Probabilistic Forecasts, Based on NCEP GEFS Reforecasts, Version2
This page presents experimental probabilistic forecasts, produced with the non-homogeneous Gaussian regression technique of Gneiting et al. (Monthly Weather Review, 2005). Two predictors are used for each variable: that variable's forecast ensemble mean and spread (e.g., T850 forecast ensemble mean and spread for the T850 predictand). Forecasts are available for 2m temperature, 850mb temperature and 500mb height at several forecast leads, ranging from 2-3 days to 8-14 days.
We follow the NOAA Climate Prediction Center definitions of forecast day. Their definition of day 1 is that it includes the 6-hourly data from forecast hours 30, 36, 42 and 48. The 6-10 day forecast is thus defined as forecast hours 150-264; the 8-14 day forecast is forecast hours 198-360. For example, for the forecast initialized 00Z January 1, day 1 includes forecast data between 06Z January 2 - 00Z January 3. Day 2 includes forecast data between 06Z January 3 - 00Z January 4, etc.
These forecasts will usually (but not always) be updated by 16 UTC each day. They likely will not be available as consistently as operational products from the National Weather Service. Also please note that this is an experimental forecast product, and is not an official forecast of NOAA or its National Weather Service.
We welcome feedback on this product. Email comments to: email@example.com .