Mapping to support the Venezuela earthquake response

Dear all,

As probably everyone already knows, two major earthquakes struck Venezuela two days ago.

I haven’t found any discussion about this yet, and I’d like to highlight how important it is for the authorities and people working on the ground to have accurate maps of the buildings in the areas affected by the earthquakes.

A couple of mapping tasks have been already activated in the HOT Tasking Manager: https://tasks.hotosm.org/explore?omitMapResults=1

The Caracas and La Guaira areas have been identified as the most affected, so I think that any mapping effort in those areas could be helpful.

5 Likes

There will probably be more projects to come using Vantor imagery now on OAM.

2 Likes

Could you please better explain what you are saying with this? Is OAM > OSM? Are there news saying that Vantor imagery will we available in Id or JOSM to be used for mapping purposes? (sorry I maybe have missed it)

All I have is what I read here Slack OAM=OpenAerialMap

In other events, such data has been processed for use in HOT projects, I expect no change, but I’ve no knowledge of the usage licence that the data is covered by. Projects are usually JOSM and ID.

A coverage-gap check: some hard-hit coastal places have no remote-sensing damage assessment yet

Hi all, and thank you to everyone mapping this response.

I’m an independent volunteer, and (full disclosure up front) I put this together using
public data and automated/scripted tooling. I’m not an experienced OSM mapper myself.
I tried to answer one narrow question: which populated places have no remote-sensing
building-damage assessment yet?
The tooling took the damage products that already exist,
unioned their footprints, and tested OpenStreetMap populated places against them. Places that
fall outside all of them are blind-spots for a satellite/aerial damage layer, though
(importantly) that is not the same as “no aid has reached them” (see caveats).

One caveat about my own method, since this is the #54357 thread: I used your task area as
one of the “covered” layers, which is a rough proxy on my part. It tells me mapping is
already underway there (thank you for that), not that a finished damage assessment exists, so
I’m treating #54357 as the softest of my inputs.

Two areas stood out that don’t appear to be covered yet:

1. The eastern La Guaira coast and the Galipán hillside. #54357 covers Maiquetía/La Guaira
and Microsoft AI4G covers Catia La Mar, but the coastal road east of Caraballeda (Naiguatá,
Camurí Grande, Anare, Todasana, Caruao, Chuspa
) and the Galipán mountain villages fall
outside every remote-sensing product I could find. That is 42 of the 109 coastal places I
tested. (The products tested were Microsoft AI4G, #54357, and Copernicus EMSR884; UNOSAT’s
only Venezuela analysis is near Caracas and doesn’t reach this coast, so it neither covers nor
rules out anything here.)

2. The Carabobo-Falcón coast, near the M7.5 epicentre. Copernicus assessed Morón (a real
grading product) and delineated Puerto Cabello, but Tucacas (press reports a 5-storey
collapse, La Mar Suites), Urama (press reports homes destroyed across Morón and Urama
parishes), Chichiriviche and Boca de Aroa sit outside every EMSR884 grading AOI.
This region has only EMSR884 as a layer, so I’m less certain here, but the epicentre proximity
and the reported collapses make it worth a look.

Method (reproducible): OSM Overpass for the places; footprints from Microsoft AI4G (HDX),
the #54357 task AOI (TM API), and Copernicus EMSR884 grading AOIs (ArcGIS feature services);
ray-cast point-in-polygon on place centroids. The 46 gap points are in the attached CSV,
and the full code plus data plus instructions are here:
Venezuela 2026 earthquake — building-damage assessment-coverage gap analysis (public data + code, no PII). Run: python3 recompute.py -> covered 67 / GAPS 42. See REPRODUCE.md. · GitHub . Clone and run
python3 recompute.py and it should print covered 67 / GAPS 42.

Honest caveats: “no remote-sensing assessment” is not “no aid”; Tucacas, for instance,
had international rescue teams. The Carabobo-Falcón result is single-source. I used place
centroids (a proxy), and OSM place completeness varies. I excluded EMSR884’s full-extent
reference envelope (AOI00) so it wouldn’t falsely mark everything covered. The Zone-2 damage
details are from press reports, not an official assessment. No personal data is involved.

Indicative AOIs (pad before drawing a real task; these just bound the gap points):

  • Eastern La Guaira coast: -66.793, 10.575, -66.292, 10.644
  • Galipán / western tip: -67.144, 10.538, -66.876, 10.598
  • Tucacas / Carabobo-Falcón coast: -68.332, 10.444, -68.271, 10.945

Since I’m not a mapper myself, I’m offering this as data and analysis you can use, not as
hands-on mapping. The 46 priority points (CSV) are attached, and the full code plus data are
in the gist linked above. I’m happy to refine the data, adjust the AOIs, or answer questions
about how it was produced.
Is this useful, and if so what would help most?

Thanks again.


Attach: assessment_gaps_priority.csv (the 46 points).
Full code + data (gist): Venezuela 2026 earthquake — building-damage assessment-coverage gap analysis (public data + code, no PII). Run: python3 recompute.py -> covered 67 / GAPS 42. See REPRODUCE.md. · GitHub

1 Like

Hi @Vittorio_Adesso There’s a lot going on, with various data sources being collated. There is also a MapSwipe project to verify AI damage assessment. You may find you can get more involved by posting on the HOT Slack channel, this post and others are covering Venezuela.

Thanks Adrian, that is helpful.

Picking up your earlier point about Vantor imagery: I checked the Vantor open-data bucket and there is now a Venezuela-Earthquake-Jun-2026 collection with 25 post-event scenes (25 to 28 June), several sub-meter at low cloud. I re-ran the same coverage test against those new footprints, and it splits my blind-spot list in a useful way. Of the 46 places (42 on the La Guaira coast, 4 on the Carabobo/Falcon coast), 29 now fall inside fresh post-event imagery, so they are mappable now. 17 are still uncovered, including Tucacas and Chichiriviche. So the new imagery closes part of the gap and leaves a smaller, sharper still-dark set. Happy to share that as a CSV or GeoJSON if it is useful to whoever is collating the sources.

On MapSwipe, that makes sense and I think it is complementary: it validates AI damage assessments where products exist, whereas these still-dark places are the ones with no product to validate yet, so the two line up rather than overlap.

I tried to join the HOT Slack via the invite form, but the link I ended up getting in the email came up as expired (“this link is no longer active”). If you are able to share a new invite, that would be a help.

Thanks again.

Invite sent. Hopefully you’ll have success with this route - I’ve not tried it before.

Hi all, as of Thursday 25 June HOT commenced coordinated mapping in response to this disaster. So far all Tasking Manager projects have focused on adding building footprints in urban and peri-urban areas, which have a very low level of completeness.

  • The technical documentation for this response can be seen on the dedicated OSM Wiki page including links to pre and post event Vantor satellite imagery scenes which have been added to OpenAerialMap
  • A dedicated HOT Portal page for this response has also been added, which you might understand as a more user friendly way to access relevant links and data related to this response

So far over 400 contributors have beed adding data to OSM via Tasking Manager in response to this event. If anyone has the time or space to get ivolved as well please do select an urgent project from this dynamic Tasking Manager list to contribute towards.

2 Likes