Forecasting India's Smog Season — Why One Air Quality Number Cannot Tell You What You Are Breathing
Every year between roughly mid-October and mid-November, air quality across the Indo-Gangetic Plain deteriorates sharply. Three things arrive together: smoke from crop-residue burning in the northwest, festival emissions, and — most importantly — a meteorological shift that stops the atmosphere from ventilating. Most reporting collapses this into a single Air Quality Index (AQI) figure. That number is useful for a headline and almost useless for a decision, because it cannot tell you whether you are looking at mineral dust or combustion smoke. The two form differently, sit at different heights, clear on different timescales, and call for different responses. This article explains what the underlying quantities actually measure, why the distinction is operationally decisive, and how VayuMet computes them.
Why this is a meteorology problem, not only an emissions problem. Emissions vary year to year, but the sharp seasonal deterioration is driven by the atmosphere. As the Intertropical Convergence Zone (ITCZ) retreats southward and a continental high builds over northwest India, surface winds slacken, nights cool rapidly, and a shallow temperature inversion forms. The same quantity of emissions produces far higher surface concentrations when the layer available for mixing collapses from a kilometre or more to a few hundred metres.
What the Quantities Actually Measure
Four fields do most of the work, and they are not interchangeable. Two describe what is near the ground; two describe the whole atmospheric column above a point.
PM2.5 and PM10 — mass near the surface
Particulate Matter (PM) is classified by aerodynamic diameter. PM2.5 covers particles up to 2.5 micrometres, PM10 up to 10 micrometres. Both are reported as a concentration in micrograms per cubic metre (µg/m³). The size distinction matters physically: PM2.5 is dominated by combustion products — vehicle exhaust, biomass and crop-residue burning, industrial emissions — and penetrates deep into the respiratory tract. PM10 includes those particles but also the coarse fraction: road dust, construction dust, and wind-blown mineral dust.
A useful first check is the relationship between the two. When PM2.5 forms a large share of PM10, the loading is combustion-weighted. When PM10 runs far above PM2.5, a coarse source — dust — is contributing heavily. That single comparison already tells you more than an index value does.
Aerosol Optical Depth — the whole column
Aerosol Optical Depth (AOD) at 550 nanometres is a dimensionless measure of how much light is scattered or absorbed by aerosol through the entire atmospheric column, from the surface to the top of the atmosphere. It is not a surface concentration, and conflating the two is the most common error in reading aerosol maps. A high AOD with clean surface air means the aerosol layer is aloft — transported smoke or dust passing overhead, which affects satellite imagery and solar generation but not necessarily what a person at ground level is breathing that hour. A low AOD with poor surface air means a shallow, concentrated layer trapped beneath an inversion, which is the classic winter morning case across the Indo-Gangetic Plain.
Dust AOD — the field that separates the two stories
Dust AOD isolates the mineral-dust component of total AOD. This is the single most useful diagnostic VayuMet carries for this season, because it lets you decompose a high aerosol loading into its causes rather than treating it as one undifferentiated problem:
| Signal | Interpretation | Typical behaviour |
|---|---|---|
| High total AOD, high dust AOD | Mineral dust dominant | Often coarse-fraction weighted; associated with wind events and dry surfaces; clears with a change in flow |
| High total AOD, low dust AOD | Combustion aerosol dominant | Smoke or urban pollution; fine-fraction weighted; clears with ventilation or rain, not with wind direction alone |
| Rising total AOD, flat dust AOD | A combustion source is growing | The signature to watch as crop-residue burning builds through late October |
Why the Distinction Matters Operationally
Aviation
Visibility, not concentration, is what constrains flight operations. Smoke and haze reduce slant visual range and can hold aerodromes below visual flight rules minima for hours around sunrise, when the inversion is strongest and mixing is weakest. An index number gives a dispatcher nothing actionable; an estimate of the visibility field, and a sense of whether the cause is likely to disperse with daytime heating or persist, does. For live aerodrome conditions the authoritative source remains the Meteorological Aerodrome Report (METAR) and Terminal Aerodrome Forecast (TAF), available on the VayuMet Aviation Dashboard.
Public health lead time
The value of a forecast over a monitor is warning time. A monitoring network tells you the air is bad now; a forecast field indicates when a ventilation collapse is likely, which is what allows a school, a hospital or an event organiser to act a day or two ahead rather than react. The fine fraction is the health-relevant one, which is why PM2.5 is reported separately rather than folded into a single figure.
Solar generation
Aerosol loading directly attenuates surface irradiance. A persistent high-AOD episode over the northern plains reduces generation across a wide area at the same time — a correlated loss, not a site-specific one. The same field that concerns a health department concerns a grid scheduler, which is why the air quality and renewable energy layers are worth reading together.
How VayuMet Computes It
VayuMet does not operate its own monitoring network or generate its own composition forecasts. It ingests, processes and visualises output from authoritative public sources.
Input Data
| Dataset | Operator | Use | Resolution | Cycle |
|---|---|---|---|---|
| Copernicus Atmosphere Monitoring Service (CAMS) | European Centre for Medium-Range Weather Forecasts (ECMWF) | PM2.5, PM10, total AOD, dust AOD | 0.25° | Daily |
| GFS (Global Forecast System) | NOAA | 2 m temperature and dew point, for the humidity correction | 0.25° | 4× daily (00, 06, 12, 18 UTC) |
The four composition fields are rendered directly as forecast maps in the concentrations and units described above — µg/m³ for the two particulate fields, dimensionless for the two optical depth fields. No index conversion is applied, deliberately: the intent is to show the physical quantity rather than a category that hides it.
Visibility — two independent estimates
Visibility is not measured by these datasets; it is derived. VayuMet runs two independent estimation methods rather than one, on the principle that where they agree confidence is higher, and where they diverge the divergence is itself informative.
The first works from column aerosol. It integrates AOD over an effective mixing height Heff, defaulting to 1,000 metres:
The second derives visibility from near-surface PM2.5 concentration using a Koschmieder relationship. Both apply the same per-pixel humidity correction f(RH), derived from GFS 2 m temperature and dew point. The correction matters because aerosol particles take up water as relative humidity rises, growing in size and scattering disproportionately more light — which is why visibility can collapse on a humid winter morning without any change in the mass of pollutant present.
Both visibility methods are experimental. They are intended for spatial pattern guidance — identifying where a haze corridor is setting up along the plain, and how it is likely to evolve — and not as a substitute for observed aerodrome visibility. Where an operational decision depends on visibility at a specific location and time, use METAR and TAF. The default 1,000 m mixing height is a simplification; in a strong shallow inversion the true mixing depth may be far lower, in which case the AOD-based estimate reads optimistically.
What This Method Cannot Tell You
Stating the limits plainly is more useful than a decimal place of false precision:
- It is not a monitor. These are model fields on a 0.25° grid — roughly 25 kilometres. A single grid cell covers an entire metropolitan area. Street-level variation, and the gradient between a roadside and a park a kilometre away, is well below what this resolution can represent.
- It does not resolve individual fires. Dust AOD separates dust from combustion aerosol; it does not attribute that combustion aerosol to a specific source, district or activity.
- Column and surface are different questions. AOD is a column quantity. On any given morning it may bear little relationship to what is trapped in the lowest few hundred metres — which is precisely why PM2.5 and PM10 are carried alongside it rather than inferred from it.
- It is a daily composition cycle. The CAMS input updates once per day, so the aerosol fields do not refresh on the four-times-daily rhythm of the meteorological fields.
Reading the Season Ahead
Through October and November, the sequence worth watching on the map is a meteorological one as much as an emissions one: the establishment of the continental high over northwest India, the slackening of surface wind, and the progressive shallowing of the mixing layer. Emissions that were tolerable in September produce markedly higher surface concentrations once that structure is in place. When total AOD rises while dust AOD stays flat, the added loading is combustion aerosol — and unlike a dust event, it is unlikely to clear on a shift in wind direction alone.
The air quality layers sit in the Air Quality group of the main map, with a standalone summary on the Air Quality Forecast page.
Open the Air Quality Map →