How the Oregon fire maps failed

The gap between the line and the ground

The perimeter published on September 3 placed the fire’s western edge a full ridge east of where it burned. Ground truth from seventeen drone flights tells a different story, one with consequences for every evacuation model built on the official line.¹

Fire seasons in the western United States have lengthened measurably over the last four decades, a shift attributed in large part to anthropogenic climate change (Abatzoglou & Williams, 2016). As seasons lengthen, the mapping systems built for shorter, slower fires are asked to do more than they were designed for.

This is a demonstration of bold, italic, and inline code, plus a standard link — the ordinary Markdown an author reaches for first.

What the drone surveys show

Burn severity was classified from multispectral imagery using the Normalized Burn Ratio, following the landscape-assessment protocol established for FIREMON (Key & Benson, 2006).

The perimeter itself, mapped across three synthetic observation days, is served as GeoJSON straight from the dataset this article ships with:

Growth over the first two weeks follows the familiar early-exponential curve before containment lines hold:

Sequence of events

1. Sep 1, 2025

   Ignition reported

   Initial attack within four hours; perimeter mapping begins the following morning.

2. Sep 3, 2025

   Official perimeter diverges

   Published line places the western edge one ridge east of drone-observed fire.

3. Sep 5, 2025

   First containment

   Containment reaches 6% as the synthetic growth curve begins to flatten.

4. Sep 14, 2025

   Survey series complete

   Seventeenth and final drone flight closes the observation record.

The data behind this article

Every investigation publishes its evidence. The perimeter dataset is documented, licensed, checksummed, and citable on its own landing page:

Reproducing the growth chart takes a few lines — code blocks get Shiki highlighting on the obsidian surface automatically:

import geopandas as gpd

perimeters = gpd.read_file("perimeters.geojson")
perimeters["area_ha"] = perimeters.to_crs(epsg=5070).area / 10_000
print(perimeters[["date", "area_ha"]])

Observation days summarize into an ordinary Markdown table:

Live context

Related monitoring lives in a self-contained dashboard, embedded here and maintained independently of the article:

A perimeter is a claim about the world. Like any claim, it can be checked.

The divergence documented here is small in hectares and large in consequence: evacuation zones, insurance assessments, and post-fire funding all inherit the official line’s errors.