the Creative Commons Attribution 4.0 License.
the Creative Commons Attribution 4.0 License.
| 15 Nov 2024
Seasonal effects in the application of the MOMA remote calibration tool to outdoor PM2.5 air sensors
Abstract. Air sensors are being used more frequently to measure hyper-local air quality. The PurpleAir sensor is among one of the most popular air sensors used worldwide to measure fine particulate matter (PM2.5). However, there is a need to understand PurpleAir data quality especially under different environmental conditions with varying particulate matter (PM) sources and size distributions. Several correction factors have been developed to make the PurpleAir sensor data more comparable to reference monitor data. The goal of this work was to determine the performance of a remote calibration tool called MOment MAtching (MOMA) for temporally varying PM2.5 sources. MOMA performs calibrations using reference site data within 0–15 km from the sensor. Data from 20 PurpleAir sensors deployed across a network in Phoenix, Arizona from July 2019 to April 2021 were used. The results showed that the MOMA calibration tool improved the accuracy of PurpleAir sensor data across Phoenix and was comparable to the EPA correction factor with a root mean square error (RMSE) of 4.19 – 7.92 µg m-3 vs. 4.23 – 9.27 µg m-3. However, MOMA provided a better estimate of daily exceedances compared to the reference data for smoke conditions. Using speciated PM data, MOMA was able to distinguish between different PM sources such as winter wood burning, and wildfires and dust events in the summer.
Competing interests: The authors declare the following financial interests/personal relationships which may be considered as potential competing interests: L.W. and G.H. are employees of Aeroqual Ltd, a sensor manufacturer. G.H. is a founder and shareholder in Aeroqual Ltd.
Publisher's note: Copernicus Publications remains neutral with regard to jurisdictional claims made in the text, published maps, institutional affiliations, or any other geographical representation in this preprint. The responsibility to include appropriate place names lies with the authors.- Preprint
(1250 KB) - Metadata XML
- BibTeX
- EndNote
Status: final response (author comments only)
-
RC1: 'Comment on egusphere-2024-3454', Anonymous Referee #1, 08 Jan 2025
Line 35-38: Based on your statement, I’m curious about how the correction process is applied. If the 3-day rolling average triggers a correction after exceeding the threshold for 4 consecutive days, would the correction only apply to data collected from Day 4 onward, or would it retroactively adjust the data for Days 1, 2, and 3 as well?
Line 111: Could you clarify whether the speciation data is collected every third day? If so, does this limit the number of days labeled as "Dust" that actually have corresponding speciation data? Given that dust events are short-lived, does the speciation data reveal any significant differences in crustal composition between days labeled as “Dust” and “Urban”? In Figure 4, the comparison of crustal composition shows little difference between the means for Dust and Urban. Could this be influenced by the disparity in the number of days analyzed (20 Dust days vs. ~600 Urban days)?
Lines 103–106 state: “The reference network consisted of 1 Beta Attenuation Monitor (BAM) at the Phoenix JLG Supersite and 8 dichotomous Taper Element Oscillating Microbalance (TEOM) monitors.” Does this mean that the Phoenix JLG Supersite has both BAM and a TEOM monitor?
Line 107-108: The authors mention: “For comparison, we also used hourly PM2.5 reference data from a reference-grade optical instrument – T640x (Teledyne API, San Diego, U.S.), which was deployed at West Phoenix from November 2018 until April 2021.” Could you clarify what specific comparison the authors are referring to in this context?"
Line 119 – 121: If multiple proxy sites are equidistant from a specific PurpleAir sensor, which one did the authors choose for comparison? Were there instances where these two proxy sites showed different concentrations?
Line 128: Please specify what does MV stands for?
Line 128: Please check if “E – variance, var (MV)” should be “E, variance, var”?
Line 139-141: The author mentions: “The impact of this threshold was tested using Purple Air data from South Phoenix, West Phoenix, and Durango and calibrating these against the nearest proxy reference, to represent the impact of calibrating against a distant proxy reference.” Are the authors emphasizing that these sites had collocated sensors, but the calibration was still performed using the nearest proxy reference?
Line 161: About RH, is the PurpleAir sensor reported relative humidity or meteorological data?
Line 168: Please correct “between smoke and dust and general urban traffic PM sources” to “between smoke, dust, and general urban traffic PM sources”
Figure 2: Please specify if it is 3-day rolling MOMA gain daily average.
Line 194-195: The authors mention that “The difference between the PurpleAir sensor and the reference data varies diurnally over winter with the largest differences observed at nighttime.” Since diurnal data is not presented in the manuscript, please clarify where this information is derived from.
Figure 4: Were the species concentrations normalized by the maximum concentration? Could you please specify how the normalization was performed?
Figure 4: Did the remaining ~600 days (between July 2019 and April 2021, excluding 20+18 for Smoke and Dust) contribute to Urban days?
Figure 4: Please provide the range of PM10 concentrations from the regulatory monitor for the marked "Dust" days. A low PM2.5 /PM10 ratio does not necessarily indicate high absolute PM10 (observed during Dust events). For example, PM10 concentrations below 50 µg/m³ can also result in low PM2.5/PM10 ratios.
General comment: Should the authors accept as a limitation that their method does not allow real-time correction, unlike the EPA's method that uses the PurpleAir parameter for real-time data correction?
Citation: https://doi.org/10.5194/egusphere-2024-3454-RC1 -
RC2: 'Comment on egusphere-2024-3454', Anonymous Referee #2, 06 Mar 2025
“Seasonal effects in the application of the MOMA remote calibration tool to outdoor PM2.5 air sensors”
General Comments
The manuscript presents research work using a remote calibration method (Moment Matching, MOMA) for a network of PurpleAir PM2.5 measurements made in Phoenix, Arizona USA between July 2019 and April 2021. While the authors showed that the MOMA approach compares well with compared the performance of the MOMA approach with US EPA correction methodology in improving the accuracy of the PM2.5 data across the PurpleAir network, the presented evidence that MOMA performed better in estimating daily exceedance of smoke levels compared to the EPA correction when both methodologies were compared to exceedances derived from reference data.
Finally, the authors further showed that they could identify different PM2.5 categories corresponding to dust, smoke and urban aerosol by using information from the MOMA gains estimated from 3-day rolling average, the PM categories were assessed using reference PM speciated measurements.
Specific comments
The authors have used specific MOMA gain threshold for the PM source category identification (MOMA gain < 0.6 ≡ smoke, MOMA gain > 2.5 ≡ dust and 0.6 < MOMA gain < 2.5 ≡ urban) based on the network averaged MOMA gains derived for the PurpleAir. It would be good to add some uncertainty to criteria. I would suggest the authors not only show the average daily MOMA gains across the network (Figure 3) but also include the standard deviation of the daily MOMA gains across the network as a measure of confidence in the derived daily MOMA gains. This is particularly important as the authors have indicated in Figure 2 that there might be sites that trigger frequent MOMA calibrations like Dysart and how does this affect the statistics (mean MOMA gain) presented in Figure 3.
Technical corrections
P.2, line 41: PM2.5 size range is not defined as was done for PM1 and PM10.
- 4, line 96: add that 70% is actually ‘70% hourly percentage difference’.
- 4, line 11: authors need to specify that the number of JLG Supersites is two.
- 7, line 178: Author needs to include the period over which the “Mean uncorrected PM2.5 concentration...” was estimated over. Was it for the entire duration of the deployment?
P.9 line 201, Looking at the interpretation of Figure 3 and the MOMA gain thresholds, I would say the short periods with MOMA gains > 2.5 are mainly June and August.
P.9 line 215, “change the PM2.5/PM10 for each PM2.5 source category” to “the reference PM2.5/PM10 for each PM2.5 source category identified by MOMA”
P.10 line 224, change “source categories, and b)…” to “MOMA source categories, and b)…”
P.10 line 224, change “… b) normalized species concentrations measured for…..” to “… b) normalized species concentrations measured at two Supersites for …”
P.12 line 260, change “Nr. “ to “Number”
P.14 line 303-305, A.D. is not listed in the author list, remove? Also, C.C is missing from this list.
P.14 line 312-314, Keep the abbreviation consistent, G.S.H should read G.H. (see lines 303-305). D.E.W. is not an author in this manuscript, delete or include if missing.
Viewed
| HTML | XML | Total | BibTeX | EndNote | |
|---|---|---|---|---|---|
| 140 | 36 | 9 | 185 | 7 | 10 |
- HTML: 140
- PDF: 36
- XML: 9
- Total: 185
- BibTeX: 7
- EndNote: 10
Viewed (geographical distribution)
| Country | # | Views | % |
|---|---|---|---|
| United States of America | 1 | 91 | 52 |
| China | 2 | 10 | 5 |
| India | 3 | 10 | 5 |
| France | 4 | 9 | 5 |
| Netherlands | 5 | 8 | 4 |
| Total: | 0 |
| HTML: | 0 |
| PDF: | 0 |
| XML: | 0 |
- 1
- 91