Model accuracy
← Model statusForecast error (RMSE) vs lead time, per model, verified against the consensus analysis as each forecast reaches its valid time. Lower is better. Numbers fill in over the first week.
Accumulating…
No verified forecasts yet. As newly-stored model runs reach their valid times and are scored against the analysis, skill curves appear here — short lead times within a day or two, the full 5-day curve over the first week.
Precip nowcast skill — CSI vs lead (rain ≥ 2 mm/hr)
Extrapolation nowcasts (IMERG / GSMaP) scored against the observed precip snapshot at each valid time. CSI (Critical Success Index, 0–1) — higher is better.
Accumulating nowcast verification…
Truth = the consensus analysis (a common reference) at each valid time. ERA5 gold-standard verification, per-location skill, and the regional models (HRRR, etc.) are coming. Each model is scored on its own forecast, so skill is attributed to the specific model.