A hazy sky can have many explanations. Fine particles may have accumulated during a period of stagnant weather. Smoke from wildfires may have travelled hundreds or even thousands of kilometres. Dust lifted from the Sahara can cross the Mediterranean and reach large parts of Europe. On hot, sunny days, ozone concentrations near the surface can increase even though ozone is not directly emitted there.
What all these situations have in common is that the substances we find in the air are closely connected to the state of the atmosphere.
Once pollutants enter the atmosphere, their behaviour depends strongly on the weather. They may be carried far from their source, remain trapped close to the ground, or become more widely dispersed. Rain can clear some particles from the air, while sunlight and temperature can trigger chemical reactions that create new pollutants from substances already present in the atmosphere.
Forecasting air quality therefore involves more than predicting the weather. Models must also describe what is present in the atmosphere, where it comes from, how it reacts chemically and where atmospheric circulation will carry it next.
From Weather Forecasting to Atmospheric Composition
Numerical weather prediction models simulate how the atmosphere evolves. Air quality models go further, also representing gases, aerosols and the processes that affect them.
The air quality forecasts visualised by meteoblue are based on data from the Copernicus Atmosphere Monitoring Service (CAMS), which combines atmospheric observations and modelling to forecast pollutants, aerosols and other atmospheric constituents. Ensemble modelling is also used to account for differences between individual forecasts and provide an indication of uncertainty.
Air quality models consider emissions, transport, chemical reactions, dispersion and processes that remove pollutants from the atmosphere. Put simply, they need to answer two questions: What is in the atmosphere, and what will happen to it next?
Where Does Air Pollution Come From?
Some pollutants enter the atmosphere directly. Traffic, industrial processes, energy production and residential heating release gases and particles. Wildfires inject smoke and large amounts of particulate matter into the atmosphere. Wind can lift mineral dust from dry surfaces, sometimes carrying it over entire continents.
Natural processes contribute as well. Vegetation releases organic compounds, volcanoes emit gases and particles, sea spray produces aerosols, and plants release enormous quantities of pollen during their flowering seasons.
Other pollutants have a more complicated origin because they form after emission.
Ground-level ozone is a good example. Unlike pollutants released directly from an exhaust pipe or chimney, ozone near the surface is produced through photochemical reactions involving precursor substances such as nitrogen oxides and volatile organic compounds. This is one reason ozone pollution is often associated with warm and sunny conditions. The meteoblue ozone map shows concentrations near the surface; this tropospheric ozone should not be confused with the stratospheric ozone layer that protects Earth from harmful ultraviolet radiation.
Particulate matter can also have several origins. PM10 and PM2.5 describe particles according to their size, rather than a single chemical substance. They can include smoke, soot, mineral dust, salts and other materials, while additional particles can form through chemical reactions in the atmosphere.
This mixture of direct emissions, natural sources and atmospheric chemistry is one reason forecasting air quality is such a complex modelling problem.
The Weather Determines What Happens Next
Knowing how much pollution has been emitted is not enough. The same emissions can produce very different air quality depending on the weather.
Consider two otherwise similar days in a large city.
On the first day, winds are light and the atmosphere near the surface is stable. Vertical mixing is weak, allowing pollutants to accumulate close to the ground. If these conditions persist, concentrations may continue to rise even without any dramatic increase in emissions.
On another day, stronger winds and a deeper mixing layer can disperse the same pollutants over a much larger volume of air. A passing front may then bring precipitation, which can remove some particles and soluble substances from the atmosphere.
Wind direction is equally important. Poor air quality observed in one place does not necessarily originate there. Smoke, dust and other pollutants can be transported across national borders and over very large distances.
Desert dust provides one of the clearest examples. Mineral particles lifted over the Sahara can travel northwards across the Mediterranean when the atmospheric circulation is favourable. At sufficiently high concentrations, desert dust can become visible as a veil in the sky.
Wildfire smoke behaves in a similar way. Depending on wind speed, direction and the height at which smoke enters the atmosphere, a fire can affect air quality far away from the burning area.
This is why an air quality map becomes particularly informative when considered together with weather maps. A plume of elevated particulate matter tells us where concentrations are high; wind fields and the larger synoptic situation can help explain where that air came from and where it may travel next.
Starting a Forecast: What Is in the Atmosphere Now?
To predict tomorrow’s air quality, a model first needs an accurate estimate of today’s atmospheric conditions, which is achieved through data assimilation.
Similar to numerical weather prediction, observations are combined with model information to build a picture of the current state of the atmosphere. These observations can come from ground-based monitoring stations as well as satellites, providing information about gases and aerosols at different spatial scales.
Monitoring stations provide direct measurements at specific locations, while satellites offer much broader spatial coverage. Forecasting systems combine these observations with atmospheric models to estimate the current distribution of pollutants and predict how their concentrations may change and move over the coming hours and days.
One Forecast, Many Variables
There is no single variable that completely describes air quality. Different maps answer different questions.
The Common Air Quality Index (CAQI) provides a simplified overview. Rather than displaying the concentration of one substance, it translates several important pollutants into an index that indicates overall background air quality. meteoblue uses the background CAQI because atmospheric models cannot reproduce the very small-scale differences found directly along individual roads.
For investigating a particular situation, however, looking at the individual components is often much more informative.
PM10 and PM2.5 show concentrations of particulate matter. PM10 represents particles smaller than 10 micrometres, while PM2.5 represents the finer fraction with diameters of 2.5 micrometres or less. Sources range from combustion and wildfire smoke to dust and particles produced through atmospheric reactions.
Nitrogen dioxide (NO₂) is strongly associated with combustion. In cities, road traffic is an important source, although NO₂ also has natural sources and participates in atmospheric chemistry leading to ozone formation.
Sulphur dioxide (SO₂) is associated particularly with the combustion and processing of sulphur-containing fuels and materials. Once in the atmosphere, sulphur compounds can undergo further reactions and contribute to particulate pollution and acid deposition.
Carbon monoxide (CO) originates from incomplete combustion and can therefore be associated with sources including fires and fossil-fuel burning. It also participates in atmospheric chemistry and indirectly affects concentrations of other gases, including ozone and methane.
Ozone (O₃) tells a different story because elevated surface concentrations can develop through photochemical reactions rather than direct emissions. Temperature, solar radiation, precursor pollutants and atmospheric circulation all influence its development and transport.
Desert dust isolates another specific aerosol source. Since dust can be transported over very long distances, viewing its forecast at different heights can help reveal the three-dimensional structure of a dust intrusion.
Then there is aerosol optical depth (AOD). Unlike PM2.5 or PM10 concentration near the surface, AOD describes how strongly aerosols reduce the transmission of light through the entire atmospheric column. High AOD can be associated with wildfire smoke, desert dust or anthropogenic pollution. It is therefore useful for identifying broad aerosol plumes, but it should not be interpreted as a direct measurement of what someone is breathing at ground level.
And atmospheric forecasting is not restricted to pollution. Pollen is also transported by the atmosphere. meteoblue Weather Maps include grass, ragweed, birch and olive pollen layers. Pollen can travel considerable distances with air currents, while precipitation can remove it from the air. Thunderstorms can complicate the picture because strong winds may initially increase airborne pollen concentrations.
Together, these layers show what is in the air, where it may have come from, and how weather conditions affect its movement.
Why Local Air Quality Is Especially Difficult to Predict
Air pollution can vary considerably over very short distances. A monitoring station beside a busy road may record much higher NO₂ concentrations than another station only a few kilometres away in a park. Street orientation, building height, traffic intensity, industrial sources and local wind conditions can all contribute to these differences.
Regional atmospheric models cannot represent every one of these small-scale effects. They may capture a pollution episode across a wider region, while conditions near a busy road, industrial site or other local source can differ from the modelled values.
Weather models face a similar issue. Higher resolution can provide more local detail, but processes occurring at scales smaller than the model grid cannot always be represented directly.
Why Use an Ensemble for Air Quality?
Air quality forecasts contain several sources of uncertainty. Emissions can change, weather forecasts are not perfect, and different models represent transport, aerosols and atmospheric chemistry in different ways.
Using an ensemble of several models helps account for these differences. Instead of relying on a single forecast, multiple model results can be compared or combined. The variation between them also gives an indication of forecast uncertainty.
As with weather forecasting, no single model performs best for every pollutant, location and atmospheric situation. When models disagree, that disagreement itself provides useful information about how confident we can be in the forecast.
Reading the Atmosphere as a System
Perhaps the most interesting aspect of air quality forecasting is that none of these variables exists in isolation.
An elevated PM2.5 forecast becomes more meaningful when compared with wind direction. A broad aerosol signal can be investigated alongside desert dust to determine whether mineral particles may be involved. Ozone concentrations can be considered together with temperature, sunshine and atmospheric circulation. Pollen forecasts can be compared with wind and precipitation.
The atmosphere continuously connects processes occurring at very different scales – from emissions along a city street to dust storms over North Africa and wildfires hundreds or thousands of kilometres away.
Modern atmospheric models attempt to connect these pieces, combining meteorology, emissions, chemistry, transport and observations into a forecast of what the air around us may contain over the coming hours and days.
Interested in exploring this topic further or sharing your own observations? Visit the meteoblue Community Forum, where our experts and community members continue the discussion.
