An environmental monitor is useful only when its variable, sensing principle, location, time basis and uncertainty are fit for the decision. Connectivity can deliver a value reliably while the measurement itself answers the wrong question.
Start with the decision: investigating comfort complaints, controlling ventilation, protecting stored goods, detecting condensation risk and studying outdoor exposure require different variables, positions and response times.
Write a point schedule before choosing devices
For each proposed point, record the room or outdoor zone, the question it answers, variable, units, sensing principle, range, accuracy specification, response time, local sampling interval, reporting interval, calibration status and owner.
Keep measured and derived quantities distinct, and record the medium and exposure represented by each sensor:
- air, surface and immersed-fluid temperatures are different measurements; select and mount the probe for the intended medium;
- relative humidity is measured, while dew point may be derived from temperature and humidity;
- measured carbon dioxide is not equivalent or inferred CO₂ derived from another sensor response;
- a volatile-organic-compound response or proprietary IAQ index is not a concentration of a named pollutant or a regulatory air-quality index.
Do not transfer a component datasheet's performance, battery condition or calibration claim to a finished instrument without evidence for that exact product and configuration.
Place sensors for the stated question
A sensor reports conditions at its own position. Representative occupied-zone measurements need representative airflow and separation from doors, supply jets, sunlight, local heat or moisture sources and people, unless one of those influences is the subject.
For workplace CO₂ investigation, the UK HSE recommends a monitor that measures CO₂ directly, commonly with a non-dispersive infrared sensor. It also recommends checking several locations, measuring at head height, keeping monitors more than 50 cm from people and interpreting readings during occupied periods with ventilation state and occupancy. A snapshot is not a representative trend.
CO₂ monitoring is less useful where people are not the main CO₂ source, including processes that release CO₂, and in spaces with very low, rapidly changing or sparse occupancy. Poorly mixed or large spaces can need several points, but extra sensors do not remove the need to decide whether CO₂ is a suitable ventilation indicator for that space.
Air cleaning can reduce some contaminants without reducing CO₂. A CO₂ trend therefore does not assess whether an air cleaner is effective.
HSE's sustained 1,500 ppm guidance is a UK workplace investigation and action indicator, not a universal safety boundary. Carbon dioxide does not establish comprehensive indoor air quality, health safety or compliance. Current ASHRAE guidance also distinguishes demand-control ventilation setpoints from IAQ indicators.
Treat room, source and outdoor points separately
An occupied-zone sensor, a supply-air sensor and an outdoor reference answer different questions. Record height, surrounding features, nearby sources or sinks and relevant activity so later users know what each value represents.
Outdoor context needs its own location rationale. A point beside a road, exhaust or loading bay measures a local influence; a background point does not. For outdoor air temperature, use a suitable ventilated radiation shield and distinguish air from surface or ground temperature. The selected enclosure, probe and operating range must suit weather exposure.
An outdoor correction or model derived at one location may not transfer indoors or to another site. Keep the original observations and the transformation used to create any corrected series.
Record calibration and uncertainty
A calibration record should identify the instrument, certificate or reference, relevant range and conditions, uncertainty, date and maintenance history. Metrological traceability is an unbroken documented chain of calibrations; it does not by itself prove that the result is fit for a particular decision.
Side-by-side comparison can expose relative offset, drift or response lag. Two sensors agreeing does not prove either is accurate. Avoid inventing a universal annual calibration interval: follow the instrument guidance and a site-specific risk plan, and recheck after changes that can affect exposure or installation.
Separate response, sampling, reporting and averaging
Four durations affect what a trend can reveal:
- the physical sensor's response time;
- local sampling interval;
- radio or network reporting interval; and
- display or alarm averaging interval.
Choose them from the event to be resolved, then validate the resulting latency with a known change. Faster delivery cannot make the physical sensor respond faster. A battery-life estimate at a particular reporting interval is not a recommendation for that interval.
Make data quality visible
Preserve raw readings and timestamp provenance. Check units, time zone, matched periods, completeness, implausible changes and stuck values. Flag suspect measurements rather than silently replacing them, and never fill an outage with invented values.
Define quality states such as fresh, stale, unavailable, outside declared range, sensor fault, communications fault and derived from incomplete inputs. State the minimum completeness required for an aggregate and show when it is not met; a percentage taken from one guide is not a universal acceptance threshold.
Protocol compatibility, whether Zigbee, LoRaWAN, Modbus or a 4–20 mA loop, does not establish measurement accuracy. Commission transport and measurement as separate evidence legs.
Separate alerts from diagnosis and control
Every alert needs a purpose, variable, source for its threshold, averaging or hold time, reset or hysteresis, occupancy applicability, recipient and response. Add a separate stale or no-data alert and test both a genuine excursion and a communications failure.
An alert identifies a condition for investigation; it does not diagnose mould, disease risk, ventilation failure or a guaranteed energy-saving opportunity. Automated control adds authority, interlocks and a safe response to stale data. Prove those requirements in the project design before allowing a monitoring point to command plant.
Commission and hand over the measurement
Record exact model and firmware where available, mounting position and photographs, point schedule, configuration, calibration evidence, raw and displayed values, clocks, quality rules, alert tests and ownership. Introduce a known environmental or simulated input only within the instrument's approved procedure, then verify the complete path from sensing through storage and display.
Review the design when room use, occupancy, ventilation, mounting, firmware, derived algorithms or alarm purpose changes. Continue with the indoor environmental application plan, 4–20 mA scaling calculator, or System Builder.
Frequently asked questions
Does a carbon dioxide reading measure indoor air quality?
No. Carbon dioxide can help investigate ventilation in occupied spaces when interpreted with occupancy, placement and time, but it does not establish comprehensive indoor air quality, health safety or ventilation compliance.
Where should an indoor environmental sensor be installed?
Install it where the stated question can be answered, with representative airflow and away from local heat, moisture, supply jets, doors or occupants unless one of those sources is the subject. Record the position, height, nearby features and operating context.
Is agreement between two sensors proof that they are accurate?
No. Side-by-side comparison can expose relative offset or lag, but two instruments can agree and both be wrong. Accuracy claims require suitable calibration, uncertainty and traceability evidence for the intended range and decision.
Working out what this needs on your own site? Answer a few questions