Catastrophe Claims Inspection With Drones and 3D Capture
After a regional storm or wildfire, adjusters are scarce and half the properties are unsafe to walk. Catastrophe claims inspection with drones turns one flight into a walkable 3D scene, so the desk team works the loss while the field team moves to the next street.
Day four after the hail line came through. The carrier has a thin roster of adjusters on the ground for a backlog measured in thousands, half the roofs on the list are still tarped, and the street you need is behind a road closure until the utility crews finish. Catastrophe claims inspection with drones does not add adjusters. It changes what one field visit produces: a single flight that becomes a walkable scene, so the desk team works the roof while the field team is already on the next street.
When the claim count outruns the roster
The volume is not a feeling. Swiss Re Institute's sigma report of March 2026 put insured losses from severe convective storms at USD 51 billion for 2025, the third-costliest year on record for hail and damaging wind. The same report logged around USD 40 billion of insured loss from the Los Angeles wildfires, the largest insured wildfire loss in its records, a figure Munich Re's July 2025 mid-year review also reached. On the ground that meant paper. The California Department of Insurance reported 33,717 residential and commercial claims filed by 5 February 2025, less than a month after the fires started.
Every one of those files needs someone to look at the property. A cat deployment solves that by moving people, and people are the scarce part. A roof with fractured decking is not one you put a ladder against. A burn zone stays closed while crews make it safe. And the photo set that does come back answers the questions the field adjuster thought to ask that afternoon, which is rarely the question the contractor's supplement raises three weeks later: was the north slope hit too, and how many squares is that really. Then someone goes back out.
What a drone pass turns into
Flying drones over losses is not new to most CAT programmes. What catastrophe claims inspection with drones changes is the output: instead of a folder of aerial stills, the footage is reconstructed as a Gaussian Splat scene: continuous, photoreal surfaces you open in a browser and move through freely, at any angle, rather than scroll through in the order the pilot happened to shoot them.
For hail and wind, the desk adjuster orbits the roof, drops low over each plane, checks the ridge caps, vents and gutters, then swings round to the elevations for siding and window wraps. Nothing has to have been framed in advance. If the question is the slope nobody photographed, they turn the camera.
Wildfire loss documentation asks something different. The structure is often a total, so the scene is about establishing what stood and what remains: foundation footprint, the chimney, detached garages, fences and outbuildings that carry their own coverage limits, and the condition of the lot before debris removal changes it for good. A single capture holds all of it in spatial relation, which a sheet of thirty stills never quite manages.
The demands are short. Coverage in the air, a reference length for scale, and somewhere between ten minutes and two hours of cloud processing before the desk can open it.
How does catastrophe claims inspection with drones work?
A pilot flies two or three orbits around the damaged structure at different heights, plus a low pass over each roof plane, recording video or overlapping photos. The footage reconstructs in the cloud as a walkable 3D scene in 10 to 120 minutes. A desk adjuster scales it to a known dimension, measures, marks damage and shares it by link.
The detail behind each of those steps is where the time is won or lost.
The workflow, from flight to file
Flying the capture
Think of the flight as coverage, not photography. When catastrophe claims inspection with drones goes wrong, it is almost always the flight that missed something, not the processing. A reliable pattern for a single dwelling is a high orbit that keeps the whole roof and lot in frame, a second orbit lower down with the camera angled at the eaves and elevations, and then a slow pass along each roof plane with the camera pointing down. Keep speed steady and let each frame overlap heavily with the next; reconstruction needs to see every surface from more than one position.
Wildfire lots want a wider first orbit to take in outbuildings, fence lines and the neighbouring exposure, and a closer one around whatever is still standing. Fly it before debris crews arrive if the schedule allows. A typical residential capture is a few minutes of airtime, which is the point: the pilot is back in the truck long before an adjuster would have finished setting a ladder.
Fly within your drone programme's existing operating rules and approvals. Nothing about the 3D output changes what the flight itself requires.
What to upload
Drone video in MP4 or MOV goes straight in, and so do sets of drone stills. A drone video converted to a 3D scene is usually the quickest route, because continuous footage gives the reconstruction dense overlap without the pilot thinking about shot spacing. Ground-level phone photos of the elevations or the interior can be added from the same visit. Where a carrier or vendor already scans large commercial losses with LiDAR, point clouds in E57, LAS, LAZ or PLY can be reconstructed the same way through the managed LiDAR to 3D service.
If you have battery for one more minute, spend it on a second low pass over the most damaged plane from the opposite direction. Gaps show up as soft or missing surfaces in the scene, and they are always on the slope you will be asked about.
Processing while the crew moves on
Processing runs in the cloud, with no GPU in the truck and nothing installed on the adjuster's laptop. A single house sits toward the faster end of the 10 to 120 minute range; a large commercial roof or a whole block toward the slower end. Upload from the first property, fly the second, and the first scene is usually ready before the third. Firms running volume can push captures through the reconstruction API instead of uploading one at a time.
Desk triage
This is where storm damage assessment moves off the roof. The desk team opens the scene in a browser and sorts the queue: clear total, clear denial, needs a closer look. Hotspots go straight onto the finding, carrying a note, a ground photo, a video clip or a link to the claim record, so the next reviewer lands on the same spot rather than hunting for it. A file that once waited on a field adjuster's next free afternoon now waits on someone at a desk, and the field team never had to stop.
Where an earlier capture of the same property exists, from an underwriting inspection or a prior claim, Multi Splat aligns the two and renders the difference, which turns "was that damage already there" from an argument into a comparison. The same desk-review pattern, applied to utility and infrastructure assets, is laid out in remote site inspection with 3D capture.
Scaling to a known dimension
A reconstruction has shape but not size until you give it one. Pick something in the capture whose true length you know: a ridge length from the prior measurement report, an eave edge taped on site, a standard garage door the field adjuster measured on the way in. Set that length in the scene and everything else scales with it. From then on, distances are read directly in 3D, so a disputed ridge run or a gutter length is a measurement off the capture rather than a return trip.
Use the longest reliable dimension you have, ideally one you measured yourself, and record what it was in a hotspot. No accuracy tolerance has been published for scene measurements, so treat them as the desk's working figure and keep your measurement report as the number of record.
Sharing with reinsurers and contractors
This is the part of catastrophe claims inspection with drones that reaches beyond the carrier. The finished scene goes out as a link, an embed in the claims system or a QR code, and opens on a phone, tablet, desktop or VR headset with no app. A reinsurer reviewing a large loss walks the same roof the desk adjuster did. A contractor disputing scope can see the slope they claim was hit and the one that was not. Enterprise accounts can serve it under the carrier's own brand and domain, and an offline viewer opens scenes without a connection for teams still working where the network is down.
Objections a CAT team will raise
Is this accurate enough to settle on? It is accurate enough to triage on and to settle most scope questions by sight. For the dimensions that decide payment, keep your measurement report as the record and use the scene to check and explain it. Precision depends on the reference and the flight, and we do not publish a tolerance we cannot stand behind.
Does it replace the field adjuster? No. Someone still has to be on site, fly or direct the flight, talk to the policyholder and see what the air cannot, like interior water damage. Catastrophe claims inspection with drones moves the review, not the visit. One field adjuster's day covers more addresses because the slow part, the looking, happens elsewhere.
What about no signal in the event zone? Capture records to the drone. Upload waits until the truck is back in range, which in a regional event might be the staging area at the end of the day. The workflow tolerates that delay; only the processing clock starts later.
Who has to learn something new? The pilot learns a flight pattern built for coverage. The desk opens a link. Neither is a new software discipline, and reinsurers and contractors need nothing at all.
What happens to the source footage? It stays where it was, unchanged. The scene is built from it, not instead of it, so the original video and photos remain available if a finding needs to go back to raw evidence.
| Field inspection with photos | Drone capture with a 3D scene | |
|---|---|---|
| Time at the property | Ladder setup, roof walk if safe, photo log | A few minutes of flight, no roof access |
| Unsafe or closed sites | Wait, or document from the street | Covered from the air within operating rules |
| Who reviews | The adjuster who visited | Anyone with the link, from a desk |
| Contractor supplement | Often a re-inspection | Turn the camera to the disputed slope |
| What a reinsurer sees | A report and selected stills | The same walkable scene the desk used |
| Measurement | Tape or a separate report | Scaled from a known dimension, read in 3D |
FAQ
How long after the flight can the desk team review the property?
Processing takes 10 to 120 minutes once the footage uploads, scaling with the size of the structure rather than the input format. A single-family roof is usually toward the faster end. In practice the gating factor during a regional event is upload, since crews often wait until they reach the staging area or a working network.
Can a drone capture document a wildfire total loss?
Yes, and it suits it well. A total loss is mostly about establishing what stood, what remains and how the lot was laid out, including detached structures and fences. A wide orbit plus a closer one around standing elements records all of it in one scene, ideally before debris removal alters the site.
Do reinsurers or contractors need an account or software to open the scene?
No. The scene opens from a link in an ordinary browser on a phone, tablet or desktop, with nothing to install. It can also be embedded in a claims platform or shared as a QR code, and enterprise accounts can serve it under the carrier's own branding and domain.
What should we use as the scale reference on a roof?
The longest dimension you trust. A ridge length from a prior measurement report, an eave edge taped during the visit, or a standard garage door measured on site all work. Record which one you used in a hotspot, so anyone reviewing the measurement later knows exactly where the scale came from.
The roof from day four still needs a decision, and a person still makes it. Catastrophe claims inspection with drones only moves the looking. What changes is where that person sits and what they are looking at: a scene of every slope, scaled to a ridge someone actually measured, open on the same screen as the claim file, while the pilot who captured it is three streets further into the list.