Experimental feature. This is not part of the standard beginner workflow and may be unavailable.
First compute the route normally. On an eligible Pro route of at least seven days in the validated Transpac or Pacific Cup area, open Ensemble / robust and choose Run ensemble analysis.
This is a second opinion, not a replacement for the model consensus. It builds a small set of alternative routes from ECMWF ensemble forecasts and checks them, plus the consensus route, against the full ensemble. Keep the tab open while it runs. Previously downloaded members are reused when possible. On a constrained connection, WX Navigator shows the exact estimated transfer and asks before downloading.
The result has four possible readings:
- Robust alternative found means a materially different route did better in the worst outcomes while staying close to the fastest route on the published forecast.
- Forecast holds means the normal consensus route was also the best hedge.
- No clear alternative means another route was too close to tell apart from the noise in the calculation. Treat the consensus route as primary.
- Analysis incomplete means one or more routes could not be scored across the full matrix. No robust verdict is published, and the analysis can be retried rather than becoming the saved verdict.
The table shows the time each route loses, on average, in its worst outcomes compared with the best candidate. It is a comparison across this candidate set, not an ETA promise or the probability of a particular weather outcome.
You can inspect a robust alternative on the chart. It is advisory and is never silently promoted to the main route, shared as the recommendation, or substituted for model consensus. Re-run the normal route when its forecasts update, then run ensemble analysis again if the decision still matters.
WX Navigator sends coarse device and phase diagnostics for this unusually heavy browser computation— such as browser family, memory bucket, cores, connection class, duration, processed members, transfer bytes and coarse peak memory when available. It does not send your route geometry, polar or forecast fields in these diagnostics. A random diagnostics identifier correlates the phases of one run; the ensemble session capability is never sent to analytics. Raw ensemble diagnostics are retained for 90 days.