GeCC-LAG-ENSEMBLES: a Generalized Calibration and Combination approach to mix in an optimum way lagged multi-model ensemble forecasts

GeCC-LAG-ENSEMBLES：一种广义校准和组合方法，以最佳方式混合滞后的多模型集合预测

• 批准号: 490941584
• 负责人: Professor Paolo Reggiani, Ph.D.
• 金额: --
• 依托单位: Scuola Superiore SantAnna
• 依托单位国家: 德国
• 项目类别: Research Grants
• 资助国家: 德国
• 项目状态: 未结题
• 来源: https://gepris.dfg.de/gepris/projekt/490941584?language=en
• 关键词: GeCC; LAG; ENSEMBLES; Generalized; Calibration

In this proposal we aim to develop a Generalized Calibration and Combination (GeCC) methodology and apply it to combine in an optimum way lagged ensembles generated by different weather forecasting models. The generalized combination technique involves weights that are determined on the basis of two Bayesian methods: Bayesian Model Averaging (BMA) and the correlation-based Model Conditional Processor (MCP) method. The GeCC-processed probabilistic forecasts can be expressed in terms of predictive statistics for meteorological variables over the spatial model grid, or at station locations. The GeCC technique could be used to merge ensembles with different characteristics to generate more accurate and reliable probabilistic products, than the ones that could be generated by a single ensemble is available. These ensembles could be, for example, lagged global medium-range, sub-seasonal and seasonal ensembles (e.g. produced by the European Center for Medium-Range Weather Forecasts), or lagged global and limited-area ensembles (e.g. produced by the Deutscher Wetterdienst). In particular, GeCC-processed probabilistic forecasts of rare events are expected to improve.The GeCC technique could also lead the way toward smarter and computationally more efficient ways of computing combination weights, and/or calibration coefficients that today are computed using re-forecast datasets that currently absorb considerable computational resources to be generated. The three main outcomes of this project are going to be: - A pre-operational weighting procedure that allows to combine multiple lagged ensemble forecasts with different forecasting horizons. - A systematic verification of the proposed method through error statistics, including meteorological extremes, for selected variables and locations in at least one geographical region in Europe. - Estimation of a calibrated predictive distributions for selected variables at selected forecast locations and/or for the selected geographical region.

在这个建议中，我们的目标是开发一个广义的校准和组合（GeCC）的方法，并将其应用到联合收割机在一个最佳的方式产生的不同的天气预报模式滞后合奏。广义组合技术涉及的权重是基于两种贝叶斯方法确定的：贝叶斯模型平均（BMA）和基于相关性的模型条件处理器（MCP）方法。GeCC处理的概率预报可以用空间模型网格或台站位置上气象变量的预测统计数据来表示。GeCC技术可用于合并具有不同特征的集合，以生成比可由单个集合生成的概率产品更准确和可靠的概率产品。例如，这些集合可以是滞后全球中期、亚季节和季节性集合（例如，由欧洲中期天气预报中心制作），或滞后全球和有限区域集合（例如，由Deutscher Wetterdienst制作）。特别是，GeCC处理的概率预测的罕见事件，预计将改善。GeCC技术也可以导致更智能和计算更有效的方法来计算组合权重，和/或校准系数，今天使用重新预测数据集计算，目前吸收相当多的计算资源来生成。该项目的三个主要成果将是：-一个操作前加权程序，允许联合收割机多个滞后集合预测与不同的预测范围。- 通过误差统计，包括欧洲至少一个地理区域的选定变量和地点的极端气象条件，对拟议方法进行系统验证。- 在选定的预报位置和/或选定的地理区域估计选定变量的校准预测分布。