On the Economic Value Of Ensemble Based Weather Forecasts

The NCEP global ensemble forecasting system, as of late 1999, comprises of 17 forecasts per day, running out to 15 days lead time. The case dependent divergence of these forecasts reflects the uncertainty of the day, related to small initial/model errors, arising due to the chaotic nature of the atmosphere. The ensemble forecasts and a number of derived products are disseminated through different channels and are used by a wide array of users on various time scales. In this paper, we will present verification results evaluating the overall performance of the global ensemble forecasting system. The impact of upgrades in the operational configuration, i. e., increased horizontal resolution and a larger ensemble size, will also be discussed. The effect of a simple calibration procedure on probabiistic forecast performance will be presented. It will be shown that most users can benefit from probabilistic forecast information derived from the ensemble, compared to using a single control forecast only. Comparison between ensemble-derived probabilistic forecasts and those generated using past verification statistics associated with a higher resolution control forecast will also be discussed. A new type of probabilistic guidance, indicating how much more or less reliable a particular forecast is compared to long term average reliability, will be interoduced. The presentation will end with a description of possible future upgrades to the NCEP ensemble forecasting system.