WeatherBlend

Multi-model forecast blending for Sennen, Cornwall

Dry-window models

Per-(station, window) P(N-hour dry block in 09–18 local). 3b LightGBM + 3p copula MC. Brier, lower better. Δ vs best single NWP — negative = blend wins.

Dry window — St Erth — 2-hour

Phase 3q · v2026-06-14_153854_phase3q

Phase 3q blender. Trained 2026-06-14. Metric: Test Brier.

Lead Blend Best single Δ vs best
+24h 0.000 — (0.000 val) —
+48h 0.000 — (0.000 val) —
+72h 0.000 — (0.000 val) —
Verify history (no runs yet)

No verify rows yet for phase 3q. Next cycle: Mon/Thu 09:30 UTC, then 5d ERA5 latency.

Dry window — St Erth — 3-hour

Phase 3b · v2026-06-14_153806

53-feature LightGBM per-(station, window). Trained 2026-06-14. Metric: Test Brier.

Lead Blend Best single Δ vs best
+24h 0.060 has_dry_window_mf (0.143) -58.0%
+48h 0.081 has_dry_window_icon (0.163) -50.6%
+72h 0.081 has_dry_window_icon (0.185) -56.0%
Verify history (no runs yet)

No verify rows yet for phase 3b. Next cycle: Mon/Thu 09:30 UTC, then 5d ERA5 latency.

Phase 3q · v2026-06-14_153854_phase3q

Phase 3q blender. Trained 2026-06-14. Metric: Test Brier.

Lead Blend Best single Δ vs best
+24h 0.000 — (0.000 val) —
+48h 0.000 — (0.000 val) —
+72h 0.000 — (0.000 val) —
Verify history (no runs yet)

No verify rows yet for phase 3q. Next cycle: Mon/Thu 09:30 UTC, then 5d ERA5 latency.

Dry window — St Erth — 4-hour

Phase 3b · v2026-06-14_153820

53-feature LightGBM per-(station, window). Trained 2026-06-14. Metric: Test Brier.

Lead Blend Best single Δ vs best
+24h 0.065 has_dry_window_icon (0.148) -55.9%
+48h 0.070 has_dry_window_icon (0.170) -58.9%
+72h 0.081 has_dry_window_jma (0.170) -52.6%
Verify history (no runs yet)

No verify rows yet for phase 3b. Next cycle: Mon/Thu 09:30 UTC, then 5d ERA5 latency.

Phase 3q · v2026-06-14_153854_phase3q

Phase 3q blender. Trained 2026-06-14. Metric: Test Brier.

Lead Blend Best single Δ vs best
+24h 0.000 — (0.000 val) —
+48h 0.000 — (0.000 val) —
+72h 0.000 — (0.000 val) —
Verify history (no runs yet)

No verify rows yet for phase 3q. Next cycle: Mon/Thu 09:30 UTC, then 5d ERA5 latency.

Dry window — St Erth — 5-hour

Phase 3q · v2026-06-14_153854_phase3q

Phase 3q blender. Trained 2026-06-14. Metric: Test Brier.

Lead Blend Best single Δ vs best
+24h 0.000 — (0.000 val) —
+48h 0.000 — (0.000 val) —
+72h 0.000 — (0.000 val) —
Verify history (no runs yet)

No verify rows yet for phase 3q. Next cycle: Mon/Thu 09:30 UTC, then 5d ERA5 latency.

Dry window — St Erth — 6-hour

Phase 3b · v2026-06-14_153833

53-feature LightGBM per-(station, window). Trained 2026-06-14. Metric: Test Brier.

Lead Blend Best single Δ vs best
+24h 0.133 has_dry_window_gfs (0.215) -38.3%
+48h 0.129 has_dry_window_mf (0.286) -54.7%
+72h 0.141 has_dry_window_gfs (0.319) -55.7%
Verify history (no runs yet)

No verify rows yet for phase 3b. Next cycle: Mon/Thu 09:30 UTC, then 5d ERA5 latency.

Phase 3q · v2026-06-14_153854_phase3q

Phase 3q blender. Trained 2026-06-14. Metric: Test Brier.

Lead Blend Best single Δ vs best
+24h 0.000 — (0.000 val) —
+48h 0.000 — (0.000 val) —
+72h 0.000 — (0.000 val) —
Verify history (no runs yet)

No verify rows yet for phase 3q. Next cycle: Mon/Thu 09:30 UTC, then 5d ERA5 latency.

Dry window — St Ives Towednack — 2-hour

Phase 3q · v2026-06-14_153852_phase3q

Phase 3q blender. Trained 2026-06-14. Metric: Test Brier.

Lead Blend Best single Δ vs best
+24h 0.000 — (0.000 val) —
+48h 0.000 — (0.000 val) —
+72h 0.000 — (0.000 val) —
Verify history (no runs yet)

No verify rows yet for phase 3q. Next cycle: Mon/Thu 09:30 UTC, then 5d ERA5 latency.

Dry window — St Ives Towednack — 3-hour

Phase 3b · v2026-06-14_153724

53-feature LightGBM per-(station, window). Trained 2026-06-14. Metric: Test Brier.

Lead Blend Best single Δ vs best
+24h 0.064 has_dry_window_mf (0.154) -58.6%
+48h 0.070 has_dry_window_icon (0.167) -57.9%
+72h 0.073 has_dry_window_icon (0.144) -49.2%
Verify history (no runs yet)

No verify rows yet for phase 3b. Next cycle: Mon/Thu 09:30 UTC, then 5d ERA5 latency.

Phase 3q · v2026-06-14_153852_phase3q

Phase 3q blender. Trained 2026-06-14. Metric: Test Brier.

Lead Blend Best single Δ vs best
+24h 0.000 — (0.000 val) —
+48h 0.000 — (0.000 val) —
+72h 0.000 — (0.000 val) —
Verify history (no runs yet)

No verify rows yet for phase 3q. Next cycle: Mon/Thu 09:30 UTC, then 5d ERA5 latency.

Dry window — St Ives Towednack — 4-hour

Phase 3b · v2026-06-14_153738

53-feature LightGBM per-(station, window). Trained 2026-06-14. Metric: Test Brier.

Lead Blend Best single Δ vs best
+24h 0.084 has_dry_window_mf (0.154) -45.2%
+48h 0.085 has_dry_window_icon (0.197) -56.9%
+72h 0.101 has_dry_window_icon (0.152) -33.3%
Verify history (no runs yet)

No verify rows yet for phase 3b. Next cycle: Mon/Thu 09:30 UTC, then 5d ERA5 latency.

Phase 3q · v2026-06-14_153852_phase3q

Phase 3q blender. Trained 2026-06-14. Metric: Test Brier.

Lead Blend Best single Δ vs best
+24h 0.000 — (0.000 val) —
+48h 0.000 — (0.000 val) —
+72h 0.000 — (0.000 val) —
Verify history (no runs yet)

No verify rows yet for phase 3q. Next cycle: Mon/Thu 09:30 UTC, then 5d ERA5 latency.

Dry window — St Ives Towednack — 5-hour

Phase 3q · v2026-06-14_153852_phase3q

Phase 3q blender. Trained 2026-06-14. Metric: Test Brier.

Lead Blend Best single Δ vs best
+24h 0.000 — (0.000 val) —
+48h 0.000 — (0.000 val) —
+72h 0.000 — (0.000 val) —
Verify history (no runs yet)

No verify rows yet for phase 3q. Next cycle: Mon/Thu 09:30 UTC, then 5d ERA5 latency.

Dry window — St Ives Towednack — 6-hour

Phase 3b · v2026-06-14_153752

53-feature LightGBM per-(station, window). Trained 2026-06-14. Metric: Test Brier.

Lead Blend Best single Δ vs best
+24h 0.110 has_dry_window_ecmwf (0.159) -31.0%
+48h 0.137 has_dry_window_icon (0.182) -24.4%
+72h 0.145 has_dry_window_gfs (0.280) -48.3%
Verify history (no runs yet)

No verify rows yet for phase 3b. Next cycle: Mon/Thu 09:30 UTC, then 5d ERA5 latency.

Phase 3q · v2026-06-14_153852_phase3q

Phase 3q blender. Trained 2026-06-14. Metric: Test Brier.

Lead Blend Best single Δ vs best
+24h 0.000 — (0.000 val) —
+48h 0.000 — (0.000 val) —
+72h 0.000 — (0.000 val) —
Verify history (no runs yet)

No verify rows yet for phase 3q. Next cycle: Mon/Thu 09:30 UTC, then 5d ERA5 latency.

Dry window — Trengwainton — 2-hour

Phase 3q · v2026-06-14_153851_phase3q

Phase 3q blender. Trained 2026-06-14. Metric: Test Brier.

Lead Blend Best single Δ vs best
+24h 0.000 — (0.000 val) —
+48h 0.000 — (0.000 val) —
+72h 0.000 — (0.000 val) —
Verify history (no runs yet)

No verify rows yet for phase 3q. Next cycle: Mon/Thu 09:30 UTC, then 5d ERA5 latency.

Dry window — Trengwainton — 3-hour

Phase 3b · v2026-06-14_153642

53-feature LightGBM per-(station, window). Trained 2026-06-14. Metric: Test Brier.

Lead Blend Best single Δ vs best
+24h 0.073 has_dry_window_mf (0.143) -48.9%
+48h 0.079 has_dry_window_icon (0.158) -50.1%
+72h 0.084 has_dry_window_icon (0.165) -49.0%
Verify history (no runs yet)

No verify rows yet for phase 3b. Next cycle: Mon/Thu 09:30 UTC, then 5d ERA5 latency.

Phase 3q · v2026-06-14_153851_phase3q

Phase 3q blender. Trained 2026-06-14. Metric: Test Brier.

Lead Blend Best single Δ vs best
+24h 0.000 — (0.000 val) —
+48h 0.000 — (0.000 val) —
+72h 0.000 — (0.000 val) —
Verify history (no runs yet)

No verify rows yet for phase 3q. Next cycle: Mon/Thu 09:30 UTC, then 5d ERA5 latency.

Dry window — Trengwainton — 4-hour

Phase 3b · v2026-06-14_153657

53-feature LightGBM per-(station, window). Trained 2026-06-14. Metric: Test Brier.

Lead Blend Best single Δ vs best
+24h 0.101 has_dry_window_icon (0.143) -29.4%
+48h 0.099 has_dry_window_icon (0.180) -45.1%
+72h 0.106 has_dry_window_icon (0.180) -41.1%
Verify history (no runs yet)

No verify rows yet for phase 3b. Next cycle: Mon/Thu 09:30 UTC, then 5d ERA5 latency.

Phase 3q · v2026-06-14_153851_phase3q

Phase 3q blender. Trained 2026-06-14. Metric: Test Brier.

Lead Blend Best single Δ vs best
+24h 0.000 — (0.000 val) —
+48h 0.000 — (0.000 val) —
+72h 0.000 — (0.000 val) —
Verify history (no runs yet)

No verify rows yet for phase 3q. Next cycle: Mon/Thu 09:30 UTC, then 5d ERA5 latency.

Dry window — Trengwainton — 5-hour

Phase 3q · v2026-06-14_153851_phase3q

Phase 3q blender. Trained 2026-06-14. Metric: Test Brier.

Lead Blend Best single Δ vs best
+24h 0.000 — (0.000 val) —
+48h 0.000 — (0.000 val) —
+72h 0.000 — (0.000 val) —
Verify history (no runs yet)

No verify rows yet for phase 3q. Next cycle: Mon/Thu 09:30 UTC, then 5d ERA5 latency.

Dry window — Trengwainton — 6-hour

Phase 3b · v2026-06-14_153711

53-feature LightGBM per-(station, window). Trained 2026-06-14. Metric: Test Brier.

Lead Blend Best single Δ vs best
+24h 0.135 has_dry_window_icon (0.218) -38.2%
+48h 0.125 has_dry_window_icon (0.271) -53.8%
+72h 0.140 has_dry_window_icon (0.241) -41.8%
Verify history (no runs yet)

No verify rows yet for phase 3b. Next cycle: Mon/Thu 09:30 UTC, then 5d ERA5 latency.

Phase 3q · v2026-06-14_153851_phase3q

Phase 3q blender. Trained 2026-06-14. Metric: Test Brier.

Lead Blend Best single Δ vs best
+24h 0.000 — (0.000 val) —
+48h 0.000 — (0.000 val) —
+72h 0.000 — (0.000 val) —
Verify history (no runs yet)

No verify rows yet for phase 3q. Next cycle: Mon/Thu 09:30 UTC, then 5d ERA5 latency.