Advances on the Predictability and Prediction Methods of 10-30 d Extended Range Forecast

The 10-30 d extended range forecast (ERF) fills the gap between traditional weather forecast and short-term climate prediction, and it plays an important role in the decision making of disaster prevention and mitigation. Therefore, ERF becomes one hot topic in both scientific research and predictive operations. The research progress and operational status of ERF are reviewed from three aspects, the source of predictability, sub-seasonal climate phenomenon and operational predictions. The research achievements on predictability of ERF and its applications are specially emphasized, and some new forecasting methods of ERF in recent years are summarized. At last, key scientific issues and technical problems are raised and some thoughts and possible ways enhancing the predictive skills of ERF are proposed.ERF exceeds time limits of traditional daily weather forecast, largely beyond the atmospheric memory of initial conditions, while it is too short to consider the variability of the ocean, which makes it difficult to beat persistence. Fortunately, recent years, some research work indicates the existence of some important sources of predictability at this time range, such as Madden-Julian oscillation (MJO), ENSO, soil moisture, snow cover and sea ice, stratosphere-troposphere interaction, ocean conditions, tropics-extratropics teleconnections, etc. Verification results of numerical model indicate that upper bounds of the prediction skill can be extended to 4 weeks. However, the complexity and diversity of mechanisms associated with the connection between source of predictability and climate variables prevent the potential predictability from being transformed into realized forecast skill. The effective forecast of most climate variables of numerical model is still limited within 2 weeks.Although the direct application of numerical dynamical model output in ERF is unsatisfactory, some research institutes and operational centers still conduct a series of scientific research and propose some practical methods. According to utilization of numerical model data, those forecast methods can be divided into two categories, i.e., statistical methods and the combination of both statistical and dynamical methods. Based on dynamical forecast model, Beijing Climate Center develops several methods, including Dynamical-Analogue Ensemble Forecasting (DAEF), statistical downscaling, ensemble forecast of ERF based on predictable components and probabilistic calibration of model biases. On the other side, based on predictable signals of extended range, such as low frequency variation of atmosphere, MJO and periodic relationship, some statistical forecast methods are proposed, which show considerable predictive skill and good prospects of application.