Lead project

Who lived inside the 2025 Los Angeles fires?

After the Palisades and Eaton fires, I wanted to know which neighborhoods actually sat inside the burned areas and how much those households earned. I built a public dataset and map that answers that for every census block group in Los Angeles County, published it on Harvard Dataverse with a permanent DOI (digital object identifier), and now run a second version that records flood and mudslide warnings over the burn scars through this winter.

Role

Sole author

When

Aug 2026 (dataset); version 2 collecting since Sep 3, 2026

Mentor

Dr. Bo Zhao, Humanistic Geographic Information Systems (GIS) Lab, University of Washington

Live map

wildfire.myrondomain.com

Dataset

doi.org/10.7910/DVN/2R57BG (Harvard Dataverse, public domain)

The problem

News coverage of the January 2025 fires talked about whole cities: Pacific Palisades, Altadena, Malibu. But a fire perimeter cuts through cities unevenly. One street burns and the next one over does not, and the people on those two streets can have very different incomes, insurance and ability to rebuild.

The Census Bureau publishes income data for small areas called block groups. A block group is a few city blocks, usually 600 to 3,000 people. That is about the smallest unit where public income numbers exist, so it is the right size for asking who was inside a fire line.

Who it's for

People who decide where recovery money and outreach go: city and county staff, nonprofits, reporters, and researchers who want a clean starting table instead of building one from scratch. The map is meant for someone who has never used mapping software.

What I built

I overlaid the 19 official 2025 fire perimeters from CAL FIRE (the California Department of Forestry and Fire Protection) on all 6,587 block groups and measured what share of each block group's land was inside a perimeter. I joined that to median household income from the American Community Survey (ACS), a vegetation index from satellite images, and the federal list of disadvantaged communities.

On the map, two sliders let you choose what counts as "high fire" and "low income". The map, a scatter plot and a ranked table all re-sort as you drag. I built it that way because any single cutoff is a judgment call, and I wanted readers to see how the answer changes when they move it.

Dr. Bo Zhao, who runs the Humanistic GIS Lab at the University of Washington, mentored the project. The framing of the map follows his earlier equity mapping work, and I simplified the interface after his feedback. He is credited on the map itself.

What happened

With the default cutoffs (10% of area burned, and income under $100,000, close to the county's block-group median of $93,240), 62 block groups count as high fire. Most of them are not poor. Nine of them sit at the Census income cap ($250,000 or more). The 9 highest-priority block groups, which have high fire plus several vulnerability flags, are almost all in Altadena, inside the Eaton fire.

I published the full table, the fire perimeters and the scripts on Harvard Dataverse in August 2026 under a public-domain license, so anyone can reuse them without asking.

Version 2: warnings over the burn scars

Burned hillsides can't hold rain, so after a fire the next danger is flooding and debris flows (fast mudslides full of rock and ash). The National Weather Service (NWS) issues these warnings for whole watersheds, the area of land that drains into one creek or channel. Watersheds follow the shape of the mountains, while neighborhoods follow streets and city lines. A single warning can cover parts of several neighborhoods and miss half of another, so it is hard to say from the warning alone who was actually told to leave.

Forecasters expected a wet El Niño winter for 2026 to 2027, so I started a collector on September 3, 2026. Every 30 minutes it saves any NWS flood or debris-flow alert that touches a 2025 burn scar and records which block groups, and which incomes, fall inside the warning area. It also logs rain forecasts against measured rain, drought conditions and El Niño indices. By early October it had captured 4 alerts, all minor flood advisories. The season has barely started, so there are no findings yet, and the data can only describe who was inside a warning area, not who received it or acted on it.