Where to find earth observation and satellite data

Almost every serious source of remote-sensing and Earth-observation data is free and publicly funded — this is one of the rare corners of analytics where the market leaders are government and intergovernmental agencies, not vendors competing on price. That changes the decision: you are not weighing cost against features, you are weighing which archive actually covers your sensor, time period and processing needs, and whether you want raw data or a service that processes it for you. This guide is for researchers, analysts and developers who need satellite, weather or climate data as an input, not for anyone looking for finished maps or dashboards — for that, look at geospatial analytics platforms built on top of these archives instead.

Start with the sensor and time period you need, not the agency

The single biggest thing separating these archives is which satellites and sensors they cover, because that determines resolution, revisit frequency, and how far back the record goes.

• Copernicus Data Space Ecosystem covers the EU's Sentinel fleet: optical imagery down to 10m resolution (Sentinel-2), radar (Sentinel-1), ocean and land monitoring (Sentinel-3), and atmospheric composition (Sentinel-5P). Revisit is around 5 days for Sentinel-2 over a given location, with the archive extending back to each mission's launch in the mid-2010s.
• NASA Earthdata is the broadest single archive by mission count — Landsat, MODIS, VIIRS, GPM, ICESat-2 and more — with resolution ranging from roughly 250m to 1km for daily global products, and multi-decade records for some sensors back to the 1970s–80s.
• USGS EarthExplorer is the dedicated home of the full Landsat record specifically, back to 1972, at 30m resolution (15m panchromatic), with roughly 16-day revisit per satellite — the longest continuous satellite land-imaging record available, and the reason it is the default choice for long-term land-cover change work.
• NOAA Open Data is the source for weather, climate and ocean data rather than land imagery specifically — real-time and forecast weather model output updated multiple times daily, radar and satellite imagery refreshed every few minutes to hours, and historical climate station records extending back over a century for some US stations.

If your question is "how has this specific patch of land changed since the 1980s," EarthExplorer's Landsat archive is the obvious start. If it is "what's the highest-resolution recent imagery I can get," Copernicus's Sentinel-2 or NASA's newer sensors will beat Landsat's 30m. If it is about weather or ocean conditions rather than land imagery, NOAA is the right family of archives regardless of how the others compare.

Raw archive or processing API

The other major decision is whether you want to search and download raw scenes yourself, or have processing happen for you.

Copernicus, NASA Earthdata, NOAA and EarthExplorer are all, at core, search-and-download portals: you draw an area, filter by date and cloud cover, and download scenes or order bulk datasets, with programmatic APIs available alongside the web interface for scripted access. NASA Earthdata additionally hosts many of its newer collections on AWS, so cloud-based analysis can run near the data rather than downloading it locally — a meaningful difference for large-volume work.

Sentinel Hub is different in kind: a cloud API and processing service, built by Sinergise and now owned by Planet Labs, that runs band math, cloud masking and index calculations like NDVI server-side through a scripting language (evalscript), returning ready-to-use images or statistics rather than raw scenes you process yourself. It draws on Sentinel, Landsat, MODIS and — through its Planet integration — commercial Planet imagery, spanning several of the archives above rather than being tied to one. It suits developers building geospatial applications or anyone who wants repeatable, scriptable processing without managing raw downloads.

Free is the norm, but "free" still has a cost

Every archive here is open data — free to access, funded as public infrastructure (the EU for Copernicus, NASA and USGS as US federal agencies, NOAA likewise) — and none require a subscription or usage-based payment for the core archive. What is not free is the compute and storage to actually work with satellite data at scale: raw scenes are large, processing them (atmospheric correction, mosaicking, index calculation) takes real compute, and downloading terabytes repeatedly is slow. That is the gap Sentinel Hub's freemium processing tier fills commercially, and it is also why NASA's AWS-hosted collections and Copernicus's on-demand cloud processing exist — both let you compute near the data instead of moving it. Budget for this regardless of which archive you choose; "free data" does not mean a free pipeline.

Where analysis code fits in

None of these platforms are analysis software themselves — they are data sources. Once you have imagery or a data extract, you still need tools to do spatial analysis, compute vegetation or land-cover indices, or produce maps, typically through a GIS. In the adjacent case of space and astronomical data specifically, Astropy is the open-source Python library most researchers reach for: it handles FITS file I/O, coordinate transformations, units and time-scale conversions that would otherwise be reimplemented from scratch in every project, and it is close to a default dependency in professional and academic astronomy software.

A shortlist by situation

• Long-term land-cover change, decades of history: USGS EarthExplorer's Landsat archive.
• Highest-resolution recent optical or radar imagery, free: Copernicus Data Space Ecosystem.
• Broadest single archive across many NASA missions, with cloud-hosted access: NASA Earthdata.
• Weather, climate or ocean data rather than land imagery: NOAA Open Data.
• Scriptable, on-demand processing without managing raw downloads: Sentinel Hub.
• Astronomical (not Earth-observation) data analysis in Python: Astropy.

Questions to ask before you commit to a workflow

• Does the archive cover the specific sensor, resolution and time range your research actually needs, not just "satellite data" in general?
• Is a cloud-hosted or on-demand processing option available, or will you need to download and process raw scenes yourself?
• What is the realistic revisit frequency for your area of interest — cloud cover and orbit geometry mean the nominal revisit is often optimistic?
• If you need imagery across multiple missions (Sentinel and Landsat together, for example), does a single service like Sentinel Hub simplify that, or is manual cross-referencing acceptable?

Common mistakes

• Assuming higher resolution is always better. A 30m Landsat record spanning 50 years answers a different question than a 10m Sentinel-2 scene from last month; match resolution and time span to the actual research question.
• Downloading raw scenes when a processing API would do the job faster. If you only need an NDVI time series for a handful of areas, Sentinel Hub's statistical API can be far less work than scripting your own pipeline against raw downloads.
• Ignoring cloud-hosted access. Repeatedly downloading NASA or Copernicus data to a laptop for compute that could run on AWS near the data wastes real time on large projects.
• Treating "free" as "zero effort." Registration, API quotas, and the compute to process what you download are all real costs even when the data itself carries no price tag.

For head-to-head detail on two common choices, see Copernicus Data Space vs NASA Earthdata and Sentinel Hub vs USGS EarthExplorer. The full list of tools in this category is at every tool in this category.