rdjarbeng/groundsource-enriched

Groundsource Enriched是一个增强版本的全球洪水事件数据集，包含2,646,302个事件，解码了原始Google Groundsource数据集中的二进制WKB几何图形，提供了可读的经纬度质心坐标。该数据集新增了纬度、经度、年份、月份、持续天数和区域等列，便于分析。数据覆盖时间范围从2000年1月1日到2026年2月3日，地理分布涵盖全球多个区域，如欧洲、东南亚、南亚等。数据集基于CC-BY 4.0许可证发布，原始数据通过Google Research使用Gemini LLM从超过500万篇新闻文章中提取。已知限制包括非洲地区代表性不足、区域分类近似等。

Groundsource Enriched is an enriched version of Googles Groundsource dataset, containing 2,646,302 global flood events with decoded coordinates from binary WKB geometry into human-readable latitude/longitude centroids. It adds derived columns such as latitude, longitude, year, month, duration_days, and region for easier analysis. The dataset covers the period from 2000-01-01 to 2026-02-03, with geographic distribution across regions like Europe, Southeast Asia, and South Asia. It is licensed under CC-BY 4.0 and originates from Google Research using Gemini LLMs to extract structured flood event data from over 5 million news articles. Known limitations include underrepresentation of Africa and approximate region classification.