Before dawn on Saturday 9 August 2025 a storm crossed south-east Nebraska, and about 70,000 people woke up with no power. The global weather record caught none of it. RE-FOCUS rebuilt the night at 3 km, hour by hour, from what people saw.
RE-FOCUS is an AI weather model that rebuilds past storms at the scale they happened: hour by hour on a 3 km grid, anchored to the station readings and storm reports each storm left behind. It doesn't replace your risk model. It gives that model a sharper history of what happened.
Here is one night it rebuilt.Watch the night →
The storm came in from the west before dawn and reached Omaha a little after six. Along the way, weather stations, storm spotters and emergency managers logged it as it passed. Here is every report, at the minute it was logged.
Every wind report NOAA's Storm Prediction Center logged in this area between 3 and 7 a.m., with the observers' own notes. Times are local.
Trees came down across Seward, Crete, Lincoln and west Omaha, on houses, cars and power lines. Barns and grain bins went in the fields between. By sunrise the storm was in Iowa.
Claims, reserves, recoveries and triggers all turn on the same thing: what the wind did at each place, and when. The place most people look first is the global weather record.
The global weather record is drawn in squares about 31 km wide. A thunderstorm is a few kilometres across. Of the 121 extreme-gust reports that night, it caught none. Drag across the map to see the same night rebuilt.
The global record · 31 kmRE-FOCUS · 3 km
Darker is a stronger gust. Circles are the storm reports: red where the record's gust was too low, blue where it was too high, white where it was close. The rebuilt night has the reports taken in.
Four records of the same storm, each scored against the same 121 extreme-gust reports. A report counts as caught when the record's gust at that spot is within 15 mph of it.
It runs as 20 versions of the storm, so you see the most likely gust and the spread around it.
Its typical gust error before any reports went in, against 50 mph in the global record.




The two-thirds is scored against the reports the run took in. Stations from ASOS, AWOS and METAR; the high-resolution model is NOAA's HRRR; the global record is ECMWF ERA5.
The public record missed every report that night. Rebuilt from what people saw, RE-FOCUS puts the storm back where it was.
The same night, scored against the same 121 reports.
The global record · 31 km
RE-FOCUS · 3 km, reports taken inWhat did the wind do at this address, and when?
The record has no storm at the address. What's left to reserve and settle on is a regional average.
An hourly gust record for every 3 km cell the storm crossed. Reserve and settle on what happened at the address, and size the loss while the claims are still coming in.
Did our event history ever count a night like this?
A night the record missed is a loss your event history never counted.
Check your event set against the storm as it actually happened. The rebuilt footprint lets you count it.
What was the gust at the site, inside the contract's window?
A wind trigger is only as good as the record it settles on, and this record never saw the storm.
Verify the trigger on what the storm did at the site, hour by hour, with the spread across 20 versions of the storm on the table for the basis-risk conversation.
Single storms, rebuilt one at a time at 3 km and scored on gusts against storm reports and stations. The 67% is scored against the reports the run took in. They verify and settle events; they are not yet a multi-decade US wind record to price on.
Hourly gusts on the 3 km grid for every version of the storm, with the mean, the spread and the extremes, in a standard gridded file. It enters your model at the hazard step; vulnerability, terms and the model itself stay yours. Your model keeps its job. It gets a sharper history.
A working session is an hour with the people who rebuild them, on an event you already know.