11. Field Deployment and Calibration Guidance

This section addresses practical calibration and deployment questions commonly raised by research teams integrating AirQo low-cost sensors with reference-grade instruments in field-based monitoring projects.

11.1 Calibration Frequency

In challenging field environments characterised by high humidity, heavy dust loading, or significant biomass burning influence, AirQo recommends the following calibration intervals:

Initial co-location period — A minimum of 4–6 weeks of co-location alongside a reference-grade monitor before network deployment is strongly recommended. This allows sufficient data to build a robust, site-specific calibration model across varying atmospheric conditions.
Recalibration interval — A 2–3 month recalibration interval is generally sufficient for most African urban environments. However, in areas with extreme seasonal shifts (e.g., pronounced dry-season dust or heavy wet-season humidity), quarterly recalibration (every 3 months) is strongly recommended to account for sensor drift and changing pollution composition.
Continuous quality monitoring — AirQo applies machine learning-based calibration corrections continuously to account for environmental drift between formal recalibration events. Researchers should document the calibration model version applied to their dataset.

11.2 Calibration Approach: Co-location vs. Remote Methods

Physical co-location with a reference-grade monitor remains the gold standard for calibrating low-cost sensors. It is necessary, particularly at the start of a project and at each formal recalibration event.

Duration — A minimum of 4 weeks of co-location is recommended per calibration event. Longer periods (6–8 weeks) yield more robust models, especially where pollution sources are highly variable across seasons.
Frequency over a 12-month period — We recommend at least 2–3 formal co-location events over a 12-month monitoring period: one at project start, one mid-year, and one toward the end of the study period. This captures seasonal variation in calibration relationships.
Remote and model-based correction — Between formal co-location events, AirQo applies machine learning-based remote calibration corrections using environmental variables (temperature, humidity) and cross-sensor network consistency checks. These methods reduce but do not eliminate the need for periodic physical co-location, which remains essential for model validation.

11.3 Sensor Movement During Calibration and Maintenance

Temporary relocation of sensors to a co-location site is standard and acceptable practice. Key operational considerations:

Document all sensor relocations with timestamps in your metadata. Data collected during transit or at the calibration site should be clearly flagged and excluded from your primary analysis dataset.
Where backup sensors are available (see Section 11.4), rotate a spare into the network while the primary sensor undergoes calibration. This preserves spatial data coverage and minimises gaps in your time series.
Where physical relocation is not feasible, model-based correction using network-wide consistency checks can serve as an interim measure; however, this should not be the primary long-term strategy.

11.4 Use of Backup Sensors

Including at least one spare sensor per network is strongly recommended, particularly for research-grade monitoring projects where data continuity is essential. AirQo's experience across African city deployments confirms that spare sensors are critical for:

Maintaining continuous data coverage during calibration rotation events.
Replacing failed sensors without waiting for procurement lead times, which can be 4–8 weeks in many African contexts.
Enabling cross-sensor validation by comparing a spare unit against deployed sensors over time.

For a network of 10 sensors, we recommend budgeting for 1–2 spare units (10–20% of network size). This is considered best practice in AirQo's operational networks and is strongly recommended for policy-grade data generation projects.

11.5 Reference-Grade Monitor Requirements

For a network of approximately 10 low-cost sensors, AirQo recommends the following approach to reference-grade instrumentation:

Minimum viable setup – 1 reference monitor: One reference-grade monitor is sufficient for sequential calibration of all low-cost sensors, provided the sensors are rotated through the co-location site in batches. This works well for networks where the monitoring environment is relatively homogeneous.
Preferred setup – 2 reference monitors: For a city with distinct micro-environments (e.g., coastal areas vs. inland hillside communities), two reference monitors placed in contrasting environments are preferable. This supports the development of environment-specific calibration models and significantly improves data reliability across the network.
Acceptable reference instruments: BAM (Beta Attenuation Monitor), TEOM (Tapered Element Oscillating Microbalance), or FEM (Federal Equivalent Method) instruments. Ensure the instrument is maintained by a certified operator and serviced per manufacturer specifications throughout the project period.

Sourcing reference-grade equipment

AirQo does not supply reference-grade air quality monitoring equipment. For projects requiring such instruments, we recommend Met One Instruments as a reputable provider. AirQo systems support remote data collection from the BAM 1022, which is widely deployed for regulatory-grade PM measurements: https://metone.com/products/bam-1022/

For research projects seeking to integrate AirQo sensors within a larger hybrid monitoring system, contact support@airqo.net to discuss partnership arrangements, including technical training support, API integration, and data hosting on the AirQo platform.

11.6 Physical Deployment Geometry for Co-location

When co-locating an AirQo sensor with a reference-grade monitor (see Section 11.5), the physical setup materially affects calibration model reliability, sometimes as much as sensor-intrinsic uncertainty. The specifications below are field-tested across AirQo co-location campaigns in Kampala and Nairobi and help ensure the low-cost sensor (LCS) and reference monitor sample a comparable air parcel.

Mounting height — Match the LCS inlet height to the reference monitor's inlet height within ±0.5 m (ideally ±0.3 m). Typical urban neighbourhood-scale sites use 2.0–4.0 m above ground level. Vertical PM gradients near traffic and dust resuspension mean height mismatches can bias results as much as sensor drift.
Horizontal separation — Position units 1–5 m apart, inlet-to-inlet. Never go below 1 m: closer spacing risks the LCS fan/pump exhaust affecting the reference monitor's inlet (or vice versa). Avoid separations beyond roughly 10–20 m so both instruments still sample the same air mass.
Orientation and clearance — Orient all inlets in the same cardinal direction, ideally facing the prevailing wind, to minimise differential rain/dust impaction. Maintain a minimum 2 m clearance radius free of walls, trees, or equipment that could create turbulence or shading; increase to 3–5 m upwind where possible.

Quick reference

Parameter	Target	Why it matters
Vertical offset	≤ ±0.5 m (ideal ±0.3 m)	Minimises bias from vertical PM gradients near traffic/dust sources
Horizontal separation	1–5 m (never below 1 m)	Same air mass, without fan/heat interference
Inlet orientation	Identical, facing prevailing wind	Reduces differential rain/dust impaction across inlets
Clearance radius	≥ 2 m (≥ 3 m upwind preferred)	Avoids turbulence, shading, and localised sources

Documentation — Record inlet heights, separation distance, orientation, and GPS coordinates at setup, and again after any reconfiguration. Photograph each setup: a wide establishing shot, an inlet close-up with a tape measure visible, and any obstacles within roughly 10 m. This geometry metadata should accompany your dataset alongside the calibration model version (see Section 11.1) so results remain auditable and reproducible across sites and seasons.

Seasonal adaptation — In high-dust (e.g., Harmattan/dry season) or high-humidity (wet season) conditions, increase inlet inspection frequency and keep enclosures ventilated but rain-tight; see Section 11.1 for corresponding calibration interval adjustments.

11.1 Calibration Frequency​

11.2 Calibration Approach: Co-location vs. Remote Methods​

11.3 Sensor Movement During Calibration and Maintenance​

11.4 Use of Backup Sensors​

11.5 Reference-Grade Monitor Requirements​

11.6 Physical Deployment Geometry for Co-location​