Planet makes available select imagery for major disaster events, including major earthquakes, floods, storms, wildfires, and human-made disasters. These are available here in a STAC Catalog under a CC-BY-NC license. See the Planet Disaster Data page for more information.
You are looking at a Portolan / STAC collection (collection.json, STAC 1.1.0, Portolan profile v0.1.0) — inside Disaster Data Releases from Planet Labs PBC: Planet Crisis Response imagery for major disaster events.
Very-high-resolution (~0.64 m) SkySat collects acquired 13 August 2026 at 12:01 UTC, three days after the mainshock, over Quibdó — the capital of Chocó and the largest population centre in the affected department, roughly 105 km north-north-west of the epicentre. Two adjacent collects covering the city and the Atrato river frontage, extending about 6 km further south than the 10 August coverage of the same target. Visual, pansharpened, and usable-data mask products for building-level and infrastructure damage assessment. 82% of this footprint was also covered by the 10 August collects, so the pair gives a same-sensor, same-target three-day repeat at sub-metre scale. Atmospherics are the limiting factor here: cloud is only 12%, but UDM2 puts light haze over 84% of the valid pixels and just 4% of each collect is fully clear. The 12:01 UTC pass is 07:01 local, giving a 13.6 degree sun elevation — long shadows across the whole scene and deep shading on west-facing walls and narrow streets. Both effects work against automated differencing; treat this as an interpretation-grade image over Quibdó rather than a radiometric one, and prefer the 10 August collects as the reference frame. 2 scenes; each Item carries the available visual, pansharpened, udm2, and udm assets.
Everything is plain HTTPS — no credentials, no signing, no request-payer.
Two caveats worth knowing before you write a fetch loop:
Python-urllib User-Agent. Send a
browser or curl UA if you use urllib; requests, curl, GDAL/rasterio, rio-tiler and
DuckDB all work as-is.aws s3 ls; navigate the STAC
links instead — every object links to its parent, children and items.Each collection publishes a STAC-GeoParquet mirror of its items — one row per item, with geometry and all properties. Query it directly, no download:
That is almost always faster than walking the rel:"item" links, which are there for
STAC clients and for humans clicking through the browser.
Every raster is a cloud-optimized GeoTIFF, so read the window you need rather than the file:
Asset keys on each item (visual used above):
visual — Visual (true-colour RGB) ortho COG (roles: data, visual)pansharpened — Pansharpened multiband ortho COG (roles: data)udm2 — Usable Data Mask (UDM2) (roles: data, snow-ice, cloud, cloud-shadow)udm — Unusable Data Mask (UDM) (roles: data, snow-ice, cloud, cloud-shadow)file:size and file:checksum (sha256 multihash, 1220-prefixed) so a download can be verified.proj:code.