Compare every model's daily-high forecast for any station, side by side
Pick a station and see how each model calls the days ahead: the ECMWF HRES anchor, our calibrated ECMWF ensemble, the NBM blend and the NWS forecaster grid — every one bias-corrected — drawn as lines you can toggle, then laid out in a model-by-day grid shaded by how far each runs from the blend. The bold consensus and its uncertainty sit on top, and the spread row shows where the models agree or split (the bust-risk tell). Tap any day for the full source breakdown and the odds of clearing a temperature.
The engine behind the number: for a chosen model run, the regime the models are aiming at, where they agree vs diverge (bust risk visible before the outcome), the scenarios in play, and which model to lean on. New here? The Method page explains every panel with live examples.
The predicted Day-5 regime across recent cycles. A flip means the models revised the pattern; a steady hold means the outlook has locked in — a confidence signal in itself.
Ensemble divergence (the Fork), known now with no outcome needed. Redder = the members disagree more = the forecast is more likely to bust. Reading down a column shows where confidence runs out. The metric's companion site: forecastfork.com.
The shadow House weights for this run's predicted regime. The base weight is a diversity-aware blend — the minimum-variance combination that penalizes correlated models (ECMWF & AIFS share the IFS error structure), then tilted toward the model that owns this pattern. Shadow only.