全球10米风速栅格数据（1973-2021）

风速数据被广泛用于科学、管理和政策领域，在评估可再生能源潜力、解决风灾、研究生物现象和探索气候变化等方面发挥着重要作用。但现有的风速产品存在很大的局限性：气象观测数据在空间和时间上存在不连续性，再分析产品和气候模型模拟虽然实现了数据的连续性，但大多未能重现观测到的风速趋势。此外，风速数据的高变异性及站点分布的不均匀和稀缺性，使得传统的统计插值方法，如克里金或主成分分析，在重构全球风速上表现不佳。因而，风速数据成为风速研究中“卡脖子”的难题。

在此，研究团队基于部分卷积神经网络算法（the partial convolutional neural network），融合了34个气候模式数据和气象站点观测数据HadISD（由Met Office Hadley Centre提供），重构了1973-2021年间共588个月的全球10米近地风速，空间分辨率为1.25°×2.5°（纬度×经度），该数据集包含了观测到的风速趋势信息。详细的重构过程请见参考文献中的方法部分。

Wind speed data is widely utilized in scientific, management and policy fields, playing a vital role in assessing renewable energy potential, addressing wind disasters, investigating biological phenomena, and exploring climate change. However, existing wind speed products face significant limitations: meteorological observational data presents spatial and temporal discontinuities. Although reanalysis products and climate model simulations achieve data continuity, most of them fail to reproduce the observed wind speed trends. Furthermore, the high variability of wind speed data, combined with the uneven distribution and scarcity of meteorological stations, leads traditional statistical interpolation methods such as Kriging and Principal Component Analysis (PCA) to perform poorly in reconstructing global wind speeds. Consequently, wind speed data has become a bottleneck issue in wind speed research.

In this study, the research team adopted the partial convolutional neural network algorithm, fused 34 climate model datasets with the meteorological station observational dataset HadISD (provided by the Met Office Hadley Centre), and reconstructed global near-surface 10-meter wind speeds for a total of 588 months from 1973 to 2021, with a spatial resolution of 1.25° × 2.5° (latitude × longitude). This dataset contains the observed wind speed trend information. For detailed reconstruction procedures, please refer to the Methods section in the cited references.

搜集汇总

背景与挑战

背景概述

该数据集是一个全球范围的月尺度10米近地表风速栅格数据，覆盖1973年至2021年，空间分辨率为1.25°×2.5°。它基于部分卷积神经网络算法，融合了多源气候模式和站点观测数据，旨在解决传统风速产品在连续性、趋势重现和空间覆盖方面的不足，并包含观测到的风速趋势信息。数据以开放获取方式共享，但使用时需注意剔除少量离群值，并遵守引用规范。