FernWX
FernWX Lab Mode is a Research Preview, and can be disabled at any time. New sessions will default to the standard FernWX experience.

Forecast Memory

Memory

Replay what providers knew, verify how forecasts changed, and inspect provenance.

Sign in to use your collected forecast memory. This view uses synthetic or empty examples.

Canonical stored values with provider-issued lead, retrieval-derived knowledge lead, NWS station reference, signed error (forecast minus reference), and absolute error.

Filter options are derived from the rows this location actually holds. Only one variable is present: the read-model refresh currently builds verification pairs for temperature alone, so no other variable exists to filter by yet.

Showing 0 matching rows

Valid-time bounds are read in your browser’s timezone, matching the Valid column. Exports follow the active filters and page through the whole matching set, up to 20,000 rows.

No persisted verification records are available for this location yet.
How Memory works

Replay freezes what each provider said at a collection time. Verify compares stored forecasts with the nearest eligible NWS station observation. Providers ranks temperature forecasts by mean absolute error over a rolling window and carries each provider’s attribution and interpretation limits. Records holds the canonical stored rows behind all three.