How do sensors perform compared to reference instruments?

Back to AQEG advice on the use of 'low-cost' pollution sensors page.

Rigorous tests of different types of commercial low-cost sensors are now appearing in the academic literature and these illustrate how important it is to evaluate air pollution sensors against accepted reference methods to understand the quality of their measurements. It is also clear that tests need to be conducted both in ‘real’ ambient air and in a lab environment. Laboratory tests can be useful to identify system-to-system differences, and issues such as hysteresis, drift and cross interference, but field tests under conditions that are close to their intended application are essential for a realistic evaluation of their performance, including over the long-term (weeks to months).

A recent example is from Castell et al., (Environment International, 2017) who provide an extensive evaluation of one of the most widely used multi-pollutant air quality sensor units. The results show some important conclusions that highlight the need for field-testing.

"We found high correlations for all the gaseous pollutants in the laboratory (r > 0.9) when the sensors were tested under steady temperature and relative humidity conditions, while in the field the correlations were significantly lower. Our results clearly show that a good performance in the laboratory is not indicative of a good performance under real-world conditions"

Taken from http://www.sciencedirect.com/science/article/pii/S0160412016309989

Some field tests of commercial sensors have also shown that sensors for different pollutants give markedly divergent quality of performance. In Lewis et al., Faraday Discussions, 2016, the sensor detecting ozone performed well compared to reference instruments, whereas the NO2 sensing component performed very poorly. A simple conclusion is that not all sensors are equal and they must be evaluated individually.

“For ozone a good overall agreement was found between reference measurements and sensors, once sensor calibration values had been linked to the in situ reference observation…..For other air pollution parameters, varying degrees of performance were found. For NO and PM, general trends in atmospheric pollution were recreated by the sensors, although with notable negative and positive biases compared to reference. NO2 sensor measurements from the commercial package did not track trends in reference NO2, and this may be responding to some different air pollution metric”

Taken from http://pubs.rsc.org/en/content/articlepdf/2015/FD/C5FD00201J (PDF)