Every hour of dangerous heat since January 2024, at every census tract in the Lower 48, from NOAA's hourly 2.5 km weather analysis, counted for the people who live there. Is the heat shared evenly between rich and poor neighborhoods, older Americans, and households without a car to get somewhere cooler?
Poorest vs. richest fifth uses national income fifths, so a state's richest fifth is its neighborhoods in the country's top fifth. States with few tracts in a fifth show "—".
The heat. NOAA's Real-Time Mesoscale Analysis (URMA) blends observations into hourly temperature and humidity every 2.5 km. The heat index is computed hour by hour with the Weather Service's formula. Each census tract takes the grid cell at its center. The same data powers the Threshold Hours exit.
The people. Census ACS 2020–2024: population, residents 65 and older, people below the poverty line, households without a vehicle, and median household income for every tract. Tracts are split into five income groups nationally, each holding a fifth of all households. Every number here is an average over people: hours at each tract times the people there.
Reading it honestly. Most of the gap between rich and poor neighborhoods is geography. The South and the desert Southwest are both hotter and, on average, poorer. The within-county view strips that out. At 2.5 km the analysis still smooths over much of the block-by-block urban heat island (pavement versus tree cover), so the true neighborhood difference is probably larger than the within-county numbers show. The heat index is also computed in the shade: full sun adds up to 15 °F.
The trees. The U.S. Forest Service's Tree Canopy Cover for the 2023 National Land Cover Database estimates the share of every 30 m pixel under tree canopy. It's averaged over each census tract's land, about 1 km at a time (tracts smaller than that read the kilometer around their center), then over residents the same way as the heat. Canopy is one of the strongest levers on street-level heat: shaded pavement runs far cooler than sunlit pavement. The 2.5 km heat analysis mostly can't see that, which is why the third chart's difference is small even where the canopy gap is large. The 30 m satellite product also undercounts scattered street and yard trees, so city canopy reads lower here than in local lidar surveys (Washington, D.C. shows about 20% against roughly 38% from lidar). The comparisons between neighborhoods hold up better than the absolute numbers.