Fire season active. Emergency resources & evacuation info →
FEATURED: Human emissions have likely delayed the next glacial period by tens of thousands of years. Read the full story.

Smoke Kills More People Than Fire

Wildfire smoke is the fastest-growing air quality crisis in the western United States.

Smoke kills more people than fire does. An estimated 1,500–2,500 people die each year from wildfire smoke exposure in the western US (Burke et al. 2021, PNAS), more than die in the fires themselves. In 2020 alone, the Labor Day fires killed 11 people directly but wildfire smoke contributed to an estimated 6,300–8,800 excess deaths nationwide (O’Dell et al. 2023, GeoHealth).

Wildfire PM2.5 is not ordinary air pollution. A Southern California study of respiratory hospital admissions found it can be up to 10x more harmful to respiratory health per microgram than equivalent ambient PM2.5 (Aguilera et al. 2021, Nature Communications), a regional finding on respiratory outcomes, not a universal toxicity multiplier. The economic toll: $11–20 billion per year in health costs (Fann et al. 2018, EPA).

2,500
Excess deaths/year
Western US, wildfire smoke
Up to 10x
Respiratory harm vs ambient PM2.5
So. California admissions study (Aguilera 2021)
$20B
Annual health costs
EPA estimate
12K+
PurpleAir sensors
Snapshot, western corridor

AQI snapshot from PurpleAir sensors. PurpleAir refreshes daily; for live conditions see fire.airnow.gov. Toggle "Smoke (historical fit)" to overlay a 2017–2024 PM2.5 transfer-function model (R²=0.19); useful for historical context, not as a current forecast.

EPA Air Quality Index (AQI)
0–50
Good
51–100
Moderate
101–150
USG
151–200
Unhealthy
201–300
Very Unhealthy
301+
Hazardous

What to do at each level

Good (0–50)Normal activity.
Moderate (51–100)Unusually sensitive people: consider reducing prolonged or heavy outdoor exertion.
USG (101–150)Sensitive groups — children, elders, pregnant people, anyone with heart or lung conditions — reduce prolonged outdoor exertion.
Unhealthy (151–200)Everyone: reduce prolonged exertion. Sensitive groups: avoid it. Consider an N95 outdoors.
Very Unhealthy (201–300)Everyone: avoid outdoor exertion. Stay indoors with filtered air. N95 if you must be outside.
Hazardous (301+)Everyone: remain indoors with filtered air, seal gaps around doors and windows, and leave the area if you cannot maintain clean air.

During multi-day smoke events, cumulative exposure matters, even at moderate levels, take clean-air breaks (a filtered room, library, or community clean-air shelter) when smoke lingers for days.

USG = Unhealthy for Sensitive Groups. During major events AQI can exceed 500. Portland hit 500+ during the 2020 Labor Day fires.

Protect Yourself From Smoke

  • Only NIOSH-approved N95 or P100 respirators filter PM2.5. Surgical and cloth masks do not protect against wildfire smoke.
  • Create a clean-air room: a HEPA purifier, or a DIY box-fan filter (Corsi-Rosenthal box, roughly $40–100), in the room where you spend the most time.
  • Set HVAC to recirculate so it stops pulling smoky outdoor air inside.
  • Check live conditions at fire.airnow.gov and see the full preparedness guide at /prepare.
Major Smoke Events (2017–2023)Peak AQI · Western corridor
2020Labor Day firesAQI 500OR, WA, CA2018Camp FireAQI 500CA2017PNW fire seasonAQI 400OR, WA2021Bootleg/Dixie firesAQI 350OR, CA2023Canadian smokeAQI 200MT, ID, WA
Worst Smoke-Impacted Counties2015–2025 · % days exceeding USG threshold
Nez Perce, ID47.7%Kootenai, ID44.7%Amador, CA43.8%Latah, ID43%Twin Falls, ID41.6%Moffat, CO40.4%Kings, CA35.4%Uinta, WY35.1%Tuolumne, CA34.7%Bonner, ID32.1%Idaho, ID30.9%Sweetwater, WY27.6%Walla Walla, WA27.5%Lincoln, WY27.4%Glenn, CA26.8%

EPA AQS annual PM2.5 summaries (parameter 88101), ~700 monitors, 11 western states. PM2.5 > 35.5 μg/m³.

Smoke Layer (Retrospective Analysis)Historical fit

The Smoke Impact layer on the dashboard map is a retrospective fit, not a forecast. It attributes historical PM2.5 exposure to nearby fires using 315,000 daily EPA PM2.5 observations (Jun–Oct 2017–2024) paired with 800 MTBS fire centroids by date and distance. Power law transfer function with weak fit (R² = 0.19, RMSE = 24.1 μg/m³), useful for historical context, not for predicting future smoke. For live smoke and AQI forecasts use fire.airnow.gov or NOAA HRRR-Smoke.

Community Monitoring Network

The AQI layer on this page is a snapshot from 12,000+ PurpleAir citizen science sensors across the western corridor, roughly 20x the density of the 700-monitor federal network, filling critical gaps in rural fire-prone communities. PM2.5 values use the EPA 2021 correction equation. For live conditions, use fire.airnow.gov.

2026 Smoke Risk OutlookAs of 2026-06-12

18 uncontained large fires burn across 8 states; 77 active incidents in the western corridor (WFIGS). Snowpack at record lows across 8 of 11 corridor states. El Niño developing with 80% probability by summer, 25% chance of Super El Niño (≥+2.0°C). Key smoke risk factors:

Oregon: Zero snowpack. Willamette and Rogue valleys are inversion-prone. Portland’s 2020 AQI 500+ is the template.
California: Central Valley inversions trap Sierra fire smoke. Sacramento and Bay Area chronically exposed.
Colorado/Utah: SWE at 0–20%. Front Range and Wasatch communities face expanding WUI fire risk.
Idaho: 43–48% smoke days in worst counties. Boise unhealthy air every year since 2017.
Policy Failures
Clean air shelters: Most rural fire-prone counties have zero. EPA recommends MERV-13; most buildings have MERV-8.
Worker protections: Only CA, OR, and WA have enforceable smoke rules. 8 of 11 corridor states have nothing.
No forecasting mandate: HRRR-Smoke exists but isn’t integrated into emergency management. PurpleAir fills the gap.
Insurance ignores smoke: No insurer accounts for cumulative smoke burden despite 2–5% property value impacts.

AQI snapshot: PurpleAir (12,000+ sensors, EPA-corrected); live AQI at fire.airnow.gov. Historical PM2.5: EPA AQS (2015–2025, 11 states, ~700monitors). Health: Burke et al. 2021, O’Dell et al. 2023, Aguilera et al. 2021, Fann et al. 2018. Smoke layer: retrospective fit, 315K daily observations × 800 MTBS fires, R²=0.19. Historical attribution only, see methods.