| 1.00 | — | — | — | |
| 1.00 | 0.99 | — | — | |
| — | 1.00 | 0.99 | — | |
| — | — | 1.00 | — |
4-hour dry window
| Target date (UTC) | +24h | +48h | +72h | Observed |
|---|---|---|---|---|
| 1.00 | — | — | — | |
| 1.00 | 0.98 | — | — | |
| — | 1.00 | 0.98 | — | |
| — | — | 0.98 | — |
6-hour dry window
| Target date (UTC) | +24h | +48h | +72h | Observed |
|---|---|---|---|---|
| 0.95 | — | — | — | |
| 1.00 | 0.97 | — | — | |
| — | 1.00 | 0.97 | — | |
| — | — | 1.00 | — |