Generation, validation and calibration of probabilistic quantitative precipitation forecasts using breeding and Ensemble Kalman methods with the global model GME

天气时间尺度为3-5天的定量降水预报(QPF)本质上是概率的。预测降水量基本概率密度函数估计的唯一可行方法是集合模拟。我们将在育种技术的基础上建立全球模式GME的集合预报系统。有了集合预报系统，第一次猜测的协方差矩阵可以从个体实现的扩散中估计出来。这些想法被称为集合卡尔曼滤波，它将与目前正在开发的DWD的3D-Var分析一起实施。整个系统将首先通过使用文献的标准分数、单一观测和集水区的平均值进行验证。标准分数将通过地统计学的一种新方法进行扩展，以说明空间代表性误差。在建立了这个系统之后，我们将在验证和集合卡尔曼滤波方面加入新的想法，这些想法源于对统计问题的贝叶斯观点。

Quantitative precipitation forecasts (QPF) with a lead time of synoptic time scales of 3-5 days are intrinsically probabilistic. The only feasible method for predicting estimates of the underlying probability density function of precipitation amount are ensemble simulations. We will set up an ensemble forecast systems for the global model GME based an the breeding technique. With an ensemble forecasting system available, the covariance matrix of the first guess can be estimated from the spread of the individual realizations. These ideas are known as Ensemble Kalman filtering which will be implemented together with the 3D-Var analyis of DWD currently under developement. The complete system will be validated first by using standard scores from literatur an single observations and averages for catchments. Standard scores will be extended through a new method from geostatistics to account for the spatial representativity error. Having set up this system we will include new ideas in validation and Ensemble Kalman filtering arising from a Bayesian view of the statistical problems.