WeatherBlend

Multi-model forecast blending for Bonehill Rocks, Dartmoor

Wind models

10 m wind speed — ERA5-truth `wind` + Dunkeswell-truth `wind_speed_lgb`; plus gust and direction (`wind_mvn`). MAE — m/s for speed/gust, ° for direction; lower better. Δ vs best single NWP — negative = blend wins.

10 m wind speed — Bonehill Rocks

Phase wind · v2026-06-14_150647

10 m wind-speed LightGBM blender, ERA5 WindSpeed10m truth. Per-NWP speed + direction (sin/cos) + ensemble spread + calendar. Trained 2026-06-14. Metric: Test MAE (m/s).

Lead Blend Best single Δ vs best
+24h 0.473 ecmwf_aifs025_single (0.569) -16.8%
+48h 0.573 ecmwf_aifs025_single (0.698) -17.9%
+72h 0.753 ecmwf_aifs025_single (0.887) -15.1%
Verify history (no runs yet)

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

Phase wind_speed_lgb · v2026-06-14_125959_wind_speed_lgb

10 m wind-speed LightGBM blender trained on Dunkeswell SYNOP obs (real-station truth) instead of ERA5. 29-feature lean+ORO+spread set. The honest real-wind predictor — see the cross-truth bake-off. Trained 2026-06-14. Metric: Test MAE (m/s).

Lead Blend Best single Δ vs best
+24h 0.962 gem_seamless (1.124) -14.4%
+48h 1.068 gem_seamless (1.288) -17.0%
+72h 1.260 ecmwf_ifs025 (1.441) -12.6%
Verify history (no runs yet)

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

Wind direction — Bonehill Rocks

Phase wind_mvn · v2026-06-14_125651_wind_mvn

PyTorch MLP predicting a bivariate normal over the (u, v) wind vector, trained on Dunkeswell SYNOP truth. Direction = atan2(−μ_u, −μ_v); also emits a calibrated direction confidence interval. Trained 2026-06-14. Metric: Test MAE (°).

Lead Blend Best single Δ vs best
+24h 45.276 gem_seamless (25.009) +81.0%
+48h 45.582 jma_seamless (27.586) +65.2%
+72h 54.628 ukmo_seamless (30.629) +78.4%
Verify history (no runs yet)

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

10 m wind gust — Bonehill Rocks

Phase wind_gust_lgb · v2026-06-14_150909_wind_gust_lgb

10 m wind-gust LightGBM blender, ERA5 WindGusts10m truth. Per-NWP gust + gust/speed ratio + ratio spread over the gust-publishing NWPs. Trained 2026-06-14. Metric: Test MAE (m/s).

Lead Blend Best single Δ vs best
+24h 1.078 icon_seamless (1.822) -40.8%
+48h 1.253 icon_seamless (2.071) -39.5%
+72h 1.534 ukmo_seamless (2.351) -34.7%
Verify history (no runs yet)

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