Spatial Verification of High-Resolution Ensemble Forecasts using Wavelet Transformation

使用小波变换对高分辨率集合预报进行空间验证

• 批准号: 381875761
• 负责人: Privatdozentin Dr. Petra Friederichs, since 3/2018
• 金额: --
• 依托单位: Meteorologisches Institut (aufgelöst)
• 依托单位国家: 德国
• 项目类别: Research Grants
• 财政年份: 2017
• 资助国家: 德国
• 起止时间: 2016-12-31 至 2021-12-31
• 项目状态: 已结题
• 来源: https://gepris.dfg.de/gepris/projekt/381875761?language=en
• 关键词: Spatial; Verification; Resolution; Ensemble; Forecasts

The aim of VeriWave is to develop a method for spatial verification of high-resolution ensemble model output with observations with non-negligible uncertainties. The proposed technique is based on wavelet transforms and bears the potential to yield sophisticated diagnostic information tailored to the physical process of interest and comparable to prominent feature based methods, while retaining the robustness to noise and the computational efficiency of scale separation methods. Particular emphasize lies on the orthogonality of the proposed verification score components intensity, location, orientation and texture, which allows one to reduce the high-dimensional data to only a few key characteristics. We propose the following research objectives: (1) Define an framework of wavelet transformations tailored to the data at hand and the physical process of interest, (2) derive a wavelet based methodology for spatial verification that yields diagnostic information on intensity, location and texture, (3) ensure that the methodology is robust with respect to varying parameters and observational uncertainties, (4) extend methodology to ensemble forecast, and (5) disseminate the proposed method by providing the software together with a user handbook.The proposed method will be studied on a hierarchy of data ranging from simple geometric test cases and synthetically disturbed forecaststo high-resolution case studies and large sets of ensemble data.

VeriWave的目的是开发一种方法，对具有不可忽略不确定性的观测结果的高分辨率集合模型输出进行空间验证。所提出的技术基于小波变换，具有产生适合感兴趣的物理过程的复杂诊断信息的潜力，可与突出的基于特征的方法相媲美，同时保留了对噪声的鲁棒性和尺度分离方法的计算效率。特别强调的是所提出的验证分数成分强度、位置、方向和纹理的正交性，这使得人们可以将高维数据减少到只有几个关键特征。我们提出以下研究目标：(1)定义适合手头数据和感兴趣的物理过程的小波变换框架；(2)推导基于小波变换的空间验证方法，产生关于强度、位置和纹理的诊断信息；(3)确保该方法对不同参数和观测不确定性具有鲁棒性；(4)将方法扩展到集合预测；(5)通过提供软件和用户手册的方式传播所提出的方法。提出的方法将在数据层次上进行研究，从简单的几何测试用例和综合扰动预测到高分辨率案例研究和大型集成数据集。