REFORECASTS An Important Dataset for Improving Weather Predictions

Environmental Prediction (NCEP)–National Center for Atmospheric Research (NCAR) reanalysis (Kalnay et al. 1996) and the European Centre for Medium-Range Weather Forecasts (ECMWF) 40-year reanalysis (ERA-40; Uppala et al. 2005) have become heavily used products for geophysical science research. These reanalyses run a practical, consistent data assimilation and short-range forecast system over a long period of time. While the observation type and quality may change somewhat, the forecast model and assimilation system are typically fixed. This facilitates the generation of a reanalysis dataset that is fairly consistent in quality over time. These reanalysis datasets have facilitated a wide range of research; for example, the Kalnay et al. (1996) article above has been cited more than 3200 times. In this article, we explore the value of a companion dataset to reanalyses, which we shall call “reforecasts.” These are retrospective weather forecasts generated with a fixed numerical model. Model developers could use them for diagnosing model bias, thereby facilitating the development of new, improved versions of the model. Others could use them as data for statistically correcting weather forecasts, thereby developing improved, user-specific products [e.g., model output statistics (MOS) Glahn and Lowry 1972; Carter et al. 1989]. Others may use them for studies of atmospheric predictability. Unfortunately, extensive sets of reforecasts are not commonly produced utilizing the same model version as is run operationally. These computationally expensive reforecasts are “squeezed out” by operational data assimilation and forecast models that are run at as fine a resolution as possible. Would the additional forecast improvement and diagnostic capability provided by reforecasts make them worth the extra computational resources they require? To explore this, we recently generated a prototype 25-yr, 15-member ensemble reforecast dataset using a 1998 version of the NCEP Medium Range Forecast (MRF) model run at T62 resolution, which is admittedly a resolution far from being state REFORECASTS An Important Dataset for Improving Weather Predictions