“In the past decade, ECMWF has improved substantially the model components and the algorithms used to estimate the initial and forecast states, expressed in terms of a probability distribution function (PDF) of earth-system states. Atmospheric processes have been made more realistic, and ocean processes, aerosols and chemical species have been included in the forecast model. These advances have led to the extension of ensemble-based, probabilistic predictions out to sub-seasonal and seasonal time ranges. How far ahead can we provide skilful forecasts? More precisely, if we consider phenomena with increasingly coarse scales (both spatially and temporally), how far ahead can we predict them? In this talk, these questions are addressed applying the forecast skill horizon framework to ECMWF ensemble forecasts. Results based on 1 year of forecasts indicate that the forecast skill horizon for instantaneous, grid-point fields is between 16‒23 days, while it is considerably longer for time- and spatial-average fields. Forecast skill horizons longer than the 2 weeks, thought to be the limit up to twenty years ago, are now achievable thanks to major advances in numerical weather prediction.”

     “过去10年，ECMWF大幅度改进了模式组件和估计初始状态和预报状态的算法，用概率分布函数（PDF）表示地球系统的状态。大气过程更加接近真实情况，而海洋过程、气溶胶和化学物质进入了预报模式。这些进展使得基于集合预报的概率预报拓展到季节内和季节时间段。那么我们提供有技巧预报的到底是多长时间？更准确地说，如果我们考虑不断增加的空间和时间粗糙尺度天气现象，我们能够提前多长时间预报它们？我的报告里，这些问题通过应用ECMWF集合预报的预报技巧范围框架（the forecast skill horizon framework）加以讨论。基于1年预报的结果表明，对于瞬时、格点场而言，预报技巧的范围在16‒23天之间，而对于时间和空间平均场而言要长得多。预报技巧的时间段长于直到20年前认为的2周预报极限，是因为在数值天气预报方面取得了重要进展。”

     ——在2016年初美国气象学会年会上，来自欧洲中期天气预报中心的Roberto Buizza以“预报技巧的时间界限”为题，在“无缝隙天气和气候预报——多尺度可预报性的期待和局限特别研讨会”上作了特约发言。