Severe-weather prediction using convective-scale ensembles

对流风暴是气候系统的一个关键要素，其影响范围从微观的气溶胶颗粒到全球环流。在区域范围内，它们提供了维持生命和农业的宝贵降雨，但也造成了包括冰雹和洪水在内的恶劣天气。这样一个至关重要的过程需要来自业务气象中心的准确预测。然而，由于其高度湍流、混乱和多尺度的性质，这些风暴非常难以准确预测。在数值天气预报模型中，即使是很小的不确定性也会导致风暴在错误的地点、错误的时间形成，或者根本不会形成。虽然某种程度的预报不确定性是不可避免的，但对流预报仍然可以通过减少这种不确定性的程度和采用量化其对模式结果影响的概率方法来改进。

Convective storms represent a key element of the climate system, with impacts on scales ranging from microscopic aerosol particles up to the global circulation. On the regional scale they provide valuable rainfall that sustains life and agriculture but also inflict severe weather including hailstorms and flash-flooding. A process of such critical importance demands accurate prediction from operational weather centres. However, because of their highly turbulent, chaotic, and multi-scale nature, these storms are tremendously difficult to accurately predict. Even small uncertainties in numerical weather prediction models cause storms to form in the wrong place, at the wrong time, or to not develop at all. Although some level of forecast uncertainty is unavoidable, convection forecasts can still improve by diminishing the magnitude of this uncertainty and by adopting probabilistic methods that quantify its impact on the model outcome.