Field notes: two storms crossed paths and the map couldn't say which hit first

Dramatic dark storm clouds rolling over a suburban neighborhood with tall trees and houses

We had a county last spring that got hit twice in six hours. A squall line came through around 9pm, dropped straight-line winds in the 70s, moved on. Then a second cell, spun off the same front, tracked almost the identical path around 2am with gusts that topped 90 in a couple of spots. Same county, same towns, same rooftops. Two separate catastrophe codes on two separate claims, and from the ground there was no way to tell which storm did what.

This matters more than it sounds like it should. Reserves get set per event. Reinsurance treaties attach per event. If a roof failed in the first pass and the second pass just finished the job, that's a different story for the carrier than if it held through the first and failed outright in the second. Adjusters walking the block three days later are looking at one pile of shingles and two incident reports, and they have to guess.

We tried the usual things. Wind field models from both cells, overlaid. Radar loops, frame by frame, trying to time the worst of the rotation against known loss addresses. 911 call logs, which are useful for "something happened here" but useless for "which storm did it." None of it gave us a clean line. The wind fields overlapped almost completely for an eight-mile stretch, which is exactly the stretch where the damage was worst, which is exactly the stretch where it mattered most to know.

What we ended up leaning on, more than I expected going in, was the imagery timing itself. If you've got a pass that landed between the two events, even a rough one, you can at least establish a before-and-after for the first storm and a different before-and-after for the second. It's an imperfect answer: a timestamp that lets you say "this tree was down and this roof was intact at 11pm," which narrows the argument considerably without closing it. On a case with two codes stacked on top of each other, a timestamp is worth more than another model run.

This ambiguity costs you on deployment decisions as much as it does on claims decisions, and that's the part nobody talks about enough. If you think it's one event, you send one wave of adjusters at one footprint and call it done. If it's actually two footprints with two different severity profiles, you've potentially got your worst crews sitting on moderate damage from event one while event two's actual bad blocks wait their turn. We didn't figure that out until day four, when a supervisor noticed that one neighborhood's claim volume kept climbing well after the "storm" was supposed to be over. Turned out it wasn't the same storm.

I don't have a clean fix for the attribution problem. Weather doesn't arrange itself for the convenience of claims handling, and two cells on the same track in one night is going to keep happening on the Plains and anywhere else storms stack up along a front. What I'd tell anyone running cat response on a multi-event night is to get eyes on the footprint as early and as often as the imagery allows, because every pass you have is another timestamp you can argue from later. We've started pulling per-building damage grades over the whole footprint within days of the event specifically so we're not relying on a single snapshot taken whenever the first adjuster happened to drive down that street.

It won't tell you which storm hit first. But it'll tell you where the worst of it landed, and that's usually the question that decides where the trucks go.

If you're staring down a footprint with more than one storm in it, see what Windstorm Damage can get you before the first truck rolls.

← Back to the blog