Reforecasts: An Important Dataset for Improving Weather Predictions featured in BAMSMarch 18, 2006
ESRL researchers Tom Hamill and Jeff Whitaker, co-authored the current cover article for the January 2006 issue of the Bulletin of the American Meteorological Society (BAMS). The article entitled "Reforecasts: An Important Dataset for Improving Weather Predictions," discusses how weather forecasts can be improved through statistical corrections of the current numerical forecast using many observations and past forecasts from the same fixed model. Improvements of probabilistic quantitative precipitation forecasts are shown to be very large, and the authors propose that reforecast-based techniques become a mainstay of the NWS production suite. The question is how to achieve the benefits of reforecasts without hindering model development.
The authors propose a scenario whereby model development would proceed much as it does now, but in addition every several years a new reforecast database would be produced offline with a new, updated version of the operational model, a version that is computationally less expensive to run than the operational models. The reforecast model would then be run in real time alongside the operational models. Forecasters would then have the benefit of both raw, high-resolution model output and high-quality statistical products based on the reforecasts.
Approximately four years ago, Hamill and Whitaker began their "reforecasting" effort, a companion effort to the NCEP/NCAR reanalysis. Fifteen-member ensemble forecasts were run over the last 25+ years using initial conditions from the reanalysis and a recent, unchanging version of NCEP's global forecasting system. The reasons for this effort included improving medium-range weather forecast products, improving probabilistic forecasts of extreme events, and diagnosing model errors. Users wishing to explore possible applications of reforecasts can download data at http://www.cdc.noaa.gov/reforecast/. Real-time probabilistic precipitation forecasts are available at http://www.cdc.noaa.gov/reforecast/narr.
Errors from past forecasts can be found and corrected, dramatically increasing forecasting skill. The potential benefits of reforecasts are so large that their regular use should become part of the numerical weather prediction process. This research supports NOAA's mission goal of serving society's needs for weather and water information.