We are currently experiencing what’s been termed the Pyrocene, a new geologic epoch characterized by a surge in wildfire activity linked to human-induced climate change. A recent study indicates that extreme fire weather events are now occurring at nearly double the rate compared to preindustrial times. This year, wildfire smoke has blanketed North America throughout the summer, disrupting the outdoor activities of millions. The situation is particularly severe in central Canada, which is undergoing its third-largest fire season on record. Meanwhile, in the Western U.S., wildfires in states like Arizona and Utah are also producing unhealthy levels of smoke.
It seems likely that smoke issues will escalate across the continent into August and possibly September. A new wave of smoke is currently affecting the Upper Midwest, expected to push further into central U.S. areas over the weekend.
Living in the Pyrocene presents unique challenges, primarily understanding the latest wildfire smoke predictions to effectively plan outdoor ventures. Below, I highlight some of the most reliable models available for surface air quality predictions:
- 1- to 2-day forecasts: NOAA’s HRRR-Smoke
- 1- to 3-day forecasts: NOAA’s Air Quality Model (AQM v7) or Environment Canada’s three-day air quality forecast
- 5- to 10-day forecasts: NASA’s GEOS-FP model
Forecasting wildfire smoke is challenging
It’s important to remember that predicting wildfire smoke is complicated, and model forecasts can be wildly inconsistent—either overestimating or underestimating levels. Since the smoke forecasting models are still relatively new, they face hurdles in accurately depicting meteorological conditions, fire behavior, and atmospheric chemistry. This includes the interaction of smoke with clouds and rain.
Your instinct might suggest that a heavy rainstorm would clear the air of smoke—often, it does—but not always. For example, during a rainstorm in southeast Michigan last June, I observed a spike in PM 2.5, reaching the “Unhealthy” level, even after a downpour of over half an inch. This unusual outcome was due to the storm’s airflow bringing down particles from a smoke plume above, which the models failed to anticipate, leading to discrepancies in air quality predictions.
Satellites can assess total smoke levels from the surface up into the atmosphere, but they struggle to determine smoke concentrations at varying altitudes—information crucial for accurate smoke forecasting. Moreover, satellite data is only updated daily. If a new fire ignites or an existing one grows significantly after satellite pass-over, that information won’t be included until the next day.
This means forecasts can miss the mark, especially more than a day in advance. New fires or diminished activity due to firefighting or rain can throw predictions off. Smoke models often do not account for horizontal fire spread and must make educated guesses about smoke plume heights, vegetation types, combustion characteristics, and other factors that are tough to quantify.
Using smoke model forecasts effectively
To better gauge your wildfire smoke risk, it’s wise to consult multiple models and check the initial smoke concentration data they utilize at the start of their runs. Trustworthiness often correlates with models that accurately depict initial conditions. Even if a forecast doesn’t pinpoint PM 2.5 levels perfectly, it can provide useful trends.
Pay close attention to areas with major wildfires. Sometimes, models can be clueless about fire locations, particularly under thick clouds, or they might falsely assume ongoing large smoke sources when fires have subsided. I’ve noticed substantial discrepancies in assumed starting concentrations of smoke, sometimes outweighing actual levels by over ten times. This is especially common during dense cloud coverage over active fires.
Also, remember that PM 2.5 particles have various sources, including vehicles and industrial facilities. It’s helpful to gauge total PM 2.5 from all contributions—not just wildfire smoke. On hot, windy days, especially during wildfires, smoke might not even be the primary source of PM 2.5, notably in urban environments.
Checking your air quality
Here are a few reliable sources for current PM 2.5 data:
- EPA’s Fire and Smoke Map, which aggregates data from EPA monitors and various low-cost sensors, notably from purpleair.com.
- Purpleair.com, offering a broad network of low-cost PM2.5 sensors. I’ve had one for about 10 years and it delivers solid results, though it reads about 10-40% higher during severe smoke episodes. To adjust your readings, use the site’s conversion tool.
- The NOAA Hazard Mapping System Fire and Smoke Product has an interactive smoke plume and fire location map, offering twice-daily text updates on the smoke status.
- IQAir features a global PM 2.5 map that’s pretty user-friendly.
- For our neighbors to the north, aqmap.ca displays PM 2.5 observations from official monitors and various backyard sensors.
Human-generated forecasts from EPA and Environment Canada
The best smoke forecasts can be found at EPA’s AirNow.gov, which provides forecasts for today and tomorrow alongside links to detailed state air quality department forecasts. These forecasts often highlight wildfire smoke impacts.
PROS: Top-notch forecasts that blend model predictions with expert human analysis. They are also available via the free AirNow app.
CONS: Current forecasts are only available for one day ahead.
In Canada, Environment Canada offers three-day forecasts via its Air Quality Health Index (AQHI), detailing wildfire impacts. Their Interactive Weather Alerts map shows where air quality warnings are in effect.
PROS: Combines model predictions with professional insight for high-quality smoke forecasts.
Computer models for smoke forecasts
Several agencies in both the U.S. and Canada provide computer models to predict smoke levels. The first operational smoke model—HRRR-Smoke—launched in 2020 and improvements continue to develop rapidly. These models express PM 2.5 output in micrograms per cubic meter, so knowing the conversion to AQI is essential. In the U.S., the EPA sets the standard for 24-hour PM 2.5 levels at 35 micrograms per cubic meter, corresponding to an AQI of 100. The threshold for reaching the red “Unhealthy” range is set at 55.4 micrograms per cubic meter, linking to an AQI of 150.
It’s important to know that many models only account for PM 2.5 emitted from wildfire smoke, which can underestimate the overall health burden from other pollutants in the air.
U.S. wildfire smoke models
For predicting wildfire smoke, my preferred model is NOAA’s AQM v7, operational since 2024. This model offers hourly forecasts for North America extending three days forward. By selecting “Bias-Corrected PM2.5,” you can see forecasts for PM 2.5 from all sources, including wildfire smoke.
PROS: Shows total PM 2.5 levels, including contributions from vehicles and industry, not solely smoke; it displays both micrograms per cubic meter and AQI.
CONS: Lower resolution (13 km) and data is only updated twice daily.
Another U.S. model I utilize is HRRR-Smoke, known for its high resolution. Meteorological data comes from the High-Resolution Rapid Refresh (HRRR) model, which also provides national weather forecasts. HRRR-Smoke releases hour-by-hour smoke forecasts for 18 hours and every six hours for 48 hours across the U.S.
PROS: Offers the finest resolution for smoke predictions.
CONS: Only includes PM 2.5 from smoke, excluding other sources of pollution.
Then there’s RAP-Smoke, a slightly coarser (13 km) model used for regional smoke transport forecasting. Its forecasts cover up to 21 hours and are generated every hour, thus providing quicker data access.
Global wildfire smoke models
The NASA GEOS-FP model stands out for its long-range forecasts of smoke across North America. It produces predictions up to 10 days out, although forecasts beyond a few days can be unreliable, particularly with new fires or changes in existing fires.
PROS: Global scope and accommodates PM 2.5 forecasts from smoke and pollution; the only model for ultra-long-range forecasting.
CONS: Coarse resolution (around 25 km) and forecasts can lag by about 11 hours.
The CAMS (European Copernicus Atmosphere Monitoring System) provides five-day forecasts for PM 2.5, including smoke. This version has a slightly higher resolution for Europe.
PROS: Global forecasts; covers PM 2.5 from all sources; offers long-range predictive capabilities.
CONS: Coarse resolution and limited daily updates.
The SILAM model from the Finnish Meteorological Institute offers a seven-day global forecast, although its resolution is also limited.
Apps are available for smartphones, some providing PM 2.5 predictions. Notably, popular apps like Plume Labs and AirVisual don’t incorporate forecasts from the aforementioned models. However, a few apps do offer useful models for wildfire smoke, including:
The Google Weather App (pre-installed on many smartphones) features PM 2.5 forecasts based on data from BreezoMeter. The iPhone weather app, by contrast, only supplies current AQI readings.
Gaia GPS (available with a free trial) includes smoke forecasts derived from the HRRR-Smoke model, offering easy-to-read maps of smoke predictions.
Windy.com provides a dedicated surface PM 2.5 forecast layer based on CAMS data, with the free version presenting four-day forecasts.
