Air Quality Exposure

Daily aerosol exposure at neighbourhood scale, for your cohort.

Satellite aerosol retrievals downscaled across the study area you define, with a table keyed to your participant geocodes.

Spec table

Platform
coarse-resolution satellite
Spectral
multispectral
Resolution
250 m or coarser, downscaled to neighbourhood scale
Cadence
daily
Thermal
none
Delivery unit
your study area and participant geocodes
Gaps
flagged and left empty

Files in the delivery

The pilot covers the study area and the follow-up window you name. Once it is running, new days arrive at the daily cadence.

DeliverableFormatWhat's in itCadence
Daily exposure layerGeoTIFFOne raster per day over your study area, holding the downscaled aerosol retrieval at neighbourhood scale. Cells with no usable retrieval carry no value. Opens in QGIS, ArcGIS, R or Python.Daily
Geocode extraction tableCSVOne row per participant geocode per day: geocode ID, date, retrieved value and quality flag. Built to join to your cohort file on ID and date, so the linkage happens in your own R, Python or SAS code.Daily
Data-quality flagsColumn in the table, band in the layerRetrieval confidence and a cloud-gap marker on every row and every cell. Filter on them in your analysis plan. A day with no clear view of a geocode shows as a gap, and the gap stays.Daily
Retrieval notesPDFA short methods note for your methods section: the satellite source, how the downscaling was done, the date range and which days have gaps. Written so a reviewer can see how exposure was assigned.Per delivery

How it works

  1. 01in

    Send the study area and geocodes

    A study-area boundary as a shapefile or GeoPackage, participant geocodes as a CSV with ID and coordinates, and the follow-up window. Names and addresses stay with you. An ID and coordinates are enough.

  2. 02process

    Daily retrieval

    The coarse-resolution aerosol retrieval for each day is pulled for your study area.

  3. 03process

    Downscale

    Each retrieval is downscaled to neighbourhood scale across the study area.

  4. 04process

    Extract and flag

    Values are read at each geocode for each day, and the quality flag is attached.

  5. 05out

    Hand-over

    Layers, extraction table and methods note come back in formats that open in your GIS and your statistics package.

Limits

These are the conditions that change what you can claim from the output.

ConditionEffect on the outputIn the delivery
Cloud and clear-sky windowsAn optical retrieval needs a clear view of the ground. A cloudy day shows as a gap with no estimate behind it, and a study area with a long cloudy season will have long runs of them.The flag column says which days.
Neighbourhood scale is the floorThe output resolution is 250 m or coarser, downscaled. It separates one neighbourhood from the next.One side of a street against the other is out of reach, and so is a per-address value.
Column, ground levelA satellite retrieves aerosol through the whole air column, so the value says nothing directly about particle mass at the participant's door.Relating it to ground concentration is a modelling step that belongs in your analysis plan.
One overpass a dayA daily retrieval is one moment, taken when the satellite passes over.It is no 24-hour average.
Bright surfaces and snowRetrievals over snow, ice or very bright ground are less reliable.Those days may be flagged or left empty.
ValidationNo validation figure against ground monitors is published on this page.Checking the layer against your own monitors is part of piloting it.

Why the nearest monitor falls short

Each participant's exposure comes from the nearest fixed monitor, or from a regional surface too coarse to separate one neighbourhood from the next. Someone who lives far from a monitor inherits a reading from another street, and the contrasts inside the cohort flatten out. You find the cost at analysis, as misclassification pulling the estimate toward the null, and again at review, when someone asks how exposure was assigned. This replaces that assignment step with a daily value for every geocode.

Morning haze over a city's rooftops, with hills fading out behind
Scenery: morning haze over a city. Photo by Ekam Juneja, Pexels.

Who it's for

For

  • Environmental epidemiologists and exposure scientists building exposure estimates for a cohort study.
  • Exposure modellers who want a daily, spatially resolved input to a larger model, with quality flags so they can handle gaps their own way.

Not for

  • Studies that require a per-address or street-level value. This works at neighbourhood scale.
  • Consumer air-quality apps, health advisories or a public index.
  • Studies that need a validated ground-level concentration out of the box. The output is an aerosol retrieval, and conversion to concentration is yours to model.
  • Health-outcome analysis. The delivery holds exposure inputs and no outcome associations.

Technical and procurement questions

How does this compare to a fixed monitor?

A monitor gives an accurate point reading at one location. This gives a spatially resolved value across the whole study area, so each participant gets a value for their own neighbourhood.

What happens on cloudy days?

They are left as gaps and flagged. Nothing is filled in or interpolated, so you see exactly where exposure is missing.

Is it validated against ground monitors?

No validation figure is published here. Bring the monitors you trust for your study area and the pilot can compare against them.

Can it resolve individual addresses?

No. It works at neighbourhood scale, and participants in the same neighbourhood can share a value.

How is a day defined across time zones?

One retrieval per overpass day for the study area. The exact timing is confirmed with you when the study area is scoped.

Does it give PM2.5 concentrations?

The retrieval is aerosol optical depth, a column quantity. Turning it into a ground concentration takes a calibration step that you own in your model.

How does it link to our cohort records?

The extraction table is keyed to the geocode ID and date you send, so you join it in your own statistics code.

Do you need participant addresses?

No. An ID and coordinates are enough, and names and addresses stay with you.

Who owns the delivered data, and what does it cost?

Use and ownership terms are written into the contract before work starts. A pilot is scoped on study area size, follow-up window and number of geocodes, and you get a written scope for your institution's procurement.

Send us the study area and we'll scope a pilot.

Or .

Start a pilot