Some problems are invisible until someone looks.
Caelus finds humanitarian and infrastructure gaps from orbit, then hands that data, and the work of responding to it, straight to the people already there.
- NASA
- Yale
- NOAA
- Cambridge
- FEMA
- MIT
- USGS
- SkyTruth
- National Weather Service
- Stanford
- Copernicus
- GFZ Potsdam
- ICIMOD
- Red Cross
- ESA
- Princeton
- HOTOSM
- German Aerospace Center
- EPA
- PISFCC
- Humanitarian Data Exchange
- Vida Water
- NOSDRA
- ISRO SAC
Found, then
handed over.
A satellite passes over every one of these places on a fixed schedule, whether or not anyone is reading what it sends back. The work is in the reading, and in getting it to someone who can use it.
- 100,000+
- properties analysed
- 13
- organisations engaged
- 5
- continents
- 200+
- researchers using Kairos
- 40+
- years of imagery read
Infrastructure loss, historical trends, and remapping for reconstruction
Counterparts, data partners, and response networks we are working with
From the Rio Grande to the Himalaya to the Pacific
Including Cambridge, Yale, GFZ Potsdam, and the German Aerospace Center
Optical back to the Landsat archive in 1984, radar from Sentinel-1 in 2014, read together
A collapsed span, a flooded block, a field that did not come back: all of it is visible from orbit within hours. Getting that picture to the people standing in it is the part nobody is paid to do.
See the open engagementsA disaster is never only about the disaster.
A magnitude 7.0 hit New Zealand in 2010 and killed nobody. A magnitude 7.0 hit Haiti the same year and the death toll ran to six figures. Across all hazards, earthquakes, cyclones, drought, wildfire, flood, people exposed in the poorest countries die at roughly seven times the rate of people exposed in the richest.
The pattern is inside countries too, not just between them. Buildings fall where codes were not enforced. Housing sits on the slope and the floodplain because that is the land nobody else wanted. Warning, drainage and evacuation routes follow the money rather than the risk.
Which would be a problem about disasters, except that the systems built to find need carry the same slope. Satellite poverty maps are trained on a thin layer of survey data and asked to predict everywhere else, and they turn out to be much better at telling a city from a village than at telling two villages apart.
So the places hardest to see are the places the models are worst at, and they are the same places that pay for it when the ground moves.
more likely to die from the same hazard in a low-income country
UNDRR, disaster mortality 1996 to 2015
of earthquake building-collapse deaths, in anomalously corrupt states
Ambraseys and Bilham, Nature, 2011
floods matched against income distribution across 67 countries
Lindersson et al., Nature Sustainability, 2023
countries where satellite poverty maps were audited for bias
Aiken, Rolf and Blumenstock, 2023
From mapping
to recovery.
Every one of these starts the same way: something measurable is happening to people, and the record that would prove it does not exist. Closing that gap is what saves the weeks, the money, and sometimes the lives that get spent waiting for someone to notice.
- Active analysis
El Paso County colonias
El Paso and Hudspeth Counties, Texaswith Vida WaterMapping household-level water infrastructure gaps across El Paso County colonias, including Hueco Mountain Estates, with Vida Water. A second strand looks at building age as a proxy for contamination risk in older housing stock.
- Active analysis
Trishuli corridor, after the flood
Rasuwa and upper Nuwakot, NepalBuilding a glacial lake outburst flood and flash-flood analysis for the Himalayan corridor after the 26 August 2026 Trishuli flood, cross-referencing two decades of flood events near glaciers against ICIMOD's historical outburst record.
- In conversation
Testimony and evidence, Pacific Islands
Vanuatu, and Niuewith Pacific Islands Students Fighting Climate ChangeEarly conversations with PISFCC about pairing the oral testimonies they collect from frontline communities with satellite evidence of the same changes: coastal erosion, island-shape change, and cyclone impact.
- Scoping
Thar Desert
Western Rajasthan, IndiaScoping a desertification and drought analysis for western Rajasthan, where the land is greening and the water underneath it is disappearing at the same time.
- Scoping
Niger Delta
Southern NigeriaScoping gas flaring and oil spill detection across the Niger Delta, where the flares are already counted nightly from orbit and the exposure underneath them is not.
Nine metres,
thirty minutes.
- 9 m
- rise in the Trishuli, in about thirty minutes
- 72 km
- of river with settlements struck
- 682
- outburst floods already catalogued across the range
On the morning of 26 August 2026 a flood came down the Lhende, crossed into Nepal at Rasuwagadhi, and struck settlements along roughly seventy-two kilometres of the Trishuli. Seismometers on the other side of the world recorded the collapse that started it before most people in the valley knew anything had happened.
The event was observable. It was observed. What it was not was delivered to anybody standing in the water.
This is not an unstudied hazard. ICIMOD and its collaborators have been cataloguing outburst floods across High Mountain Asia for years, and the same corridors keep appearing in the record. The map below is a selection from it, 1981 to 2026.
HOVER A POINT, OR A ROW BELOW, TO READ THE EVENT
- 2026Lhende / TrishuliLhende / Trishuli · China and Nepal
- 2024Thame / ThyanboBhote Koshi (Khumbu) · Nepal
- 2023South Lhonak TshoTeesta · India
- 2022Shishper / HassanabadHunza · Pakistan
- 2021Ronti Gad / ChamoliRishiganga / Dhauliganga · India
- 2021MelamchiMelamchi Khola · Nepal
- 2016GongbatongshacuoBhote Koshi · China and Nepal
- 2013Chorabari / KedarnathMandakini · India
- 1998Tam PokhariDudh Koshi · Nepal
- 1994Luggye TshoPho Chhu · Bhutan
- 1985Dig TshoLangmoche / Bhote Koshi · Nepal
- 1981ZhangzangboPoiqu / Bhote Koshi · China and Nepal
TWELVE DOCUMENTED EVENTS · 1981–2026 · SOURCE: ICIMOD HMAGLOFDB
Writing for people
doing the work.
Short technical pieces on what each instrument can and cannot resolve, and the conversations behind the work. Not briefs, and not marketing.
What twenty-five years of this work actually looks like
A call with John Amos, who founded SkyTruth in 2001 and has spent longer than I have been alive watching the ocean from orbit. Funding, patience, one celebrity's satellite project, and the case for building the boring, public, attributable tool instead of the impressive one.
Nine metres, thirty minutes
A glacier collapsed above the Nepal border on 26 August 2026. Seismometers on the other side of the planet knew about it before the valley did. This is what that gap is made of, and what closing it would actually require.
The coastline you need is older than the radar
A community remembers a shoreline from the 1990s. Free SAR does not exist before 2014. Which instrument answers the question is not a preference, and the honest answer changes what you can promise.
Caelus is not a closed
program.
Everything we find is meant to be picked up. If you want to build a response to something on this site, or you know a place that deserves a closer look, there is a way in.
No application to read, no equity taken, no ownership claimed over what you build.