In many urban areas the population is exposed to elevated levelsof air pollution. However, real-time air quality is usually only measured atfew locations. These measurements provide a general picture of the state ofthe air, but they are unable to monitor local differences. New low-costsensor technology is available for several years now, and has the potentialto extend official monitoring networks significantly even though thecurrent generation of sensors suffer from various technical issues.

Citizen science experiments based on these sensors must be designed carefullyto avoid generation of data which is of poor or even useless quality. Thisstudy explores the added value of the 2016 Urban AirQ campaign, which focusedon measuring nitrogen dioxide (NO₂) in Amsterdam, the Netherlands.Sixteen low-cost air quality sensor devices were built and distributed amongvolunteers living close to roads with high traffic volume for a 2-monthmeasurement period.

Each electrochemical sensor was calibrated in-field next to an air monitoringstation during an 8-day period, resulting inR² ranging from 0.3 to 0.7.When temperature and relative humidity are included in a multilinearregression approach, theNO₂ accuracy is improved significantly,withR² ranging from 0.6 to 0.9. Recalibration after the campaign iscrucial, as all sensors show a significant signal drift in the 2-monthmeasurement period. The measurement series between the calibration periodscan be corrected for after the measurement period by taking a weighted average of the calibrationcoefficients.

Validation against an independent air monitoring station shows goodagreement. Using our approach, the standard deviation of a typical sensordevice forNO₂ measurements was found to be7 µg m⁻³, provided that temperatures are below30 ^∘C. Stronger ozone titration on street sides causes anunderestimation ofNO₂ concentrations, which 75 % of the timeis less than 2.3 µg m⁻³.

Our findings show that citizen science campaigns using low-cost sensors basedon the current generations of electrochemicalNO₂ sensors mayprovide useful complementary data on local air quality in an urban setting,provided that experiments are properly set up and the data are carefullyanalysed.

• Article (PDF, 5081 KB)

• Supplement(152 KB)

• Article(5081 KB)
• Full-text XML
• Supplement(152 KB)
• BibTeX
• EndNote

Because air pollution is difficult to measure, instrumental and operationalcosts of official measurement stations are usually high. Air quality networksin cities, if present at all, are therefore usually sparse. Diffusivesampling is a common addition to these real-time measurements and aresuccessfully used to monitor local differences (see, e.g., Cape, 2009).However, these differences are poorly attributed to an emission source due tothe long averaging time of these measurements (usually monthly). Emerginglow-cost sensor technology has the potential to extend the officialmonitoring network significantly, and improve our understanding of localurban air pollution. Miniaturized and affordable sensors potentially enablecitizens to measure their environment in more detail in space and time (Kumaret al., 2015). Most commercially available sensors, however, suffer fromvarious technical issues which limit their applicability. Despite theirlimitations many experiments are done with air quality devices containingthese sensors, often by motivated but not necessarily scientifically trainedpeople. Comprehensive calibration and validation of these devices is crucial(see, e.g., Lewis and Edwards, 2016; Lewis et al., 2016), but often overlooked.The resulting poor data quality is of concern to health authorities,scientists, and citizens themselves.

Several studies have been done to explore the performance of low-cost airquality sensors (e.g. Jiao et al., 2016; Duvall et al., 2016; Meadet al., 2013; Moltchanov et al., 2015). ForNO₂ monitoring, mostlymetal oxide and electrochemical sensors are used (Borrego et al., 2016;Spinelle et al., 2015b; Thompson, 2016). Typical ambient concentrations ofNO₂ are at parts-per-billion (ppb) level. The main problemsencountered inNO₂ sensor evaluations in these real-worldenvironments are low sensitivity, poor selectivity, low precision andaccuracy, and drift. Metal oxide sensors are especially not very stable(Spinelle et al., 2015b; Thompson, 2016) and suffer from lower selectivity.Therefore, in this study, we opted for electrochemical sensors to measureNO₂.

Mead et al. (2013) already noted the strong interference of ozone and otherambient factors in electrochemicalNO₂ sensors. The performance canbe increased significantly when adding additional measurements of, forexample, temperature and humidity in a regression model or neural network, as shownby, for instance, Piedrahita et al. (2014), Spinelle et al. (2015b), and Masson et al. (2015).Coping with sensor degradation remains a serious issue. Some studies, such asJiao et al. (2016), include an additional temporal term in their linearregression which improves the predictedNO₂ slightly.

In the following sections we assess the data quality of the 2016 Urban AirQcampaign. As with many similar initiatives depending on participating citizens,this campaign was not set up as a strictly controllable scientific experimentsuch as in the previously mentioned studies. However, we will demonstratethat citizen air quality monitoring using the current generation ofelectrochemicalNO₂ sensors may provide useful data of urban airquality, by using a practical method for field calibration and correcting for sensor degradation in retrospect.

Figure 1Locations of the sensor devices during the citizen measurementcampaign. The green marker indicates the calibration location at GGDVondelpark. In the circle the location of SD04 and the GGD station at OudeSchans (in red). The location of Valkenburgerstraat is highlighted inyellow.

The Urban AirQ project explores the added value of alternative air qualitymeasurements in the city by addressing citizens' questions about their localair quality. It focusses on a 2 km × 1 km areaaround Valkenburgerstraat, a primary road in the east-central part ofAmsterdam (see Fig. 1). Its dense traffic causes regular exceedances of theEuropean annual limit value for nitrogen dioxide (40 µg m⁻³).

Two town hall meetings were organized in which residents of this area wereinvited to raise their concerns about air pollution in their neighbourhood andto formulate related research questions. Topics included the relation betweentraffic density and air pollution, the difference between main roads and sidestreets, the front side of an apartment compared to its backside, theinfluence of apartment height, and the influence of cut-through traffic atnighttime. The residents were invited to participate in finding answers totheir questions by measuring their outdoor air quality with 16 experimentallow-cost sensor devices (labelled SD01 to SD16), built for this purpose byWaag Society.

Measurements were done from June to August 2016. Beforehand, the sensordevices were calibrated using side-by-side measurements next to an officialair quality measurement station. With a second calibration period after thecampaign, individual sensor drift was assessed and compensated for in retrospect.

The Urban AirQ experiment is unique in the sense of the used number ofdevices, the duration of the experiment, the direct involvement of citizens,and the use of open hardware and generation of open data.

The approach used in the Urban AirQ project is to build a sensor device with low-costelectronic components which is easy to operate so that citizens can take their ownair quality measurements. It builds on the basic design described by Jianget al. (2016), having an improved power supply, weather resistant housing,WiFi connectivity, and additional sensors for temperature, relative humidity,and particulate matter. The sensor development is part of an open hardwareproject; detailed technical information can be found athttps://github.com/waagsociety/making-sensor.

Figure 2Hardware modules of a sensor device (a), and theintegration in the casing: open (b) and closed (c).

Download

The microcontroller board (Arduino UNO), which handles the readingof the sensors and sends the data to the WiFi module (ESP8266), is central in the design (see Fig. 2).

ForNO₂ measurements, an electrochemical cell is used fromAlphasense Ltd (Essex, UK). The cell contains four electrodes. The targetgas,NO₂, diffuses through a membrane where it is chemicallyreduced at the working electrode, generating a current signal. This electriccurrent is balanced by a opposite current from the counter electrode. Thereference electrode sets the operating potential of the working electrode.The sensor also includes an auxiliary electrode, which is used to compensatefor baseline changes in the sensor. To get full sensor performance, low-noiseinterface electronics are necessary. An individual sensor board withamperometric circuitry, also provided by Alphasense, is used to guaranteea low noise environment and to optimize the sensor resolution at low ppblevels. The sensor signal is read by a 16 bit analogue-to-digital(A∕D) converter (ADS1115). Of the 16 devices, 2 (SD01 and SD02) use modelNO2-B42F forNO₂ measurements and the other 14 use the newerNO2-B43F sensor.

Of the 16 sensor devices, 12 are also equipped with a Shinyei PPD42NS sensorin order to measure particulate matter optically. The present paper, however,will focus only on the assessment of theNO₂ measurements. Alldevices measure internal temperature and relative humidity (RH) with a DHT22sensor from Aosong Electronics.

The system is supplied with a 7.5 V voltage output adapter anda regulator board which generates 5 V for the Arduino and thesensors. The microcontroller consumes 10 mA current (measured). ThePM sensor needs up to 80 mA (measured), theNO₂ sensorabout 10 mA (measured), and the DHT22 less than 1 mA. TheWiFi module peaks periodically at 350 mA when establishing aninternet connection.

Figure 3Raw sensor data, unfiltered but hourly averaged, from the 16 sensorsduring the first calibration period, 2–10 June 2016. The data gap around5 June is due to a connectivity problem to the central database.

Download

3.1 Averaging and filtering

Raw sensor measurements are stored in a central database on a 1 minbase. However, the calibration analysis is based on hourly averages to enabledirect comparison between the ground truth (also provided as hourly values),and to improve the signal-to-noise ratio.

TheNO₂ sensor measurements are done at the working electrode(S_WE) and the auxiliary electrode (S_AE). They areprovided as counts from the A∕D converter. Sensor readings of temperatureand RH are converted according to the indication of the manufacturer todegrees Celsius and percentages respectively.

Raw, hourly averaged sensor data are shown in Fig. 3. The spread intemperature and RH displayed in the raw data is partly explained by thesensor-to-sensor variability. By looking at nighttime temperatures (toeliminate the effect of local heating by exposure to direct sunlight) we seethat the internal sensor temperatures are 2–5 ^∘C higher thanambient temperature. The devices are not actively ventilated, which meansthat the energy dissipation of the electronics influences their internaltemperature. The variable position of the temperature sensors with respect tothese heat sources further explain the variance in temperature and relativehumidity.

Careful filtering is needed before the data can be further processed. We haveapplied the following rules:

• Raw, minute-basedS_WE andS_AE measurements outside a ±10 % range of their meanvalue during the entire measuring period are considered outliers. Thisfilters out 0.33 % of all measurements. This criterion was used for itssimplicity and effectiveness. Note that, due to the large offset in the rawS_WE andS_AE signal, realisticNO₂ peakvalues are still detectable as the corresponding sensor response is stillwithin a 10 % bandwidth.

• All readings at sensor temperatures above 30 ^∘C are discarded to avoid non-linear temperaturedependence of the electrochemicalNO₂ sensor (seeSect. 4.4). This filters out 4.53 % of the measurements during theentire period.

• At least 20 valid minute-based measurements are required to calculate a representative hourly mean. This criterionwas found to be a good trade-off between noise reduction by averaging and notlosing too many hourly measurements.

During the first calibration period, the sensors took measurements 79 % ofthe time on average. After applying the criteria above, this resulted in70 % valid hourly measurements. During the measurement campaign, thesensors produced 79 % valid hourly measurements on average, with theuptime dropping to 50 % in places were sensors experienced connectivityproblems due to limited range of the participant's WiFi network.

Figure 4Box-and-whisker diagrams of hourly ambient parameters during the twocalibration periods and the measurement campaign. The box edges indicate the25th–75th percentile; the whiskers the minimum and maximum values. Themedian is indicated in red. Temperature and RH are based on the averagevalues of all sensors devices,NO₂ and ozone are taken from thereference station at Vondelpark. For comparison,NO₂ from thereference station at Oude Schans (OS) is also shown.

Download

Electrochemical sensors such as the Alphasense NO2-B series are known to besensitive to interfering species and ambient factors. Ozone,temperature, and relative humidity, in particular, influence the sensor reading (see,e.g.,Spinelle et al., 2015a).

4.1 Explaining theNO₂ sensor signal

To understand better the behaviour of theNO₂ sensor, we study itssensitivity to different ambient factors. We use the first calibration periodto test the correlation of the measuredS_WE andS_AEsignal withNO₂, ozone, temperature, and humidity by making a bestfit though the hourly time series:

$$\begin{array}{}\text{(1)}& {\displaystyle }{\displaystyle }{S}_{\text{WE}}\left(t\right)=c_{\mathrm{0}}+c_{\mathrm{1}}{\mathrm{NO}}_{\mathrm{2}}\left(t\right).\end{array}$$

Temperature and RH were not readily available from the GGD Vondelpark stationdata. We take temperature and RH from the average readings from the DHT22sensors instead, which better reflect the internal sensor conditions thanambient air measurements.

Figure 5Typical sensor performance (SD10) explained as a linear regression of respectivelyNO₂,O₃, T, RH,and all variables.(a) The results for the working electrode and(b) for the auxiliary electrode. The axes represent the A∕Dconverter counts, which are proportional to the currents generated by thesensor at the corresponding electrode.

Download

Table 1Fit results for regression model A. Older NO2-B42F sensorshighlighted in bold.

Download Print Version |Download XLSX

Table 2Regression models forNO₂.

Download Print Version |Download XLSX

Figure 5 shows scatter plots for an average performing sensor and theR², the coefficient of determination. The measuredS_WEsignal can be explained by ambientNO₂ (R² = 0.20), butbetter by its anti-correlation with ozone (R² = 0.49). Temperaturealone is an even better predictor for the sensor signal (R² = 0.73),because of the sensors' direct dependence on temperature, and indirectdependence on temperature (being a reasonable proxy for bothNO₂andO₃ concentrations). The correlation with relative humidityis also very strong (R² = 0.73). The measuredS_WE signal canbest be explained as a linear combination ofNO₂,O₃, T, and RH together, resulting in a correlation of 0.98(R² = 0.96).

TheS_AE signal is practically insensitive toNO₂. Thissuggests that a combination ofS_WE andS_AE is moresensitive toNO₂ and less to the other interfering factors, asintended by the manufacturer.

4.2 NO₂ calibration models

ForNO₂ measurements, the sensor manufacturer suggest correctingboth working electrode and auxiliary electrode for a zero offset withS_WE,0 andS_AE,0 respectively. Then a sensitivityconstant s is applied to convert from mVto ppb NO₂:

$$\begin{array}{}\text{(2)}& {\displaystyle }{\displaystyle }{\mathrm{NO}}_{\mathrm{2}}\left[\mathrm{ppb}\right]={\displaystyle \frac{(S_{\text{WE}}-S_{\text{WE},\mathrm{0}})-(S_{\text{AE}}-S_{\text{AE},\mathrm{0}})}{s}}.\end{array}$$

In practice, the factory-supplied constantsS_WE,0,S_AE,0, and s do not result in realistic values ofNO₂; see, e.g., Cross et al. (2017). As an alternative, we proposea linear combination of the signalsS_WE andS_AE(calibration model A):

$$\begin{array}{}\text{(3)}& {\displaystyle }{\displaystyle }{\mathrm{NO}}_{\mathrm{2}}\left[\mathrm{\mu }\mathrm{g}{\mathrm{m}}^{-\mathrm{3}}\right]=c_{\mathrm{0}}+c_{\mathrm{1}}{S}_{\text{WE}}+c_{\mathrm{2}}{S}_{\text{AE}}.\end{array}$$

The coefficientsc₁ andc₂ are determined with data from thecalibration period using ordinary least squares (OLS). As can be seen fromthe fit results in Table 1, within the batch of sensors there is a largevariability of direct sensitivity to ambientNO₂.

During the calibration period, hourly ozone values (also taken from theVondelpark station) happened to be a good proxy for the ambientNO₂concentration:NO₂(t) = 44.6 − 0.40 ⋅ O₃(t) inµg m⁻³, withR² of 0.49.

When compared with Table 1, it can be seen that direct sensor readings froma fair part of the sensors cannot outperform this result. To improve theresults we use additional measurements and their statistical relation toNO₂. We fit different calibration models with multiple linearregression (using OLS). The calibration models which were tested are listedin Table 2.

Temperature and RH are taken from the DHT22 sensor. Note that there is noneed to calibrate the individual T and RH sensor signals beforehand; thecalibration coefficients forNO₂ are determined for the specificset of all sensors in the box. However, this means that if an individualsensor is replaced, new calibration parameters for the sensor box have to bederived.

Table 3Fit results for regression model D. Older NO2-B42F sensorshighlighted in bold.

Download Print Version |Download XLSX

4.3 Calibration results

A complete overview of the regression coefficients and their error estimatesfor all models can be found in the Supplement. The sign of the calibrationparameters can be easily understood. As the electrochemicalNO₂sensor loses sensitivity at higher temperatures (see the negative slope inFig. 7b for temperatures below 30 ^∘C), coefficients c₃are positive to compensate for this effect. The additional sensor responsedue to cross-sensitivity with ozone is compensated for by negative values forc₅.

From the fit results we see that model B (including RH) performs better thanmodel A, but model C (including T) outperforms model B. When both RHand T are included (model D) the results of model C are marginallyimproved. This can be understood in terms of strong sensor dependence ontemperature, weak dependence on RH, and the collinearity betweentemperature and RH. Note that measuring RH is essential for guarding the dataquality of electrochemical sensors, as these sensors are very sensitive tosudden changes in RH (see, e.g., Alphasense, 2013; and Panget al., 2016).

The best calibration results (i.e.R² values closer to 1) are obtainedby including ozone (model E). The ozone values were obtained from the GGDVondelpark station, as the sensor devices do not measure ozone themselves.

As local ozone measurements were only available during the calibrationperiods, we used model D for the Urban AirQ campaign, i.e. generating anNO₂ value based on a linear combination ofS_WE,S_AE, T, and RH. The regression analysis of model D andcorrelation with theNO₂ ground truth can be found in Table 3.

Figure 6(a) Calibration model results for an average performingsensor (SD15). Bottom row shows the recommended calibration by model D(left), and the results when ozone is included (right).(b) Time series compared to ground truth with calibration parametersof model A and D.

Download

The two worst-performing sensor devices (SD02 and SD01) contain the olderNO2-B42F sensor. The newer NO2-B43F model is designed to have highersensitivity toNO₂ and less interference of ozone. The old sensormodel has indeed smaller coefficients forS_WE and largercorrection terms for ozone (see thec₁ andc₅ coefficients ofmodel E in the Supplement). This, however, can also be related to theirlonger operating time, as both sensors have been used in previous experimentsfor more than a year. Again, it can be seen that even within the same batchof sensors there is a significant spread in performance, around a medianvalue forR² of 0.83. Figure 6 shows the results for the differentcalibration models for the average performing sensor SD15. The time series inFig. 6b shows clearly how the performance of a typical sensor device improveswhen temperature and humidity are included in the calibration analysis. TheadjustedR², which correctsR² for the number of explanatoryvariables, increases from 0.29 to 0.82. Note that$R_{\text{adj}}^{\mathrm{2}}$ isonly slightly smaller thanR², as the number of observations(n ≈ 150) is relatively high compared to the number of regressionvariables (k = 2…5).

Figure 7(a) Examples of negative spikes in the calibratedNO₂ measurements (solid line) due to internal sensor temperatures(dotted line) exceeding 30 ^∘C.(b) Variation of zerooutput of the working electrode caused by changes in temperature fora typical batch of electrochemical sensors. Image taken from Alphasense DataSheet for NO2-B43F (Alphasense, 2016).

Download

4.6 Predictivity, sensor drift, and uncertainty estimation

Almost all electrochemical sensors have some degree of drift because of agingand poisoning (Di Carlo et al., 2011; Hierlemann and Gutierrez-Osuna, 2008).This becomes a serious complication when the drift is of the order of thestrength of the signal of interest. The idea of keeping sensor SD03 next tothe reference station during the whole campaign was to study sensordegradation in more detail. Unfortunately, the sensor was removed temporarilyfrom 10 to 14 July for service, when it was decided to add a PM module to thedevice. The increased energy dissipation after the modification (the ShinyeiPPD42NS module uses a heater resistor to force a convective flow of samplingair) caused an increase of the internal device temperature by2.5 ^∘C on average. This sudden jump in temperature disruptedthe reference time series.

Table 4Descriptive and short-term predictive error of model Din µg m⁻³.

Download Print Version |Download XLSX

Instead, to assess the short-term stability of the calibration model, we usethe first 60 % of the measurements from the calibration period(2–7 June) to derive the regression coefficients, and predict theNO₂ values for the remaining 40 % (8–10 June; see Table 4).The average RMSE increases from 6.5 to 7.0 µg m⁻³when the regression is used for prediction.

Figure 8Sensor drift during 2 months of operation, shown as thedistribution of residuals (in 2 µg m⁻³ bins) with thereference measurements during the first calibration period (black bars) andduring the second period (red bars).

Download

We assess the long-term stability of the sensors with a second calibrationperiod after measurement campaign, again at the Vondelpark calibration site.As can be seen from the distribution of the residuals in Fig. 8, most sensorsdrift significantly in the intermediate 2-month period. We describe thisdegradation effect as a bias b between the mean of the hourly estimatedNO₂ values${\widehat{x}}_i$ and the mean of the hourly trueNO₂ x_i during the calibration period:

$$\begin{array}{}\text{(4)}& {\displaystyle }{\displaystyle }b={\displaystyle \frac{\mathrm{1}}{N}}\sum _{i=\mathrm{1}}^N{\widehat{x}}_i-{\displaystyle \frac{\mathrm{1}}{N}}\sum _{i=\mathrm{1}}^N{x}_i,\end{array}$$

and the root-mean-square error (RMSE) of the difference between the bias-corrected calibrated measurement and the ground truth. The latter is the sameas the standard deviation of the residuals (SDR)${\widehat{x}}_i-x_i$:

$$\begin{array}{}\text{(5)}& {\displaystyle }{\displaystyle }\text{SDR}=\sqrt{{\displaystyle \frac{\mathrm{1}}{N}}{\sum }_i{\left(\left({\widehat{x}}_i-b\right)-x_i\right)}^{\mathrm{2}}}.\end{array}$$

Table 5Bias and random error inµg m⁻³ when calibrated inthe first period with model D.

Download Print Version |Download XLSX

As can be seen in Table 5, the bias is mostly positive. Note that sensor SD16and SD01 had a limited uptime in the second period, which makes their biasand RMS calculation not very representative.

The strongest bias after 2 months is found for SD02 and SD01. Both are ofmodel NO2-B42F and have been used in others experiments for more than 1year. These sensors also have the largest RMSE in the first calibrationperiod (see also Table 3), which is another indication of their poorperformance. The range in RMSE of the remaining sensors is4.5–7.2 µg m⁻³ for the first period. The bias-corrected RMSEincreases to 5.3–9.3 µg m⁻³ for the second period. Thelatter is a more conservative yet more realistic estimation of the precisionof theNO₂ estimates, as they are based on measurements which werenot used for calibration. Based on our results listed in the last columns ofTables 4 and 5, we take 7 µg m⁻³ as a typical uncertainty forthe estimatedNO₂ values.

The increase of SDR is also due to a loss of sensitivity over time. The agingof the sensors can be further investigated by recalibrating the devices, i.e.determining the coefficients of regression model D, using the data of thesecond calibration period (see the Supplement). All calibration coefficientsofS_WE (the only component which has direct sensitivity toNO₂) decrease in value, showing that all sensors suffer fromsensitivity loss toNO₂. This results in lowerR² values,although the performance loss is partly compensated for by the other componentsin the regression. The older Alphasense NO2-B42F sensors suffer the largestsensitivity loss, which (although the regression tries to compensate withincreased temperature dependence) results in the worst performance loss interms of R².

Figure 9(a) Comparison of sensor SD04NO₂ time serieswith the nearby Oude Schans station (8-day snapshot), and the effect of biascorrection. For comparison, measurements of Vondelpark station are alsoshown.(b) Distribution of residuals ofNO₂ measurementsbetween sensor SD04 and Oude Schans station during the campaign period, withand without bias correction.

Download

4.7 Weighted calibration

Taking 18 µg m⁻³ as a typicalNO₂ concentration inan urban environment (Fig. 4), the sensor drift as listed in Table 5 isa significant error component, even after a 2-month period. It isimpossible to predict the progressing bias for an individual sensor. However,using the second calibration period we can compensate for signal drift after the measurement period. If${\widehat{x}}_{\mathrm{1}}\left(t\right)$ represents the estimatedNO₂ valueat time t based on the first calibration period (starting at t₁), and${\widehat{x}}_{\mathrm{2}}\left(t\right)$ the estimatedNO₂ value based on the secondcalibration period (ending at t₂), then we take for intermediate times$t_{\mathrm{1}}\le t\le t_{\mathrm{2}}$ as a weighted average of both calibrations:

$$\begin{array}{}\text{(6)}& {\displaystyle }{\displaystyle }\widehat{x}\left(t\right)=(\mathrm{1}-f(t\left)\right){\widehat{x}}_{\mathrm{1}}\left(t\right)+f\left(t\right){\widehat{x}}_{\mathrm{2}}\left(t\right).\end{array}$$

Assuming that the sensor degradation is linear in time we select

$$\begin{array}{}\text{(7)}& {\displaystyle }{\displaystyle }f\left(t\right)=(t-t_{\mathrm{1}})/(t_{\mathrm{2}}-t_{\mathrm{1}}),\end{array}$$

such thatf(t₁)=0 andf(t₂)=1.

Table 6Comparison of sensor SD04 with Oude Schans station during thecampaign period, according to different calibrations.

Download Print Version |Download XLSX

The Alphasense NO2-B4 sensor is used to measure ambientNO₂ in many low-cost air qualitysettings. As all electrochemicalNO₂ sensors, it is not very selective regarding the target gas. The sensorresponse can be explained well by a linear combination ofNO₂,O₃, temperature, and relative humidity signals(R² ≈ 0.9).

As a consequence, a linear combination of the working electrode and theauxiliary electrode alone gives a poor indication of ambientNO₂concentrations. The accuracy varies greatly between different sensors(R² between 0.3 and 0.7). For the Urban AirQ campaign, temperature andrelative humidity were included in a multilinear regression approach. Theresults improve significantly withR² values typically around 0.8. Thiscorresponds well with the findings of Jiao et al. (2016), who find anadjustedR² = 0.82 for the best-performing electrochemicalNO₂ sensor in their evaluation, when including T and RH.

Best results are obtained by also including ozone measurements in thecalibration model:R² increases to 0.9. Spinelle et al. (2015b) useda similar regression and foundR² ranging from 0.35 to 0.77 forfour electrochemicalNO₂ sensors during a 2-week calibration period,but dropping to 0.03–0.08 when applied to a successive 5-month validationperiod. LowNO₂ values at their semi-rural site partly explainthis poor performance, but it is most likely that there were also unaccounted-for effects such aschanging sensor sensitivity and signal drift.

The sensor devices were tested in an Amsterdam urban background insummertime, withNO₂ values ranging from 3 to78 µg m⁻³, and median values around15 µg m⁻³. During the 3-month period most sensors show lossof sensitivity and significant drift, ranging from−9 to21 µg m⁻³. After bias correction we found a typical value forthe accuracy of theNO₂ measurements of 7 µg m⁻³.

This error consists of several components. The reference measurements by theNO∕NO_x analysers have an estimated hourly error of3.65 % (certified validation at a 200 µg m⁻³NO₂ concentration), which would contribute to0.5 µg m⁻³ under typical conditions. The low-cost DHT22sensor has a reported error of 0.5 ^∘C for temperature and2–5 % for RH. For a single measurement, this would contribute toa propagated regression error of approximately 1 and0.5 µg m⁻³ respectively. It should be noted, however, thatbinning minute-based measurements to hourly averages removes a large part ofthe variability, while determining the best fitting regression model for eachsensor device removes large part of the remaining systematical biases. Thelargest part of the error term is therefore introduced by the linearregression model itself, which does not include all interfering species ormeteorological quantities, and is not able to describe non-lineardependencies of its variables. One should therefore be careful extrapolatingthe calibration model for conditions different than the calibration period.

The validation results from Sect. 4.8 show that the calibration holdswell for urban locations with similarNO₂∕O₃ ratios.NeglectingO₃ as a regression parameter, however, will introducea bias at locations with differentNO₂∕O₃ ratios found,e.g. closer to emission sources. To get a better understanding of thepossible impact, we compared hourly ozone measurements from the GGDauthorities at Van Diemenstraat (VDS, classified as street station) againstNieuwendammerdijk (NDD, classified as urban background station) duringJune–August 2016. The relation can best be described by[O₃]_VDS = 0.87[O₃]_NDD + 0.85 (with 0.93 correlation), which meansthat ozone levels at the street station are typically 13 % lower, due totitration ofO₃ with NO. Due to the sensor's cross-sensitivity forozone, larger values must be subtracted from its signal when the ozoneconcentration increases. This explains the negative sign of the ozonecoefficient c₅ of model E (see Supplement). Calibration with model Dovercorrects (i.e. subtracts too much) for locations which have lowerozone concentrations than at the calibration site, resulting in anunderestimation ofNO₂ concentrations. Using typical valuesc₅ = −0.3 and [O₃] = 60 µg m⁻³(75th percentile of the distribution during the measurement camping,according to Fig. 4), we estimate the underestimation of road-sideNO₂0.3 × 13 % × 60 = 2.3 µg m⁻³.

The found sensor accuracy after weighted calibration is good enough toprovide some complementary spatial information on local air quality betweenreference stations. When looking at the difference between Vondelpark stationand Oude Schans station (both classified as city background stations) in theperiod June–August 2016, 22 % of the hourly measurements differ morethan 7 µg m⁻³, and 6 % of the hourly measurements differmore than 14 µg m⁻³. These differences increase further whenconsidering road-side stations. From this perspective, even sensor deviceswith an accuracy around 7 µg m⁻³ can contribute to animproved understanding of spatial patterns. However, it must be furtherinvestigated if the calibration method used here can provide realisticestimates for peak values (such as the EU hourly limit value,200 µg m⁻³).

In this study, we examined low-cost electrochemical air quality sensors forcitizen urban air quality monitoring. In other words, we evaluated animperfect air quality sensor in an imperfect scientific experiment. Ingeneral, we found that low-cost electrochemical sensors have the potential tocomplement official environmental monitoring data to help answer questionsfrom the public, which usually cannot be fully answered from official dataalone. To reach full potential, however, proper measurement set-up,calibration and recalibration, and data analysis should be guaranteed.

The current generation of low-costNO₂ sensors has some seriousissues which make straightforward application difficult. To make electrochemicalNO₂ sensor measurements accurate, careful filtering of the raw datais necessary. There is a strong spread in sensor performance, even if thesensors come from the same batch, which makes individual calibrationessential. A practical calibration method is to measure side by side with an airmonitoring station. The accuracy of the measurements can be improved byincluding temperature and humidity measurements from other low-cost sensorsin a multilinear regression approach. It is worth noting that more advancedcalibration algorithms such as by Cross et al. (2017) and Muelleret al. (2017) could give better results, but this is not the focus of thispaper. It is hard to quantify an optimal length of a calibration periodwithout having a proper understanding of the sensor degradation ratebeforehand. The measurement period should be at least a few days to capturethe sensors behaviour under a wide range of pollution levels andmeteorological conditions. Very long calibration periods (of the order ofmonths) will cause sensor degradation issues to interfere with thecalibration results.

Startup time of sensors is estimated to be 4 h. To avoid nonlinear response of theelectrochemical sensor at elevated temperatures, we filter out measurementsabove 30 ^∘C. This is not a serious restriction forapplicability in moderate climates such as in the Netherlands, provided thatthe sensor is protected from direct sunlight. However, for warmer regions orduring heatwaves this may reduce the data stream considerably, unless thetemperature dependencies are better captured by more advanced regressionmodels.

The calibration seems to be location independent, as long as theNO₂ ∕ O₃ ratio is comparable. Road-side applicationis likely to introduce a small positive bias. Calibration coefficients arenot constant in time. During the 3-month period most sensors suffer fromsignificant sensitivity loss and drift. The strongest drift and largestuncertainty are found for the older NO2-B42F sensors. It remains unclear ifthe worse performance is related to the sensor model or to longer usage infield experiments.

The sensor degradation makes practical applications in operational urbannetworks difficult. Smart re-calibration programs, such as bringing back sensorsto a calibration facility on a regular basis or recalibrating on the spotwith a travelling reference instrument, are essential. New data-driven techniques, such asBayesian networks (e.g. Xiang et al., 2016), might offer a solution for thisproblem.

On the hardware side, we recommend including active ventilation to guaranteeconstant air flow over the gas sensor and suppress unwanted internaltemperature changes due to heating of electronic components. To improve theNO₂ measurements further we recommend including an additionallow-cost ozone sensor, e.g. Ox-B431 by Alphasense. It is likely that thelinear regression approach is able to resolve a significant part of thecross-sensitivity to ozone andNO₂. The RH sensor signal should beused more intelligently to detect and filter sudden changes in relative humidity.Adding a local data logger is also recommended to be able to recover datafor periods when the WiFi connection to the central database is lost.

A complete overview of fit results for all models can befound in the Supplement. The hourly Urban AirQ sensor data, calibratedafter the measurement period by interpolating the calibration in time between two calibrationperiods, can be downloaded athttps://github.com/waagsociety/making-sensor (KNMI-Waag Society, 2016).

The supplement related to this article is available online at: https://doi.org/10.5194/amt-11-1297-2018-supplement.

The authors declare that they have no conflict of interest.

The Urban AirQ project was partly funded by a 2016 Stimulus Grant from AMS(Advanced Metropolitan Solutions). The project is also part of Making Sense,funded by European Union's Horizon 2020 research and innovation programme.Qijun Jiang is supported by the China Scholarship Council for his PhDresearch. The authors would like to thank Emma Pareschi from Waag Society, whowas responsible for the hardware development.

Edited by: Piero Di Carlo
Reviewed by: David Ramsay and twoanonymous referees

Alphasense: AAN 105-03, Alphasense Application Note: Designing a Potentiostatic Circuit,March 2009, available at:http://www.alphasense.com/WEB1213/wp-content/uploads/2013/07/AAN_105-03.pdf(last access: 1 March 2018),2009. 

Alphasense: AAN 110, Alphasense Application Note on Environmental Changes: Temperature,Pressure, Humidity, available at:http://www.alphasense.com/WEB1213/wp-content/uploads/2013/07/AAN_110.pdf(last access: 1 March 2018), 2013. 

Alphasense: ADS, Alphasense Data Sheet for NO2-B43F, April 2016, available at:http://www.alphasense.com/WEB1213/wp-content/uploads/2017/07/NO2B43F.pdf(last access: 1 March 2018), 2016. 

Borrego, C., Costa, A. M., Ginja, J., Amorim, M., Coutinho, M., Karatzas, K.,and Penza, M.: Assessment of air quality microsensors versus referencemethods: the EuNetAir joint exercise, Atmos. Environ., 147, 246–263,https://doi.org/10.1016/j.atmosenv.2016.09.050, 2016. 

Cape, J. N.: The use of passive diffusion tubes for measuring concentrationsof nitrogen dioxide in air, Crit. Rev. Anal. Chem., 39, 289–310,https://doi.org/10.1080/10408340903001375, 2009. 

Cross, E. S., Williams, L. R., Lewis, D. K., Magoon, G. R., Onasch, T. B.,Kaminsky, M. L., Worsnop, D. R., and Jayne, J. T.: Use of electrochemicalsensors for measurement of air pollution: correcting interference responseand validating measurements, Atmos. Meas. Tech., 10, 3575–3588,https://doi.org/10.5194/amt-10-3575-2017, 2017. 

Di Carlo, S., Falasconi, M., Sanchez, E., Scionti, A., Squillero, G., andTonda, A.: Increasing pattern recognition accuracy for chemical sensing byevolutionary based drift compensation, Pattern Recogn. Lett., 32, 1594–1603,https://doi.org/10.1016/j.patrec.2011.05.019, 2011. 

Duvall, R., Long, R., Beaver, M., Kronmiller, K., Wheeler, M., andSzykman, J.: Performance evaluation and community application of low-costsensors for ozone and nitrogen dioxide, Sensors, 16, 1698,https://doi.org/10.3390/s16101698, 2016. 

Hierlemann, A. and Gutierrez-Osuna, R.: Higher-order chemical sensing, Chem.Rev., 108, 563–613,https://doi.org/10.1021/cr068116m, 2008. 

Jiang, Q., Kresin, F. Bregt, A. K. Kooistra, L., Pareschi, E., van Putten, E. Volten, H., and Wesseling, J.: Citizen sensing for improved urban environmental monitoring,J. Sensors, 2016, 5656245,https://doi.org/10.1155/2016/5656245, 2016. 

Jiao, W., Hagler, G., Williams, R., Sharpe, R., Brown, R., Garver, D.,Judge, R., Caudill, M., Rickard, J., Davis, M., Weinstock, L.,Zimmer-Dauphinee, S., and Buckley, K.: Community Air Sensor Network(CAIRSENSE) project: evaluation of low-cost sensor performance in a suburbanenvironment in the southeastern United States, Atmos. Meas. Tech., 9,5281–5292,https://doi.org/10.5194/amt-9-5281-2016, 2016. 

KNMI-Waag Society: UrbanAirQ NO₂ final, available at:https://github.com/waagsociety/making-sensor/blob/master/data/urbanairq_no2_final.csv(last access: 1 March 2018), 2016. 

Kumar, P., Morawska, L., Martani, C., Biskos, G., Neophytou, M., DiSabatino, S., and Britter, R.: The rise of low-cost sensing for managing airpollution in cities, Environ. Int., 75, 199–205,https://doi.org/10.1016/j.envint.2014.11.019, 2015. 

Lewis, A. and Edwards, P.: Validate personal air-pollution sensors, Nature,535, 29–31,https://doi.org/10.1038/535029a, 2016. 

Lewis, A. C., Lee, J. D., Edwards, P. M., Shaw, M. D., Evans, M. J.,Moller, S. J., and White, A.: Evaluating the performance of low cost chemicalsensors for air pollution research, Faraday Discuss., 189, 85–103,https://doi.org/10.1039/c5fd00201j, 2016. 

Masson, N., Piedrahita, R., and Hannigan, M.: Quantification method forelectrolytic sensors in long-term monitoring of ambient air quality, Sensors,15, 27283–27302, 2015. 

Mead, M. I., Popoola, O. A. M., Stewart, G. B., Landshoff, P., Calleja, M.,Hayes, M., and Jones, R. L.: The use of electrochemical sensors formonitoring urban air quality in low-cost, high-density networks, Atmos.Environ., 70, 186–203,https://doi.org/10.1016/j.atmosenv.2012.11.060, 2013. 

Moltchanov, S., Levy, I., Etzion, Y., Lerner, U., Broday, D. M., andFishbain, B.: On the feasibility of measuring urban air pollution by wirelessdistributed sensor networks, Sci. Total Environ., 502, 537–547,https://doi.org/10.1016/j.scitotenv.2014.09.059, 2015. 

Mueller, M., Meyer, J., and Hueglin, C.: Design of an ozone and nitrogendioxide sensor unit and its long-term operation within a sensor network inthe city of Zurich, Atmos. Meas. Tech., 10, 3783–3799,https://doi.org/10.5194/amt-10-3783-2017, 2017. 

Pang, X., Shaw, M. D., Lewis, A. C., Carpenter, L. J., and Batchellier, T.:Electrochemical ozone sensors: a miniaturised alternative for ozonemeasurements in laboratory experiments and air-quality monitoring, Sensor.Actuat. B-Chem., 240, 829–837,https://doi.org/10.1016/j.snb.2016.09.020, 2016. 

Piedrahita, R., Xiang, Y., Masson, N., Ortega, J., Collier, A., Jiang, Y.,Li, K., Dick, R. P., Lv, Q., Hannigan, M., and Shang, L.: The next generationof low-cost personal air quality sensors for quantitative exposuremonitoring, Atmos. Meas. Tech., 7, 3325–3336,https://doi.org/10.5194/amt-7-3325-2014,2014. 

Spinelle, L., Gerboles, M., and Aleixandre, M.: EUROSENSORS 2015: performanceevaluation of amperometric sensors for the monitoring ofO₃ andNO₂ in ambient air atppb level, Procedia Engineer., 120,480–483, 2015a. 

Spinelle, L., Gerboles, M., Villani, M. G., Aleixandre, M., andBonavitacola, F.: Field calibration of a cluster of low-cost availablesensors for air quality monitoring. Part A: Ozone and nitrogen dioxide, Sens.Actuat. B-Chem., 215, 249–257,https://doi.org/10.1016/j.snb.2015.03.031, 2015b. 

Thompson, J. E.: Crowd-sourced air quality studies: a review of theliterature and portable sensors, Trends in Environmental AnalyticalChemistry, 11, 23–34,https://doi.org/10.1016/j.teac.2016.06.001, 2016. 

Xiang, Y., Tang, Y., and Zhu, W.: Mobile sensor network noise reduction andrecalibration using a Bayesian network, Atmos. Meas. Tech., 9, 347–357,https://doi.org/10.5194/amt-9-347-2016, 2016. 

• Article(5081 KB)
• Full-text XML

• Supplement(152 KB)
• BibTeX
• EndNote

• Abstract
• Introduction
• The Urban AirQ project
• Urban AirQ sensor devices
• NO₂ calibration
• Discussion
• Conclusions and outlook
• Data availability
• Competing interests
• Acknowledgements
• References
• Supplement