Questions, corrections, or feedback on the data? Drop a line.
What you're looking at, where the numbers come from, and how the map is built.
Every point comes from the NHTSA Fatality Analysis Reporting System (FARS), the U.S. census of crashes in which someone died. It covers fatal crashes only — crashes with no fatality are not in FARS and are not shown here. This map uses FARS years 2019–2023 for crashes located on US Interstate (I-) and US numbered (US-) highways. FARS is a public-domain U.S. government dataset.
Each highway's through-line is built from OpenStreetMap geometry: the route is walked along its straightest continuation and divided carriageways are merged into one line, so a divided interstate is a single road rather than two. Every FARS crash is then projected onto that line, giving its distance along the route. That distance — not raw latitude and longitude — is what positions a crash on the chart.
The route is split into equal-length bins. Each bar is a stack of person figures — one figure per death — for the fatal crashes whose distance falls in that bin; busy stretches stand their figures in several columns, like a crowd in rows. The busiest bin sets the height scale. Bin length adapts to how much road is on screen (the Auto setting), or you can fix it with the slider. Click any bar to drill into that stretch: a breakdown by year, time of day, day of week, crash type, light and weather, plus a Street View of the location.
The year buttons and the time-of-day dial filter the crashes before they are binned, so the bars and color always reflect the slice you have selected.
Bars are counts of deaths, not a risk rate. They are not adjusted for traffic volume, so a busy interstate shows more crashes than a quiet one simply because more vehicles pass through. This is a one-dimensional view along each route, not a two-dimensional geographic heatmap, and it excludes every non-fatal crash.