WeatherBlend

Multi-model forecast blending for Bonehill Rocks, Dartmoor

Temperature models

Phase 2b (lean, 13 features) and Phase 2c (rich, 88 features). Trained against ERA5 reanalysis, scored as MAE in °C — lower is better. The Δ column compares the blend to the best single NWP on the same test slice; negative means the blend wins.

Temperature

Phase 2b · v2026-04-28_232613

Lean blender — six per-NWP temperatures, their mean/std/range, and cyclical hour/day-of-year encodings (~13 features). Trained 2026-04-28. Metric: Test MAE (°C).

Lead Blend Best single Δ vs best
+24h 0.533 temp_aifs (0.563) -5.3%
+48h 0.642 temp_aifs (0.661) -2.8%
+72h 0.792 temp_aifs (0.829) -4.4%
+96h 0.963 temp_aifs (0.943) +2.1%
+120h 1.129 temp_aifs (1.140) -1.0%
Verify history (1 run)

Twice-weekly Brier/MAE on the held-out rolling window — one row per verify run, drift flag in the last column. Metric: MAE (°C). Version column names which trained model the row's numbers came from — a freshly retrained champion shows zero rows here for ~5-9d (one verify cycle plus 5d ERA5 latency), so a row labelled with an older version is the previous lineage's history under the same phase.

Run (UTC) Version +24h+48h+72h Drift
v2026-04-21_201231_phase2redo 0.9590.9430.888 ⚠

Phase 2c · v2026-04-28_232637_phase2c

Rich blender — adds per-NWP dew point, RH, cloud {total/low/mid/high}, wind speed/direction/gusts, surface pressure, plus cross-model aggregates (~88 features). Trained 2026-04-28. Metric: Test MAE (°C).

Lead Blend Best single Δ vs best
+24h 0.457 temp_aifs (0.538) -15.1%
+48h 0.601 temp_aifs (0.641) -6.3%
+72h 0.697 temp_aifs (0.806) -13.6%
+96h 0.858 temp_aifs (0.899) -4.5%
+120h 1.034 temp_aifs (1.040) -0.6%
Verify history (1 run)

Twice-weekly Brier/MAE on the held-out rolling window — one row per verify run, drift flag in the last column. Metric: MAE (°C). Version column names which trained model the row's numbers came from — a freshly retrained champion shows zero rows here for ~5-9d (one verify cycle plus 5d ERA5 latency), so a row labelled with an older version is the previous lineage's history under the same phase.

Run (UTC) Version +24h+48h Drift
v2026-04-23_132453_phase2c 0.8340.904 ⚠