AI News

Google DeepMind Open-Sources the SatWeave and PARDIG Datasets Used by WeatherNext 3

English text is being prepared.Updated Nov 13 21:47

English text is being prepared.

AI summary

Google DeepMind has released SatWeave and PARDIG, the two foundational datasets underpinning the WeatherNext 3 weather model, providing researchers worldwide with real-time data streams and historical archives. SatWeave integrates imagery from five geostationary weather satellites—GOES-East, GOES-West, Meteosat 0, Meteosat Indian Ocean, and Himawari—updating hourly, with a historical archive spanning more than ten years. PARDIG is a global precipitation reanalysis dataset based on a spherical graph-grid Transformer architecture, with a spatial resolution of about 10 kilometers (0.1°), a temporal resolution of 15 minutes, and an average latency of about 5 hours; Google says that in evaluations benchmarked against ground-based radar networks in the United States, Europe, and Japan and more than 36,000 weather stations worldwide, it outperforms baselines such as NASA IMERG-Final across all rainfall intensity tiers, with hourly detection skill for light to moderate precipitation improving by more than 30%. The data is hosted on Google Cloud Storage and released under the CC BY 4.0 license, with the PARDIG historical archive from 1999 to 2026 also made available.

Why read this

AI score 67

The original text provides the resolution, latency, and comparative evaluation results of the two datasets, which can be used to assess changes in the data threshold for weather modeling.

Topics

Tags

Body

谷歌开源 WeatherNext 3 气象模型所用数据集:含 SatWeave 卫星影像拼接数据集与 PARDIG 全球降水信息

2026-11-13 20:04 IT之家 - 小泵

IT之家 11 月 13 日消息,谷歌 DeepMind 今天正式开放 WeatherNext 3 气象模型所依托的两大基础数据集 SatWeave 与 PARDIG,相应数据集面向全球研究者与开发者提供实时数据流及历史归档,可用于天气、水文、气候及相关下游应用建模。

据介绍,SatWeave 是一套整合多颗静止气象卫星的全球影像拼接产品,聚合了 GOES-East、GOES-West、Meteosat 0、Meteosat 印度洋星以及 Himawari 共 5 颗卫星的观测数据,解决了不同卫星在地图投影、文件格式、传感器与扫描节奏上的差异难题。该数据集每小时更新一次,并提供跨越十年以上的历史存档,能够让 WeatherNext 3 基于实时云图与风暴动态生成逐小时全球预报,而传统分析数据通常每六小时才刷新一次。谷歌还希望研究社区将其拓展至太阳能功率预测、野火监测等场景。

而 PARDIG 是一套基于 AI 架构的全球降水再分析数据集,采用与 WeatherNext 3 相似的球形图网格 Transformer 架构,直接以 NASA 全球降水测量核心观测台的稀疏星载雷达观测为训练目标,输入使用 SatWeave 影像与欧洲中期天气预报中心的大气场。其空间分辨率约 10 公里(0.1°),时间分辨率 15 分钟,平均延迟约 5 小时。

谷歌通过对照美国、欧洲、日本三大陆基雷达网络及全球 3.6 万余座气象站的评估显示,PARDIG 在所有雨强档位上均优于 NASA IMERG-Final 等主流基线产品;对轻到中度降水的小时级识别技巧较 IMERG-Final 提升 30% 以上,对极端降水的捕捉能力尤为突出。由于模型直接从卫星观测学习,其在缺乏地面雷达与雨量计的地区同样具备可靠精度。

除此之外,谷歌本次还同步开放 1999 年至 2026 年的 PARDIG 历史档案以及 SatWeave 与 PARDIG 的实时数据流。相应数据托管在 Google Cloud Storage 上,以 CC BY 4.0 许可开放。

Source

English text is being prepared.