An environmental agency is replacing a folder of satellite exports with a portal used by analysts and the public. Analysts search by place and acquisition date, inspect two source bands or calculate a temporary index, and cite the exact scene version behind a published map. The archive will grow continuously while most map views read a small window of one scene.
The pilot source file is supplied as sentinel2: a two-band, 20 m Sentinel-2 extract whose delivered object is 29,582,479 bytes. Treat it as the measured size of a representative source asset for this design exercise. The agency's planning rate is 100 new scenes per day for a 30-day month. Use decimal GB (10⁹ bytes) and report month_source_gb for the new source assets alone, before replicas and derivatives.
The portal must answer bbox-plus-date searches, provide 256 px previews of small viewports, support user-selected red/NIR expressions, let specialists download original pixels, and reproduce a map even after a corrected scene arrives. Choose one option for each design decision and write a short justification for each. The justifications are recorded and shown with the official reasoning but are not machine-scored; the choices and sizing value are scored.
The attached pilot is one regional extract, not a claim that every scene has identical size. The arithmetic is a capacity estimate under the stated planning assumption; production sizing must sample the actual scene-size distribution.
These are where the teaching is. Read them twice.
USGS Landsat documents STAC search and downloadable COG assets (https://www.usgs.gov/landsat-missions/spatiotemporal-asset-catalog-stac). The STAC specification defines geospatial/time Items and asset links (https://stacspec.org/en/about/stac-spec/). OGC API Tiles supports tiled imagery and other spatial resources (https://docs.ogc.org/is/20-057/20-057.html). The pilot size is measured from the supplied Kharita Sentinel object.
The decision is graded. What you write beside it is kept and shown next to the official reasoning when you are done — it is never scored.
How should the portal answer spatial-and-date scene searches as the archive grows?
What should be the durable source asset for interactive previews and full-resolution download?
Where should the selected scene be turned into map tiles and band expressions?
How should a corrected scene coexist with a published analysis that cites the original?
Work the problem in whatever tool you like, then enter the answers here.
month_source_gb5 scored, 0 informational
Scene discovery serves space and time
23%Raster source supports windowed reads
23%Interactive previews preserve source bands
15%Corrections preserve published provenance
23%First month of source-asset storage
15%