Scaling accuracy across sensor networks
MCERTS certified low-cost sensors require careful ongoing QA/QC to ensure measurements remain fit for purpose – as highlighted previously. While periodic co-location might be feasible for a few sensors, how do you ensure data quality across tens or hundreds of sensors, without the logistical burden?
We’ve alluded to a solution before and our probabilistic AI framework for network calibration and quality assurance, MODUS+, is that solution. Through a three-year co-location study and operational deployment within the Greater Manchester (GM) sensor network, we’ve demonstrated that indicative class measurements can be maintained at scale without ongoing co-locations.
How does it work?
MODUS+ integrates diverse inputs – including nearby reference monitors, satellite data, and local meteorology – to generate probabilistic pollution predictions at each sensor location that serve as a proxy for a co-located reference. Where inputs lack predictive power, prediction intervals widen, providing an explicit quantification of uncertainty in space and time
Comparing predictions with sensor measurements, the system derives simple linear calibrations with confidence intervals on the slope and intercept. This enables an evidence-based decision on whether and how to correct individual sensors, while preserving traceability to the underlying measurements. Once the sensor has been calibrated, individual measurements can also be compared to the prediction band to flag anomalies. The framework is pollutant and sensor-agnostic and can be applied across diverse networks and operating conditions.
How does it compare to conventional approaches? Let's look at the data.
Twelve low-cost PM sensors were co-located at four reference sites between 2022 and 2025, and for rolling 12-week periods we compared relative expanded uncertainty (Figure 3) from: (i) uncorrected data, (ii) calibration using short-term co-location (10 days), (iii) calibration using full co-location data, and (iv) MODUS+ network calibration. For both PM2.5 and PM10, MODUS+ significantly improves performance compared to uncorrected data and short-term co-location and achieves the 50% relative expanded uncertainty criterion for indicative measurements.
The keen-eyed reader will notice the orange outlier in the ‘corrected with ten days of reference’ pane, which highlights another issue with the conventional approach: the temporary co-location period might not capture representative conditions. In this instance, low measurement variability during the ten-days produced a slope calibration factor that failed to scale later peaks accurately. The usual remedy, extending the co-location, adds to data loss and still offers no guarantee that conditions at the reference site will match the deployment location.
What has this enabled for the GM sensor network?
Greater Manchester required a robust understanding of how pollution levels change across the region in order to explore the impact of local pollution sources, such as domestic wood burning, and aid public engagement. With 45 low-cost sensors and a 2-year measurement campaign, GM needed to maintain indicative-quality measurements at scale.
With MODUS+ deployed, we can independently verify in real time the performance of each sensor in the network. Where drift is detected, a remote calibration is applied – no co-location required, no interruption to data collection.
As an example, the image (right) tracks the estimated bias at the limit value, derived using the slope and intercept, through time and flags when a correction is required to achieve indicative measurements.
Detection of the same pattern across multiple sensors revealed a systematic humidity issue – an insight that was fed back to the manufacturer to further improve future data. Once the humidity issue was resolved, the slope settled around one as expected.
So, what does this mean for LCS users?
The quality of LCS measurements continues to be a concern for users, particularly if used to inform potential policies to improve air quality at a local and national government level. As discussed above, using the “traditional” co-location-based approach is not practicable as the number of LCS in a network increases; the logistics and costs become prohibitive, which flies in the face of “low-cost” and the portrayed ease of use. MODUS+ is a solution: firstly, it provides a traceable workflow from raw measurement to scaled concentration with a quantified quality stamp enabling users to make informed decisions on how to utilise the data; secondly, it does this without the need to carry out regular co-locations for a large number of LCS, reducing the costs for operating such networks and increasing network data capture rates.
MODUS+ integrates with Ricardo’s existing MODUS data management platform to provide a fully comprehensive package that can be applied to all air quality monitoring networks.
For more information on further QA/QC or if you would like advice on how to approach procuring and operating sensors for a particular project or application, get in touch.
