Air Quality Prediction Map: How to Read the Forecast

Learn how to read an air quality prediction map, its forecast layers and time ranges, and how to plan your day around pollution if you get migraines.

In this guide
  1. What an Air Quality Prediction Map Shows and Where It Stops
  2. Forecast Layers and Time Ranges at a Glance
  3. Reading AQI Categories and Reference Levels
  4. Why Grid Size and Forecast Error Limit Precision
  5. Outdoor Readings Versus the Air You Breathe Indoors
  6. What the Evidence Shows
  7. Planning Your Day and Tracking Your Own Pattern

Does a green dot on an air quality prediction map tell you what your lungs, and your head, will meet today?

Not quite, and the gap matters when you live with migraine. An air quality prediction map is a forecast of outdoor pollution for a specific place and a future hour, assembled from atmospheric chemistry models, weather prediction, satellite data and ground monitors. It isn't a sensor reading from your street, and it isn't a statement about your risk of an attack.

That distinction changes how you should use one. A map works best as an exposure signal with a stated uncertainty, something you compare against your own logged symptoms over weeks rather than a verdict you accept in a single glance. What follows explains what sits under the colours, how to read the numbers, and where the outdoor forecast stops describing the air you breathe.

What an Air Quality Prediction Map Shows and Where It Stops

Every colour on that map began as a number inside a model. Forecast means a projection for a future hour, not an observation. A model cell is the patch of ground each grid square covers, where one value stands in for everything inside it, from the park to the parking garage. The Air Quality Index (AQI) converts pollutant concentrations into health-oriented bands so you don't have to reason in micrograms per cubic metre.

The pollutants behind those bands are worth knowing by name. PM2.5 is fine particulate matter under 2.5 micrometres across, small enough to reach deep into the lungs. PM10 is the coarser dust fraction. Nitrogen dioxide comes mostly from traffic and combustion, while ozone forms in sunlight when other pollutants react.

Think of a weather map. The colour shows a predicted state, the time control shows when that state is expected, and the model underneath decides how much weight the whole picture deserves. Air quality works the same way. Services such as the Copernicus Atmosphere Monitoring Service run chemistry models alongside weather models, so published maps reflect both pollutant transport and shifting weather, refreshed on a schedule rather than captured once.

A diagram explaining factors involved in air quality prediction maps, including weather inputs and modeling limitations.

One clarification, because the two words get swapped constantly. A headache is a symptom with many possible causes. Migraine is a recurrent neurological condition, and an attack may arrive with a prodrome (mood, appetite or energy changes in the hours before), aura (visual or sensory disturbances), photophobia (light sensitivity) and a postdrome that leaves you depleted the following day. Pollution is one possible influence among many, and no map diagnoses anything.

Here's what the map is not:

  • Not your personal exposure. It describes outdoor ambient air at a coarse scale, not the air in your bedroom or on your train.
  • Not a diagnosis. A high reading doesn't mean pollution caused your last attack, or that it will cause the next one.
  • Not a guarantee. A clean forecast doesn't promise a symptom-free day, and a bad one doesn't promise an attack.

Practical rule: use the colour to decide whether to look closer, never as the final answer.

This article is for informational purposes and is not medical advice. Consult a healthcare provider for personalized guidance.

Forecast Layers and Time Ranges at a Glance

A colour on a map is only as useful as the clock and the grid behind it. Two products can both read "moderate" and still answer different questions: one about the next two days near city scale, another about a coarser global field several days out. Treat each layer as an outdoor exposure signal with a stated window of doubt, not as a verdict on one person.

ProductTime coverageSpatial resolution
CAMS European regional serviceHourly analyses and a daily four-day forecastApproximately 10 km
CAMS global systemAnalyses every six hours; two five-day forecast cycles every 12 hoursCoarser than the regional grid, with 137 vertical levels after the 2019 upgrade
NOAA National Air Quality Forecast CapabilityForecasts up to about 48 hours aheadOn the order of 12 km
NASA GEOS composition forecastGlobal composition fields on the model's own cycle; the issue stamp on the layer shows the lead timeCoarser than a neighbourhood, set by the model's own grid

The Copernicus Atmosphere Monitoring Service runs both CAMS rows: a European regional service near 10 km, with hourly analyses and a daily four-day forecast, and a global system that issues analyses every six hours plus two five-day forecast cycles every 12 hours. Its July 2019 global upgrade raised the atmospheric model from 60 to 137 vertical levels and added nitrate and ammonium aerosol species, which improved how the system represents particulate matter. NOAA's National Air Quality Forecast Capability, as a general characteristic of that product, pairs a weather forecast with CMAQ chemistry and looks about 48 hours ahead at a resolution on the order of 12 km. NASA's GEOS composition forecast publishes global fields on the model's own cycle. The issue stamp tells you the lead time, and the grid is coarser than a neighbourhood.

Analysis or Forecast, and Why the Difference Matters

An analysis estimates conditions now by blending model output with observations. A forecast projects forward from that starting point. Analyses sit closer to what monitors recorded, because the model has been pulled toward those measurements. Forecasts drift as lead time grows. The fourth day of a four-day panel deserves less confidence than its first six hours.

A forecast panel works like a weekly weather strip. The near hours are a sharper picture than the far end, even when both use the same colour scale. Error is part of the product, not a footnote you can ignore.

If you spend most of the day indoors, even a sharp outdoor hour can mislead you. The layer describes ambient air outside, not the room where you sit. Use the colour as a signal to look closer, with that mismatch in mind.

How to Read One Layer Without Guessing

Before you change a plan because of a colour, walk four checks.

  1. Find the issue time. A stamp of 06:00 UTC is when the model ran, not the clock in your kitchen.
  2. Pick the hour closest to time you will be outside. A commute or a run matters more than the day's headline value.
  3. Confirm which pollutant is displayed. A PM2.5 layer and an ozone layer can disagree on the same afternoon. The colour belongs to the one selected.
  4. Check whether you are looking at a forecast or a past analysis. Earlier hours estimate what already happened. Later hours are the projection, and they carry more doubt.

A product that reaches about 48 hours can speak to tomorrow morning. A four-day regional forecast can speak to the weekend, with wider error the further you go. Both remain outdoor estimates. Neither one says what will happen to you.

Reading AQI Categories and Reference Levels

Two things decide the colour you see: which pollutant the map is showing, and the averaging window behind the number. A category label on its own hides both.

MeasureValueSource or meaning
AQI 0 to 50 (Good)24-hour PM2.5 of 0.0 to 9.0 micrograms per cubic metreU.S. EPA AQI breakpoint
AQI 51 to 100 (Moderate)24-hour PM2.5 of 9.1 to 35.4 micrograms per cubic metreU.S. EPA AQI breakpoint
WHO guideline, PM2.515 micrograms per cubic metre over 24 hours; 5 as an annual meanWHO 2021 population-health goal
WHO guideline, PM1045 micrograms per cubic metre over 24 hours; 15 as an annual meanWHO 2021 population-health goal

The AQI is a communication scale, not a personal risk score. In the U.S. EPA's AQI technical assistance document, 0 to 50 is labeled "Good" and 51 to 100 "Moderate", with higher bands signaling increasing health concern. For PM2.5, the technical breakpoint for Good is a 24-hour concentration of 0.0 to 9.0 micrograms per cubic metre, and Moderate runs from 9.1 to 35.4. Those figures are population-level air-quality standards. None of them was set with migraine in mind, so none of them marks the point where your head starts to ache.

Public health goals sit somewhere else again. The World Health Organization's 2021 air quality guidelines recommend an annual mean PM2.5 of 5 micrograms per cubic metre and a 24-hour level of 15, with PM10 at 15 annually and 45 over 24 hours. These describe long- and short-term protection targets for whole populations. They aren't thresholds where an individual attack begins.

Why Two Maps Disagree About the Same Street

One app shows yellow, another shows green, and both can be technically correct. They may be displaying different pollutants, averaging over different windows (one hour versus 24 hours), or using their own colour ramp. Before you conclude that one is wrong, check what each is plotting.

When you log a day, write down the category, the pollutant name and the concentration together. A note reading "Moderate, PM2.5, 22 micrograms per cubic metre" is something you can compare with your symptoms later. A note reading "yellow day" isn't. For a practical walkthrough of measuring conditions where you physically are, our guide to how to measure air quality covers the handheld and at-home options.

A Good or Moderate label doesn't predict an attack. It describes a band, and bands are population tools.

Why Grid Size and Forecast Error Limit Precision

How can a colour sitting on your address still miss the air on your street?

A forecast map works like weather radar painted in large tiles. Each tile is one calculation for a patch of land, not a sensor on your roof. Regional models commonly use horizontal grid cells of about 4 to 36 km, with 5 to 50 vertical layers, and the AirNow air quality forecasting guidance notes that shrinking the grid improves spatial detail and can improve accuracy, at a real computing cost. A cell that covers a town is a regional average. Global composition forecasts often run on cells tens of kilometres wide while still representing ozone, carbon monoxide, nitrogen oxides and PM2.5. That average can hide a busy road on one edge and a park on the other.

Error also depends on the pollutant. Particulate forecasts usually hold up better than short-lived gases such as nitrogen dioxide and ozone, which swing sharply near roads, cities and industrial sources. A daily particle field can look steady across a cell. Nitrogen dioxide can spike at a junction and fall a few blocks away, inside the same coloured square. Ozone often builds downwind of the traffic that fed it, so the high value is not always where the cars are.

Before you treat a value at your address as a verdict, ask three questions:

  1. What cell size is this? If the grid is 12 to 25 km wide, the number describes a region, not a rooftop.
  2. When was it issued? An older issue time means more drift between the model's starting point and now.
  3. Is this value at my location, or interpolated for display? Interpolation smooths colours across a map. It does not create local measurements.

A careful layer shows its native resolution, interpolates only so the map can be drawn, and flags extra uncertainty near coastlines, mountains, wildfire boundaries and urban emission hotspots. Read the square as an exposure signal with a margin around it. The margin is part of the forecast, not a footnote you can ignore when the colour looks precise.

Outdoor Readings Versus the Air You Breathe Indoors

The forecast says unhealthy. Now look at your actual day: two hours on a train, seven in a sealed office, an evening on the sofa with the windows shut because it's cold outside. The map is describing the air outside your building, and almost none of your day happens there.

A woman wondering about indoor air quality while checking an unhealthy air quality forecast on her phone.

Exposure happens indoors, in vehicles, at work and across shifting microenvironments. A rain gauge in the park tells you nothing about how wet the room under the roof is.

The research draws the same line. Researchers modelling the 2022 and 2023 wildfire seasons in British Columbia built separate indoor and outdoor PM2.5 models, which carried different prediction errors, and concluded that personal time-activity and building characteristics have to be part of any dynamic exposure estimate. A 2025 California study went further, combining outdoor observations, indoor sensors, building characteristics, ventilation modelling and behaviour, and found that home type, HVAC conditions and proximity to fires all shifted personal exposure, as reported in ACS ES&T Air.

Here's a short walkthrough of why the outdoor number and your dose part ways:

How Indoor Air Quality Affects Your HealthChris Kresser · plays from YouTube

For your own log, note the things that change your dose rather than chasing a perfect number. Windows open or shut during the hours you were home. Whether the building runs mechanical ventilation or a filter. Hours spent inside a vehicle, and on which route. Your floor level in a tall building, since a street-level reading doesn't describe a twelfth-floor flat. These are observations to test against your symptoms, not guaranteed fixes.

The question no ambient map answers is the one that matters most: if the map says unhealthy, how much pollution am I breathing where I spend my day? Opening a window, commuting at rush hour or switching on filtration changes that number in ways the forecast can't see. The nearest monitor is not your dose.

What the Evidence Shows

How much should a coloured square on an air quality map change what you expect of your own migraine risk? Research treats pollution as a real signal, and a noisy one.

A systematic review of migraine and air pollution reported that most point estimates suggested positive associations for PM2.5, nitrogen dioxide, ozone and carbon monoxide across the included studies, while results varied by pollutant and by study design. The same pollutant can look important in one population and weak in another. Which days are compared, which lag is tested, and whether heat is in the model all shift the result.

The clearest published ratios come from emergency-department visits. For an interquartile-range increase in exposure, a case-crossover study of emergency-department visits for migraine reported odds ratios from its best-fitting lag models of 1.031 for PM2.5, 1.032 for PM10, 1.053 for nitrogen dioxide, 1.034 for ozone and 1.029 for carbon monoxide. That study also found that pollution-related migraine risk appeared particularly relevant on high-temperature days.

What the Numbers Mean for How You Read the Map

Those ratios sit just above 1. Across a population, a modest rise in a pollutant lines up with a slightly higher chance of an emergency visit for migraine. They do not give the chance that today's outdoor reading will start your attack. A city rainfall map can show that wet days bring more hospital slips, and still say nothing about whether you will fall on your street. Heat changes the picture again, so one combined air quality score flattens pollutants that do not move together.

Read the map as a set of signals to compare against your own record, not a warning label for your body.

A pollutant-by-pollutant view keeps that separation. The guide to weather and air quality as migraine influences walks through temperature and pressure alongside the air itself. Air pollution and heat are common triggers some people identify, not universal ones. Migraine varies widely between people and between attacks, so a pattern in a study population may never show up in your log, and a pattern in your log may never show up in a study. The square on the map remains an outdoor estimate with a wide error band.

Planning Your Day and Tracking Your Own Pattern

Plan around the safest available time window given the forecast's uncertainty, not the single colour shown for one hour.

Most maps answer "what is the AQI here?" and stop. The more useful question is "how reliable is this forecast, and when should I act?" A 2023 wildfire-season evaluation, posted as a preprint on air quality forecasting, found that forecasts carried useful information about the timing of PM2.5, yet no single model consistently predicted which days would be high or which hours would be lowest. Longer-range forecasts and heavy-pollution events remain the hardest cases.

Five Steps for Planning Around a Forecast

  1. Name your window. A commute, a run, a morning of outdoor work. Without a window, any forecast is just trivia.
  2. Check the issue time and how far ahead it reaches. A forecast issued yesterday for this afternoon carries more doubt than one issued at dawn.
  3. Compare the category across the whole window. Two hours at Moderate with an improving trend is a different decision from two hours heading toward a higher band.
  4. Look for a confidence note, and for the source of the data. Measured, modelled and satellite-derived values carry different weight, and a map that says which is which is telling you something.
  5. If confidence is low, treat the forecast as a loose guide and keep your personal, clinician-agreed plan within reach. A low-confidence prediction is not reassurance. It's an absence of information.

Those five steps are easier to hold onto as a picture than as a list, so here they are laid out end to end.

A five-step infographic showing how to plan your day based on air quality and health tracking.

That last step is where people get caught. Green on the map feels like permission, and permission isn't what the model offered.

What to Write Down, and What to Look For

Keep the logging simple enough to survive a bad day. Each day, record the forecast category, the pollutant behind it, and the decision you made, such as walking at 7am instead of 6pm. Each attack gets its own entry: start time, severity, symptoms, and anything you suspect. If you want a structure for turning those entries into something readable, our guide to how to identify your migraine triggers sets out the comparisons worth making. After two to four weeks, read the two lists side by side and look for repeated overlap on the same days, whether with pollution, temperature or pressure swings. Then take that pattern to a healthcare provider for personalised guidance.

Seek immediate medical care if you have a sudden severe headache that peaks within seconds to minutes, a headache with fever and a stiff neck, neurological changes such as weakness, confusion or trouble speaking, or a headache following a head injury. These are red flags, not forecast questions.

Relief sends informational forecasts of higher migraine risk based on environmental factors such as weather, atmospheric pressure and air quality, and those forecasts are probabilistic estimates, so an absence of a warning doesn't mean an attack won't occur.

Relief lines your attack log up against the environmental conditions around each one, so the overlap between the forecast and your own history becomes visible instead of something you try to remember. You log an attack in seconds, and over time the app surfaces which triggers track with your migraines rather than which ones you assumed.

This guide is general information, not medical advice, and it isn’t a diagnosis. Talk to your doctor about your own headaches.