Wildfire Smoke Resilience: Forecasting AQI, Clean-Air Shelters, and Public Health Response

Smoke has become a distinct hazard from the fire itself

Wildfire smoke can travel hundreds of kilometres from an active burn and blanket cities that face no direct fire risk at all — the 2023 Canadian wildfire smoke events that pushed New York City's Air Quality Index above 400 are the clearest recent illustration. Because the hazard is airborne pollutant exposure rather than flame, the response infrastructure is entirely different from evacuation planning: instead of moving people away from the danger, cities build clean-air refuges, distribute filtration, and time public warnings around smoke plume arrival.

The core pollutant of concern is PM2.5 — particulate matter smaller than 2.5 micrometres, small enough to penetrate deep into lung tissue and cross into the bloodstream. Wildfire smoke also carries carbon monoxide and a mix of volatile organic compounds, but PM2.5 concentration is what drives most AQI-based public health guidance.

Estimating downwind AQI and warning lead time

A simplified downwind smoke-exposure estimate scales fire size, attenuates with distance, and adds a wind-driven transport term: smoke intensity rises with the burned area (more fuel consumed, more particulate emitted) and falls off with distance from the city (dispersion and settling reduce concentration over range), while wind speed independently raises the AQI estimate because faster-moving smoke plumes deliver pollutant to a downwind city with less time to disperse vertically.

For a 320 km² fire, 18 km/h wind, and a city 140 km downwind, a screening-level estimate puts AQI in the 260s — solidly in the 'very unhealthy' band (dividing the fire-size contribution by distance-scaled dispersion and adding a wind term of roughly 8 AQI points per km/h). Warning lead time is estimated as distance divided by wind speed: 140 km at 18 km/h gives roughly 7–8 hours of notice before peak smoke arrival — enough time to activate clean-air shelters and issue a mask advisory, but not enough to relocate vulnerable populations at scale, which is why the response toolkit here is centred on shelter-in-place rather than evacuation.

Real operational systems (such as the US AirNow network and NOAA's HRRR-Smoke model) use satellite plume tracking, ground-based PM2.5 monitors, and dispersion modelling rather than a single formula, but the same three drivers — source strength, distance, and wind — dominate the physics.

Sizing a clean-air shelter and filtration network

A HEPA-filtered clean-air centre's throughput is a straightforward function of its air-handling capacity: air changes per hour (ACH) target for effective particulate removal in an occupied space is typically 4–6, and total filtration capacity in cubic metres per hour, divided by that target ACH rate, gives an estimate of the population that can be adequately served (since 'people served' scales with how much filtered air volume is available per person per hour of occupancy).

For a network of 26 clean-air centres delivering a combined 185,000 m³/hour of filtered air against a metro population of 420,000 people at risk, dividing the hourly filtered-air throughput by an assumed per-person air volume requirement yields coverage in the range of two-thirds of the at-risk population — a level at which planners typically flag the remaining uncovered population (often lower-income neighbourhoods with less access to private air conditioning and filtration) for targeted portable air cleaner distribution rather than centre expansion, since new centre siting takes months while distributing standalone HEPA units can happen within days.

Health outcomes and the payoff of early response

Smoke exposure drives measurable increases in emergency respiratory and cardiac visits, and the relationship is roughly dose-and-time dependent: longer warning lead time gives residents more opportunity to reduce exposure (staying indoors, sealing windows, running filtration) before peak concentration arrives, mask stockpile availability determines whether N95-grade protection is accessible during outdoor exposure that can't be avoided, and the volume of proactive health consultations (telehealth advisories, pharmacy outreach to asthma and COPD patients) correlates with reduced hospitalization rates during smoke events.

A composite resilience index blending warning lead time, mask stock relative to population, and consultation volume gives planners a single number to track season over season, and it maps reasonably well to a projected hospitalization rate per 100,000 residents — a resilience score in the 0.8+ range is associated with meaningfully lower hospitalization than a score near 0.3, giving city health departments a concrete argument for filtration and outreach budget ahead of fire season rather than reactive spending during it.

Who bears the exposure burden, and building the response network

Smoke exposure is not distributed evenly. Outdoor workers, people without air conditioning, unhoused populations, and residents of older housing stock with poor building envelope sealing all face disproportionately higher effective exposure even under identical ambient AQI. Effective resilience programs pair the technical monitoring stack — low-cost PM2.5 sensor networks (such as PurpleAir), satellite plume tracking, and AI-based short-term AQI forecasting — with targeted outreach to schools, senior centres, and community organizations that can reach these higher-risk groups faster than a citywide broadcast alert.

Because wildfire seasons are lengthening in many regions due to climate change, cities that historically treated smoke as a rare nuisance are increasingly building standing clean-air infrastructure — permanent filtration in schools and libraries doubling as public shelters — rather than emergency-only response.