How Do Air-Quality Sensors Work, and How Accurate Are They?

Reference monitors versus lower-cost sensors

Reference and equivalent methods use specified instruments, procedures, siting, calibration, maintenance, and quality systems to support regulatory decisions. Instruments such as beta attenuation monitors, or BAMs, and gravimetric methods measure particulate matter using well-defined approaches. Research-grade optical instruments, including devices from GRIMM and other manufacturers, are often used for detailed particle measurements and evaluation.

Lower-cost sensors make dense neighborhood networks and real-time community information more affordable. Their data can be excellent for trends, source discovery, education, and supplementary mapping when the system is designed and evaluated for that purpose.

How optical PM sensors estimate particles

Many PM sensors pull air through a small chamber and shine light across it. Particles scatter the light, and the detector uses that signal plus assumptions about particle size, shape, density, and optics to estimate particle number or mass.

Because the instrument does not directly weigh every collected particle, its estimate can change with aerosol composition and humidity. Smoke, urban aerosol, and mineral dust may produce different responses at the same true mass concentration.

Gas, temperature and humidity sensors

Compact gas sensors may use electrochemical, metal-oxide, or other sensing principles. They can respond to temperature, humidity, interfering gases, and baseline drift. Temperature and relative humidity are also measured because meteorology shapes pollution and helps interpret sensor behavior.

Accuracy, precision and reliability

• Accuracy describes closeness to an accepted reference
• Precision describes how consistently repeated measurements agree
• Bias is a systematic high or low difference
• Completeness describes how much expected data was successfully captured
• Reliability includes stable operation, maintenance, power, communications, and metadata

Why nearby sensors show different values

Real air pollution varies over short distances near roads, cooking exhaust, generators, construction, waste burning, and wind channels. Placement differences can be as important as instrument differences. Even co-located units may disagree because of manufacturing variability, dirt, inlet blockage, moisture, aging, clock drift, or a failing component.

Calibration and co-location

Co-location means operating a sensor beside a quality-assured reference instrument under the conditions where it will be used. The paired data reveal bias and support a correction equation. Evaluation should cover a useful range of concentrations and weather, then be repeated to detect drift.

A correction developed in one city or smoke type may not transfer perfectly to another. Keep the raw measurement, correction version, training dates, performance statistics, and uncertainty traceable.

Quality assurance and outlier removal

Automated checks can flag impossible values, repeated flat lines, sudden jumps, missing data, internal conflicts, and readings outside a defensible physical range. An extreme value should not be deleted only because it is inconvenient. Compare neighboring instruments, logs, weather, known events, and reference data, then preserve an audit trail of the decision.

Raw data are the sensor output before correction. Corrected data apply a documented adjustment. Both labels should be visible, and neither should be presented as exact personal exposure.

What sensor data are suitable for

• Understanding hourly and neighborhood patterns
• Finding potential hotspots for further investigation
• Testing whether a local intervention changed a pattern
• Public information and education with clear uncertainty
• Supplementing, not silently replacing, reference networks

The VayuDrishti data-quality approach

VayuDrishti aims to show source, time, location, validation status, corrections, and limitations beside the decision guidance. Values beyond a defensible range can be withheld from the public map while remaining flagged for technical review. Transparency matters because useful data should empower action without creating false precision.

Now act

Practical recommendations

• Match the sensor and data-quality target to the question you need to answer.
• Record location, height, nearby sources, airflow, shelter, power, and maintenance history.
• Co-locate sensors with a trusted reference before and periodically during deployment.
• Check clocks, duplicates, missing intervals, impossible values, sudden steps, and flat-lined data.
• Review humidity and temperature when optical PM readings behave unexpectedly.
• Publish raw and corrected data definitions, calibration periods, coverage, and limitations.