ECMWF IFS High-Resolution Operational Forecasts

ECMWF has implemented a significant resolution upgrade and methodology for high-resolution forecasts (HRES) and ensemble forecasts (ENS) beginning January of 2016. HRES is now performed via a transform grid with a nominal grid point spacing of 9 kilometers (0.08 degrees), and is carried out with IFS (Integrated Forecast System). Improvements in computational efficiency and effective resolution have been brought about by implementing a triangular cubic octahedral reduced Gaussian grid in which the shortest spatial wavelength is represented by at least four grid points anywhere on the globe, as opposed to the former linear arrangement whereby the shortest wavelength was represented by two grid points, while at the same time retaining the same number of spherical harmonics and triangular truncation. (The term "cubic" is due to the ability of the grid to represent cubic products in the dynamical equations.) In addition, the reduction of grid points along latitude circles as one approaches the poles is achieved using a triangular to octahedral mapping which corresponds to a poleward reduction of four points per latitude circle and an optimization of the total number of grid points and their local mesh resolution. ECMWF has documented superior filtering properties at higher resolution, an improved representation of orography, improved global mass conservation properties, substantial efficiency gains, and more scalable locally compact computations of derivatives and other properties that depend on nearest-neighbor information only. More details may be found in the publications cited below.
NCAR's DECS is performing and supplying a grid transformed version of HRES IFS, in which variables originally represented as spectral coefficients or archived on a reduced Gaussian grid are transformed to a regular 5120 longitude by 2560 latitude N1280 Gaussian grid. In addition, DECS is also computing horizontal winds (u-component, v-component) from spectral vorticity and divergence...

自2016年1月起，欧洲中期天气预报中心（European Centre for Medium-Range Weather Forecasts，ECMWF）对高分辨率预报（High-Resolution Forecasts，HRES）与集合预报（Ensemble Forecasts，ENS）开展了重大分辨率升级与方法优化。目前HRES采用标称网格点间距为9公里（0.08度）的变换网格运行，并基于综合预报系统（Integrated Forecast System，IFS）实现。

通过采用三角立方八面缩减高斯网格，实现了计算效率与有效分辨率的双重提升：该网格在全球任意区域的最短空间波长均可由至少4个网格点表征，而此前的线性布局仅能以2个网格点表征最短波长；同时，该方案保留了相同数量的球谐函数与三角截断项。（此处“立方”一词源于该网格可在动力方程中表征立方乘积项的特性。）

此外，在向极地靠近时，纬度圈上的网格点缩减采用三角至八面映射实现：每一条纬度圈向极地方向每圈缩减4个网格点，同时优化全球总网格点数与局地网格分辨率。ECMWF已验证，升级后的系统具备更优异的高分辨率滤波特性、更精准的地形表征能力、更完善的全球质量守恒特性，同时计算效率大幅提升，且仅依赖近邻信息的导数与其他属性的局地紧凑计算可扩展性更强。更多细节可参阅下文引用的文献。

美国国家大气研究中心（National Center for Atmospheric Research，NCAR）的DECS项目正在处理并提供HRES IFS的网格变换版本：该版本将原本以谱系数形式表征或存储于缩减高斯网格的变量，转换为规则的5120×2560（经度×纬度）N1280高斯网格。此外，DECS还基于谱涡度与散度计算水平风场（u分量、v分量）……

搜集汇总

背景与挑战

背景概述

该数据集是欧洲中期天气预报中心（ECMWF）提供的高分辨率业务预报数据，覆盖2016年1月至2025年12月的全球范围，每日更新，包含气温、降水、风速等多种气象变量。其特点是采用9公里高分辨率网格和三角立方八面体缩减高斯网格技术，提高了预报精度和计算效率，数据以GRIB1和netCDF4格式提供，总容量达265.58 TB。