Experimental Week 2 Forecast Products


latest forecast maps   |   forecast map archive   |   verification statistics |   pentad 2 forecasts (6-10d)

latest calibrated station forecasts: precip (lower 48ak and hitext), temp (lower 48ak and hitext

NEW:  Operational NCEP products derived from the CDC ensemble

NEW:  Probabilistic precip forecasts downscaled to 32 km resolution using forecast analogs

If you use these products, and you would like to seem them continue, please let me know how you use them and why.

These forecast products are derived from an experimental real-time ensemble run at CDC.  A 15-member ensemble is run every day at 00 UTC using a frozen version of the operational MRF, which was operational between January and June 1998.  Forecasts with the same model and ensemble configuration are being run retrospectively for every day from November 1978 to the present as part of the MRF Reforecast Project. Having a sufficiently long historical dataset of forecasts with the same model used to produce the real-time ensemble will allow us to correct for model systematic errors and calibrate forecast probabilities more accurately than can be done with the operational NCEP ensemble.

Download recent real-time forecast data and station forecasts here.
Download historical re-forecasts (197811-200212) here

The forecast plots are displayed as follows:

Lower tercile probability
Upper tercile probability
Climatological Lower Tercile
Climatological Upper Tercile
Ensemble Mean Forecast Anomaly
Verifying (CDAS) Anomaly - when available

Five fields are currently available (500 mb height, 250 mb zonal wind, 850 mb temperature, sea-level pressure and precipitation).

Anomalies and terciles are relative to a 1979-2001 daily climatology (smoothed with a 31-day running mean).  Forecast anomalies are calculated relative to the climatology of the forecast model (i.e the mean bias is removed). Tercile probabilities are calculated using the retrospective forecast database (see here for a description of the method) and are generally quite reliable (as evidenced by the cross-validated forecast verification statistics).

For more information, see the manuscript submitted to Monthly Weather Review.