SATELLITE DATA PORTALS · 1/9 · OVERVIEW
Satellite data portals�for policy research
What exists, what it costs, how you sign up, and whether you get pictures or measurements.
Source archives (NASA, USGS, Copernicus, ISRO, China), derived products (GHSL, WorldPop), and compute platforms (Google Earth Engine)
Abhijit Vaidya · Panel: Role of Alternative Data for Policy Research · Alt Data Workshop 2026
SATELLITE DATA PORTALS · 2/9 · TILES OR FULL IMAGES?
Four ways you get the data
1 · Viewer tiles. PNG or JPEG map tiles served over WMTS or WMS. Instant, usually no login. But these are rendered pictures, not measurements. (NASA Worldview, Bhuvan viewers)
2 · Full scenes and granules. The scientific product: GeoTIFF, HDF5 or NetCDF files with metadata and quality flags. This is what analysis needs. (Earthdata, EarthExplorer, Copernicus, Bhoonidhi)
3 · Cloud-native. Cloud-Optimized GeoTIFF plus STAC catalogues. You stream only the area and bands you need, so you get full data with the convenience of tiles. (Landsat and Sentinel on open buckets)
4 · Compute-to-data. Your code runs next to the archive and you export results. The data never leaves. (Google Earth Engine)
A mature archive offers all three surfaces: a viewer for people, a catalogue and API for machines, and scientific files for reproducibility.
SATELLITE DATA PORTALS · 3/9 · PICK BY TASK
Pick by what you actually need
If you want to see something (an event, a place, a change): use the viewers. NASA Worldview, Copernicus Browser, Bhuvan, MOSDAC RAPID. No login, minutes.
If you want to measure something (radiance, extent, an index): go to the source archives. Earthdata, EarthExplorer, Copernicus Data Space, Bhoonidhi. Registration, full granules.
If you want to count people, buildings or built-up area: use derived products. GHSL, WorldPop, HRSL, building footprints. Often no signup at all.
If you want to scale the analysis (national, multi-year): use a compute platform. Google Earth Engine, Planetary Computer, open cloud buckets.
Most policy questions get answered at the third level, with a documented derived product, without ever touching raw imagery.
SATELLITE DATA PORTALS · 4/9 · GLOBAL SOURCE ARCHIVES
Global portals and platforms
Portal | What's there | Cost / signup | Tiles or full images |
NASA Earthdata + Worldview | All NASA EO: MODIS/VIIRS, rainfall, fires, nighttime lights (Black Marble) | Free; browse without login, downloads need free Earthdata Login | Both. Worldview serves tiles (GIBS); the archive serves full granules (HDF5, NetCDF). |
USGS EarthExplorer | Landsat archive back to 1972, the longest Earth observation record. Free since 2008. | Free; free account for downloads; M2M API | Full scenes, now Cloud-Optimized GeoTIFF; also on AWS open buckets |
Copernicus Data Space | Sentinel-1 SAR, Sentinel-2 10m optical, S-3, S-5P air quality | Free account; quota-controlled | Both. Browser plus full products via STAC, S3, OData or openEO |
NOAA on AWS | GOES, Himawari full-disk weather imagery, minutes after acquisition | Free. Open S3 buckets, no signup | Full granules |
EOG (Colorado School of Mines) | VIIRS nighttime light monthly and annual composites. The standard source for lights-and-economics work. | Free; registration required | Full analysis-ready GeoTIFF rasters |
SATELLITE DATA PORTALS · 5/9 · INDIA
Indian portals (ISRO and IMD)
Portal | What's there | Cost / signup | Tiles or full images |
Bhuvan (NRSC) | India geoportal: thematic layers (LULC, water, hazards), 2D/3D viewers | Free browse; download terms vary by product | Mainly tiles / WMS; selected layers downloadable |
Bhoonidhi (NRSC) | The archive: Resourcesat, EOS, RISAT, OceanSat + Landsat/Sentinel mirrors | Free registration. Products at 5 m and coarser are free. Finer than 5 m (Cartosat) is priced for non-government users. | Full scenes; HTTPS + API delivery |
MOSDAC (SAC) | INSAT-3D/3DR/3DS weather, ocean missions; near-real-time products | Free registration; some NRT products need approval | Both. The RAPID viewer needs no login; HDF5 and NetCDF granules need an account. |
VEDAS (SAC) | Application dashboards: agriculture, water, environment | Mostly open; some apps/APIs need a key | Dashboards / WMS, not a raw-scene archive |
IMD (mausam.imd.gov.in) | 40+ Doppler radars, nowcasts, warnings | Free, no signup | Neither. PNG images on a webpage. No API, no archive. |
SATELLITE DATA PORTALS · 6/9 · CHINA
Chinese open-data programmes
Portal | What's there | Cost / signup | Tiles or full images |
Fengyun / NSMC | FY-4 geostationary and FY-3 polar weather satellites. FY-2H sits at 79 degrees E, over the Indian Ocean. | Free registration (WMO open-data commitment) | Full-disk L1 granules (HDF); CMACast broadcast |
CASEarth Big Earth Data | CAS open-science datasets: global 30m land cover (GLC_FCS30) etc. | Free registration | Full raster datasets |
SDGSAT-1 Open Science (CBAS) | SDG-dedicated satellite: 40m Glimmer night imager, 10m multispectral, 30m thermal | Free for approved research. You submit a proposal, a committee reviews it, and you sign a data agreement. Redistribution is restricted. | Full Level-4 scene packages (300×300 km) |
CNSA-GEO | Gaofen GF-1/GF-6 16m wide-field optical, opened internationally | Free registration | Full scenes |
SATELLITE DATA PORTALS · 7/9 · DERIVED PRODUCTS
Analysis-ready derived products
Product | What it is | Access | Typical policy use |
GHSL (EC-JRC) | Global built-up surface, population, settlement classes, 1975 to present | Free, no signup; GeoTIFF downloads | Urbanization trends, settlement growth |
WorldPop (U. Southampton) | 100m gridded population - census disaggregated via Random-Forest covariates | Free, no signup; GeoTIFF | Denominators for per-capita indicators |
HRSL (Meta + CIESIN) | ~30m population spread over AI-detected buildings | Free via HDX / AWS | Fine-scale settlement location |
Google Open Buildings | Building footprints across the Global South, per-polygon confidence score | Free CSV/parquet, no signup | Housing stock, infrastructure planning |
Microsoft footprints / Overture | Global building polygons, places, roads | Free (GitHub / open parquet on S3) | Base layers for any urban analysis |
OpenStreetMap (Geofabrik) | Crowd-mapped streets, POIs, buildings; completeness varies by area | Free extracts, no signup | Street networks, amenities, accessibility |
The trade-off: these are easier to use, but there are more modelling assumptions inside. Ask for the published lineage, the version, and the uncertainty before you build on one.
SATELLITE DATA PORTALS · 8/9 · COMPUTE PLATFORMS
Compute-to-data platforms
Platform | Model | Access (2026) | Caveats |
Google Earth Engine | Petabyte catalogue (Landsat, Sentinel, VIIRS, GHSL, WorldPop) + server-side compute | Non-commercial unpaid: verified Google Cloud project + quota tier; commercial paid | Quotas, account governance, platform dependence; pin dataset versions |
Microsoft Planetary Computer | STAC catalogue of COGs + hosted compute | Free; light signup / API key | Catalogue breadth below GEE; evolving service tiers |
AWS / GCP open-data buckets | Sentinel-2, Landsat, Himawari, GOES as raw objects (COG/HSD) | Free, no signup; bring your own compute | You pay for your own compute/egress next to the data |
Copernicus openEO / processing | Server-side processing APIs inside the Copernicus Data Space | Free account; quota-controlled | Sentinel-centric; quotas for heavy jobs |
Compute-to-data removes the download problem and replaces it with account, quota and governance problems. Reproducibility needs pinned dataset versions and exported metadata.
SATELLITE DATA PORTALS · 9/9 · SUMMARY
The gates
Every archive sits somewhere on this ladder:
• no-login bucket (NOAA on AWS)
• free registration (Earthdata, EOG, MOSDAC)
• quota or verification (Google Earth Engine, Copernicus)
• priced tier (Cartosat below 5 m)
• proposal review (SDGSAT-1)
Each gate is a policy choice. “Open” without the gate stated is not a description.
Three things to hold on to:
• A dashboard is not a dataset. Research needs data, metadata, versions, and an API underneath it.
• Resolution is not accuracy. Publish the scale at which the quantity was actually observed.
• Derived products (GHSL, WorldPop, building footprints) answer many policy questions without raw imagery, as long as lineage and uncertainty are published.