Strategies of method selection for fine-scale PM<sub>2.5</sub> mapping in an intra-urban area using crowdsourced monitoring

Xu, Shan; Zou, Bin; Lin, Yan; Zhao, Xiuge; Li, Shenxin; Hu, Chenxia

doi:https://doi.org/10.5194/amt-12-2933-2019

Articles | Volume 12, issue 5

https://doi.org/10.5194/amt-12-2933-2019

© Author(s) 2019. This work is distributed under
the Creative Commons Attribution 4.0 License.

https://doi.org/10.5194/amt-12-2933-2019

© Author(s) 2019. This work is distributed under
the Creative Commons Attribution 4.0 License.

Articles | Volume 12, issue 5

Research article

|

28 May 2019

Research article |

| 28 May 2019

Strategies of method selection for fine-scale PM_2.5 mapping in an intra-urban area using crowdsourced monitoring

Shan Xu, Bin Zou, Yan Lin, Xiuge Zhao, Shenxin Li, and Chenxia Hu

Download

Final revised paper (published on 28 May 2019)
Supplement to the final revised paper
Preprint (discussion started on 07 Jan 2019)
Supplement to the preprint

Interactive discussion

Status: closed

AC: Author comment | RC: Referee comment | SC: Short comment | EC: Editor comment

- Printer-friendly version

- Supplement

RC1: 'review for amt-2018-402', Anonymous Referee #4, 30 Jan 2019
- AC5: 'Response to the comments RC1', Bin Zou, 05 Mar 2019
RC2: 'Review of amt-2018-402', Anonymous Referee #3, 04 Feb 2019
- AC2: 'Response to the comments RC2', Bin Zou, 04 Mar 2019
RC3: 'Strategies of Method Selection for Fine Scale PM2.5 Mapping in Intra-', Anonymous Referee #2, 04 Feb 2019
- AC3: 'Response to the comments RC3', Bin Zou, 04 Mar 2019
RC4: 'Review of Xu et al', Anonymous Referee #1, 05 Feb 2019
- AC4: 'Response to the comments RC4', Bin Zou, 04 Mar 2019
AC1: 'Response to the comments RC1', Bin Zou, 04 Mar 2019

Peer-review completion

AR: Author's response | RR: Referee report | ED: Editor decision

AR by Bin Zou on behalf of the Authors (06 Mar 2019) Author's response Manuscript

ED: Referee Nomination & Report Request started (06 Mar 2019) by Pierre Herckes

RR by Anonymous Referee #2 (21 Mar 2019)

Suggestions for revision or reasons for rejection

After major revisions I am still not convinced that this paper and the supporting research meet the standards of AMT. I think there are a number of major flaws the authors must address in order to make their paper suitable for publication.
In this revision of the paper I feel the authors have revealed a potential fatal flaw of their study:
Page 23 line 12-15: This is very surprising that volunteers just looked at their sensors 3 times in an hour and reported the value to the app. So the concentrations are not hourly averages from each monitor. This makes me question the usefulness of this study. How often does the screen update with the concentration? Is it a 1-minute average? Rolling average of the past hour? Something else? If you pulled 3 random minutes of data from an hour you would expect they might be significantly different than the average for the full hour.
Were all the performance metrics calculated in the same way just reading 3 random minutes off the screen? If not the performance metrics (Figure 1 and +/- 20% reported by Tsinghua) are not really good indicators for the expected accuracy of this work. I think the implications of this sampling methodology on the results need to be discussed.
See Zheng et al. 2018 for discussion of averaging time and low-cost sensor performance: https://www.atmos-meas-tech.net/11/4823/2018/
My other major concern is the overestimate by the sensors during period 1, and the over estimate of the sensors during period 2. During one of the evaluations periods the RH was 94% at maximum while during one of their sampling campaigns the RH was reported from 95% to 98%. Much previous work has shown that nephelometer type devices greatly over estimate during high RH events and in a nonlinear way (including: https://www.atmos-meas-tech.net/11/4823/2018/). There is no evaluation data at the RH that occurred during this project and they present no figures showing the relationship between RH and sensor error and no past work showing how this sensor performs at high RH. Maybe having accurate sensor readings is not important for the main goal of this paper (generating different models) but they have tried to draw a number of large conclusions based on the higher average from the sensor data (people should stay indoors even when government monitors say it is clean out, pollution at road level is higher). I have a similar concern during the second period where the monitors underestimate since they have presented no evaluation data and included no references that show that the sensor provides a linear response above 160 ug/m3. Low-cost sensors may show non-linear responses (http://www.aaqr.org/article/detail/AAQR-17-10-OA-0418) and it is very important to evaluate them under the conditions they will be evaluated under. Any supporting literature the authors can find about the monitor they used or the internal sensing component may support their findings but currently the authors have not cited any previous work with their monitor.
Additional comments:
Figure 1. It’s not clear to me what is on the right scatter plot versus the left? Does each plot show the data from all three stations? How was the data recorded for these tests? Since for the experiment it was read off the screen is this also what was done for these? What is the averaging time for the points shown on this plot. It would be helpful to label them as c and d so you can reference them in the text.
Page 22 line 4: please define small difference.
Abstract line 20: Not sure what all the numbers are in the parenthesis
Abstract Line 16: suggest mentioning how many hours each period was.
Abstract line 22: I don’t think you can make any generalizations based on this dataset suggest: “During this project”, OK interpolation performs… or something similar
Page 20 line 24: What do you mean by “fine” exposure control?
Page 21: Line 9 and 10: suggest including a citation about underestimation/mis calculation of risk
Page 21: Lines 22-27: These results from Tsinghua are super important since you are reporting data up to 260 ug/m3 but you only have a comparison range up to 160 ug/m3. Please include a citation or a figure in the supplement. Is this an ambient comparison or in the lab?
Page 21 Lines 29-30: Is this conversion factor calculated by you or by the company?
Page 21 line 25: At what averaging time is this +/- 20%
Page 27 line 32: Suggest citing previous literature that has seen significant gradients from 2 to 15 m.
Page 25 Line 15: I don’t think it’s helpful/scientific to state “Generally, the statistics differed”. I think this is not needed
Page 27 Line 20-22: To explore the spatial variation in the… This sentence is unclear I don’t understand any point you are trying to make
Page 28 line 14: You can’t say this “guaranteed” their reliability
Page 28: Lines 4-7: I think this is too large of a finding based on the results you have presented here

Referee Report: PDF

Hide

ED: Reconsider after major revisions (21 Mar 2019) by Pierre Herckes

AR by Bin Zou on behalf of the Authors (12 Apr 2019) Author's response Manuscript

ED: Referee Nomination & Report Request started (17 Apr 2019) by Pierre Herckes

RR by Anonymous Referee #2 (30 Apr 2019)

Suggestions for revision or reasons for rejection

After a second round of major revisions I still think the authors have some flaws and inconsistencies in their discussion that need to be addressed. In the last draft I had 2 major concerns, 1) the sensor “hour averages” are actually a few much shorter time periods averaged and assumed to be equivalent to the 1-hour measurement and 2) that the results from the high RH period may be biased high.
The first concern remains unclear as they now report that 9-12, 1-second sensor measurements were used to calculate a 1-hour average, but this is not described consistently through the paper where initially they are reported as 1 second measurements and then they are referred to as 1-minute measurements for the rest of the paper.
Essentially in this draft the authors claim they will not correct for RH because although it will bias their measurements it should be a uniform bias across the study area. I think that this is acceptable.
Please see my specific comments below.
Major concerns:
In this draft the authors reveal that the sensor has 2 operating modes. During evaluation sensors collected a minute of data and then slept for 5 minutes. During the crowdsourced collection the screen updates concentrations every second and saves 1-second data. This means that a few 1-second measurements were used for the crowdsourcing campaign. The authors have not provided us reason to believe taking 10-minutes of measurements per hour as was done during the evaluation stage would be comparable to 9 seconds of measurements during the crowd sourced campaign.
Specific concerns:
Page 10 line 26: The link provided is not accessible (“403 Forbidden error”)
Figure S1b: What is X and Y in the equation? A quick check shows that the equation printed on the plot is not the equation printed (y=-.001x+1.508 at 40% y should be 1.45 and it is <1.1)
Page 12 line 18: You say 3 minutes but on page 10 line 20-21 you say that the measurement updates and automatically stores every second. So, this means they are recording 3 one-second readings not 3 one-minute readings.
Page 17 line 11: Again, this contradicts page 10 line 20-21.
Page 17 line 14: But if one is second measurements and one is minute measurements this may not be a good comparison
Editorial comments:
Page 17 line 7: should be monitors (not monitor)
Page 18 line 2: Suggest using lightly polluted instead of light polluted so it is clear you aren’t talking about light pollution from lights or the sun.

Hide

ED: Publish subject to minor revisions (review by editor) (01 May 2019) by Pierre Herckes

AR by Bin Zou on behalf of the Authors (09 May 2019) Author's response Manuscript

ED: Publish as is (09 May 2019) by Pierre Herckes

AR by Bin Zou on behalf of the Authors (10 May 2019) Author's response Manuscript

Short summary

This study presents strategies of method selection for 100 m scale PM_2.5 mapping using a crowdsourced sampling campaign. Interestingly, PM_2.5 concentrations in micro-environments varied significantly in intra-urban areas. These local PM_2.5 variations can be effectively revealed by crowdsourcing sampling rather than national air quality monitoring sites. The selection of models for fine-scale PM_2.5 mapping should be adjusted with the changing sampling and pollution circumstances.

Strategies of method selection for fine-scale PM2.5 mapping in an intra-urban area using crowdsourced monitoring

Download

Interactive discussion

Peer-review completion

Strategies of method selection for fine-scale PM_2.5 mapping in an intra-urban area using crowdsourced monitoring