Assessing the calibration of multivariate ensemble forecasts (Ref:4659)

本项目旨在创建强大的统计方法来评估多元集合预测的校准。新的特征将是现有方法的扩展，使它们适合于集合预测，并利用包含在多视界预测中的信息。这些方法将提高预报员监测和了解其预测效果的能力，识别表现不佳的情况以及表现往往不佳的情况，量化预测中的偏差，以及识别预测信息的新来源。

This project aims to create powerful statistical methods for assessing the calibration of multivariate ensemble predictions. Novel features will be the extension of existing methods to make them suitable for ensemble predictions, and the exploitation of information contained in multi-horizon predictions. These methods will enhance the ability of forecasters to monitor and understand the performance of their forecasts, to identify cases of poor performance as well as conditions in which performance tends to be poor, to quantify biases in their forecasts, and to recognise new sources of predictive information.