Summary

Air quality microsensors[i] are necessary to overcome the issue of managing ground-level pollution. However, extensive scientific research meant to compare air quality microsensors with reference stations has showed that when used in real-life environments, microsensors can be influenced by ambient conditions (temperature, relative humidity) and show significant drifts over time[ii], therefore harming the projects led by the users of air quality microsensors. Organizations that are willing to track local pollution with microsensors often find themselves confronted with the natural ageing of sensors and their deterioration, which makes them useless. eLichens has developed a novel recalibration method which continually guarantees the most reliable air quality measurements over a long period of time.

Depending on the pollutants they monitor, air quality microsensors measure variations of voltage, current, electrical impedance, or light attenuation which are converted by a calibration function into a mass or volume concentration, expressed in the units used for air quality (µg.m-3, ppb, etc.). This function is usually determined in laboratory conditions by comparison with a standard. In outdoor conditions, local government agencies monitor air quality with expensive, bulky but very accurate certified instruments – reference stations –and according to a regulated measurement protocol. These measurements are considered as reference measurements.  When deployed in outdoor environments near reference stations, the calibration function of air quality microsensors can be revised to better match sensor measurements to those of reference stations. This is often referred to as recalibration functions.

Many academic publications define recalibration functions (see for a review[iii]).  Although the shapes of these functions differ (from linear functions to more complex algorithms such as support vector machine), the kind of parameters or correction factors  remain largely the same: correction factors for temperature and humidity and determination of offset and sensitivity parameters. Another common feature in these recalibration functions is that their parameters date back from the comparison with a collocated reference station or initial calibration where the microsensor has been placed next to a reference instrument. When the microsensors are deployed during a measurement campaign, the microsensors are not next to a reference instrument allowing to obtain reference measurements with which to update the parameters of the calibration function: the parameters are fixed which leads to two issues:

1. When an air quality microsensor is moved into a new location, e.g. city without reference stations, their performance (concentration accuracy) is uncertain.  

2. As most microsensors drift over time, their performance will deteriorate[iv].

As organizations and communities that invest time and money to improve local air quality need to benefit from the most accurate information easily, everywhere and at all times, these issues strongly limit the use of air quality microsensors. Indeed, recurring needs for manual recalibration to make up for drifts contradict some differentiators generally associated with microsensors such as low-cost maintenance and ease of use.

eLichens has developed eLos, a state-of-the-art microsensors station, which ensures optimal stability by using IoT technology to estimate calibration function parameters dynamically over time.

eLichens’ cloud-based patented solution is shown in Figure 1. This solution is already operational for NO2 and O3 measurements and under development for fine particle measurements (PM10, PM2.5).  The innovation of this method is to obtain frequently (at least once a week) concentration values that can be considered as a reference value at eLos stations even if they are not located near a reference station. This signal, allowing the parameters of the calibration function to be updated with the temperature and humidity measurements recorded in real time by eLos, is based on the existence of periods when the concentration of pollutants can be considered homogeneous over a city. These periods are determined automatically when the concentration measurements of the reference stations of a city or area are within a range of minimum difference.  The average of concentration value is then extracted and is used to re-estimate offset and sensitivity parameters. Thisdynamic recalibration model of eLos is called hyperlocal calibration.

The objective of this solution is to complete the already existing networks of air quality reference stations which are located in large urban areas and to offer ultra-local monitoring to government agencies, local communities or any organization willing to track and improve air quality. However, in the case of a city or area without a refence station, the microsensors are recalibrated using the concentration values on a larger spatial scale.

The colocalization of three eLos stations with an urban background reference station (from the local government air quality agency ATMO AuRA) in Grenoble (France) for more than 17 months has enabled an extensive performance assessment of our solution over time. The two calibrations called hyperlocal and large scale were tested. Conventional indicators (such as correlation coefficient mean absolute error) were used to characterize the degree of agreement of the measurements with those of the reference station. With a data presence rate of more than 95% over the duration of the campaign, Figure 2 shows that the different temporal cycles of NO2 and O3 concentrations (seasonal, weekly, and daily) are very well reproduced with the hyperlocal calibration. Thus, this new technology makes it possible to use microsensor data almost as reference station data in territories where a network of reference stations already exists.

[i] The generic name of “microsensor” corresponds to a sensor or combo sensors that are highly compact and of reduced cost in comparison to instruments certified in regulatory fixed stations. The price generally does not exceed one tenth of the price of regulatory instruments.

[ii] See for a review: Karagulian et al. 2019. Review of sensors for air quality monitoring. JRC Report. Available at ec.europa.eu/jrc

[iii] Maag et al. 2018. A Survey on Sensor Calibration in Air Pollution Monitoring Deployments. IEEE Internet of Things Journal, 5(6), 4857-4870

[iv] See for example: Mijling et al. 2018. Field calibration of electrochemical NO2 sensors in a citizen science context. Atmos. Meas. Tech. 11, 1297-1312.