Lecture Notes on Numerical Weather Prediction Predictability, Probabilistic Forecasting and Ensemble Prediction Systems

Ensemble techniques have been used to generate daily numerical weather forecasts since the 1990’s in numerical centers around the world due to the increase of computation ability. One of the main purposes of numerical ensemble forecast tends to assimilate initial uncertainty (both observation and analysis errors) and forecast uncertainty (model errors) by applying either initial perturbation method, ensemble assimilation, or multimodel/multi-physics method, and stochastic physics. In fact, the mean of ensemble forecasts is offering better forecast than deterministic (or control) forecast after a short lead-time (1-3 days) for the global model application. There is about a 1-2 day improvement in the forecast skill when using ensemble mean instead of a single forecast for longer lead-time. The skillful forecast (65% and above of an anomaly correlation) could be extended to 8-10 days (or longer) by present state-of-the-art analysis and ensemble forecast system. It is most important that ensemble forecast can deliver the probabilistic forecast directly, which is based on probability density function (PDF), instead of a single value forecast from traditional deterministic system to the users. It has long been recognized that the ensemble forecast is not only improving our weather forecast predictability but also offering a remarkable forecast for a future uncertainty, to help us making right decision, such as relative measure of predictability (RMOP), economic value (EV) and probabilistic quantitative precipitation forecast (PQPF). Not surprisingly, the success of ensemble forecast and its wide application are greatly increasing the confidence of model developers and research communities. Note: This article is mainly expansion of published article “Ensemble Forecast: A New Approach to Uncertainty and Predictability” (Zhu, 2005) which contributes to the lecture notes for WMO/RTC publication. Corresponding author address: Yuejian Zhu, Environmental Modeling Center/NCEP, 5200 Auth Road, Camp Springs, MD 20746, USA; e-mail: [email protected]