Ozone is an invisible air pollutant that can cause serious lung damage — and Wisconsin occasionally sees dangerous levels when Canadian wildfire smoke drifts south on hot summer days. But 2023 saw 33 days where ozone exceeded the federal safety limit, while 2025 had only 5 days, even though both years had record wildfires. Why the difference? This project found the answer: smoke alone isn't enough — you need heat too. On a smoky day, Wisconsin can reach dangerous ozone levels at just 83°F, a temperature that would be completely safe on a clear day. And communities along Lake Michigan face this risk at three times the rate of inland areas.
Two massive wildfire years — vastly different outcomes
Both 2023 and 2025 had enormous Canadian wildfires that sent smoke into Wisconsin, with similar levels of particulate matter (PM2.5) measured at the surface. Yet their ozone impacts were completely different. To understand why, the analysis looked back across 16 ozone seasons from 2010 to 2025 to find what really separates a bad ozone year from a manageable one.
The key difference wasn't the fires — it was the weather. 2025 had 18 fewer days above 85°F and 10 fewer days above 90°F than 2023. The analysis confirms: smoke without heat doesn't push ozone over the limit. 2025 had plenty of smoke but a cooler summer, and the model predicts exactly what happened.
16 years of data from across Wisconsin
Five separate datasets were combined into a single record covering every day of Wisconsin's ozone season (May–September) from 2010 to 2025 — about 2,400 days total. For each day, the analysis tracked the highest ozone reading anywhere in the state, the highest smoke particle (PM2.5) reading, the highest temperature, and whether large Canadian wildfires were burning within roughly 900 miles.
The analysis then used two types of statistical models. The first estimates how many ppb of ozone each degree of heat or unit of smoke adds — like a recipe with measured ingredients. The second calculates the probability that ozone will exceed the federal limit on any given day, based on the forecast temperature and smoke levels.
Temperature is the main driver — smoke is a powerful multiplier
Temperature alone explains about a third of the variation in daily ozone levels. Each additional degree Fahrenheit adds roughly 0.8 ppb of ozone. Smoke particles (PM2.5) add another 0.24 ppb per unit — weaker on its own, but when combined with heat, the two factors amplify each other.
Simply having Canadian wildfires burning within 900 miles adds almost no predictive value. What matters is whether smoke actually reaches Wisconsin and shows up in PM2.5 readings. Distant fires with smoke that stays north of the border are irrelevant to Wisconsin's ozone levels.
Crucially, smoke and heat don't just add together — they multiply. On a 90°F day, the same amount of smoke pushes ozone much higher than it would on a 70°F day. A cool summer with heavy wildfire smoke is far less dangerous than a hot summer with the same smoke.
The model can predict dangerous ozone days with 89% accuracy
A second model was built to answer a simpler yes-or-no question: will ozone exceed the federal safety limit today? Using just temperature and smoke levels, the model correctly identifies dangerous ozone days 89% of the time.
On a clear day, ozone typically doesn't cross the safety limit unless temperatures reach around 99°F. On a smoky day, that danger threshold drops to about 83°F — a temperature that feels like a pleasant summer afternoon. When smoke is present, days that would normally be safe become risky.
When a day is classified as "smoke likely" — meaning fires are burning nearby AND elevated particles are measured at the surface — the odds of exceeding the ozone limit nearly triple compared to a similar-temperature clear day. This is the core finding that makes the model practically useful for forecasters.
The lakefront is a distinct, high-risk zone
Site-level analysis (54,197 site-days across 39 monitors) revealed a stark geographic divide. Lakefront sites along Lake Michigan show 3× higher exceedance rates and respond more strongly to both heat and smoke.
The reason is a phenomenon called lake-breeze recirculation. On summer afternoons, cooler air from Lake Michigan pushes onshore, trapping a shallow layer of warm, polluted air against the shore. Ozone and the chemicals that form it build up in this trapped layer — and when wildfire smoke arrives on top of the heat, the effect compounds. On smoky days, lakefront monitors record nearly 15 ppb more ozone than inland monitors seeing the same smoke.
The risk is also higher in the south than the north. Ozone decreases steadily from Kenosha/Racine up to Door County, likely because southern lakeshore communities sit downwind of the Chicago and Milwaukee urban areas, which pump out the chemical precursors that smoke and heat convert into ozone. Two sites stand out most:
| Site | Exc Rate | R² | AUC | Smoke OR |
|---|---|---|---|---|
| Racine | 4.9% | 0.604 | 0.936 | 10.7× |
| Bayside (Milwaukee) | 4.1% | 0.447 | 0.930 | 5.3× |
| Chiwaukee Prairie | 6.8% | 0.413 | 0.911 | 5.2× |
| Sheboygan – Kohler Andrae | 6.4% | 0.353 | 0.901 | 4.4× |
| Newport Park (Door Co.) | 3.2% | 0.220 | 0.887 | 3.1× |
Racine is the most predictable site in the study — temperature and smoke alone explain 60% of its ozone variation and correctly predict dangerous days 94% of the time. When smoke is present, Racine's exceedance risk increases more than ten-fold compared to smoke-free days — the strongest smoke response of any monitoring site in Wisconsin.
What the data shows
- Heat is the main driver, smoke is the amplifier. Temperature alone explains about a third of ozone variation. Smoke nearly triples the odds of a dangerous day when combined with heat.
- Where the fires are doesn't matter — only the smoke that arrives. Just knowing wildfires are burning upwind adds almost no predictive value. Measured smoke particles at the surface are what count.
- Medium smoke is more dangerous than heavy smoke. Very thick smoke can actually block the sunlight needed to form ozone. The most dangerous combination is moderate smoke on a hot day, not the heaviest smoke events.
- 2025 proves the model. Despite massive wildfires, 2025 had far fewer dangerous ozone days than 2023 — because the summer was cooler. The model predicted this correctly without any adjustment.
- Lakeshore communities face three times the risk. Lake Michigan's shoreline traps heat and pollution in a way that dramatically amplifies the smoke-plus-heat effect compared to inland areas.
- The danger decreases as you go north along the lake. Communities near Kenosha and Racine are more at risk than those near Green Bay or Door County, partly due to proximity to Chicago-area pollution sources.
National studies exist — Wisconsin's geography requires its own analysis
A 2024 study by Lee and Jaffe looked at 600 ozone monitoring sites across the entire U.S. using a more complex model with many more variables. Their model explains about 60% of ozone variation nationally.
This project explains about 36% statewide using just two variables — but reaches 60% at individual lakefront sites like Racine. More importantly, national analyses can't capture what makes Wisconsin distinct: the Lake Michigan shoreline effect, the south-to-north gradient, and the specific way the lake-breeze mechanism concentrates risk in shoreline communities. That geographic specificity is the unique contribution of this work.
Why this matters — concrete benefits
Same-day air quality alerts. The model can take a weather forecast (temperature) and a morning smoke reading (PM2.5) and calculate the probability that ozone will be dangerous by afternoon — with 89% accuracy. This gives public health agencies a simple, early-morning tool to decide whether to issue an air quality action day, triggering advisories for people with asthma, the elderly, and outdoor workers.
Targeted lakeshore warnings. The analysis shows that Racine, Kenosha, and southern Milwaukee-area communities face dramatically higher risk than inland areas on smoke-plus-heat days. Alerts could be issued specifically for shoreline zip codes rather than statewide — which would be more accurate and less likely to cause "alert fatigue" in areas that aren't actually at risk.
Supporting EPA "exceptional event" designations. When ozone exceeds federal limits due to wildfire smoke — something outside a state's control — states can apply to have those days excluded from their compliance records. This analysis provides quantitative evidence of the smoke-ozone link that supports those applications for Wisconsin.
Planning for a warming future. As summers get hotter due to climate change, the heat-smoke interaction means that even moderate wildfire seasons will produce more dangerous ozone days. This model gives Wisconsin DNR a baseline to estimate how exceedance frequency will change as temperatures rise — important for long-range air quality planning.