NASA Earthdata

NASA Earthdata is the single front door to NASA's Earth science holdings: more than 12,400 datasets spanning the atmosphere, ocean, land, and cryosphere (the frozen parts of the planet), plus records of how people interact with the environment. The data comes from dozens of satellite missions, airborne campaigns, and ground stations, organised into collections managed by NASA's network of distributed data centres, the DAACs.

You discover and pull data programmatically through the Common Metadata Repository, the search API that lets you filter every dataset by area, time, and keyword. A free Earthdata Login is required to download, and the earthaccess Python library smooths over the authentication and search steps. Much of the archive now lives in the cloud on AWS, so you can work against it there without shifting terabytes around.

Here is the practical reality for RAG: most of the actual measurements are gridded scientific arrays in formats like HDF, NetCDF, and GeoTIFF, which are numbers, not prose, and do not embed into a vector database in any useful way. The retrieval value sits in the surrounding text: dataset descriptions, user guides, algorithm documentation, and mission summaries. A strong Earthdata RAG system indexes that metadata and documentation to help people find and understand the right dataset, then hands off to proper geospatial tooling for the numbers themselves.

On licensing there is little to worry about. As a US Government product the data is public domain, so you can use it commercially with no attribution requirement, though crediting NASA is good manners and some third-party or near-real-time products carry their own terms.

For wider coverage it pairs well with NOAA's open data and Europe's Copernicus programme. Earthdata is the natural starting point whenever your questions centre on satellite Earth observation and climate science.