Evaluation and calibration of Aeroqual Series 500 portable gas sensors for accurate measurement of ambient ozone and nitrogen dioxide

Low-power, and relatively low-cost, gas sensors have potential to improve understanding of intra-urban air pollution variation by enabling data capture over wider networks than is possible with ‘traditional’ reference analysers. We evaluated an Aeroqual Ltd. Series 500 semiconducting metal oxide O 3 and an electrochemical NO 2 sensor against UK national network reference analysers for more than two months at an urban background site in central Edinburgh. Hourly-average Aeroqual O 3 sensor observations were highly correlated ( R 2 = 0.91) and of similar magnitude to observations from the UV-absorption reference O 3 analyser. The Aeroqual NO 2 sensor observations correlated poorly with the reference chemiluminescence NO 2 analyser ( R 2 = 0.02), but the deviations between Aeroqual and reference analyser values ([NO 2 ] Aeroq – [NO 2 ] ref ) were highly significantly correlated with concurrent Aeroqual O 3 sensor observations [O 3 ] Aeroq . This permitted effective linear calibration of the [NO 2 ] Aeroq data, evaluated using ‘hold out’ subsets of the data ( R 2  0.85). These field observations under temperate environmental conditions suggest that the Aeroqual Series 500 NO 2 and O 3 monitors have good potential to be useful ambient air monitoring instruments in urban environments provided that the O 3 and NO 2 gas sensors are calibrated against reference analysers and deployed in parallel.


Introduction
Ozone (O 3 ) and nitrogen dioxide (NO 2 ) are very important air pollutants subject to mandatory air quality limits in many jurisdictions. Road traffic and static combustion are major sources of the NO x gases (NO and NO 2 ) leading to pronounced spatiotemporal gradients in NO 2 in urban areas (Cyrys et al., 2012). As a consequence of the fast photochemical cycling between NO x and O 3 , concentrations of O 3 also exhibit strong spatiotemporal variability in urban areas (McConnell et al., 2006;Malmqvist et al., 2014). At present, NO 2 and O 3 are measured using expensive, but traceably-calibrated, fixed-site monitors in sparse networks, or via passive diffusion samplers (Martin et al., 2010;Matte et al., 2013). The former lack spatial resolution, whilst the latter lack temporal resolution. The development of low-power gas-sensitive semiconductor and electrochemical technology has potential to improve understanding of intra-urban air pollution variation by enabling simultaneous data capture, at lower net cost, over wide urban networks (Mead et al., 2013;Williams et al., 2013;Bart et al., 2014), and via peripatetic and mobile sampling designs (Abernethy et al., 2013;Saraswat et al., 2013). However, the quality of the data generated by these monitors compared with established techniques remains a concern (Snyder et al., 2013), in particular interference in the sensing of NO 2 by O 3 (Williams et al., 2009;Mead et al., 2013). One such type of monitor is the Aeroqual Ltd. Series 500 ENV portable gas monitors (www.aeroqual.com/category/products/handheld-monitors). These are relatively compact and lightweight (460 g), and can be operated from an inbuilt battery (for ~8 h) or from mains power. Interchangeable metal oxide semiconductor and electrochemical sensors permit continuous monitoring of a range of gases at low mixing ratios (Williams et al., 2009). The Aeroqual monitors are a factor of approximately 5 to 10 times lower cost than standard air quality monitoring instrumentation for these gases.
In this study, we evaluated the capabilities of two Aeroqual Series 500 portable gas monitors, one with a semiconductor oxide O 3 sensor (OZU 0-0.15 ppm) and one with an electrochemical NO 2 sensor (GSE 0-1 ppm), to measure ambient concentrations of these gases in Edinburgh, UK. We demonstrate the applicability of a linear calibration for the NO 2 sensor using parallel measurements of the O 3 sensor and deployment of both against reference instruments.

Methods
The two Aeroqual monitors were placed under a weatherproof plastic shelter at ~1.5 m elevation above the ground on a post adjacent to the cabin housing the O 3 and NO 2 reference gas analysers of the Edinburgh St. Leonard's air quality monitoring station (55.946 N, 3.182 W). The site is near the centre of the city of Edinburgh, UK, and is classified as urban background in the UK national network (http://uk-air.defra.gov.uk/data). The air inlet for the reference analysers was approximately 1.8 m horizontal distance from and 1.2 m higher than the Aeroqual monitors. The Aeroqual sensor inlets were positioned so that the sensor heads were level with the lower edge of the waterproof shelter and sampled freely flowing ambient air in close vicinity to the reference analysers. The monitoring location was approximately 30 m from the nearest road (with no other primary pollutant sources nearby) hence any differences in pollution concentrations resulting from the small separation distance between the reference analyser and Aeroqual monitor inlets were anticipated to be minor in the comparison of observed concentrations. The Aeroqual units were used as received, with mains power; the waterproof enclosure available from Aeroqual was not used. An Onset HOBO U23 Pro v2 External Data Logger (with solar radiation shield) was also attached to the shelter to record ambient T and RH at 1 min resolution.
The Aeroqual monitors were programmed to record 5-min average concentrations of NO 2 and O 3 continuously between 7 th June and 15 th August 2013. Data were downloaded to a laptop every two weeks, at which time the internal clocks of both monitors were synchronised via the Aeroqual software with the laptop, which was in turn regularly synchronised with Internet Time Servers.
Time stamps for the 5-min averages downloaded from the Aeroqual monitors were adjusted from BST to GMT. The 5-min averages were aggregated to hourly means, denoted as [NO 2 ] Aeroq and [O 3 ] Aeroq . No data capture threshold was set for the averaging.
Both instruments were maintained and calibrated in accordance with the QA/QC protocol for the UK ambient air quality monitoring network (http://uk-air.defra.gov.uk/networks/networkinfo?view=aurn). All data from the reference analysers were subject to the network data review and ratification process. Hourly-averaged NO 2 and O 3 derived from these instruments were downloaded from www.scottishairquality.co.uk, and are denoted as [NO 2 ] ref and [O 3 ] ref .

Results and Discussion
The ambient hourly T (range: 10-33°C; mean ± sd: 19 ± 4°C) in this study was within the operating range of the Aeroqual sensors (5 to 45°C). The vast majority of the hourly RH measurements (2997%; 69 ± 17%) were also in the sensor operating range of 0-95% (<3% of hourly RH measurements were in the range 95-97%). In contrast, the time series and scatter plot in Figure 2 show very limited agreement between the Aeroqual NO 2 sensor and the reference NO 2 chemiluminescence analyser (R 2 = 0.02, and sensor overestimation compared with the reference analyser by approximately 3-fold on average). In contrast, a closer correspondence of an Aeroqual gas-sensitive semiconductor (GSS) NO 2 sensor and reference analyser observations was reported in a similar comparison by Delgado Saborit (2012)  Some sensitivity of gas sensors to ambient water vapour has previously been noted (Bart et al., 2014). Figure 3 shows the relationships between the deviations in the observations of both Aeroqual sensors from their respective reference analyser values and the ambient RH recorded by the HOBO logger. Although the deviations of both sets of Aeroqual values appear to show some trends with RH, these are very weak and the correlations correspondingly poor (R 2 = 0.02 and 0.01, for NO 2 and O 3 , respectively), and over a range in ambient RH from ~30% to almost 100%. The negative relationship with RH for the O 3 sensor is consistent with the observations of Bart et al. (2014), although the latter present a slightly greater negative trend, albeit with considerable scatter as is the case with our data. We observe a small, but again non-significant, positive trend between Aeroqual NO 2 deviations and RH. Overall, we conclude that any systematic impact of RH on our sensor bias and imprecision is limited. In particular, there is no obvious systematic relationship of Aeroqual electrochemical NO 2 sensor observations with RH that might account for the limited agreement between NO 2 sensor and NO 2 reference analyser observations. There were similar lack of associations between 'Aeroqual -reference analyser' O 3 and NO 2 deviations and ambient T (data not shown).
Instead, we examined whether the substantial deviation of Aeroqual electrochemical sensor NO 2 measurement from the reference measurement may have been driven by interference from ambient O 3 . We used the first two-thirds of the measured data (between 7 June and 24 July) as a 'test' dataset to investigate this. Figure  indicates that the measurements used to derive both the calibration relationship and its subsequent application were not subject to long-term drifts on the timescales of the data collection in this study.
The proportion of the full dataset assigned to derivation of calibrated Aeroqual NO 2 values above was arbitrary. suggests that any other cross-interference on the NO 2 sensor is much smaller than that of O 3 .
Finally, it is noted that a potential operational disadvantage of these portable low-power instruments is the minimum ambient operating temperature of 5C currently specified.

Conclusions
An Aeroqual Series 500 ENV O 3 semiconductor oxide gas sensor yielded close agreement with hourly-averaged observations from a reference UV-absorbance O 3 analyser in temperate ambient conditions. Although an Aeroqual NO 2 electrochemical sensor appeared to suffer considerable co-sensitivity to O 3 (to the point of the NO 2 sensor evaluated in this study being inadequate as a measure of NO 2 on its own), it was demonstrated that the O 3 interference can be corrected for by co-deployment with an Aeroqual O 3 sensor plus prior calibration 9 alongside an NO 2 reference instrument. Individual sensor heads may vary in performance so further tests with different instruments at different locations are clearly required to confirm the findings. Overall, however, this study suggests that the Aeroqual Series 500 NO 2 and O 3 monitors could be potentially useful ambient air monitoring instruments.